Embed Size (px)
Citation preview
Reply Attention of: Matthew D. KeenDirect Phone: 604.641.4913 Direct Fax: 604.646.2551 E-Mail: [email protected]
Our File: 15-4171 Date: February 3, 2016 BY E-MAIL
British Columbia Utilities Commission 6th Floor – 900 Howe Street Vancouver, BC V6Z 2V3 Attention: Erica Hamilton, Commission Secretary
Dear Sirs/Mesdames:
Re FortisBC Energy Inc. Common Equity Component and 2016 Return on Equity Association of Major Power Customers of BC (AMPC), British Columbia Old Age Pensioners’ Organization et. al. and the Commercial Energy Consumers Association of British Columbia (Utility Customers) Revised Evidence of Dr. Booth
We are legal counsel to AMPC in this matter, and write on behalf of AMPC and the Utility Customers. After filing his evidence, Dr. Booth discovered that he had misread a table. This error does not affect his overall conclusions, but does affect some of the underlying calculations, and he has therefore corrected his evidence.
Accordingly, please find enclosed the corrected evidence of Dr. Booth, as well as a blackline comparing it to the version filed on January 26 that clearly shows where the changes have been made. The blackline document does not contain appendices, as no changes were made in those sections.
Again, bookmarks have been embedded in the document to assist with navigation.
Please contact the undersigned if you have any questions.
Bull, Housser & Tupper LLP
Matthew D. Keen
;
311 Pages C7-7-1
FAIR RETURN AND CAPITAL STRUCTURE FOR FORTIS
BC ENERGY
EVIDENCE OF
Laurence D. Booth
BEFORE THE
BC Utilities Commission
JanuaryCorrected February 3, 2016
TABLE OF CONTENTS
EXECUTIVE SUMMARY 1
I INTRODUCTION AND OVERVIEW 3
II FINANCIAL AND ECONOMIC OUTLOOK 8
III RISK PREMIUM ESTIMATES OF THE FAIR ROE 36
IV DCF ESTIMATES OF THE FAIR ROE 50
V CAPITAL STRUCTURE & BUSINESS RISK 64
VI FINANCING AND CONCLUSIONS 81
APPENDIX A: PROFESSOR BOOTH’S CURRICULUM VITAE
APPENDIX B: MARKET RISK PREMIUM ESTIMATES
APPENDIX C: RELATIVE RISK FOR A BENCHMARK UTILITY
APPENDIX D: DCF ESTIMATES
APPENDIX E: ALTERNATIVE MODELS TO THE CAPM
i
1
EXECUTIVE SUMMARY
1. The Canadian economy has stalled somewhat as resource prices have collapsed following
a slowdown in China and with them the value of the Canadian $. This has hurt the resource
intensive sectors like Alberta, but added extra stimulus to others, in particular, Ontario and
Quebec. As the economy adjusts to this change short term growth has weakened, but is expected
to get back on track in 2017. The BC economy is one of the strongest in Canada in terms of
economic growth.
2. Although the US Federal Reserve has stopped its bond buying program (quantitative
easing or Operation Twist), others are still very active and there is now a huge block of
government bonds that have been taken off the bond market increasing prices and lowering
yields from where they would otherwise have been. This has led to continuing low long Canada
bond yields causing corporate credit spreads over government bond yields to remain high at 191
bps. This is mainly due to unusually low government bond yields not higher credit risk, since all
the standard stress indicators show normal capital market conditions.
3. Canadian utilities have issued 40 and in some cases 50 year bonds at extremely low
interest rates and FEI recently issued 30 year MTNs at a little over 1.35% spread over equivalent
maturity long Canadas..
4. I have been recommending a credit spread adjustment of 50% of the change in the credit
spread from the normal spread of 100bps to my ROE recommendations. This adds 45 bps to
simple CAPM estimates, but should even out over the business cycle. I regard this as simply
converting the simple CAPM into a Contingent CAPM or CCAPM as required by theory.
5. However, the credit spread adjustment does not adjust for the overall drop in bond yields
as partly indicated by the behaviour of preferred share yields. Since the US embarked on its third
round of quantitative easing the traditional spread of preferred share yields over both
government of Canada and generic A rated corporate bonds has changed. This indicates that
current long Canada bond yields are at least 1.30% too low based on traditional, spreads
confirming my real bond yield model that indicates an even higher 2.5%.
6. For 2016 I continue to recommend an ROE of 7.50% for a benchmark utility. This
recommendation includesreflects a 0.45% adjustment for credit spreads and, 1.3% for the impact
of global bond buying programs and evidence from DCF equity costs. For 2017 and later years I
recommend the BCUC continue with its ROE adjustment model with the 3.80% floor to the long
Canada bond yield. However, I do not expect this floor to be triggered over the next three years.
7. I accept FEI’s broad assessment that its business risk is largely unchanged since 2012,
but would point out there are signs of a marginal decrease in risk. Consequently, I see no reason
to change my standard common equity ratio recommendation of 35% which is slightly below a
comparator benchmark of 37%.
8. I disagree with the company’s assessment that it might have problems with the interest
coverage ratio (ICR) restriction in its bond indenture and therefore capital market access
problems. Its ICR analysis is deficient and with its own filing with securities regulators its recent
ICR was over 2.2. Currently I judge there to be no capital market or access problems for high
quality utilities and given the state of the financial system FEI should have no problems
financing itself with my recommended financial parameters.1
9. I would recommend no adjustment for the amalgamation of FEI which I do not judge to
materially alter its risk or financial parameters. This conclusion is in line with the decision of
the rating agencies, which seem to regard the amalgamation as not material.
2
1 TD’s “action note” of May 6, 2015 notes that FEI’s ultimate parent Fortis has an ROE of 7.10% and an investment grade bond rating.
I INTRODUCTION AND OVERVIEW
Q. PLEASE DESCRIBE YOUR NAME, QUALIFICATIONS AND EXPERIENCE.
A. Laurence Booth is a professor of finance in the Rotman School of Management at the
University of Toronto, where he holds the CIT Chair in Structured Finance. Dr. Booth appeared
before the BCUC most recently in a 2012 “generic” hearing, as well he has appeared before most
of the major utility regulatory boards in Canada including the CRTC, the National Energy Board,
the Ontario Energy Board (OEB), the Alberta Utility Commission (AUC), the Nova Scotia
Utilities and Review Board, the New Brunswick Public Utilities Board, the Manitoba Public
Utilities Board, the Board of Commissioners of Newfoundland and Labrador, the Regie de
l’Energie du Quebec and the Prince Edward Island Regulatory and Appeals Commission. He has
also filed testimony before the Ontario Securities Commission and in a variety of civil suits
pertaining to financial matters. A detailed resume is filed as Appendix A. Further information
and copies of working papers by Dr. Booth can be can be downloaded from his web site at the
University of Toronto at http://www.rotman.utoronto.ca/~booth.
Q. PLEASE DESCRIBE THE PURPOSE OF YOUR TESTIMONY
A. The Association of Major Power Customers of BC (AMPC), the British Columbia Old
Age Pensioners Organization (BCOAPO) and the Commercial Energy Consumers Association of
British Columbia (CEC), collectively referred to as the Utility Customers asked me to offer an
opinion on the fair rate of return on common equity (ROE), and to recommend an appropriate
common equity ratio for FortisBC Energy Inc., (FEI) for the BCUC to use as benchmarks for BC
utilities.
Q. DO YOU HAVE SOME OVERALL REMARKS?
A. Yes. It is three years since the BCUC last reviewed its benchmark ROE and the capital
structure of FEI. Over this period we have very much been in a holding pattern waiting for the
US and Europe to work their way through the after effects of their Great Recessions. Further,
just as they have proceeded to recover, Chinese growth has slowed and sent jitters through
commodity markets and triggered a stock market correction. The upshot is that the stronger
3
markets that were expected at that time have not developed as anticipated and Alberta, in
particular has been badly hit. Further conditions in the bond market have become even looser
than they were in 2012 as the massive amount of liquidity in global markets continues to
increase, depressing bond yields. The result is that the forecast long Canada bond yield has yet to
hit the 3.8% trigger that I recommended as justifying a change in FEI’s allowed ROE.
In terms of FEI’s business risk the major factor is the amalgamation of FEI with its Whistler and
Vancouver Island subsidiaries. This amalgamation has removed most of the additional risk
factors faced by those companies, such that, as even FEI admits, FEI’s business risk post
amalgamation is similar to that of FEI in 2012 (Appendix C, page1). Although FEI suggests that
this risk is “trending higher” the fact is that in their business risk assessment (Appendix C, page
3) 19 of the 24 categories are ranked the same as in 2012. The only areas where FEI judges
business risk to have increased is in energy price volatility, political initiatives on greenhouse
gas (GHG) emissions, aboriginal rights, and the security of energy supply. In my judgement,
none of these “higher” risk factors are substantial enough to justify an increase in the common
equity ratio to 40%.
The upshot is that with largely stand-pat financial markets and business risk, FEI’s case for an
increase in allowed ROE and common equity is non-existent. In the 2012 GCOC hearings FEI
presented a large number of expert panels and the BCUC thoroughly evaluated both the allowed
ROE and FEI’s capital structure. In contrast, in the current hearing FEI has only presented one
external witness and requested an abbreviated hearing on financial matters be held over a very
short time period. The obvious implication is that FEI judges that not much has changed as well,
and what changes there are would indicate a lowered, not increased ROE. It is therefore
appropriate to first consider what in fact has changed since 2012.
Q. CHANGES SINCE 2012 GCOC HEARING
A. I will present my own evidence shortly, but it is first useful to present some information
that FEI’s witness (Concentric) provided in a public document circulated by the Canadian Gas
Association.2 This is the data that Concentric regarded as most pertinent as a backdrop to the
financial parameters for regulated utilities. The data is as of May 2015 so is not the most recent,
4
2 Canadian Gas Association, Vol 111, May 1, 2015.
but as I will show it is quite similar to current data. Note that in 2013 when the GCOC decision
was made the average Government of Canada long term bond yield (LTC) averaged 2.82% and
by May 2015 it had dropped to 2.05% or decreased by 0.77%. Similarly, US government
Treasuries had dropped from 3.45% to 2.55% or by 0.90%. This drop in bond market interest
rates also affected utility borrowing costs. In Canada, the Bloomberg synthetic A rated utility
borrowing cost dropped from 4.24% to 3.50%, or by 0.74%, to all intents and purposes the same
decline as the LTC yield. Similarly, the Moody’s A rated utility bond yield dropped from 4.28%
to 3.67% or by 0.81%, slightly less than the drop in US Treasuries.
Why this is important is that FEI accepts3 that while it does not recommend an ROE adjustment
formula it does accept the two factor model adopted by the BCUC in its 2012 GCOC
proceedings. In this instance the long Canada bond yield decrease of 0.77% indicates a 0.385%
decline in the allowed ROE, while the credit spread has shown a trivial change from 1.42%
1.45%, suggesting a 0.015% increase in the allowed ROE. Overall, the BCUC’s two factor
model and Concentric’s data would suggest a 0.37% decrease in FEI’s allowed ROE from
8.75% to 8.38%.
The BCUC’s ROE adjustment model would not result in a 0.38% decrease in the allowed ROE
since it is subject to a 3.8% floor. However, it does indicate the direction in FEI’s allowed ROE
and that this should be down from the current 8.75%, not up to the 9.50% recommended by
Concentric. Further, Concentric provided the following table of currently allowed ROEs for
Canadian gas companies.
5
3 Main evidence page 33.
I regard the premier local gas distribution companies in Canada as being Gas Metro, Union Gas,
Enbridge Gas Distribution, ATCO Gas and FEI. These are the largest gas distribution companies
in Canada and dominate their provincial markets. Traditionally, I have regarded a risk ranking as
being ATCO Gas and EGDI as lowest risk with Union and FEI marginally higher and Gaz Metro
as the highest risk. This ranking is mainly based on their size, the diversity of the provincial
economies and the competitive nature of natural gas in each province. This assessment is broadly
in line with that of Ms. McShane who testified on behalf of FEI in the 2012 GCOC hearings.
Currently, the Alberta generic ROE is 8.30% and ATCO Gas is allowed 8.30% on 38% common
equity where 1% of that is a generic premium for market conditions. EGDI is allowed 9.30%,
since its ROE is based on a 2009 formula that the OEB has not recalibrated, and Union Gas
8.93%, both with 36% common equity.4 Gaz Metro is allowed 8.90%, but that includes an
additional 0.25-0.35% risk premium to reflect Gaz Metro’s risk, while the Regie’s current base
6
4 For 2016 EGDI’s ROE has dropped to 9.19% based on a forecast LTC rate of 2.706% and a 1.83% credit spread. OEB Letter, October 15, 2015. Note the Ontario government is in the process of selling off a significant part of Hydro One Transmission (HONI) which earns the OEB allowed ROE and which may explain the delay in reviewing the OEB ROE methodology. Shortly after the initial sale of HONI shares, they were selling at $21.86 whereas the book value is $15.76, that is, the market to book ratio was 1.39. This implies that the market found the HONI allowed ROE very attractive and excessive for such a low risk pure play utility.
allowed ROE for electric distribution and transmission is 8.20%. Gaz Metro’s allowed common
equity ratio is 38.5%.
My conclusion from this is that since the GCOC hearing the macroeconomic evidence of
Concentric, and the judgement of FEI on the two factor ROE adjustment model, together
indicate that the allowed ROE for FEI should be lowered. Further, that FEI’s requested financial
parameters of 9.5% ROE on 40% common equity exceed that of its peers. The more recent
litigated hearings in Quebec, as well as that in Alberta, indicate a generic ROE of 8.20-8.30%
for an energy distributor applied to a 35-38% common equity ratio.5
I develop my own recommendations in the subsequent sections. First, I consider the general
financial and economic outlook, since this is what underpins the current GCOC formula ROE.
Second, I consider fair ROE estimates derived from risk premium and discounted cash flow
(DCF) models. Here it is important to point out that the recommendation is for 2016 and forecast
long Canada rates are not expected to get close to the levels that Mr. Coyne forecasts and uses in
his CAPM estimates. Finally I discuss FEI’s business risk and the question of financial access
and bond ratings, and make my recommendations.
I have relegated most of the more detailed and technical discussion to a series of appendices. A
includes my resume; B has a discussion of the market risk premium; C has a relative risk
assessment for a benchmark utility; D discusses DCF estimates and E is a new Appendix that
discusses alternative risk premium models to the CAPM motivated by the 2012 GCOC decision.
I have not repeated the “comparable earnings” appendix since unlike Ms. McShane, Concentric
has not produced such evidence. These appendices are written to be free standing references
from the main text and are mainly standard updates from 2012.
II FINANCIAL AND ECONOMIC OUTLOOK
Q. WHY DO YOU START BY CONSIDERING CAPITAL MARKET
CONDITIONS?
A. Because the legal standard for a fair rate of return stemmed from “altered conditions in
the money market” where we would now understand the money market to mean the capital
7
5 Hydro Quebec Distribution is allowed 35% common equity.
8
market. The Supreme Court of Canada determined a fair rate of return in BC Electric Railway
Co Ltd., vs. the Public Utilities Commission of BC et al ([1960] S.C.R. 837), where the Supreme
Court of Canada had to interpret the following statute:
The Commission shall consider all matters which it deems proper as affecting the(a)
rate:
The Commission shall have due regard, among other things, to the protection of(b)
the public interest from rates that are excessive as being more than a fair and
reasonable charge for services of the nature and quality furnished by the public
utility; and to giving to the public utility a fair and reasonable return upon the
appraised value of the property of the public utility used, or prudently and
reasonably acquired, to enable the public utility to furnish the service:
This statute articulated the "fair and reasonable" standard in terms of rates, and that the
regulatory body should consider all matters that determine whether or not the resulting charges
are "fair and reasonable." To an economist, "fair and reasonable" means minimum long run
average cost, since these are the only costs which satisfy the economic imperative for regulation
and by definition do not include unreasonable and unfair cost allocations. The statute also
articulated the “prudently and reasonably acquired” test in terms of the assets included in the rate
base.
Most statutes also allow the regulatory authority to examine all factors that enter into the rates to
ensure that the rates are “fair and reasonable.” This includes the firm’s capital structure decision,
since this has a very direct and obvious impact on the overall revenue requirement. To allow the
regulated utility to freely determine its capital structure will inevitably lead to rates that are
unfair and unreasonable, otherwise the management of the regulated firm is not fulfilling its
fiduciary duties to act in the best interests of its stockholders.
In terms of financial charges, in Northwestern Utilities vs. City of Edmonton (1929), it was
stated that a utility's rates should be set to take into account ‘altered conditions in the money
market.’ A fair rate of return was further confirmed in the BC Electric decision when Mr. Justice
Lamont's definition of a fair rate of return, put forward in Northwestern utilities, was adopted:”
9
"that the company will be allowed as large a return on the capital invested in theenterprise as it would receive if it were investing the same amount in othersecurities possessing an attractiveness, stability and certainty equal to that of thecompany's enterprise."
This definition is referred to as an opportunity cost, in that the fair return is what could be earned
by investing in similar securities elsewhere; only if the owners of a utility earn their opportunity
cost will the returns accruing to them be fair, i.e., they will neither reward the owners with
excessive profits, nor ratepayers by charging prices below cost.
To any modern financial economist Mr. Justice Lamont’s definition of a fair rate of return as an
opportunity cost means a risk adjusted discount rate or expected rate of return. This is the rate
that is determined in the capital market as conditions constantly change.
Q. HOW HAS MONETARY EVOLVED IN CANADA?
A. Basic macroeconomic data since 1987 is provided as background in Schedule 1. Into
2008 we had good economic growth and for a time the unemployment rate was actually below
the natural or non-accelerating inflation rate of unemployment (NAIRU) of 6.0%. Consumer
spending was strong as low interest rates supported the purchase of consumer durables and new
housing as starts exceeded 200,000 for the sixth year in a row. The strong investment position in
Canada was partly due to a dramatic improvement in Canada’s terms of trade as commodity
prices increased. This created a perception that Canada was a “petro,” or at least a “raw
materials” based, economy as commodity prices reached record highs in summer 2008. This
perception allied to the continuing strength of the current account surplus resulted in a
strengthening Canadian dollar and incipient inflationary pressures. The result was that starting in
September 2005 the Bank of Canada increased its overnight rate from 2.5% to reduce the
stimulus being injected into the economy.
The following graph shows the impact of this tighter monetary policy from January 2006, just
before the first signs of the financial crisis appeared. Throughout 2006 and up until December
2007, the Bank of Canada set the target rate to try and slow down the economy and reduce
inflationary pressures. Of importance is that consistent with the Bank of Canada’s 2% inflation
target the overnight rate should be at least 3.0%; so 4.5% up until December 2007 was
restrictive. The Bank pays interest on deposits that the chartered banks keep with it at 0.25% less
than the overnight rate and the banks can borrow at 0.25% more than the overnight rate; a rate
that is called the Bank Rate. Bank Prime is then about 2.0% more than the overnight rate.
Consequently up until December 2007 the Bank was actively trying to increase borrowing costs
to slow interest sensitive demand. This policy stance was reversed due to the impact of the
sub-prime mortgage crisis coming out of the United States.
Canadian Overnight RateCansim 39079
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
1/2/
2006
5/2/
2006
9/2/
2006
1/2/
2007
5/2/
2007
9/2/
2007
1/2/
2008
5/2/
2008
9/2/
2008
1/2/
2009
5/2/
2009
9/2/
2009
1/2/
2010
5/2/
2010
9/2/
2010
1/2/
2011
5/2/
2011
9/2/
2011
1/2/
2012
5/2/
2012
9/2/
2012
1/2/
2013
5/2/
2013
9/2/
2013
1/2/
2014
5/2/
2014
9/2/
2014
1/2/
2015
5/2/
2015
9/2/
2015
The above graph shows that the Bank conservatively lowered the overnight rate to 3.0% in May
2008 and it kept it there throughout the summer. It was then forced to dramatically and rapidly
cut the overnight rate to 0.25% in response to the financial crisis triggered by the failure of
Lehman Brothers. 0.25% has traditionally been thought to be the lowest the Bank can set the
overnight rate, since otherwise it would mean negative deposit rates for the settlement balances
the chartered banks keep with it.6
10
6 The European Central bank (ECB) currently pays negative interest on settlement balances, so there is now a history of this actually happening so the overnight rate could be cut below 0.25%.
Canada recovered quickly from the “Great Recession” coming out of the United States and the
Bank of Canada started “normalising” by increasing the overnight rate in June 2010 as there
were obvious signs of recovery in the Canadian economy. The Bank of Canada increased the
overnight rate on three separate occasions each time by 0.25% to bring it to 1.0% and with it
Prime increased to 3.0%. Expectations in 2011 were that the Bank would resume increasing the
overnight rate as the economy continued to strengthen as it was still below the “equilibrium” or
“normal” rate.
Increasingly the Bank of Canada and the Federal Government started to worry that at 1.0% the
overnight rate would encourage too much personal borrowing and lead to levels of indebtedness
that would have negative implications as interest rates returned to their normal level. They were,
and still are, very worried about a housing bubble in Toronto and Vancouver7 where house prices
increased strongly in response to both lower interest rates and a stronger economy. In response
on July 8, 2012 the Federal Government announced a third round of tightening in the mortgage
market for anyone with less than a 20% down payment. The restrictions include limiting the
maximum amortisation period to 25 years, capping household debt ratios, limiting CMHC
insurance to homes with a purchase price less than $1 million and the maximum on a home
equity line of credit to 65%. They also moved responsibility for Canadian Mortgage and
Housing Corporation (CMHC) to the Department of Finance, so that it would be subject to OSFI
supervision and in June 2015 issued guidelines that CMHC has to follow.8
The conundrum faced by the Bank of Canada and the Federal Government is that while it wants
to stimulate the economy by maintaining low interest rates, it does not want a US style
debt-fuelled housing bubble. The additional problem is that the Canadian economy is not an
island and increasingly the Bank of Canada has been concerned about the transfer of events from
the Eurozone and the US into Canada as both the Federal Reserve (Fed) and the European
Central Bank (ECB) continue to follow “easy” money policies due to weaknesses in their
economies.
11
7 In September 2015 housing starts increased to an annualized pace of 216,924 indicating a strong housing market in Canada.
8 See the Canadian Bankers Association web site http://www.cba.ca/en/media-room/50-backgrounders-on-banking-issues/657-changes-to-canadas-mortgage-market
We can see the impact of events outside Canada in the following graph of the capacity utilization
levels in both the Canadian manufacturing and non-farm sectors.
Capacity Utilisation
65
70
75
80
85
90
1987
-03
1988
-03
1989
-03
1990
-03
1991
-03
1992
-03
1993
-03
1994
-03
1995
-03
1996
-03
1997
-03
1998
-03
1999
-03
2000
-03
2001
-03
2002
-03
2003
-03
2004
-03
2005
-03
2006
-03
2007
-03
2008
-03
2009
-03
2010
-03
2011
-03
2012
-03
2012
-03
2013
-03
2014
-03
2015
-03
Non-Farm Manufacturing
The sharp drop in capacity utilization during the recession in the early 1990’s is evident as
Canadian firms restructured in response to the Free Trade Agreement as well as a normal
cyclical downturn. We can see this again in the slowdown after the financial crisis in 2009-2010.
In both cases there followed a normal rapid recovery out of recession and a movement towards
stabilisation. However unlike earlier periods, Canada stagnated in 2012-2013 as the recovery did
not continue apace. Instead, we were hit with the aftereffects of the Euro crisis and the slow
recovery of our major trading partner the United States. Then just as the US recovery started to
gather speed we were hit with the slowdown in China during 2015 and the dramatic drop in
commodity prices.9
The following graph shows the Bank of Canada’s commodity price index.
12
9 China’s GDP growth rate has dropped to 6.9% from just over 7%, much of this was targeted by government authorities.
13
Bank of Canada Commodity Price Index
0.00
100.00
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
1000.0019
72-0
1
1973
-09
1975
-05
1977
-01
1978
-09
1980
-05
1982
-01
1983
-09
1985
-05
1987
-01
1988
-09
1990
-05
1992
-01
1993
-09
1995
-05
1997
-01
1998
-09
2000
-05
2002
-01
2003
-09
2005
-05
2007
-01
2008
-09
2010
-05
2012
-01
2013
-09
2015
-05
We can see the strong increase in commodity prices that started in 2002 as China started to
seriously industrialize. The Great Recession in the United States in 2009 caused these prices to
collapse, but they quickly recovered until the sharp sell-off in Summer 2014 as fears of a China
slowdown proved correct. It is this drop in commodity prices that has severely affected Canada’s
resource sector and triggered a technical recession in 2015Q2. We can see this in the weakened
capacity utilization at the start of 2015 and the reaction of the Bank of Canada, which surprised
markets by cutting the overnight rate twice in early 2015 to the current level of 0.50%.
The upshot is that we continue to have spare capacity in the Canadian economy. The following
graph shows the overall unemployment rate. Although the current rate of 7.0% is low compared
to the average levels of the last thirty years, it is still higher than it was in 2007 immediately
before the financial crisis when it was below 6.0%. More to the point there is concern that the
quality of many of the new jobs created is low. We also see a pick up in the unemployment rate
over the last year in response to the weakened commodity price hit to our resource sector.
14
Canadian Unemployment Rate
4
5
6
7
8
9
10
11
12
13
1987
M01
1988
M07
1990
M01
1991
M07
1993
M01
1994
M07
1996
M01
1997
M07
1999
M01
2000
M07
2002
M01
2003
M07
2005
M01
2006
M07
2008
M01
2009
-07
2011
-01
2012
-07
2014
-01
2015
-07
Q. HOW DOES THIS RELATE TO GDP?
A. The following graph has the annual change in real GDP since 1961.
Real GDP Growth
-4.00
-2.00
0.00
2.00
4.00
6.00
8.00
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
The start date reflects the initial iteration of the 1961 GDP accounts. These annual changes
clearly show the recessions of the early 1980s, 1990’s and in 2009/10. The graph also shows the
quick recovery in Canada from that recession, particularly as compared to the FTA recession of
the 1990’s. In 2015 the Canadian economy has slowed in response to the decline in commodity
prices. The following graphic from the Bank of Canada illustrates the differential impact of the
decline in commodity prices.10
There is a serious retrenchment going on in both the oil and gas sector and non-energy
commodity price related sectors of the economy. However, the rest of the Canadian economy is
chugging along with 1.4% growth, which the Bank indicates is 83% of GDP. The main reason
for this is the decline in the value of the C$. The following graph shows the value of the C$ in
terms of US currency back to 2000.
15
10 Monetary Policy Report, October 2015.
FX Rate (US:CDN)
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
2000
M01
2000
M07
2001
M01
2001
M07
2002
M01
2002
M07
2003
M01
2003
M07
2004
M01
2004
M07
2005
M01
2005
M07
2006
M01
2006
M07
2007
M1
2007
M7
2008
M01
2008
M07
2009
M1
2009
M7
2010
M1
2010
M7
2011
M01
2011
M07
2012
M1
2012
M7
2013
M1
2013
M7
2014
M01
2014
M07
2015
M01
2015
M07
The reason for going back to 2000 is to show the impact of the commodity price increases
consequent on China’s industrialisation. Back in 2002 the C$ was worth just 65 cents, but with
the dramatic increase in commodity prices it went to a premium to the US$ and peaked at a
monthly high of 1.06. However, since the slowdown in China and the US recovery it has
dropped to under 71 cents at the start of 2016. In turn this has stimulated the manufacturing
industries of central Canada that previously found it very difficult to compete with a C$ at par
with the US$.
The Bank of Canada expects that this adjustment to low commodity prices will continue to exert
a drag on Canadian growth in 2015 and 2016Q1. However, the Bank also expects the growth rate
to pick up to reach 2.6% by 2017Q1 and then for both the US and Canadian economies to grow
at about that rate.
Q. WHAT IS YOUR OUTLOOK FOR INFLATION?
A. The Bank of Canada’s 2.0% target rate of inflation was renewed with the Government of
Canada in the Fall of 2011 as part of a five year pact. The Bank of Canada increases the
overnight rate when it judges the forecast core inflation rate11 to be above this target and likely to
16
11 Core inflation is the all items CPI minus energy and food.
17
go to the top of its 1.0-3.0% operating band. Conversely, it drops the overnight rate when it fears
that inflation will drop to the bottom of its range and as a result it needs to stimulate the
economy. The inflation rate data in Schedule 1 clearly shows the inflationary pressures in 2008
prior to the recession as well as the dramatic drop in 2009 and recovery in 2010.
Since 1991, the Federal Government has been issuing two types of bonds: a nominal bond where
the interest rate is fixed and a real return bond, which guarantees the investor protection from
inflation. The difference between the nominal yield and the yield on the real-return bond is
called the break-even inflation rate (BEIR), since if actual inflation is higher than this after the
fact you would have been better off in the real bond and vice versa. Consequently the BEIR is
often taken as one measure of the market’s inflationary expectations. The following graphs the
BEIR (as a %) since 1991.
Break-Even Inflation Rate (BEIR)
0
1
2
3
4
5
6
1991
M11
1992
M05
1992
M11
1993
M05
1993
M11
1994
M05
1994
M11
1995
M05
1995
M11
1996
M05
1996
M11
1997
M05
1997
M11
1998
M05
1998
M11
1999
M05
1999
M11
2000
M05
2000
M11
2001
M05
2001
M11
2002
M05
2002
M11
2003
M05
2003
M11
2004
M05
2004
M11
2005
M05
2005
M11
2006
M05
2006
M11
2007
M5
2007
M11
2008
M5
2008
M11
2009
-05
2009
-11
2010
-05
2010
-11
2011
-05
2011
-11
2012
-05
2012
-11
2013
-05
2013
-11
2014
-05
2014
-11
2015
-05
2015
-11
We can clearly see the collapse in inflationary expectations in the late 1990’s as the market
finally believed the Federal Government’s intentions not to inflate its way out of its deficit
problems. Since then the BEIR has generally been slightly above the middle of the Bank of
Canada’s operating range for inflation of 2.0%, but never above the 3.0% upper limit set by the
Bank. We can also see the impact of the traumatic events of 2008Q3 when the BEIR dropped
from its “normal” level of just above 2.0% to 1.26% in November 2008. During this period the
18
fears of a deep recession and deflation were so strong that the BEIR essentially halved in the
space of a few months. These deflationary fears subsided once economic growth got back on
track, but have dropped recently as the decrease in commodity prices has rippled through the
economy. The Bank of Canada estimates that through 2015 the actual rate of inflation will be
just above 1.0%, while core inflation that strips out the impact of these volatile items will be
2.1% or slightly above its 2.0% target. Going forward the Bank estimates inflation both core and
actual at 2.0%, whereas the BEIR is more bearish.
Q. WHAT HAS BEEN THE RECENT HISTORY OF THE LONG CANADA BOND
YIELD?
A. Schedule 2 provides data on the full range of interest rates across the broad maturity
spectrum as of January 8, 2016. The interest rate on the 30 year Government of Canada bond at
2.07% is 1.62% higher than the yield on 91 day Treasury Bills. This is referred to as a normal
yield curve as typically this yield difference or yield spread is about 1.25%. However, currently
it is marginally higher indicating that the market expects short term interest rates to increase in
the future as the Canadian economy improves and with it the demand for loans so that the Bank
of Canada no longer follows a stimulative monetary policy.
Normally yields on long term Canada (LTC) bonds are not as affected by current monetary
policy as short term interest rates, since monetary policy works at the short end of the yield curve
via the overnight rate; its influence then weakens as the maturity of the bond increases. The
following graph shows that the LTC yield stayed at about 4.5% from 2005 until December 2007,
when the Bank of Canada started to cut interest rates after which it stayed at around 4.0% until
November 2008 when it dropped by 0.50%, as the market began to understand the severity of the
recession. However, as these fears receded the LTC yield recovered to the 4.0% level it was at
immediately prior to the financial crisis. The expectation through 2011 was that long Canada
bond yields would increase as the economy recovered, consistent with the Bank of Canada
increasing the overnight rate.
19
Canadian Interest Rates
0
1
2
3
4
5
6
7
2000
M01
2000
M05
2000
M09
2001
M01
2001
M05
2001
M09
2002
M01
2002
M05
2002
M09
2003
M01
2003
M05
2003
M09
2004
M01
2004
M05
2004
M09
2005
M01
2005
M05
2005
M09
2006
M01
2006
M05
2006
M09
2007
M1
2007
M5
2007
M9
2008
-01
2008
-05
2008
-09
2009
-01
2009
-05
2009
-09
2010
-01
2010
-05
2010
-09
2011
-01
2011
-05
2011
-09
2012
-01
2012
-05
2012
-09
2013
-01
2013
-05
2013
-09
2014
-01
2014
-05
2014
-09
2015
-01
2015
-05
2015
-09
LTC T. Bill (91 d)
Note that when we look at interest rates, although the normal yield spread is about 1.25%, it
narrows when the central bank is trying to slow down the economy by increasing the overnight
rate and increases when it is trying to stimulate the economy. We can see, for example, the very
narrow yield spread at the end of 2007 and the dramatic increase during the 2008/9 financial
crisis. However, in 2011/12 as we headed into the last hearing short term rates had already
increased and the yield spread had decreased. For example, the Royal Bank of Canada (Financial
Markets Monthly June 3, 2011) had the following forecast in June 2011.
20
At that time RBC was forecasting that long term Canadian interest rates would increase from the
then 3.75% level to a “normal” level of 4.55% and US rates from 4.55% to 5.05%. This did not
happen!
In 2011Q4 the US Federal Reserve embarked on the most dramatic third round of “quantitative
easing” (QE3), which was simply the central bank buying bonds with freshly minted money: the
more bonds they buy the higher the bond prices and the lower the interest rate. QE3 amounted to
an open-ended commitment by the US Federal Reserve to buy $85 billion of US government
bonds and Federal Agency backed mortgages every month while maintaining the Federal funds
rate at 0%. As we can see from the previous graph of interest rates the effect was a significant
drop in long term Canada yields as investors hunted for higher yields in other markets than the
US government bond market.
QE3 started to run down in 2013Q3 and again the market expected interest rates to recover.
However, although QE3 ended in the US in October 2014 the recovery in interest rates has still
not occurred. Instead long term interest rates have essentially been flat through 2015. There are
two main reasons why interest rates have still not recovered in Canada. The first is that the US
Federal Reserve has accumulated at least $3.5 trillion of securities. The following graphic from
the Federal Reserve Bank of Cleveland shows the growth in the Fed’s balance sheet.
21
As it buys securities it pays for them with US money and by the end of 2014 it had bought about
$3.5 trillion of mainly US government and agency (mortgage) bonds. This is $3.5 trillion of
securities that have been issued, but are not in the public markets: if they were the huge increase
in supply would have depressed prices and increased interest rates. Although this bond purchase
program has now ended there is still this $3.5 trillion that is not in the market.
This is like filling a bath (liquidity): the US Fed may have turned off the taps so the bath is no
longer filling, but the bath is already incredibly full. What this means is that although QE3 has
stopped, its impact will last as long as the Fed does not sell its $3.5 trillion hoard of securities.
Further as the US government has got closer to balancing its books, and reducing its deficit and
need to issue bonds, the supply of US government bonds has also dropped. The problem now is
how to drain the bath and here the actions of the Federal Reserve is critical. The bath will be
drained as the Fed sells government securities, thereby depressing their price and increasing
interest rates. In December 2015 the Federal Reserve did increase its target rate by 0.25% and
also indicated that there will be four more similar increases in 2016 to move the Federal Funds
rate to 1.50% by the end of 2016. However, this policy change is directed at the short end of the
yield curve, not the long end. The Fed seems to be in no hurry to run down its balance sheet by
22
selling its longer term bond holdings, if it did, long term interest rates would increase
dramatically.
The second reason is that the Fed is not the only central bank to inject money into the global
financial system or buy long dated government securities. The Bank of England is in no hurry to
increase its target rate and even though it stopped its QE program in July 2012, and like the Fed
it is in no hurry to sell its holdings of UK government bonds. Further the European Central Bank
did not start serious quantitative easing until January 2015 and currently has a negative target for
its policy rate. Finally in October 2014 Japan announced its own expanded QE program.
The result is that while the US and UK baths have stopped filling up, but are almost full, there
are baths filling up in Europe and Japan. The result is twofold: the supply of liquidity (money)
used to buy the bonds has enormously increased, while the supply of bonds has decreased, since
trillions have been taken off the market by central banks. Canada has been a by-stander in all this
as the Bank of Canada never had to initiate a bond buying program and after it recovered from
its own 2008/9 recession started to increase the overnight rate in 2010. Since then Canada has
been waiting for the rest of the world to sort out its problems. However, with open capital
markets Canada is affected by what has happened in the US, UK, Japanese and European
financial markets.
As global interest rates have dropped the “search for yield” has become very important and
investors have started to take note of Canada. Of importance is that Canada remains one of a
small number of AAA rate countries so is a particularly attractive location to invest government
reserves. The following graphic comes from the Bank of Canada and indicates that almost 30%
of the Canadian government bond market is now owned by non-residents. As non-residents have
invested in the Canadian government bond market they have driven up market prices and driven
down government bond yields far below where they would have been but for the massive bond
buying programs elsewhere in the world.
23
Q. WHAT IS YOUR FORECAST FOR THE LONG CANADA BOND YIELD FOR
2016-7?
A. The current (January 8, 2015) RBC forecast is below.
What is clear is that the enormous amount of liquidity in global markets has caused RBC to
dramatically lower its interest rate forecast. Whereas in June 2011 it was forecasting a quick
rebound to the 4.55-5.05% level for long Canada and US Treasury bonds this has now been
scaled back to 3.65%-4.25%. These forecasts are largely keyed off the forecast increases in the
over-night and Federal Funds rates where RBC takes the Fed at their word and sees a 1.0%
increase in 2016 and then, quite aggressively, further increases to 3.5% by the end of 2017. In
Canada RBC sees the Bank of Canada standing pat for 2016 and then increasing the overnight
rate in 2017 to end at 1.75%.
The RBC forecast is aggressive compared to the general consensus, however, and is reported
mainly because I usually report it. In their December 2015 report Consensus Economics places
the average (from 18 forecasters) for the over ten year bond forecast yield three months out at
1.80% and for one year out at 2.2%. If the current spread from Schedule 2 between the 30 year
(2.07%) and over ten year (1.951.32%) bond is added to the end of 2016 forecast this implies
that the 30 year bond yield at the end of 2016 will only be 2.322.95%. The average of the March
and December Consensus forecasts which is a proxy for the average for the year as whole, and
consistent with the application to an average forward test year rate base, is 2.122.75%.12
There has been much discussion into whether there has been a permanent decline in “neutral”
interest rates, that is, rates that are neither designed to stimulate nor slow down the economy.
Recently Reza and Sarker13 refer to Ms. Yellen’s judgement that the “neutral” or normal interest
rate is now a real rate of 1.75%, whereas the neutral rate in Canada has dropped from the
2.5-3.5% level of the mid 2000’s to a current level of 1-2%. If correct, with a 2% inflation target
these estimates imply a neutral, nominal interest rate of 3.0-4.0% in Canada and a slightly higher
3.75% in the US or slightly below RBC’s longer run forecast. However, what happens to long
term interest rates depends critically on when quantitative easing ends in Japan and Europe and
24
12 RBC is the second highest in their forecast; the highest is from my colleagues at the University of Toronto with a 3.1% forecast yield, which translates into 3.22% for the 30 year bond.
13 “Is slower growth the new normal in advanced economies,” Bank of Canada Review, Autumn 2015.
when, if ever, the central banks start seriously selling their stocks of long term government
bonds.14
What is important to note is that interest rates are not, and probably will not for the foreseeable
future, be set by private investors. Instead, they are being set by what has been termed the
“global policy maker.” As a result, forecasting interest rates for a small country like Canada in a
global financial system depends critically on central bank decisions elsewhere.
Q. WHAT HAS HAPPENED IN THE CORPORATE FIXED INCOME MARKET?
A. The following graphs the spread between the yield on 91 day Treasury Bills (TB) and
those on three month Bankers Acceptances (BA) and Commercial paper (CP).
Money Market Spreads
0
50
100
150
200
250
300
1/4/
2006
4/4/
2006
7/4/
2006
10/4
/200
6
1/4/
2007
4/4/
2007
7/4/
2007
10/4
/200
7
1/4/
2008
4/4/
2008
7/4/
2008
10/4
/200
8
1/4/
2009
4/4/
2009
7/4/
2009
10/4
/200
9
1/4/
2010
4/4/
2010
7/4/
2010
10/4
/201
0
1/4/
2011
4/4/
2011
7/4/
2011
10/4
/201
1
1/4/
2012
4/4/
2012
7/4/
2012
10/4
/201
2
1/4/
2013
4/4/
2013
7/4/
2013
10/4
/201
3
1/4/
2014
4/4/
2014
7/4/
2014
10/4
/201
4
1/4/
2015
4/4/
2015
7/4/
2015
10/4
/201
5
1/4/
2016
CP BA
Treasury Bill yields are close to the rate that the chartered banks get from their deposits at the
Bank of Canada when they have excess cash and are effectively default free, since they are
obligations of the Treasury of the Government of Canada. In contrast, the Bankers’ Acceptance
rate is the rate the market requires on short term investments in the main chartered banks,
whereas the Commercial Paper rate is the rate that large Canadian companies with the best credit
25
14 Note central banks are in no hurry to sell their holdings of long term bonds, since they are financed by issuing cash. As a result, they generate huge profits. The economic reason to sell them will be when there is a pickup up in inflation and the central bank needs to drain the system of liquidity to avoid rampant inflation. Currently, there is no likelihood of this in the foreseeable future.
rating can get by issuing notes in the money market. As a result, the spreads between these two
private rates and that on Treasury Bills is indicative of the state of the short term lending
market15 and the willingness of large investors to lend to the banks and very low risk, stable,
Canadian companies
Before discussing these spreads, it is important to note that investors in the money market are
mainly “parking” their money, rather than investing, since their main concern is security of
principal, that is, getting their money back. Consequently, with any hint of default the market
seizes up. This happens periodically in the CP market as seemingly low risk institutions default
and investors panic and refuse to roll over or reinvest the money invested in CP for fear of
further losses due to an inability to distinguish between good and bad risks.
For example, for the last 20 years the money market has been very quiet with spreads at 10-20
basis points. This changed in July 2007 with the US sub-prime problems spilling over into
Canada, where we can see the large spike and again with the Bear Stearns bailout in March
2008. This got much worse in September 2008 as Lehman Brothers failed and contagion hit the
world’s financial markets and spreads in the Canadian money market went close to 3.0%.
However, of importance is that the measures taken by central banks to stabilise the financial
system worked. The BA and CP spreads have been at normal levels for the last six years.
At the current point in time with the low overnight rate, 91 day Treasury Bills are at 0.45% and
high grade commercial paper is at 0.82% or a premium of 0.37%. In contrast, at the time of my
GCOC evidence in 2012 the cost of commercial paper was 0.34% higher at 1.16%. There is no
question that top quality credits can now access funds in the short term money market more
cheaply than in 2012.
The following graphs the generic credit or default, spreads between corporate and government
long term bonds using the AA, A, and BBB indexes maintained originally by Scotia Capital
markets.16
26
15 The main banks are generally rated R-1 (Mid) equivalent to an AA bond rating while CP is a mixture of R-1 (Mid) and R-1 (low), so generally these spreads are very similar.
16 The most recent data is from Datastream, which updates original data from Scotia Capital’s Handbook of Debt Market Indices.
Default Spreads Since Dec 1979
0
50
100
150
200
250
300
350
400
450
500
12/3
1/19
79
12/3
1/19
80
12/3
1/19
81
12/3
1/19
82
12/3
1/19
83
12/3
1/19
84
12/3
1/19
85
12/3
1/19
86
12/3
1/19
87
12/3
1/19
88
12/3
1/19
89
12/3
1/19
90
12/3
1/19
91
12/3
1/19
92
12/3
1/19
93
12/3
1/19
94
12/3
1/19
95
12/3
1/19
96
12/3
1/19
97
12/3
1/19
98
12/3
1/19
99
12/3
1/20
00
12/3
1/20
01
12/3
1/20
02
12/3
1/20
03
12/3
1/20
04
12/3
1/20
05
12/3
1/20
06
12/3
1/20
07
12/3
1/20
08
12/3
1/20
09
12/3
1/20
10
12/3
1/20
11
12/3
1/20
12
12/3
1/20
13
12/3
1/20
14
12/3
1/20
15
AA A BBB
Corporate bonds have default risk, since companies can run into financial difficulty whereas
governments borrowing in their own currency like Canada cannot.17 These yield spreads usually
behave in a predictable manner. In a recession as the risk of bankruptcy increases, investors sell
off default-risky corporate debt and their liquidity drops. As a result their bond prices fall and
their yields increase relative to the long Canada bond yield causing a wider spread. Conversely
as the economy recovers and this risk recedes the spread narrows. We can see this clearly in the
high spreads during the long recession of the early 1990s, the panic of the Asian crisis and the
bursting of the Internet Bubble and in particular the financial crisis of 2008-9. Note also that
usually the spread increases most for the BBB bond which is the riskiest.
At the current point in time generic A spreads are at 1.90% and have marginally increased since
the summer when they were just under 1.60%. In contrast, at the time of my 2012 GCOC
evidence these spreads were a tad lower at 1.75%. I don’t regard the increase as that significant
mainly because they are quite volatile and currently reflect a minor flight to quality as the stock
market has swooned since the summer on China fears. More importantly firms do not “pay” the
spread they pay the full interest cost. At the time of my GCOC evidence the long Canada rate
was 2.41% and the generic A yield was 4.16%, whereas currently with a long Canada rate of
2.01% the yield on generic A bonds is 3.91% or 0.25% lower.
27
17 This assumes they simply print more money to pay off their debts. The US can do this, but it has been the behaviour of Tea Party Republicans in Congress arguing that the US should default that has frightened global investors since 2011.
It is also interesting to compare Canadian interest rates with the US. The following graph is the
yield on the generic BBB bond in Canada and the yield on Moody’s BBB rated utility bond18
since January 2012 to encompass the period since the last generic hearing. What is clear is that
typically the Canadian BBB yield has been below that in the US, particularly since the start of
2013. Currently, there is a yield difference of 0.81%, that is, the generic BBB yield in Canada is
at 4.70% (January 1) versus the utility BBB yield in the US of 5.51%. This is further
confirmation that the financing cost (cost of capital) to Canadian utilities is significantly lower
than for US utilities.
US and Canadian BBB rated debt
3.5
4
4.5
5
5.5
6
1/6/
2012
4/6/
2012
7/6/
2012
10/6
/201
2
1/6/
2013
4/6/
2013
7/6/
2013
10/6
/201
3
1/6/
2014
4/6/
2014
7/6/
2014
10/6
/201
4
1/6/
2015
4/6/
2015
7/6/
2015
10/6
/201
5
CAN BBB
US BBB
In the past I have also stressed that in Canada the yield on utility debt rated A or BBB has traded
on lower yields than equivalent generic debt. This tendency has been particularly true during
recessions when the rating agencies seem to be slow to downgrade non-utility debt whereas the
market for non-utility debt seems to sell off. I used to use two series from the Canadian Bond
Rating service (CBRS) but this was discontinued after it was taken over by S&P. Fortunately,
Bloomberg has created a synthetic A rated utility yield series (C29530Y). The following graphs
28
18 The US yield data is from the Federal Reserve Economic Data bank (FRED) and the BBB (actually Moody’s BAA) is the only corporate series available, probably because the typical US utility is rated BBB.
the yields from the Scotia Capital generic A bond index and the Bloomberg utility synthetic
series.19
Note that utility yields were consistently lower than generic A yields as the financial crisis
started to emerge and remained so until two years or so ago when they merged. Currently the
market seems to be valuing similarly rated utility and non-utility debt the same. This is further
support for a return to normality in the bond market as the extra “safety” implicit in A rated
utility bonds is not valued to the same degree that it is during the financial crisis.
Utility A vs Generic A yields
3.00
3.50
4.00
4.50
5.00
5.50
6.00
6.50
7.00
7.50
8.00
3/1/
2006
6/1/
2006
9/1/
2006
12/1
/200
6
3/1/
2007
6/1/
2007
9/1/
2007
12/1
/200
7
3/1/
2008
6/1/
2008
9/1/
2008
12/1
/200
8
3/1/
2009
6/1/
2009
9/1/
2009
12/1
/200
9
3/1/
2010
6/1/
2010
9/1/
2010
12/1
/201
0
3/1/
2011
6/1/
2011
9/1/
2011
12/1
/201
1
3/1/
2012
6/1/
2012
9/1/
2012
12/1
/201
2
3/1/
2013
6/1/
2013
9/1/
2013
12/1
/201
3
3/1/
2014
6/1/
2014
9/1/
2014
12/1
/201
4
3/1/
2015
6/1/
2015
9/1/
2015
Utility Scotia A
Q. WHAT HAS BEEN THE GENERAL STATE OF CAPITAL MARKETS?
A. As indicated previously the bond market has been heavily influenced by the actions of
central bankers, it is interesting therefore to look at broader measures of the stress in the
financial system. In the US the Federal Reserve Bank of Kansas City has developed the Kansas
City “Financial Stress” Index (KCFSI) which is graphed below.
29
19 Note the Scotia A index includes a large number of utility bonds so the distinction, by construction, is not as clear as for the two old CBRS series.
KCFSI"+" is tough and "-" is loose
-2
-1
0
1
2
3
4
5
6
7
Feb
-90
Feb
-91
Feb
-92
Feb
-93
Feb
-94
Feb
-95
Feb
-96
Feb
-97
Feb
-98
Feb
-99
Feb
-00
Feb
-01
Feb
-02
Feb
-03
Feb
-04
Feb
-05
Feb
-06
Feb
-07
Feb
-08
Feb
-09
Feb
-10
Feb
-11
Feb
-12
Feb
-13
Feb
-14
Feb
-15
This index is designed to capture a variety of financial indicators in addition to the two which I
have traditionally focussed on, which are the short term spreads in the money market and longer
term spreads in the bond market. The additional indicators include the stock market volatility
index, the state of bank share prices, and the behaviour of stock and bond returns. When the
KCFSI is above 0 it indicates that capital markets are under stress or that access to markets is
“tough” similarly when it is below 0 it indicates relatively easy or “stress-free” capital market
conditions. The value of the KCFSI is simply that it captures in one number the impact of a
variety of capital market indicators.20 The major insight of the KCFSI is that it emphasises the
enormous pressure in the US financial system during the financial crisis. Unlike the internet
bubble crash in 2001 the crisis in 2008/9 struck at the very core of the US financial system,
which is the banking system, where liquidity, that is, the ability to trade securities at close to
their true market value, dried up in many parts of the capital market and the US government had
to intervene on a massive scale. Since the 2012 GCOC market conditions have been easy, but
with the ending of QE3 conditions in the US tightened somewhat as markets returned to average
or “normal” conditions.
30
20 Technically it captures the common element in all these indicators by using principal components analysis.
The work by the Kansas City Fed follows pioneering work done by researchers at the Bank of
Canada who developed a simpler financial conditions stress index,21 which is graphed below.
Canadian Financial Conditions Index"-" is loose and "+" is tough"
-3
-2
-1
0
1
2
3
4
5
10/1/199
9
4/1/20
00
10/1/200
0
4/1/20
01
10/1/200
1
4/1/20
02
10/1/200
2
4/1/20
03
10/1/200
3
4/1/20
04
10/1/200
4
4/1/20
05
10/1/200
5
4/1/20
06
10/1/200
6
4/1/20
07
10/1/200
7
4/1/20
08
10/1/200
8
4/1/20
09
10/1/200
9
4/1/20
10
10/1/201
0
4/1/20
11
10/1/201
1
4/1/20
12
10/1/201
2
4/1/20
13
10/1/201
3
4/1/20
14
10/1/201
4
4/1/20
15
The Bank of Canada indicator similarly tracks the enormous stress in the financial markets
during the financial crisis and similar to the KCFSI, the index reflects the recent easy or stress
free financial market conditions. The Bank has now discontinued the stress index and
emphasises a broader array of stress indicators. One of the most important indicators is the
Bank’s survey results from senior lending officers, which reflects their assessment of the lending
conditions to Canadian non-financial firms. The latest survey results from the bank are below
and similarly we see the tightening credit market conditions during the financial crisis and the
loose conditions thereafter with a gradual return to normal pricing and credit conditions in 2015,
similar to the indications from the KCFSI.
31
21 The Bank of Canada index is actually the inverse of this, I multiplied it through by -1 to get the same interpretation as for the KCFSI
32
Q. HOW DOES THE STATE OF THE ECONOMY AFFECT PROFITS AND THE
CAPITAL MARKET?
A. The following graphs the average annual ROE against the spread between the yield on
BBB debt and long Canada bonds since 1980. The data for 2015 is for the first half of the year
since full year data is not yet available. The graph illustrates the basic inverse relationship that
spreads normally increase during recessions when corporate profitability drops and then reduce
when the economy recovers and corporate profits increase. We can clearly see this in the
recession of the early 1990s and again to a lesser degree during the financial crisis. However,
even during the financial crisis basic corporate profitability in Canada remained healthy due to
strong commodity prices, indicating that much of the pressure was coming from events in the US
and elsewhere.
Corporate ROE and BBB Spread
0
50
100
150
200
250
300
350
400
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
bas
is p
oin
ts
0.00
2.00
4.00
6.00
8.00
10.00
12.00
14.00
16.00
Per
cen
t
BBB Spread ROE
Corporate profitability in Canada has recently dropped from its pre-recession highs as
commodity prices have weakened and this has kept spreads marginally high. We can see in the
first half of 2015, for example, that the drop in profitability has been matched by a minor
increase in spreads over the level for 2014.
Q. WHAT ARE YOUR CONCLUSIONS ABOUT CONDITIONS IN THE “MONEY
MARKET”?
A. The Canadian economy is suffering from two effects. On the one hand it is adjusting to
the drop in commodity prices, which is hurting the materials and energy sector. On the other
hand, it is being helped by growth in the U.S and a strengthening of traditional manufacturing,
particularly in central Canada. Three years ago the Bank of Canada anticipated that the
remaining spare capacity in the economy would be used up by mid-2015, while the financial
system was firing on all cylinders. The slowdown in China has deferred this growth forecast and
caused serious problems in the resource sector. As a result, we are still a couple of years away
from the peak of the business cycle with plenty of growth to come. Whether we can use up these
resources depends on continued growth in the U.S22 and whether the slowdown in China causes
33
22 The latest US job market growth statistics indicated an additional 292,000 in December 2015, which is more than enough to soak up new entrants into the workforce and significantly exceeded expectations.
34
more problems. Of particular concern is whether there are significant non-performing loans to
the Canadian resource sector resulting from the lower commodity prices.
In terms of the equity markets, the following graphs the performance of the TSX composite and
the S&P 500 since December 31 2011. In both cases the level is standardised with the December
31, 2011 value so we can see the changes since then. What is clear is the outperformance of the
US market, since it is not as resource sensitive as the Canadian market, with a 60-70% increase
in value over the four year period. In contrast, the Canadian market has only increased by 13%
and peaked in Summer 2014 at 15,625, after which it has drifted lower with the impact of
commodity prices on the resource sector. This has continued in the first week of 2016 where data
is not yet available in the data tapes.
Stock market IndexesSince December 2011
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
2012
M01
2012
M03
2012
M05
2012
M07
2012
M09
2012
M11
2013
M01
2013
M03
2013
M05
2013
M07
2013
M09
2013
M11
2014
-01
2014
-03
2014
-05
2014
-07
2014
-09
2014
-11
2015
-01
2015
-03
2015
-05
2015
-07
2015
-09
2015
-11
TSX SP500
Despite the recent equity market weakness, general Canadian stock prices are still the same as
they were in 2012/2013 and mainly reflect the impact of the resource sector.
Overall the stock market is valuing utilities favourably, credit is still easy and utilities continue
to have access to the 40 and 50 year debt market at very low interest rates. Market conditions
remain much as they were in 2012 very receptive to good credits like Canadian utilities.
35
Q. THESE COMMENTS ARE FOR CANADA IS THERE ANYTHING DIFFERENT
ABOUT BC?
A. While almost all the capital market data is relevant for Canada as a whole there are
important differences in the BC provincial economic performance. One of the key “risks” for the
Bank of Canada is the state of the housing market and the tendency for borrowers to get too deep
into debt due to the current low level of interest rates. Although this seems to have kept central
bankers awake at night for several years it has been a boon to the BC economy. Currently, retail
sales have been exceptional strong as homeowners “feel good” about the equity in their houses.
Further the high house prices have stimulated housing starts as developers see strong profit
potential in residential real estate. The result has been housing starts running at over 30,000
units. The combination of strong retail sales and housing starts has given BC the highest GDP
growth rate in Canada with growth over 3.0% forecast for 2016.
In terms of industry performance in BC there are mixed messages. On the one hand, the
weakness in commodity prices, such as copper and coal, has led to temporary shut downs and
severely affected major employers like Teck. On the other hand, the strong US growth and the
weak C$ have combined for strong demand for lumber exports to the US. Overall, the impact on
BC exports has been minimal. The result is that strong consumer demand gives BC the third
lowest unemployment rate in Canada at 6.3% ahead of Manitoba and Saskatchewan, but much
lower than the overall Canadian rate of 7.0%. This good provincial economic performance
provides the backdrop to FEI’s business risk assessment.
36
III RISK PREMIUM ESTIMATES OF THE FAIR ROE
Q. WHAT IS THE MOST COMMON WAY OF ESTIMATING THE FAIR ROE?
A. The capital asset pricing model or CAPM remains the most common way of estimating
the fair rate of return. It is a special form of risk premium model which simply says,
*MRPRK F
In words, the investor’s required or fair rate of return (K) is equal to the risk free rate (RF) plus a
risk premium. Where the CAPM differs from other risk premium models is that it specifies that
the risk premium is comprised of the market risk premium (MRP) times the security’’s relative
risk or beta coefficient (β). In this regard any fair ROE can always be decomposed into a risk
free rate and a risk premium, so the CAPM is perfectly general: its contribution is simply to
relate an individual risk premium to the overall market risk premium and its relative risk
coefficient.
Why the CAPM is so widely used is because it is intuitively correct. It captures two of the major
“laws’ of finance: the time value of money and the risk value of money. I will discuss the third
law of finance the tax value of money later, but the time value of money is captured in the long
Canada bond yield as the risk free rate. The risk value of money is captured in the market risk
premium, which anchors an individual firm’s risk. As long as the market risk premium is
approximately correct the estimate will be in the right “ball-park.” Where the CAPM normally
gets controversial is in the beta coefficient; since risk is constantly changing so too are beta
coefficients. This sometimes casts doubt on the model as people find it difficult to understand
why betas change. Further it also makes testing the model incredibly difficult. However, the
CAPM measures the right thing: which is how much does a security add to the risk of a
diversified portfolio, which is the central idea of modern portfolio theory. It also reflects the fact
that modern capital markets are dominated by large institutions that hold diversified portfolios.
Currently, the CAPM remains the most important model used by a company in estimating their
cost of equity capital. The following table comes from a survey of 392 US Chief Financial
officers by Graham and Harvey in the Journal of Financial Economics 2001:
70% of US CFOs use the CAPM and a further 30% use a multi-beta approach similar to the two
factor model I often use. Note in this respect that multi-beta models, while slightly more general,
don’t alter the average rate of return which is anchored by the risk free rate and market risk
premium: all they do is generate slightly different estimates for individual firms.
In response to comments on the Decision for the 2012 GCOC I have developed a new appendix
(E) that discusses alternatives to the CAPM. At that time I noted that I was not enthusiastic about
such models since they remain extremely controversial. I also don’t think they add much except
to add another area that experts can argue over. In particular, as I just mentioned they do not
generally have a significant effect on the equity cost estimates. To illustrate this, the current
standard in multi-factor models is to extend the one factor CAPM to include two additional
factors: one for the size premium (the return difference between small firms minus big or SMB)
and the other for the value premium (the return difference between value versus growth stocks).
This is the Fama-French three factor model which was raised in the GCOC in 2012. The
Fama-French equations states,
37
HMLSMBMKTRRE jjjFj ,3,2,1)(
In this case as well as the market risk premium (MKT) an investor requires a premium for small
firms and value stocks. Why the FF3 factor model is controversial is that some believe that small
value stocks are riskier and thus deserve a premium, whereas others believe that the market
consistently miss-prices large growth stocks, since they are faddish and sexier for financial
advisors to sell (think Apple, Facebook etc) and are thus over-valued and consistently earn lower
rates of return.
I tend to believe the faddish argument, but regardless for individual stocks using the FF3 factor
model versus the CAPM rarely makes much of a difference. For example, Estrada (2011)23
estimated the equity cost for the Dow 30 firms using both models where the average equity cost
using the CAPM was 9.70% versus 9.50% from using the FF3 factor model. The complete
estimates are in Schedule 3, but the general point is that we are just allocating the stock’s return
to different components, so that the sum of the parts should always (approximately) add up to the
same number. The Dow 30 stocks have a beta close to 1.0 since they are a portfolio of stocks and
this should be close to average; as we would expect they have negative exposure to the size
premium, since they are all large stocks and positive exposure to the value premium, since they
are generally value stocks. In this respect they are similar to utilities that tend to be relatively
large value stocks.
In terms of the “error” in using one model versus another, the difference ranges from +1.5% to
-1.6% or a range of 3.0%. This is not an insignificant difference but it stems from the confluence
of the size and value premiums.24 The +1.5% difference is for American Express, which has a
17.7% FF3 Factor equity cost estimate versus the 16.2% for the CAPM. This difference stems
from the observation that AmEx is a relatively small value stock and generates a premium for
both these factors, which is greater than the lower beta estimate. In contrast, Merck is a large
growth stock and its much higher FF3 factor beta coefficient is not enough to offset the negative
size and growth premiums. As a result, its CAPM equity cost at 9.1% is higher than its FF3
factor cost at 7.5%. The closest to a regulated utility would be AT&T, where the CAPM equity
38
23 Estrada, Journal of Applied Corporate Finance (Spring 2011). Estrada’s estimates are for illustration only I do not recommend them or the process he used to get them.
24 Note also that the range of equity cost estimates is from 4.80% to 17.7% for the FF3 factor model and a slightly smaller 5.3% to 17.5% for the CAPM.
cost is 7.80% versus a FF3 factor estimate of 7.30%; again its higher beta is more than offset by
the impact of the size and value premiums.
Q. IS THERE ANY OTHER SUPPORT FOR THE CAPM?
A. Yes. Levy and Roll have recently revisited the question of the empirical support for the
CAPM. Richard Roll in a ground-breaking paper25 pointed out that as long as the market
portfolio is ex post efficient then by definition the CAPM will work in empirical tests and all
securities will lie along a straight line relating returns to betas. In the Levy and Roll paper26 they
reverse engineer the tests and consider how much the sample parameters can vary to make sure
the normal proxies for the market portfolio are efficient. Levy and Roll find that even slight
variations, within the normal estimation bounds, make the proxies efficient. As they conclude:
“This article shows that a small variation of the sample parameters, well within theirestimation error bounds, can make a typical market proxy efficient. Thus, the empiricallymeasured return parameters and the market portfolio weights are perfectly consistentwith the CAPM using a typical proxy..... Hence, minor changes in estimation errorreverse previous negative and disappointing findings for the CAPM.”
The point is that the normal criticism of the CAPM is usually based on very old empirical tests
that did not recognize estimation error while the Levy and Roll results show that even slight
changes in the parameters will nullify those results.
Q. WHAT IS YOUR SIMPLE CAPM ESTIMATE FOR A BENCHMARK UTILITY?
A. In Appendix B I estimate the market risk premium of common equities over long term
Canada bonds at 5.0-6.0%. This estimate is drawn from the Canadian capital market history
going back to 1924 so encompasses periods very similar to today, such as the bleak 1930s of
slow growth and falling prices, as well as booms and serious inflation problems such as the
1970’s. While the Canadian data points to a market risk premium of under 5.0%, I give weight to
the US evidence for three main reasons. First, most of the restrictions on keeping Canadian
39
25 Richard Roll, “A Critique of the Asset Pricing Theory’s Tests: Part 1: On Past and Potential Testability of the Theory”, Journal of Financial Economics 4: 129-76, 1977.
26 Moshe Levy and Richard Roll, “The Market Portfolio May be Mean Variance Efficient After All,” Review of Financial Studies, 2010.
40
capital within Canada have been removed resulting in significant capital outflows and higher
expected returns on Canadian investments. Second, Canadian governments have moved to a
primary surplus on their budgets. The primary surplus is the actual surplus after stimulus
expenditures and the impact of an economic slow-down have been removed. Third, the Canadian
bond market has received significant foreign capital inflows depressing yields below where they
would have been historically with a segmented or closed capital market. The result has been
lower interest rates in Canada than the United States for the last five plus years, which has
removed the historic bias of a smaller Canadian market risk premium over a higher and riskier
Canadian government bond yield. Finally, I give significant weight to survey results by
Professor Fernandez, who now annually surveys thousands of academics, financial analysts and
corporate executives making investment decisions. This acts as confirmation that my market risk
premium range is well within the normal range of other finance professionals.
My Appendix C discusses relative risk adjustments or betas. The recent history of Canadian
utilities is of beta coefficients in an approximate range of 0.30-0.45. The higher estimate
represents their beta estimate post the financial crisis and internet bubble. These estimates are
consistent with the price behaviour of Canadian regulated utilities and estimates by external
outlets such as the Royal Bank of Canada, Yahoo and Google, so I regard them as reasonable. It
is indisputable that as low risk investments the relative risk of Canadian utilities has been
significantly lower than for an average Canadian equity security.
As robustness checks I also look at beta estimates for a sample of US utilities, where the
estimates are somewhat higher, but this could be due to unique effects such as merger activity. I
also look at Canadian beta estimates against a US market index to check whether the “low”
estimates are due to the hollowing out of the Canadian stock market through acquisitions and
find no such evidence. I also look at what would happen if I adjusted my beta estimates using the
Blume adjustment methodology toward their grand mean of 0.50 and find it makes very little
difference. However, any estimates reflect the time period over which they are estimated and
once a unique event falls out of the estimation window it no longer affects the estimate.
Consequently on a going forward basis I judge the relative risk of Canadian utilities to move
back to their historic range reflecting normal market risk. This is why I continue to judge the
relative risk of a Canadian utility to be 45-55% of that of the market as a whole.
I would therefore judge the going forward utility risk premium to be 2.25% to 3.30%
representing the combination of the low end of the relative risk adjustment and the low end of
the market risk premium (.45 and 5%) combined with the top end of both (.55 and 6%). If this is
added to a 2.122.75% consensus forecast for the average long term Canadian bond yield for 2016
and a 0.50% flotation cost allowance, I would judge a “simple” CAPM fair return estimate for
2016 to be as follows:
Simple CAPM Estimates 2016
Low end 4.875.50%High end 5.926.55%
The average estimate of 5.406.03% is 0.780.25% lower than the simple average of 6.28% I used
in 2012 simply because the forecast long Canada bond yield is that much lower.
Of importance is that some forecasters, such as RBC, have been predicting a rapid increase in
long Canada interest rates for some years, but the forecast has to bear some relationship with the
actions of investors in the bond market. If the current bond market yield of about 2.0% does
increase to the 2.85% as forecast by RBC over the next year, then a bond market investor would
lose on their investment.27
Q. DO YOU USE THIS SIMPLE CAPM ESTIMATE?
A. No. The CAPM estimate is appropriate under “normal” or average markets, since it uses
a normal or average market risk premium and assumes that conditions in the bond market
affecting the long Canada bond yield are also driving conditions in the equity market, that is, that
the correct “opportunity cost” for an equity investor is the bond market plus a risk premium.
However, at the current point in time conditions in the Canadian bond market are largely being
driven by external factors and are still not “average” market conditions. To adjust for this I
make two adjustments: the first is simply to make the CAPM estimate conditional on the state of
the markets thereby converting it into a conditional CAPM or CCAPM, the second is to adjust
41
27 If RBC’s forecast was generally accepted then current yields would be a lot higher, since the one year Treasury Bill yield is only 0.50%. By the unbiased expectations theorem the loss from investing in one year as opposed to 30 year Canadas is about 1.5% which has to be captured in a higher bond yield next year. This means that the yield has to increase by approximately 0.05%, not 0.85%. This is more consistent with the Consensus forecast.
for the abnormally low Canada bond yields resulting from rampant bond buying programs by
central banks.
In terms of the CCAPM there are a variety of financial variables to use to condition on the state
of the markets. For example, I have previously discussed the Kansas City financial stress index
which attempts to do this by looking at a composite stress index, which should be correlated with
investor risk aversion. Another popular index is the Chicago Board of Exchange’s (CBOE)
volatility index (VIX). This calculates the expected standard deviation of the overall stock
market return from at the money call options.28 The standard deviation, as explained in my
Appendix B, is a measure of the overall risk or volatility of the stock market and has averaged
18-20% since 1926. Below is a graph of the VIX since 2004.
CBOE Volatility Index (VIX)
0
10
20
30
40
50
60
70
80
90
1/2/
2004
7/2/
2004
1/2/
2005
7/2/
2005
1/2/
2006
7/2/
2006
1/2/
2007
7/2/
2007
1/2/
2008
7/2/
2008
1/2/
2009
7/2/
2009
1/2/
2010
7/2/
2010
1/2/
2011
7/2/
2011
1/2/
2012
7/2/
2012
1/2/
2013
7/2/
2013
1/2/
2014
7/2/
2014
1/2/
2015
7/2/
2015
1/2/
2016
The average value for the VIX over the period 2004-2015 was 19.4%, close to the average from
stock market data going back to 1926.29 However, consistent with the KCFSI we can see the
huge increase in uncertainty during the financial crisis as the VIX hit a peak value of 80% or 4X
the average value. The VIX reflected the huge panic during the financial crisis, but as always
42
28 Call options pay off only if a certain event (usually the stock price) reaches a specific value and the only “unknown” to price them is the variability or standard deviation of the price. As a result, we can work back from the market price to estimate the implied standard deviation or volatility.
29 It may surprise some but there is no evidence of any structural increase in uncertainty in the stock market: it has always been highly volatile.
43
the panic subsides and since January 2011 (the last five years) has been below average at 16.2%
despite periodic attacks of nerves in August 2015 as the first China fears hit the stock market and
the first week of 2016 when they returned.
The problem with the VIX and the KCSFI is they are not direct estimates of a rate of return and
cannot be used to reliably alter a CAPM expected return estimate. An alternative is to use the
default or credit spread between risky corporate debt and default free Government of Canada
bonds. This is the corporate spread discussed earlier and similar to the VIX and KCSFI shows
the huge impact of the financial crisis when corporate A rated spreads peaked at 3.70% versus a
pre-crisis average of 0.94% (1979-2007).
In a series of hearings into the fair rate of return in the period after the financial crisis several
utility tribunals used the credit spread to adjust the CAPM estimate. For example, the AUC
stated in its generic decision (Decision 2009-216, November 12, 2009)
In arriving at their reasonable CAPM range of 7.13%-8.62% the AUC used a market risk
premium range of 5.00-5.75% a relative risk (beta) coefficient of 0.50-0.63 and a forecast long
Canada yield of 4.13-4.50%. Together with a 0.50% flotation allowance these result in a
bottom-to-top range of 7.13-8.62%. The AUC then adjusted this estimate by a credit spread
adjustment.
Q. DO YOU AGREE WITH A CREDIT SPREAD ADJUSTMENT?
A. Yes. Before several boards in 2009 I stated that much of the increase in credit spreads
was caused by liquidity problems in the market making function of investment banks. Since then
research at the Bank of Canada has helped to disentangle the liquidity from the pure credit risk
components in the corporate spread.30 This research concludes (page 29):
“our results show that for investment grade firms, the majority of the spread corresponds toliquidity: on average, the liquidity component accounts for 63% of the spread.”
Although it remains very difficult to disentangle the liquidity from the pure credit component in
credit spreads, since 2010 as Canadian capital market conditions were returning to normal I have
been recommending a 50% adjustment to changes in credit spreads. Over a normal business
cycle this adjustment should average out to zero as capital market conditions fluctuate around
average levels. This adjustment was incorporated into ROE adjustment models adopted by the
OEB, Regie and the BCUC.
At the current point in time “A” spreads are at 1.91% or 91bps more than the typical average for
the business cycle, which I take to be 1.0%. Consequently I add 0.45% for this credit market
effect. I regard this sort of adjustment as converting the CAPM into a conditional CAPM where
the CAPM holds conditional upon the state of the financial markets. My estimate for the
CCAPM is as follows:
CCAPM Estimates 2016
Low end 5.325.95%High end 6.377.00%
The average of these two is 5.856.48% and would be appropriate as an estimate for a fair ROE at
this stage of the business cycle in a capital market that is also typical for this stage in the
business cycle. However, I still regard the resulting ROE as an under estimate at the current
point in time.
Q. WHY IS THIS CCAPM AN UNDER-ESTIMATE AT THE MOMENT?
44
30 A. Garcia and J. Yang, “Understanding Corporate Bond Spreads Using Credit Default Swaps,” Bank of Canada Review, Autumn 2009
45
A. In Appendix B Schedule 6, I develop a model to explain the behaviour of the real yield
on long Canada bonds, defined as the nominal yield minus the average of past, current and future
CPI inflation. Ignoring the indicator or dummy variables for WW2 and the 1970s, when there
was huge liquidity during the petro dollar recycling period, the model essentially says that the
real LTC bond yield is 1.34% plus a premium based on bond market uncertainty and a premium
based on the size of the government deficit. The model does well in explaining the very high
yields when there was huge volatility in the bond market and Canada was running deficits
approaching 10% of GDP.
However, since the financial crisis while we have seen bond market uncertainty go down, the
aggregate deficit in Canada has gone from a surplus to just below 2% of GDP. Normally this
would cause an increase in government debt pushing down prices and pushing up yields.
Plugging numbers into the regression model would predict a real long Canada yield of about
3.88% for 2014, rather than the actual 1.35% (2.60% average 2014 Canada yield minus 1.25%
inflation). However, the flood of government debt is being bought in part by non-residents due to
the impact of loose monetary policy elsewhere in the world. Once an indicator (dummy) variable
is added for the years since 2010, the 2014 real yield estimate is reduced by 2.53%. What this
indicates is that the current and forecast long Canada bond yields are severely depressed below
where they would have been had Canada remained partially segmented from external events.
The model also indicates that current bond yields are not being determined by ordinary investors
trading off risk versus return as assumed in standard risk premium models.
An additional insight is from looking at preferred yields. In old testimony I (with my late
colleague Dr. Berkowitz) presented four ROE estimation methods. One of them estimated the
fair ROE by looking at the premium of the earned ROE over the yield on traditional fixed rate
preferred shares and how this premium was valued by investors in terms of the market to book
ratio. The reason for doing this was that preferred shares are an equity instrument taxed at the
same rate as dividend income from ordinary shares. As a result, the tax bias from comparing the
fair ROE for a regulated utility with the yield on long Canada bonds is removed, since interest
income is fully taxed whereas dividend income via the dividend tax credit is not.
46
This tax effect is well known in capital markets. BMO-Nesbitt-Burns used to produce a
Preferred Share Quarterly that tracked the performance of the preferred share market. Prior to the
distortions introduced by the financial crisis, Nesbitt Burns provided the following yields as of
June 2004 when the stage in the business cycle was close to neutral:
June 2004
Retractable Preferreds (%)Dividend yield 4.01Mid Canada yield 4.09After tax spread (corp) 1.77After tax spread (indiv) 0.63
Straight Preferreds (%)Dividend yield 5.48Long Canada yield 5.34After tax spread (corp) 2.54After tax spread (indiv) 0.98
Floating Rate Preferreds (%)Dividend yield 3.42BA (3 month) 2.12After-tax spread (corp) 2.25After-tax spread (indiv) 1.22
The retractable preferreds are compared to mid Canada bonds, since the retraction feature
shortens their maturity as compared to a long bond. The traditional straight preferreds are
compared to long Canada bonds, while the floating rate preferreds are compared to 90-day
Bankers acceptances (BAs), since their dividends are usually reset quarterly.
The important point about the comparison is that what we observe in the capital market is a
pre-tax yield. This is determined by both risk and taxes. Take the straight preferreds, for
example, in June 2004 the long Canada bond had a yield of 5.34%, while straight preferreds had
a yield of 5.48%. Clearly, the preferreds would be regarded as riskier than the long Canada bond,
since the corporate issuer can default. However, the yield on the preferred shares was only
0.14% higher. The reason is that the dividend income gets more favourable tax treatment than
interest income. The correct comparison is the after-tax yield difference, which
BMO-Nesbitt-Burns gives as 2.54% in favour of the preferred shares for corporate investors and
0.98% for individuals. The after tax yield difference gives the correct result: that on an after tax
basis the riskier preferreds give a higher yield.
In June 2004 the long Canada bond yield is given by BMO as 5.34% and the preferred share
yield at 5.48%. At the end of June 2004 the Scotia Capital “A” yield was 6.34% for a spread of
100 basis points over the long Canada bond yield, which is about “average” for a complete
business cycle. This is partly why I the 2004 data is useful as a starting point, since in most
respects it was “normal”. The Nesbitt-Burns data is no longer available but Standard and
Poors/TSX have published a preferred share index and the spread of the yield on this index along
with that on the Scotia Capital “A” bonds over equivalent maturity long Canada bonds is
graphed below.31
Preferred and A Spreads
100
150
200
250
300
350
400
450
1/1/
2010
2/1/
2010
3/1/
2010
4/1/
2010
5/1/
2010
6/1/
2010
7/1/
2010
8/1/
2010
9/1/
2010
10/1
/201
011
/1/2
010
12/1
/201
01/
1/20
112/
1/20
113/
1/20
114/
1/20
115/
1/20
116/
1/20
117/
1/20
118/
1/20
119/
1/20
1110
/1/2
011
11/1
/201
112
/1/2
011
1/1/
2012
2/1/
2012
3/1/
2012
4/1/
2012
5/1/
2012
6/1/
2012
7/1/
2012
8/1/
2012
9/1/
2012
10/1
/201
211
/1/2
012
12/1
/201
21/
1/20
132/
1/20
133/
1/20
134/
1/20
135/
1/20
136/
1/20
137/
1/20
138/
1/20
139/
1/20
1310
/1/2
013
11/1
/201
312
/1/2
013
1/1/
2014
2/1/
2014
3/1/
2014
4/1/
2014
5/1/
2014
6/1/
2014
7/1/
2014
8/1/
2014
9/1/
2014
10/1
/201
411
/1/2
014
12/1
/201
41/
1/20
152/
1/20
153/
1/20
154/
1/20
155/
1/20
156/
1/20
157/
1/20
158/
1/20
159/
1/20
1510
/1/2
015
11/1
/201
512
/1/2
015
1/1/
2016
Pref A
On January 1, 2010 long Canada bonds yielded 4.14%, “A” bonds 5.86% and TSX’s preferred
share series 5.44%. So the spreads were 130 bps for the preferreds, and 172 bps for the generic
“A” bonds. Compared to the June 2004 Nesbitt-Burns data these spreads had increased. The
preferred share spread from 14 bps to 130 bps and the “A” spread from 100 bps to 172 bps but
some of the change in the preferred share yield is due to the different index composition. Overall
these spread increases were due to the after effects of the financial crisis, but the graph then
indicates two things. First, the generic “A” yields moved broadly in tandem with the long
47
31 This index is not comparable to Nesbitt-Burns index, which separated out the difference types of preferred shares.
Canada bond yield without any trend. This is the change that the corporate credit spread
adjustment picks up. Second, while the preferred yield spread moved in tandem with the bond
spreads until August 2011, afterwards the spread increased dramatically. Another way of saying
this is that by December 2011 the preferred share yield had dropped 0.44% from 5.44% to
5.00%, while the A bond yield had dropped by 1.53% from 5.86% to 4.33%.
I judge the change in yields from 2010 to 2012 as reflecting the impact of the Federal Reserve’s
third round of quantitative easing (QE3) or what I have called Operation Twist.32 It appears that
there was a direct impact in Canada as foreign capital slipped into Canada in search of higher
yields. This affected both the government and to a lesser extent the corporate bond market, since
yields on both came down after September 2011. However, yields in the preferred share market
did not come down to the same degree causing the preferred share yield spread to widen. This is
probably because preferred shares are unattractive to foreign investors, since the dividends
attract with-holding taxes and non-residents do not get the dividend tax credit.
I placed the “Operation Twist” impact on the Canadian bond market as approximately 80 bps
from August 2011 through to May 2013, which was my judgment in the last generic cost of
capital hearing. However, in May 2013 Governor Bernanke spooked the markets by indicating a
withdrawal plan for the Federal Reserve’s bond buying program (QE3) and interest rates started
to back up in the expectation that the withdrawal would depress prices and increase yields. From
May 2013 long term government interest rates increased from 2.20% to 3.07% by October 2013
or by 0.87% and my forecast long Canada bond yield increased almost proportionately from
3.0% to 3.60%. Of more importance the spread of the preferred share yield over long Canada
bonds dropped from 2.54% at the end of April 2013 to 2.01% and was at times even lower. In
comparison the spread of the yield on A bonds over long Canadas only dropped from 1.68% to
1.66%, so was basically constant. Consequently, I reduced my “Operation Twist” adjustment.
However, although the US bond buying program finished in 2014, new programs were started or
expanded elsewhere and a huge amount of government debt has been taken off bond markets
around the world. The result is that long Canada bond yields have backtracked and decreased
from 3.0% in October 2013 to the current level of barely over 2.0%, while the A spread has
48
32 Since the objective was to “twist” the US yield curve and lower the yield on long term bonds.
increased from 1.66% to the current level of 1.90% and the preferred spread has increased from
2.01% to 3.78% and been highly volatile
An easier way of looking at the change in spreads is the following graph which shows the
change in the spread between the preferred share spread and the A spread. If all securities
reacted in an identical way to the changes in the long Canada bond market then this spread
would average out to zero. However, the change in the spreads indicates that this is not true.
Pref-A Spread
-100
-50
0
50
100
150
200
250
1/1/
2010
4/1/
2010
7/1/
2010
10/1
/201
0
1/1/
2011
4/1/
2011
7/1/
2011
10/1
/201
1
1/1/
2012
4/1/
2012
7/1/
2012
10/1
/201
2
1/1/
2013
4/1/
2013
7/1/
2013
10/1
/201
3
1/1/
2014
4/1/
2014
7/1/
2014
10/1
/201
4
1/1/
2015
4/1/
2015
7/1/
2015
10/1
/201
5
1/1/
2016
The increase in the preferred share yield spread from 0.80% in 2012 to the current 2015 average
of 1.3% or an increase of 0.50% reinforces my 2012 judgement that the long term Canada bond
yield is not as useful as it has been in the past for estimating the cost of equity capital and the fair
ROE and that the simple CAPM estimate underestimates the fair return.33
If the 1.30% spread of the preferred share yield over the A bond yield is added to the CCAPM
estimate as a current “Operation Twist” adjustment similar to 2012 we get the following:
Risk Premium Estimate 2014
49
33 On August 7, 2015 Canadian Utilities issued a straight perpetual preferred share at a 5.0% coupon when the TSX preferred share index had a yield of 5.18%. However, many of the issues are from banks and their recent yields seem to reflect the fear of bank loan losses to oil and gas companies.
50
Low end 6.627.25%High end 7.678.30%
This would indicate a 2016 fair ROE of 7.157.78% for a benchmark utility, which is the
mid-point between the high and low estimates. Taking into account the volatility of current
yields on utility preferred shares, which are part of shareholder’s equity, but rank ahead of the
common shareholders I would tend to be conservative and recommend the same 7.5% as in
2012. However, I note that consistent with Concentric’s review of allowed ROE’s prepared for
the Canadian Gas Association, the statistical evidence of lower bond yields is for lower allowed
ROEs than in 2012 and that allowed ROEs in both Alberta and Quebec have subsequently been
lowered.
IV DCF ESTIMATES OF THE FAIR ROE
Q. WHAT ARE YOUR DCF ESTIMATES?
A. In appendix D I review the DCF model and apply the model to the market as a whole and
highlight the problems in applying it to individual stocks. For the market as a whole I would
estimate the fair return as being 8.50-9.00% in Canada and slightly higher in the US. With
forecast long Canada bond yields at 2.12% for 2016, this implies a current market risk premium
of over 6.38-6.88%, which is in excess of both the historic evidence in Canada and the judgment
of the respondents to Fernandez’ survey. For the S&P gas and electric index the historic utility
risk premium is about 3.4%, which slightly exceeds the top of my Canadian utility risk premium
range, but is based on lower ten year US bond yields and not the 30 year bond. Subtracting the
spread between these two would indicate it is consistent.
I have traditionally viewed my DCF estimates as “checks” on my CAPM estimates, since in my
view CAPM estimates have usually been in the right “ball-park.” However, the recent very low
long Canada bond yields have forced me to re-evaluate this and look at what drives the
difference between the DCF and simple CAPM estimates. This is because they should be
consistent. The CAPM equation is as follows:
*MRPRK F
In words, the required (fair) return is the risk free rate (RF) plus the risk premium comprised of
the market risk premium (MRP) times the beta coefficient (β). For the market as a whole we can
simply set the beta to one since this is definitional.
The risk free rate is directly observable since the practise in Canada is to use the long Canada
bond yield as the risk free rate, while the market risk premium is reasonably objective,
particularly now that we have Fernandez’ survey data from thousands of professionals in the
area. Consequently, the major area of dispute is the relative risk or beta coefficient, and even
here there is not much doubt that utilities are lower risk than the market. Hence the big
advantage of the CAPM is that it is difficult to make big mistakes. The CAPM also avoids one of
the big problems with DCF estimates in that the forecast inflation rate is automatically
incorporated into the long Canada bond yield, since we use the nominal rather than the real yield.
This is currently not a significant problem, since inflation is so low, but part of the reason the
DCF model fell out of favour was that it was giving bad signals when applied mechanically in
the 1990s, when there was a structural break in the forecast inflation rate.
51
The classic Gordon growth model,34 often referred to as the DCF model in most testimony before
regulatory bodies, is as follows:
gP
dK 1
In words, the required rate of return is the forecast dividend yield plus the long run growth rate.
Conceptually the DCF model and CAPM should give exactly the same values but, of course,
since they approach it from a different perspective there is always estimation error. For the
market as a whole the forecast dividend yield can be estimated with very little error, so the
estimation error is with the forecast long run growth rate, which also is easier to estimate for the
market than for an individual stock. As a result, if the CAPM and DCF estimates differ
significantly, then it is mainly due to the difficulty in estimating the growth rate in the DCF
model and the risk premium in the CAPM.35
We can assess the relative value of the DCF and CAPM by graphing the “known” parts of both
models for the overall market, which are the long Canada bond yield and the TSX dividend
yield. Note in the following graph that the current period is quite unusual since dividend yields
on the TSX are higher than the long Canada bond yield. This implies that either the long Canada
bond yield is unusual or the pricing in the equity market is unusual.
52
34 The model is named after the late Professor Myron Gordon of the University of Toronto.35 Note since for the CAPM we are dealing with the market return the following analysis is general for
any risk premium model
53
Dividend and Long Canada Bond Yields
0
2
4
6
8
10
12
14
16
18
1956M01 1961M01 1966M01 1971M01 1976M01 1981M01 1986M01 1991M01 1996M01 2001M01 2006M01 2011M01
TSXYield Canadas
Since both the DCF model and CAPM should give the same answer, we can set them equal to
each other, which indicates that for the market as a whole
gMRPRPdCAPMDCF F 1
Or in words the directly observable spread between the long Canada bond yield and the TSX
dividend yield is equal to the long run dividend growth rate minus the market risk premium.
From the above graph we can see that there is generally a very large difference between the two
indicating that the expected growth rate in the stock market was much higher than the market
risk premium but that this growth rate has varied over time. The reason for this difference is
mainly the gradual increase and then decrease in the CPI inflation rate over the period since
1956. This inflation rate is directly captured in the long Canada bond yield since it is a nominal,
and not a real yield, and yet is in the “unobserved” growth rate in the DCF model. Note for
example, that the increasing and high rates of inflation in the 1960-1980 period (next graph)
coincides with the increasingly large difference between the LTC and TSX dividend yields.
54
Average Annual CPI Inflation
0
2
4
6
8
10
12
14
1956
1958
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
It is possible to come up with a simple or naïve estimate of the market return by adjusting for
this inflation/real yield bias to the estimates. For example, we can assume that for the DCF
model the forecast growth rate is the actual CPI inflation rate at the time, based on year over year
changes, and then add a 3.50% real growth rate. This gives a simple growth rate forecast to add
to the dividend yield and thus a simple or naïve DCF estimate for the market as whole. Similarly,
we can add a long run market risk premium of 3.5% to the long Canada yield for a simple CAPM
estimate. For the entire period 1956-2014 the average naïve DCF estimate is 10.58%, while the
average naïve CAPM estimate is 10.52%, or a difference of only 0.06%between the two, so “on
average” these assumptions seem to make sense.
To see how robust this simple procedure is, the following graphs the difference between the two
estimates for every month since 1956 until October 2015. The graph indicates that the difference
was very large from the mid 1970’s until the late 1990’s. The reason for this difference is
twofold. First, in the 1970s inflation was increasing and bond yields captured the expected
inflation component directly. However, investors did not believe that the Bank of Canada and the
Government would allow these high levels of inflation to continue. This resulted in very low real
yields on LTC bonds leading to a positive difference between the DCF and CAPM estimates.
55
Difference between Naive DCF and Risk Premium estimates for the Market
-8.00
-6.00
-4.00
-2.00
0.00
2.00
4.00
6.00
8.00
10.00
12.0019
56M
01
1957
M10
1959
M07
1961
M04
1963
M01
1964
M10
1966
M07
1968
M04
1970
M01
1971
M10
1973
M07
1975
M04
1977
M01
1978
M10
1980
M07
1982
M04
1984
M01
1985
M10
1987
M07
1989
M04
1991
M01
1992
M10
1994
M07
1996
M04
1998
M01
1999
M10
2001
M07
2003
M04
2005
M01
2006
M10
2008
M07
2010
M04
2012
M01
2013
M10
2015
M07
Once investors caught up to the impact of high inflation the reverse set in, as the budget deficits
at the Federal level convinced the market that the government would inflate its way out of its
deficit problems, rather than bring down inflation. As a result, while the year over year inflation
rate dropped dramatically, LTC bond yields did not at first similarly drop, leading to very high
real yields and simple CAPM estimates exceeding their DCF equivalents. It is this phenomenon
of low real bond yields in the 1970s and 1980s and high real bond yields in the 1990s that is the
major reason for the positive deviations from 1970-1982, and the negative deviations afterwards.
The second reason is simply that the real GDP growth rate and the market risk premium have not
remained constant since 1956. I testified extensively in the 1990s to the effect that the market
risk premium was very low due to the high real interest rates and risks attached to government
bonds. Subsequently, I have increased my estimates of the MRP as this risk has been removed
and currently use 5.0-6.0%, rather than 3.5% as my market risk premium estimate. Similarly, the
long run real growth rate may have dropped and might be lower than the 3.5% used in the simple
model.
However, the point is that we can “ballpark” the broad range for the DCF estimate for the market
just as we can for risk premium models like the CAPM. At the end of October 2015 the TSX
dividend yield was 3.17% and the year over year inflation rate 1.36%, so with the 3.5% real
growth rate the simple DCF estimate is 8.18%.36 Similarly with the October 2015 long Canada
yield of 2.12% and a 3.5% market risk premium the simple CAPM estimate is 5.62%. As a
result, there is currently a positive 2.56% difference when we subtract the CAPM estimate from
the DCF estimate. Further note from the graph that this difference between the two has gradually
gone from negative to positive over the last 15 years as long Canada bond yields have gone
down. The reason for this is simply the fact that the real yield on the long Canada bond has
dropped, which has penalised the risk-premium estimate relative to the DCF estimate.
Of course current DCF and risk premium estimates are not the naïve ones graphed above, but
instead allow for differences in the market risk premium and growth rates, but this analysis
confirms the implications of the current problems in the bond market on the CAPM estimates.
As a result it supports my adjustments to the CAPM estimates and the value of currently looking
at DCF estimates.
Q. WOULD YOU USE THESE ESTIMATES?
A. No. These are very simple or naive estimates that use average numbers for a very long
period of time. They are presented simply to show that while the DCF and CAPM estimates are
consistent over long periods of time, they both have problems when used mechanically during
periods of very high or low real Canada bond yields. The analysis also helps explain why DCF
estimates fell out of favour in the 1990s while the validity of recent CAPM estimates has
recently been questioned.
Q. IS THERE ANY OTHER EVIDENCE ON THE VALIDITY OF THESE
ESTIMATES?
A. Yes. What is important is that there is another side to estimating the fair ROE and cost of
equity capital. This is that the required rate of return on the part of the investor (cost of equity
capital) is also the expected rate of return. Defined benefit pension funds need this expected rate
56
36 This is 1.0317*(1.0136+.035)-1
of return to determine whether a fund is in deficit or surplus. On October 19, 2012 TD
Economics produced its own analysis of the long run returns of the type needed in defined
benefit pension plans.37
The important point about the TD Economics forecast is that the going forward risk premium for
equities minus bonds is 4.00%. This is not the market risk premium, since adjustments need to be
made but it is certainly in the right ballpark.
Q. WHAT ADJUSTMENTS NEED TO BE MADE?
A. As TD Economics notes its return forecast is for ten year geometric returns so they have to
be converted to arithmetic returns. To make this adjustment for very long returns we add half the
variance of the arithmetic return as explained in my Appendix B, with data in Schedule 8.
Historically the standard deviation of equity returns has been about 20% (0.20) so the variance is
0.04 and half this is 0.02 or 2.0%. Similarly, the volatility of the long Canada bond return has
been about 9% (0.09). I would suspect that this overstates the future volatility, since it is unlikely
we will see LTC yields at almost 20% again, but this means a variance of 0.0081 and half this is
0.4%.
57
37 TD Economics, An Economic Perspective on Long-term Financial Returns.
58
So converting these long run returns to short run arithmetic returns means market risk premium
of 5.60% as follows:
Long run 1/2 the variance Arithmetic
Equities 7.0% 2.0% 9.0%
Bonds: 3.00% 0.40% 3.40%
However, the TD Economics forecast is over the yield on the DEX universe bond index and not
over long Canada bonds. The universe of bonds would have lower duration or maturity than the
30 year long Canada bond, but can be expected to earn more since they have default risk. Given
the prior long Canada forecast of 2.12% for 2016, this would increase the market risk premium
estimate closer to 6.00%, but is within my 5.0-6.0% range.
Q. WHAT IS YOUR FAIR ROE FOR A BENCHMARK UTILITY?
A. I would judge a fair ROE for 2016 to be in a range 6.62-7.677.25-8.30% for 2016. My
estimates are based on the following:
Risk Premium
Base LTC forecast: 2.122.75%Normal utility risk premium: 2.25%-3.30%Issue costs: 0.50%Normal Fair ROE 5.406.03%Credit Spread Adjustment 0.45%Operation Twist Adjustment 1.30%Fair ROE: 7.157.78%
DCF:
Canadian equity market return: 8.50-9.00%US SP500 Electric Utility risk premiums: 3.40%Low risk US sample Median DCF: 7.02%
Given my increasing skepticism surrounding the interest rate forecasts of some of the main
economic forecasters I would judge the fair ROE to be less than the 7.50% I recommended in
2012. There is no question that there is objective evidence of a decline in overall interest rates
since 2012. However, since the long Canada yield has yet to hit my 3.80% trigger I would keep
my recommendation at 7.50%.
Q. IS THERE ANY OTHER EVIDENCE SUPPORTING 7.50%?
A. Yes. Ultimately stock market returns are driven by the returns earned by companies, that
is, what is sometimes referred to as “comparable earnings.” In 1925 John Maynard Keynes
pointed out38 that there were two sources of returns from investing in the stock market. The first
is called the investment return which Keynes defined as “forecasting the prospective yield of an
asset over its entire life.”39 In modern terminology this would be the internal rate of return on the
firm’s cash flows, or an approximate average return on equity. The second component he called
the speculative return, which involved forecasting the psychology of the market and what
Keynes referred to as the change in the basis of valuation. In modern terminology this would be
a change in the price earnings ratio. Keynes discussed this speculative return as being generated
by the “state of confidence” and “animal spirits” but he also pointed out it is affected by the level
of interest rates.40
Keynes’ point would be that a firm may earn 10%, but if the valuation of that firm changes by
10% then the investor would earn both a speculative return as well as an investment return. This
total return is then what we look at when we look at the returns over long periods of time on the
TSX Composite or the SP500. However, in aggregate the change in the basis of valuation cannot
go on forever. We cannot continue to have a state of high confidence any more than interest rates
can continue to increase or decrease: both of them will tend to revert back to some long run
average. However, professional investors according to Keynes are mainly concerned with
speculative returns or forecasting the change in the basis of valuation six months out. In contrast
59
38 Quoted in John Bogle, The Lessons of History, September 12, 2011, John Maynard Keynes, 1925, Review of Common Stocks as Long Term Investments, Edgar Lawrence Smith
39 This definition comes from chapter 12 of the General Theory of Employment Interest and Money, Macmillan London, 1936
40 Page 149 of the General Theory
buy and hold or fundamental investors are mainly concerned with the investment return: finding
good companies and holding them regardless.
Warren Buffet is probably the most successful fundamental investor of the last fifty years. He
repeated Keynes’ argument by stating:41
“The most the owners in aggregate can earn between now and judgment day is what theirbusinesses in aggregate earn.(italics in original) True by buying and selling that isclever or lucky, investor A may take more than his share of the pie at the expense ofinvestor B. And yes, all investors feel richer when stocks soar. But an owner can exitonly by having someone take his place. If one investor sells high, another must buy high.For owners as a whole, there is simply no magic - no shower of money from outer space– that will enable them to extract wealth from their companies beyond that created by thecompanies themselves.”
Buffet’s main criticism was for the financial professionals who help individuals to trade so that
in aggregate investors lose part of the pie in fees. However, Keynes, Bogle and Buffet all point
out the basic fact that short run returns can deviate from the returns generated by the economy
and earned by firms, the investment return, but in the long run this is all there is.
This discussion of what generates stock market returns is provided since in the long run the
average stock market return should approximate the average investment return,42 that is the
speculative return should average out to zero. There are two ways in which we can look at the
investment return; the first is to look at average rates of return on equity and the second to look
at a Gordon growth model for the economy as a whole.
In Schedule 3 is the average ROE for Corporate Canada since 1987 as reported by Statistics
Canada (Table # 1800003). Over this period the average ROE has been 9.56%. The third column
reports the annual return on the TSX Composite which over the same period has been 10.39% or
0.83% more. However, the rough equality over this period hides the significant year to year
variation where speculative returns have been significantly high or low. For example, in 1987
Corporate Canada earned 11.19% but the TSX Composite only 5.88%, so there was a short term
speculative loss of 5.31%. It wasn’t until 1989 than the TSX returned 21.37%, compared to
Corporate Canada’s ROE of 11.79%, that the speculative return turned positive. In each year we
60
41 Warren Buffet’s comments in Berkshire Hathaway’s 2006 Annual Report as reported in Fortune, March 20, 2006.
42 It is an approximation since it depends on the market to book ratio at the start of the period.
61
can see that the speculative return is highly volatile and on average 5-6 times more volatile than
the investment return.
The second way of looking at the investment return is that used by Jack Bogle, the founder of
Vanguard Mutual funds. He estimated the investment return using the Gordon model, where at
the start of each year he added the subsequent five year earnings growth to the dividend yield.
He then took this analysis back to 1900 and provided the graph in Schedule 4. This marginally
understates the investment return since he should have used the forecast dividend yield, but as he
noted it did not materially affect the results. He estimated this investment return at 8.8% or
slightly less than the average US stock market return of 9.1%. However, since he underestimated
the investment return the difference in reality is de minimis. Just like Keynes, Bogle also noted
the persistent tendency for reversion towards the mean, which is another way of saying that high
or low stock markets and PE multiples do not last. As Bogle noted (page 11)
“Over the long run it is the durable economics of enterprise – enterprise – that hasdetermined total return: the evanescent emotions of investing – speculation –so importantover the short run, has ultimately proven to be meaningless.”
The approach of Keynes, Buffet and Bogle is a standard approach used by fundamental investors
who look at individual stocks, rather trying to time the market. However, it is sometimes used to
time the market over a long horizon. In a US Equity Strategy Report (July 18, 2012) RBC
pointed out that historic long run equity returns in the US had been 9.4% nominal or 6.2% real
since 1900. However, going forward they used what they termed a “Grinold-Kroner-Siegel”
supply side model, which is actually just the modified DCF model, where the long run equity
return (R.) was equal to:
This equation says that the long run return is equal to the dividend yield minus share dilution,
caused by stock issues, plus inflation and real earnings growth, plus the change in the PE
multiple.
RBC placed the dividend yield at 2.1% and while they judged average share dilution from new
issues to be 2.0% going forward they expect increased share buybacks to cause this to drop to
-0.50% so the adjusted dividend yield is 1.60%. They anticipated inflation in the US at 2.1% and
real growth of 3.8% for a nominal growth forecast of 5.9% almost identical to my estimate for
the Canadian market. RBC then forecast that the normalised Shiller PE ratio will move back to
16-18X partly due to changing demographics in the US as the proportion of peak savers
(demanders of equities) drops as the baby boomers age. They forecast that this would cause a
change in the basis of valuation reducing the long run (ten year) equity market return by 1.0%.
RBC’s long run forecast is therefore for a 4.9% US equity market return. The following graphic
pulls together the RBC forecast:
I am not as pessimistic as RBC and do not put much faith in the Shiller PE model,43 but the share
dilution adjustment is one that many make since increasingly senior executives are taking some
of the price appreciation through the exercise of executive stock options. However, it indicates
that my own estimate of 8.50-9.0% for the Canadian market as a whole is not “low”.
With the stock market as a whole earning about 9.0% and Canadian regulated utilities
unambiguously lower risk a 7.5% ROE looks reasonable.
62
43 The problem is the Shiller PE uses a ten year backward average of real earnings, which in the US includes the huge losses from the Great Recession and before that the Internet Bubble. I don’t think these scale of losses will happen until we forget the lessons of the Stock Market crash, probably in another 30 years.
63
Q. WOULD YOU RECOMMEND THE USE OF AN ROE ADJUSTMENT
MECHANISM?
A. Yes. I agree with the BCUC’s decision on an ROE adjustment model. Adjustment models
offer the potential to keep the ROE current without extensive hearings. However, as indicated at
that time, and confirmed by my current analysis, keying the fair ROE off the long Canada bond
yield forecast currently causes significant problems due to the distortions in that forecast. My
judgment is that the fair ROE has not decreased to the same extent as the decrease in the long
Canada bond yield, since they are currently not being set by investors but “global policy makers”
that is, central banks. In this respect, it is not the normal application of monetary policy that is
the problem, but the bond buying programs adopted by the most important central banks. As a
result, adjusting the ROE by 50-75% of the decrease in long Canada bond yields I judge to
potentially under-estimate the fair ROE. In 2012 I stated:
I would recommend an ROE adjustment model where the ROE adjusts by 75% of theforecast change in the long Canada bond yield and 50% of the change in the creditspread. This would be subject to a minimum forecast long Canada bond yield of 3.80%and my going in ROE recommendation. Similar models are in use by the Regie and OEB.
In the hearing I also indicated that I currently had no problem with a 50% adjustment to changes
in the long Canada bond yield since the inflation risk premium has dissipated. However, at that
time I expected that the end of QE3 in the US would lead to some normalisation in the long bond
yield, but the expansion/introduction of similar programs elsewhere, as well as the sheer scale of
the bonds purchased by the US Federal Reserve, has made me more pessimistic.
I still regard 3.8% as a minimum long Canada bond yield consistent with investors trading off
risk and return, since this equates to a negligible real after tax rate of return for a taxable
investor. RBC and other forecasters expect this minimum to be reached in 2018, but I am not as
optimistic and much depends on when, if ever, the Federal Reserve starts to sell its stock-pile of
government debt. Consequently, I am also happy for the BCUC to set a fixed rate for the period
2016-2018 and if this is considered would recommend a fixed rate of 7.50%, which is generous,
since I don’t think one year ahead forecast long Canada bond yields will increase to the 3.80%
trigger in the immediate future (next three years).
V CAPITAL STRUCTURE & BUSINESS RISK
Q. WHAT IS YOUR OVERALL JUDGMENT ON FEI’s CAPITAL STRUCTURE?
A. As indicated in my full analysis three years ago44 I judge the best way to handle capital
structure as the approach adopted by the National Energy Board, the Alberta Utilities
Commission, the Regie and the Ontario Energy Board, which is to determine the capital structure
based on the business risk of the utility. Utilities with higher business risk should then have more
common equity, so that less financial risk offsets higher business risk to equalise total risk. A
board can then award the same allowed ROE from a generic cost of capital hearing. This
normally works, but sometimes an adjustment to both the common equity ratio and the ROE is
necessary, particularly if it results in an inefficient capital structure. For FEI I see no material
changes since 2012 and make the same recommendation for 35% common equity as I made then,
but will address three issues: changes in FEI’s business risk; 2) the implications of amalgamation
and 3) financing issues and the capital attraction standard.
Q HOW WOULD YOU ASSESS BUSINESS RISK?
A. I agree with the NEB where in RH-4-2001 they differentiated between short run and long
run risk. Short run risk is the ability to earn the allowed ROE and reflects the return on capital.
Long run risk is the return of capital and reflects the ability of the utility to recover its
investment in plant and equipment, that is, capital recovery risk.
64
44 I have not repeated pages 6-26 of that evidence which is largely generic, but my current recommendation is based on the same principles.
65
The NEB stated that for the TransCanada Mainline (Page 24 of the Decision)
At that time the NEB pointed out that the Mainline had been able to earn its revenue requirement
(and allowed ROE) but that the possibility that it may not be able to do this in the future had
increased. The NEB has subsequently increased the Mainline’s common equity ratio from 30%
to 40% in several hearings to reflect its increased capital recovery risk
However, long term risks must eventually become short run risk to have any impact. As I stated
before the NEB in RH-4-2004,
“If problems occur, then firms bring these problems to the regulator and frequently“compromises” are worked out. This is part of the regulatory bargain and only regulatedfirms have this capability. For example if a competitive firm suffers a supply shock thenthe stockholders are directly affected, but in contrast a regulated firm can have losses putin a deferral account and allocated to future customers or apply to the regulator for othermeans of protecting the stockholders from loss. Consequently it is unreasonable to expectno action on the part of the regulator to the increased risk after year 11 in the aboveexample.
The increased risk after year 11 is referring to the present value of the cash flows beyond year
10, which I arbitrarily referred to as long run risk. The point is that when serious risks do arise it
is extremely rare for a Canadian utility not to come before the regulator to ask for some
reallocation of costs to keep the shareholder whole. This is what happened with the Mainline in
the RH-03-2011 hearing before the NEB, where TransCanada asked for costs to be reallocated
from the Mainline to customers of NGTL; a reallocation of depreciation across different zones to
avoid stranded costs; and significant changes in its rate design. The NEB did not allow all of the
66
changes asked for, but the fact is there was a hearing and policy measures to deal with the
Mainline’s problems.
Q. IS THE BCUC AN UNRESPONSIVE REGULATOR?
A. No, not in my judgement, in fact exactly the exactly the opposite. In 2002 the BCUC had
to decide how to protect Pacific Northern Gas (PNG) from significant load losses. In response it
allowed a special Industrial Customers Deliveries Deferral Account to capture the difference
between forecast and actual sales to Methanex and some other big industrial customers. It also
approved a longer term contract with Methanex at reduced tolls to keep it on PNG’s system. This
is the normal regulatory response in Canada, which is to set up a deferral account to capture hard
to estimate items to make sure that they are passed on to customers and not born by the utility’s
shareholders. No regulator can protect a utility from basic underlying economic forces, such as
the disappearance of demand, but it can alleviate them. I would assume that this is also why the
BCUC largely denied PNG’s request for more common equity and a higher allowed ROE:
increasing these and with them the revenue requirement and tolls does not make a utility more
competitive.
In a current hearing before the Newfoundland and Labrador Board of Commissioners of Public
Utilities, in Appendix A to their October 16, 2015 report Concentric Energy rates the BCUC as
the “best” Canadian regulator based on a point system allocation of their DBRS ratings on the
following factors:
Deemed1)Allowed ROE2)Energy cost recovery3)Cost of service vs incentive rate making4)Capital cost recovery5)Political interference6)Retail rate7)Stranded costs8)Rate freeze9)Market structure10)
DBRS is Canada’s premier rating agency and the rating seems to be based on their credit rating
support, not protecting the public. Each factor is rated on a scale from 1-5 with 5 being the best
and then seemingly added up. Although, this seems to treat all factors equally, BC with 40 is
regarded as Excellent and is the highest rated regulator in Canada. What this implies is that
Concentric sees the BCUC as the most supportive regulator in terms of its DBRS bond rating.
The following graphic reproduces Concentric’s full list of scores:
I don’t necessarily agree with Concentric’s analysis, since there is significant overlap in some of
the categories and the ones that count are the ability of the utility to earn their allowed ROE and
the treatment of stranded costs. However, I would agree with the general assessment that the
BCUC is a very supportive regulator.
Why this assessment is important is that Moody’s reported on how it determined its bond ratings
in August 2009.45
The report refines their assessment into four major areas as follows:
Regulatory framework: 25%
67
45 Infrastructure Finance; Regulated Electric and Gas Utilities, August 2009.
Ability to recover costs and earn profits: 25%Diversification: 10%Financial strength and liquidity: 40%
The weights in the last column are the percentage allocation to each of the four categories.
Moody’s states very clearly “for a regulated utility the predictability and supportiveness of the
regulatory framework in which it operates is a key credit consideration and the one that
differentiates the industry from most other corporate sectors.” A quick glance at Moody’s
weights indicates that fully 50% of the weighting is based on the first two criteria which both
reflect the supportiveness of the regulatory environment.
My judgement would be that the BCUC is a very supportive regulator and that this is the most
important aspect of its risk: certainly from the bond rating agencies perspective. I would judge
that FEI would be treated fairly if any of its risks ever materialise and that the company and
utility analysts make a fundamental mistake when they ascribe risks to the utility shareholders
since if these risks ever materialise as far as possible I would judge that the BCUC would allow a
new deferral account and the costs will be allocated to rate payers. At Schedule 6 is a list of
FEI’s deferral accounts.
Q. WHAT HAS BEEN THE HISTORY OF FEI’S ABILITY TO EARN ITS ROE?
A. The following graphs FEI’s allowed versus actual ROE, this is essentially where the
“rubber meets the road.” In a dictionary sense, risk is defined to be “the probability of incurring
harm.” In financial terms, harm is generally meant to mean a financial loss, so that we can
examine the ability of FEI to earn its ROE as an objective measure of risk.46
68
46 FEI did not provide full data for 2002 I think because it changed its year end.
69
FEI ROE
8.00%
8.50%
9.00%
9.50%
10.00%
10.50%
11.00%
11.50%
12.00%
12.50%
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
Allowed ACTUAL
Over the period since 1994 the difference between FEI’s allowed and earned ROE, prior to any
sharing from performance based regulation, has been 0.86%. This average is slightly high due to
a 0.92% under earning in 1994. The more recent performance since 2003 when there is
uninterrupted data is for a 1.17% over earning. To all intents and purposes FEI’s shareholders
have not suffered any losses or experienced any risk.
I first testified in an ROE hearing before the BCUC with my late colleague Dr. Berkowitz in
1993 and since then have never seen FEI (TGI, BC Gas), as the graph of actual versus allowed
ROEs indicates, suffer any material losses. This is despite the fact that at every hearing I listen to
FEI’s expert witnesses as to how risky the company is, or how that risk is invariably increasing.
Sooner or later FEI should experience some significant losses to offset the fact that it continues
to earn a significant risk premium. You have to go back to 1998 to see any loss that could be
regarded as material and even the 0.59% loss at that time has to be compared with the minimum
2.25% risk premium I recommend.
Q. FUTURE RISK
A. The answer always given to the historic absence of Canadian utilities actually
experiencing any material risk is that risk is a forward concept and it could happen. This is
correct and in the case of the TransCanada Mainline some of those risks have materialised in
terms of load losses and pipe on pipe competition. In the case of FEI the experience of the
Mainline is useful since the major concern is by pass of the utility’s assets leaving them stranded
and taken out of the rate base, since they are no longer used and useful. Whether this happens
depends on alternatives available to FEI’s customers and intervention by the government.
In terms of the basic supply and demand for natural gas, the Canadian Gas Association produces
an annual report on the industry.47 In their latest 2015 analysis the CGA indicated that Canada
has over 200 years of supply coverage at current rates of production. As with any natural
resource how much is commercially available depends on market prices, so low prices reduce
supply even though it is a boon to consumers. As a result, the 200 year estimate underestimates
actual supply since as the resource is used up we might expect prices to rise and what are
currently marginal supplies to come on stream with more drilling. Currently drilling is well
below the peaks in 2005. The CGA indicates for example that current low prices have retarded
some drilling in traditional “dry” gas areas. Regardless, the CGA data from the National Energy
Board does not indicate any supply problems. As the following graphic indicates supply has
shifted to shale gas and tight gas.
With such a plentiful cheap resource to distribute we should expect an expansion in supply
which is what has happened as there has been significant expenditure on the distribution system.
70
47 http://www.cga.ca/wp-content/uploads/2015/05/CGA-Annual-Natural-Gas-Report-2015_EN-Final.pdf
71
The following graphic indicates the total capital investment in the natural gas industry in 2014.
As the CGA explains, upstream extraction invested $19 billion, pipeline expansion another $1
billion and there was a further $2.6 billion in distribution spending. The significant amount spent
on distribution indicates that the industry does not see any long run market problems.
The reason for the significant expenditures in the natural gas industry is that it is plentiful and
cheap as the following CGA graphics indicates,
The graphic indicates that natural gas has a significant cost advantage over propane, electricity
and heating oil in the residential space heating and water heating markets. Currently the CGA
estimates that the “typical” user can save upwards of $2,000 by using natural gas rather than
competing fuels and as a result natural gas residential users have increased by an average of
100,000 users a year for the past ten years.
Q. DOES THIS GENERAL ASSESSMENT APPLY TO FEI?
A. Yes. The cost competitiveness of natural gas for residential users is particularly important
for FEI since unlike Union Gas and Gaz Metro it has very limited industrial load and revenues
from industrial users typically account for only about 7% (Table C3 company evidence) of total
revenues.48 This depends on the stage in the business cycle but residential and commercial
72
48 A recent report from Simon Fraser University in Burnaby for the CGA indicates that natural gas is particularly important for the following industries: Food and Beverage, Wood Products, Pulp and Paper, Chemical Products, Cement and Lime, and the Primary Metals industries. Most of these are also important for the BC economy, which generally ranks third behind Ontario and Quebec.
73
revenues are very important for FEI. In answer to BC (AMPC) IR 1.2.12 FEI provided the
following cost comparison for someone in the Lower Mainland using natural gas for heating.
In 2012 I pointed out that the cost advantage of natural gas over electricity had turned around
from the time of the 2009 hearing. At that time FEI claimed that natural gas had a 28% cost
advantage over electricity and its risk had increased. However, by 2012 that cost advantage had
increased to 45% and it is now 57%. Whatever “increased” risk existed at the time of the 2009
hearing, when FEI’s common equity ratio was substantially increased from 35% to 40%, has
since dissipated. In terms of fundamental risk natural gas is highly competitive for residential
space heating and water heating and is expected to continue to be so.
Q WHAT ABOUT THE HIGHER CAPITAL COST OF NATURAL GAS VS
ELECTRICTY INSTALLATIONS?
A. I don’t think there is any reason to doubt that there is a higher capital cost attached to a
natural gas installation than electricity, but this has always been the case and is recovered from
the lower running costs. As I indicated in 2012 this has been a major factor in the low
penetration rate of natural gas in the two Maritime, greenfield, natural gas distributors in Nova
Scotia and New Brunswick. Both Heritage Gas and Enbridge Gas New Brunswick have failed to
meet their initial penetration targets for this reason. However, the same argument works in
reverse: FEI has a huge installed residential customer base and they are not going to rip out their
natural gas systems and replace them with electricity given that their systems are a sunk cost and
74
natural gas is so much cheaper than electricity. Moreover the capital cost is essentially the same
across Canada, as is the declining customer use as retail customers move to high efficiency
furnaces. Despite these long known facts, recently the Ontario Energy Board has confirmed the
36% common equity ratios for both Enbridge Gas Distribution (EGDI) and Union Gas.
Consequently I do not regard them as either new or material.
Q. WHAT IS FEI’S GROWTH PROSPECTS?
A. As in 2012 FEI indicates that a risk factor is its capture rate in the residential market.
There is no denying that the Greater Vancouver area is condifying, the same as downtown
Toronto, and that natural gas capture rates in this market are lower than for traditional single
family dwellings. To some extent the amalgamation may reduce some of this impact as it reduces
some of the impact of the lower Mainland.
The following graphic from FEI indicates this.
However, I do not regard this as a risk factor. FEI continues to capture 77% of the traditional
single family and 59% of the semi-detached market, but it is inevitable that multiple dwellings
are better served by electricity.49 More to the point it is difficult to see how this affects FEI’s
ability to earn its allowed ROE and it would only affect its long run risk if it increases its cost
recovery risk. This in turn requires not that it fail to capture more new customers but that
existing residential customers drop off the system at an increasing rate and there does not seem
to be any evidence for this.
75
49 Normally one would expect single family housing to sue more energy for space heating.
76
The graphic below indicates that FEI continues to add customers.
Q. HOW DOES FEI COMPARE TO THE OTHER MAJOR CANADIAN LDCS?
A. In Table C-12 of the company’s application they provided the following comparison of
the cost competitiveness of natural gas versus electricity.
A similar chart was provided in 2012 in answer to an information request from the BCUC. The
data is consistent with that provided above for the lower Mainland. In 2012 natural gas had a
53% competitive advantage in BC and now for Vancouver it is 59%. The chart provides three
natural comparison groups: Alberta, Ontario and Quebec since each has at least one very large
major natural gas local distribution company (LDC). As a result I have normally used ATCO
Gas in Alberta, Gaz Metro in Quebec and EGDI and Union in Ontario as comparators for FEI.
The Regie regards Gaz Metro as above average risk since Quebec Hydro is a formidable
competitor as the above data demonstrates. Natural gas at the current low prices has only a 31%
competitive advantage over electricity in the home use market and in 2012 that was only 9%. So
consistent with the argument of FEI for themselves, the capital cost of switching to natural gas
has historically not justified the switch and electricity has been the fuel of choice in Quebec. As
a result, Gaz Metro has focussed on the new construction market and industrial load. To
compensate for the higher risk of Gaz Metro the Regie allows it a 0.25% to 0.35% premium to
its ROE over the low risk company for an 8.90% allowed ROE.50 In addition, the Regie allows a
38.5% common equity and an additional 7.5% in deemed preferred shares. Like the Regie I have
77
50 By implication the Regie regards the benchmark as an 8.60% ROE. This has not changed for some time as the ROE has been settled through negotiation.
traditionally viewed Gaz Metro as the highest risk of the very large gas LDCs. Notably the Regie
allows Hydro Quebec Distribution 35% common equity and an 8.2% ROE.
Traditionally the other slightly higher risk gas LDC has been Union Gas which has been allowed
a 0.15% premium to the ROE allowed EGDI. This was also justified by its acquisition of Centra
Gas Ontario when it assumed a large industrial load, in addition to its existing industrial load.
However, in recent negotiations and settlements this premium seems to have disappeared and
Union Gas currently is allowed an 8.93% ROE on 36% common equity.
The final two gas LDCs I regard as the lowest risk in Canada. ATCO Gas has primarily
residential load since most of ATCO’s industrial customers are served directly off ATCO
Pipelines. It is currently allowed an ROE of 8.3% on 38% common equity, but that includes a
1% premium for the impact of the “financial crisis” and its aftermath. There is a hearing
scheduled by the AUC for the Summer and this premium might finally be removed, but the
AUC’s underlying judgement is that ATCO Gas has a 37% common equity ratio. Generally
EGDI is regarded as Canada’s premier utility since it serves the Greater Toronto area with 5.5
million people and has mainly residential load. It had a business risk hearing in 2013 when its
common equity ratio was confirmed at 36%, but has not had a litigated ROE hearing since 2004.
It currently has an allowed ROE of 9.19%.51
Relative to these gas distributors I continue to rate FEI as in the same risk bucket, or perhaps
slightly riskier, than EGDI and ATCO Gas and lower risk than either Union Gas or Gaz Metro.
This seemed to also be the judgment of Ms. Kathy McShane, FEI’s long-time expert witness in
cross examination in 2012. However, to a great extent these risk differences are tiny relative to
the fact that they all distribute a very competitive fuel source that is an abundant supply.
One final comparator is Nova Scotia Power Inc. (NSPI), which is the provincial electricity
provider in Nova Scotia. NSPI has a monopoly on the supply of electricity in Nova Scotia as the
former crown corporation and has recently been allowed a comprehensive fuel adjustment
mechanism that acts much like the RSAM for FEI in removing commodity price exposure. The
“competition” from Heritage Gas in Nova Scotia is also limited and not material. NSPI
78
51 OEB letter dated October 15, 2015.
negotiated a settlement on September 14, 2012 where it maintained its 37.5% common equity
ratio in return for a 9.0% ROE.
Based strictly on these comparators, and not my own judgment, FEI’s common equity ratio
should be about 37% in a range from 36% (Union and EGDI) to 38.5% (Gaz Metro). However, I
continue to recommend the same 35% common equity I have traditionally recommended for
both gas and electricity distribution companies. My normal ranking places distribution
companies as slightly higher risk than transmission companies for which I recommend a 30%
common equity ratio. The Regie allows Hydro Quebec Distribution 35% common equity and
Hydro Quebec Transmission 30% common equity.
Q. DOES FEI FACE HIGHER POLITICAL RISK?
A. In its own assessment of its business risk FEI indicated that almost all categories used by
the BCUC to assess its business risk indicated that there were no changes since 2012. In total 19
categories indicated the same risk, one lower and four higher. Two of these I can dismiss since
lower commodity prices indicate lower risk, but the volatility of them higher risk. In my
judgment there is limited potential for increasing commodity prices, given the previously
mentioned assessment of supply, and volatility is only important if it means that natural gas
prices can materially affect the competitive margin against electricity, which I doubt.52 I also
dismiss aboriginal rights since this is a governmental and not an FEI factor and I judge that the
BCUC will continue to protect FEI. Finally, security of supply has not changed for FEI and I
will deal with amalgamation issues separately.
This only leaves higher political risk. There is no question that the reduction in greenhouse gas
(GHG) emissions is something that may affect everyone in BC. However, many parties accept
that natural gas is a solution to the GHG problem, not the cause. This is true in transportation, for
example, as well as in electricity generation. In terms of the distribution system itself, the
Canadian Gas Association provided the following graphic.
79
52 FEI confers with the BCUC every three months to make sure that the cost of gas charged to ratepayers is sufficient to cover its cost and associated pipeline service and storage. The volatility of natural gas prices only affects FEI, if at all, to the extent that a very high price may encourage some to use alternative fuels. This is highly unlikely at the moment and any impact is not material.
Despite the continued expansion of the natural gas distribution system, green-house gas
emissions from the system (GHG) are declining. What this means is that the distribution system
itself is getting cleaner.
I do not see the provincial government requiring that residential users modify their natural gas
space heating in order to replace them with more expensive electricity. It would be easier and
more effective to use a cap and trade system, where the province buys credits from the reduction
of GHG emissions elsewhere to meet its targets. Further, I do not see slower growth prospects as
a risk factor, since it does not affect the value or the risk of assets in place. If FEI does see the
provincial government requiring the removal or modification of natural gas heating systems, the
correct response is a depreciation study to depreciate the assets more quickly and reduce any
stranded asset risk. In this way t keeps FEI whole in terms of its risk exposure.
In my judgement FEI is marginally lower risk than in 2012 and I do not see any basis for the
conclusion that its risk has increased.
VI FINANCING AND CONCLUSIONS
Q. DO YOUR RECOMMENDATIONS SATISFY THE FAIR RETURN STANDARD?
A. Yes. My recommendations are based on the fair return standard. The most basic thing to
remember is that my recommendation for an ROE of 7.5% is approximately 3.5% over the
company’s long term borrowing cost, 5.50% over the current long Canada bond yield and
approximately 2.0% over the preferred share yield should FEI decide to issue preferred shares.53
In my judgement it can attract capital at my recommended financial parameters.
In terms of its “financial metrics” I am extremely reluctant to benchmark my recommendations
against guidelines issued by the rating agencies, such as Moodys, for two reasons. First, DBRS
has long maintained the exact same “A” rating on FEI and its predecessor companies through
periods when it had a 33% common equity ratio, a 35% common equity ratio, a 40% common
equity ratio and most recently a 38.5% common equity ratio. The following graphic from DBRS
(March 18, 2014) indicates its credit history back to 2001.
Also as DBRS indicates FEI has had an interest coverage ratio below 2.0 on many occasions in
the past and it is a variant of this ratio that FEI focusses on in its evidence. The fact is that DBRS
has given FEI an A rating for the last 15 years even during periods when it had 33% common
equity and an interest coverage ratio below 2.0. The idea that it would change this rating if the
BCUC were to give it the same common equity ratio as ATCO Gas, EGDI and Union, or the
33% it used to have, is not reasonable.
81
53 Fortis, FEI’s ultimate parent issues preferred shares on a regular basis and there would be no problem “mirroring” those costs down to FEI as is done with the subsidiaries of Canadian Utilities such as ATCO Gas, having FEI issue preferred shares or the BCUC simply deeming them as is done by the Regie for Gaz Metro.
82
I would also note that DBRS has maintained the A rating despite the introduction of performance
based regulation (PBR) that previously FEI and its witnesses have introduced as a risk factor. In
fact, Moody’s explicitly stated in their July 15, 2014 credit report.
“TRANSITION TO PBR EXPECTED TO HAVE MINIMAL CREDIT IMPLICATIONSFortisBC utilities have submitted detailed PBR proposals for both FEI and FBC for theperiod 2014-2018. We have assumed that it does not represent a material change in riskand that the company continues to earn its allowed ROE. The proposed PBR plan isbroadly similar to the previous PBR plan and would have both an annual and mid-termreview. FEI's proposal would set controllable O&M and non-CPCN (CPCN includeslarge capital projects that currently require regulatory pre-approval) capex by formulawith substantial costs remaining as pass through items. The proposal contains aproposed symmetrical earnings sharing mechanism on up to 200bps and is subject tomeeting service quality targets. Performance above or below the allowed ROE by morethan 200bps would trigger an automatic review of the PBR plan. There are no proposedchanges to key deferral accounts. While we don't expect it, a key risk to the proposal isthat the regulator adopts very difficult efficiency targets within the formula. The PBRplan does not propose to modify support for CPCN capex. A final decision on the PBR isexpected from the regulator in Q3 or Q4 2014. FEI previously”
The second reason is that the guidelines are heavily based on the degree of regulatory protection,
where 50% of the weight applied by Moodys is explicitly for this and not the financial metrics.
Consequently, the metrics are not the most important issue. As Moody’s also stated in its July
15, 2-14 credit note
“FEI's investment grade rating is driven by its credit supportive regulatory environmentand its monopoly position. Rates are typically set using a cost of service framework and aforward test year that enables the company to recover its costs and earn an allowedreturn established by the regulator, resulting in stable cash flow. The company has atrack record of passing through its commodity costs in rates and has no direct exposureto commodity price risk and limited volume risk. To the extent that these and many othercosts differ from forecast values, deferral or true up mechanisms limit exposure toforecast error. As a result the company has a long track record of earning the return onequity (ROE) established by the regulator.”
And most importantly
“The company benefits from a monopoly position. We believe that its customers, who areprimarily residential, continue to have the capacity and willingness to pay their bills.”
I therefore judge regulation by targeting what the rating agencies say they want in one part of
their analysis to the exclusion of the most important issues in other parts as being incorrect. I
would also add that recently Moody’s has changed its view of US regulatory protection. In a
request for comment on September 23, 2013 Moody’s stated:
“Our revised view that the regulatory environment and timely recovery of costs is in mostcases more reliable than we previously believed is expected to lead to a one notchupgrade of most regulated utilities in the US, with some exceptions. This evolving view isindependent of the proposed changes in the methodology that are highlighted in theSummary section that follows, and would have taken place even if the 2009 methodologywere to remain in place without modification.”
The comment basically says that since the regulatory protection afforded US utilities seems to
have increased it will pretty much apply a one notch upgrade to their credit ratings independent
of their credit metrics. To the extent that Moody’s has traditionally viewed Canadian regulation
as more protective than that in the US, this comment indicates that we can take the US guidelines
and add a notch for Canadian utilities, rather than just reading off from the guidelines. This
vitiates their usefulness in indicating what guidelines produce what ratings in a mechanical
manner. 54
I also agree with the BCUC statement in its 2013 decision when it stated
“The Commission Panel is supportive of maintaining an “A” category credit rating butonly to the extent that it can be maintained without going beyond what is required by theFair Return Standard.”
The fair return standard trumps the relevance of a particular bond rating since there are other
ways of ensuring market access without giving the shareholder an unfair rate of return.
Q. ARE SOME FINANCIAL METRICS IMPORTANT?
A. Yes. FEI is to some extent constrained by the prospectus of the actual entity that issues
its medium term notes (MTNs) (unsecured long term debt). To issue MTNs a 2.0X new issue test
based on its interest coverage ratio (ICR) has to be met. FEI analyses in depth (evidence pages
27-31) the implications of this restriction as it refinances its purchase money mortgages and the
implications of the BCUC reducing its financial parameters to 37% common equity and an
8.25% allowed ROE. Note that these financial parameters are similar to what other premium
83
54 Note this is not to say that Moody’s views Canadian and US utilities as equivalent in risk. This is a request for comment and reflects Moody’s “evolving” view.
LDCs are allowed in Alberta and Quebec so their choice does not seem to be arbitrary, but
instead an acknowledgement that these parameters are reasonable.
To understand how the interest coverage restriction (ICR) works I reproduce below the filing
with the Ontario Securities Commission that FortisBC Energy made on November 6, 2015,
Note that for the ICR the consolidated interest was $140 million for 2014 and consolidated
earnings $317 million. So the ICR is simply 317/140 or 2.26. This had decreased marginally to
2.21 by September 2015. Note that in the ICR neither the interest charge nor the tax charge are
the actual FEI expenses. FEI’s total interest in 2014 as in the application was $189 million which
includes $136 in interest on the long term debt (MTNs), as well as $55 million in interest on
short term debt. However, whereas consolidated earnings includes the total interest expense, the
ICR only looks at funded or long term interest. Similarly the ICR includes the notional tax
charge, since it specifically excludes the impact of FEI’s “tax loss utilisation plan” or TLUP. In
2014 the use of TULP effectively lowered FEI’s tax expense from $35 million to $23 million but
it is the notional $35 million tax expense that is used to generate the consolidated earnings in the
ICR test.55
84
55 TULP is a tax loss sharing scheme between controlled affiliates, where one sells preferred shares to another and receives non-taxable dividend income whereas the other buys debt where the interest is taxable. I assume the TULP has been approved by the BCUC and only actual interest expense is included in the revenue requirement.
There are some timing differences in the numbers used in the ICR as there are some smoothing
options, but the net result is that FEI has considerable financing flexibility and is not currently
constrained by the ICR in issuing MTNs. For example, in August 2015 FEI negotiated a
syndicated $700 million credit facility of which approximately half is currently unused. On April
8, 2015 FEI issued $150 million of 30 year MTNs at 3.375% using the proceeds to repay short
term debt. At the time the 30 year long Canada bond (Cansim V39056) was yielding 2.03% for a
1.35% credit premium.56 In September 2015 it repaid $75million of the PMMs with short term
notes.
Q. SHOULDN’T THE COMMON EQUITY RATIO BE INCREASED DUE TO THE
AMALGAMATION?
A. No. The BCUC approved the amalgamation effective January 1, 2015 and the financial
parameters of the amalgamated FEI continued at FEI’s allowed 38.5% common equity and
8.75% ROE. The effects of the amalgamation should have been reflected in the interest coverage
ratio filing of 2.21 on November 15, 2015, so the amalgamation per se has not had a material
effect on FEI’s market access. Further both DBRS and Moody’s reviewed FEI after the
amalgamation and did not change their credit rating. FEI did note in its November 6, 2015
management and discussion analysis (MD&A) filed with the OSC that the impact of the
amalgamated financial parameters was $3 million for the first nine months of 2015 and in answer
to BC (AMPC) IR # 1.1 that it would have been about $4 million in 2014. I do not regard such
numbers as material in affecting FEI’s financial flexibility for three reasons:
1) FEI has significant financial flexibility under its current financial parameters2) Under PBR I expect FEI to continue to over-earn its allowed ROE, so that its ICRis improved beyond the values implied by the regulated ROE and common equity ratios.3) Other financial strategies such as TULP may generate savings that also enhancethe ICR.
FEI concludes this analysis by stating (Company evidence page 31)
“Under certain circumstances, and in a period of higher debt capital requirements, debtissuance may be constrained. An increase in equity and/or allowed ROE would bereasonable to support issuance capacity going forward.”
85
56 The MD&A does not indicate the price received. RBC reports the spread on FEI seasoned debt on that day at 1.44%.
I disagree with this statement. First, the conclusion is based on highly speculative assumptions
and if such circumstances do obtain FEI can come to the BCUC for relief. Second, the analysis
does not take into account how the ICR actually works, where on page 25 of the company
evidence it specifically only includes $96.31 million in funded debt interest, whereas the ICR
uses all interest in calculating consolidated earnings.57 Third and most importantly, the equity
holder should not be awarded a higher common equity ratio or ROE due to issue problems in the
bond market. Put another way FEI should not be allowed more than the fair return standard due
to bond market problems. If any such problems exist, and I don’t think they currently do, they
can be addressed with a short term solution, such as issuing term preferred shares. However,
since other Canadian gas LDCs also have this interest coverage restriction, as well as similar
financial metrics that satisfy the fair return standard, I am not persuaded that FEI’s shareholder
should get such special treatment.
Q. WHAT ARE YOUR ROE AND COMMON EQUITY RATIO
RECOMMENDATIONS?
A. I judge a benchmark ROE for 2016 to be 7.157.78%, but consistent with my judgment in
2012 would not change the allowed ROE until the forecast long Canada bond yield exceeds
3.80%. I therefore recommend a continuation of both my ROE and ROE adjustment mechanism
recommendations from 2012 for 2016. In terms of the common equity ratio my personal
recommendation is for 35%, which is the same that the Regie allows a distributor in Quebec and
would point out that a comparison with other Canadian LDCs would indicate a 37% common
equity ratio, which is marginally higher than that allowed EGDI and Union Gas in Ontario.
Q. DOES THIS CONCLUDE YOUR TESTIMONY?
A. Yes, at this time.
86
57 I don’t know why there is a different approach between FEI’s statutory filings and its current evidence.
Schedule 1
Macro data (* indicates year to date) Unemployment Real CPI T Bill Canada FX Rate Average
Rate Growth Inflation Yield Yield US$ ROE1987 8.81 4.04 4.42 8.17 9.93 0.75 11.191988 7.77 4.74 3.94 9.42 10.23 0.81 12.971989 7.58 2.38 5.06 12.02 9.92 0.84 11.791990 8.16 0.13 4.81 12.81 10.81 0.86 7.481991 10.32 -2.12 5.61 8.83 9.81 0.87 3.531992 11.24 0.85 1.45 6.51 8.77 0.83 1.561993 11.42 2.61 1.90 4.93 7.88 0.78 3.691994 10.43 4.55 0.12 5.42 8.58 0.73 6.571995 9.54 2.74 2.22 6.98 8.35 0.73 9.551996 9.73 1.68 1.48 4.31 7.54 0.73 10.291997 9.16 4.25 1.69 3.21 6.47 0.72 10.861998 8.35 4.14 1.00 4.74 5.45 0.67 8.831999 7.58 5.00 1.75 4.70 5.68 0.67 9.822000 6.85 5.12 2.69 5.48 5.92 0.67 10.922001 7.23 1.69 2.52 3.85 5.79 0.67 7.412002 7.66 2.80 2.25 2.57 5.67 0.65 5.692003 7.61 1.93 2.80 2.87 5.29 0.72 9.652004 7.18 3.14 1.85 2.27 5.08 0.77 11.622005 6.77 3.16 2.21 2.71 4.41 0.83 12.702006 6.32 2.62 2.00 4.02 4.29 0.88 13.952007 6.03 2.01 2.14 4.17 4.32 0.94 12.872008 6.15 1.18 2.37 2.62 4.06 0.94 9.442009 8.23 -2.71 0.30 0.40 3.85 0.88 8.062010 7.99 3.37 1.78 0.50 3.71 0.97 9.972011 7.46 2.96 2.39 0.94 3.22 1.01 9.922012 7.29 1.92 2.03 0.96 2.35 1.00 10.682013 7.07 2.00 0.94 0.98 2.71 0.97 9.822014 6.90 2.44 1.91 0.91 2.65 0.91 10.372015 6.90 2.03* 1.11* 0.54 2.06 0.78
Cansim V13682111 v62305752 v41690973 V122484 V122501 V37426 V634672/V634628
Schedule 2
1
CANADA BOND YIELDS
Overnight money market rates 0.52
Benchmark bonds
Canada 91 day Treasury Bill yield 0.45
Canada Six month Treasury Bills 0.44
Canada One year Treasury Bills 0.44
Canada Two year 0.41
Canada Three year 0.43
Canada Five year 0.66
Canada Seven year 0.96
Canada Ten year 1.32
Canada Long term (30 year) 2.07
Canada Real return bonds 0.63
Marketable Bond Average yields
Canada 1-3 year 0.42
Canada 3-5 year 0.55
Canada 5-10 1.06
Canada Over tens 1.95
Source: Bank of Canada’s web site at http://bankofcanada.ca/en/securities.htm, for January 8, 2016.
Schedule 3
Fama-French Application
Schedule 4
Investment and Speculative TSX Returns back to 1987
ROE Spec TSX1987 11.19 -5.31 5.881988 12.97 -1.89 11.081989 11.79 9.58 21.371990 7.48 -22.28 -14.801991 3.53 8.48 12.021992 1.56 -2.99 -1.431993 3.69 28.86 32.551994 6.57 -6.75 -0.181995 9.55 4.98 14.531996 10.29 18.06 28.351997 10.86 4.12 14.981998 8.83 -10.42 -1.581999 9.82 21.90 31.712000 10.92 -3.51 7.412001 7.41 -19.98 -12.572002 5.69 -18.12 -12.442003 9.65 17.08 26.722004 11.62 2.86 14.482005 12.70 11.43 24.132006 13.95 3.31 17.262007 12.87 -3.04 9.832008 9.44 -42.44 -33.002009 8.06 26.99 35.052010 9.97 7.64 17.612011 9.92 -18.63 -8.712012 10.68 -3.49 7.192013 9.82 3.18 13.002014 10.37 0.18 10.55
Schedule 5
Jack Bogle’s Investment and Speculative Returns in the US back to 1900
Schedule 6
FAIR RETURN AND CAPITAL STRUCTURE FOR FORTIS
BC ENERGY
EVIDENCE OF
Laurence D. Booth
BEFORE THE
BC Utilities Commission
Corrected February 3, 2016
i
TABLE OF CONTENTS
EXECUTIVE SUMMARY ....................................................................................................................... 1
I INTRODUCTION AND OVERVIEW ....................................................................................... 3
II FINANCIAL AND ECONOMIC OUTLOOK ........................................................................... 8
III RISK PREMIUM ESTIMATES OF THE FAIR ROE ........................................................... 36
IV DCF ESTIMATES OF THE FAIR ROE .................................................................................. 50
V CAPITAL STRUCTURE & BUSINESS RISK ........................................................................ 64
VI FINANCING AND CONCLUSIONS........................................................................................ 81
APPENDIX A: PROFESSOR BOOTH’S CURRICULUM VITAE
APPENDIX B: MARKET RISK PREMIUM ESTIMATES
APPENDIX C: RELATIVE RISK FOR A BENCHMARK UTILITY
APPENDIX D: DCF ESTIMATES
APPENDIX E: ALTERNATIVE MODELS TO THE CAPM
1
EXECUTIVE SUMMARY 1
1. The Canadian economy has stalled somewhat as resource prices have collapsed following 2
a slowdown in China and with them the value of the Canadian $. This has hurt the resource 3
intensive sectors like Alberta, but added extra stimulus to others, in particular, Ontario and 4
Quebec. As the economy adjusts to this change short term growth has weakened, but is expected 5
to get back on track in 2017. The BC economy is one of the strongest in Canada in terms of 6
economic growth. 7
2. Although the US Federal Reserve has stopped its bond buying program (quantitative 8
easing or Operation Twist), others are still very active and there is now a huge block of 9
government bonds that have been taken off the bond market increasing prices and lowering 10
yields from where they would otherwise have been. This has led to continuing low long Canada 11
bond yields causing corporate credit spreads over government bond yields to remain high at 191 12
bps. This is mainly due to unusually low government bond yields not higher credit risk, since all 13
the standard stress indicators show normal capital market conditions. 14
3. Canadian utilities have issued 40 and in some cases 50 year bonds at extremely low 15
interest rates and FEI recently issued 30 year MTNs at a little over 1.35% spread over equivalent 16
maturity long Canadas.. 17
4. I have been recommending a credit spread adjustment of 50% of the change in the credit 18
spread from the normal spread of 100bps to my ROE recommendations. This adds 45 bps to 19
simple CAPM estimates, but should even out over the business cycle. I regard this as simply 20
converting the simple CAPM into a Contingent CAPM or CCAPM as required by theory. 21
5. However, the credit spread adjustment does not adjust for the overall drop in bond yields 22
as partly indicated by the behaviour of preferred share yields. Since the US embarked on its third 23
round of quantitative easing the traditional spread of preferred share yields over both government 24
of Canada and generic A rated corporate bonds has changed. This indicates that current long 25
Canada bond yields are at least 1.30% too low based on traditional, spreads confirming my real 26
bond yield model that indicates an even higher 2.5%. 27
2
6. For 2016 I continue to recommend an ROE of 7.50% for a benchmark utility. This 1
recommendation reflects a 0.45% adjustment for credit spreads 1.3% for the impact of global 2
bond buying programs and evidence from DCF equity costs. For 2017 and later years I 3
recommend the BCUC continue with its ROE adjustment model with the 3.80% floor to the long 4
Canada bond yield. However, I do not expect this floor to be triggered over the next three years. 5
7. I accept FEI’s broad assessment that its business risk is largely unchanged since 2012, 6
but would point out there are signs of a marginal decrease in risk. Consequently, I see no reason 7
to change my standard common equity ratio recommendation of 35% which is slightly below a 8
comparator benchmark of 37%. 9
8. I disagree with the company’s assessment that it might have problems with the interest 10
coverage ratio (ICR) restriction in its bond indenture and therefore capital market access 11
problems. Its ICR analysis is deficient and with its own filing with securities regulators its recent 12
ICR was over 2.2. Currently I judge there to be no capital market or access problems for high 13
quality utilities and given the state of the financial system FEI should have no problems 14
financing itself with my recommended financial parameters.1 15
9. I would recommend no adjustment for the amalgamation of FEI which I do not judge to 16
materially alter its risk or financial parameters. This conclusion is in line with the decision of the 17
rating agencies, which seem to regard the amalgamation as not material.18
1 TD’s “action note” of May 6, 2015 notes that FEI’s ultimate parent Fortis has an ROE of 7.10% and an investment grade bond rating.
3
I INTRODUCTION AND OVERVIEW 1
Q. PLEASE DESCRIBE YOUR NAME, QUALIFICATIONS AND EXPERIENCE. 2
A. Laurence Booth is a professor of finance in the Rotman School of Management at the 3
University of Toronto, where he holds the CIT Chair in Structured Finance. Dr. Booth appeared 4
before the BCUC most recently in a 2012 “generic” hearing, as well he has appeared before most 5
of the major utility regulatory boards in Canada including the CRTC, the National Energy Board, 6
the Ontario Energy Board (OEB), the Alberta Utility Commission (AUC), the Nova Scotia 7
Utilities and Review Board, the New Brunswick Public Utilities Board, the Manitoba Public 8
Utilities Board, the Board of Commissioners of Newfoundland and Labrador, the Regie de 9
l’Energie du Quebec and the Prince Edward Island Regulatory and Appeals Commission. He has 10
also filed testimony before the Ontario Securities Commission and in a variety of civil suits 11
pertaining to financial matters. A detailed resume is filed as Appendix A. Further information 12
and copies of working papers by Dr. Booth can be can be downloaded from his web site at the 13
University of Toronto at http://www.rotman.utoronto.ca/~booth. 14
Q. PLEASE DESCRIBE THE PURPOSE OF YOUR TESTIMONY 15
A. The Association of Major Power Customers of BC (AMPC), the British Columbia Old 16
Age Pensioners Organization (BCOAPO) and the Commercial Energy Consumers Association of 17
British Columbia (CEC), collectively referred to as the Utility Customers asked me to offer an 18
opinion on the fair rate of return on common equity (ROE), and to recommend an appropriate 19
common equity ratio for FortisBC Energy Inc., (FEI) for the BCUC to use as benchmarks for BC 20
utilities. 21
Q. DO YOU HAVE SOME OVERALL REMARKS? 22
A. Yes. It is three years since the BCUC last reviewed its benchmark ROE and the capital 23
structure of FEI. Over this period we have very much been in a holding pattern waiting for the 24
US and Europe to work their way through the after effects of their Great Recessions. Further, just 25
as they have proceeded to recover, Chinese growth has slowed and sent jitters through 26
commodity markets and triggered a stock market correction. The upshot is that the stronger 27
4
markets that were expected at that time have not developed as anticipated and Alberta, in 1
particular has been badly hit. Further conditions in the bond market have become even looser 2
than they were in 2012 as the massive amount of liquidity in global markets continues to 3
increase, depressing bond yields. The result is that the forecast long Canada bond yield has yet to 4
hit the 3.8% trigger that I recommended as justifying a change in FEI’s allowed ROE. 5
In terms of FEI’s business risk the major factor is the amalgamation of FEI with its Whistler and 6
Vancouver Island subsidiaries. This amalgamation has removed most of the additional risk 7
factors faced by those companies, such that, as even FEI admits, FEI’s business risk post 8
amalgamation is similar to that of FEI in 2012 (Appendix C, page1). Although FEI suggests that 9
this risk is “trending higher” the fact is that in their business risk assessment (Appendix C, page 10
3) 19 of the 24 categories are ranked the same as in 2012. The only areas where FEI judges 11
business risk to have increased is in energy price volatility, political initiatives on greenhouse gas 12
(GHG) emissions, aboriginal rights, and the security of energy supply. In my judgement, none of 13
these “higher” risk factors are substantial enough to justify an increase in the common equity 14
ratio to 40%. 15
The upshot is that with largely stand-pat financial markets and business risk, FEI’s case for an 16
increase in allowed ROE and common equity is non-existent. In the 2012 GCOC hearings FEI 17
presented a large number of expert panels and the BCUC thoroughly evaluated both the allowed 18
ROE and FEI’s capital structure. In contrast, in the current hearing FEI has only presented one 19
external witness and requested an abbreviated hearing on financial matters be held over a very 20
short time period. The obvious implication is that FEI judges that not much has changed as well, 21
and what changes there are would indicate a lowered, not increased ROE. It is therefore 22
appropriate to first consider what in fact has changed since 2012. 23
Q. CHANGES SINCE 2012 GCOC HEARING 24
A. I will present my own evidence shortly, but it is first useful to present some information 25
that FEI’s witness (Concentric) provided in a public document circulated by the Canadian Gas 26
5
Association.2 This is the data that Concentric regarded as most pertinent as a backdrop to the 1
financial parameters for regulated utilities. The data is as of May 2015 so is not the most recent, 2
but as I will show it is quite similar to current data. Note that in 2013 when the GCOC decision 3
was made the average Government of Canada long term bond yield (LTC) averaged 2.82% and 4
by May 2015 it had dropped to 2.05% or decreased by 0.77%. Similarly, US government 5
Treasuries had dropped from 3.45% to 2.55% or by 0.90%. This drop in bond market interest 6
rates also affected utility borrowing costs. In Canada, the Bloomberg synthetic A rated utility 7
borrowing cost dropped from 4.24% to 3.50%, or by 0.74%, to all intents and purposes the same 8
decline as the LTC yield. Similarly, the Moody’s A rated utility bond yield dropped from 4.28% 9
to 3.67% or by 0.81%, slightly less than the drop in US Treasuries. 10
11
Why this is important is that FEI accepts3 that while it does not recommend an ROE adjustment 12
formula it does accept the two factor model adopted by the BCUC in its 2012 GCOC 13
proceedings. In this instance the long Canada bond yield decrease of 0.77% indicates a 0.385% 14
decline in the allowed ROE, while the credit spread has shown a trivial change from 1.42% 15
1.45%, suggesting a 0.015% increase in the allowed ROE. Overall, the BCUC’s two factor 16
model and Concentric’s data would suggest a 0.37% decrease in FEI’s allowed ROE from 8.75% 17
to 8.38%. 18
The BCUC’s ROE adjustment model would not result in a 0.38% decrease in the allowed ROE 19
since it is subject to a 3.8% floor. However, it does indicate the direction in FEI’s allowed ROE 20
and that this should be down from the current 8.75%, not up to the 9.50% recommended by 21
Concentric. Further, Concentric provided the following table of currently allowed ROEs for 22
Canadian gas companies. 23
2 Canadian Gas Association, Vol 111, May 1, 2015. 3 Main evidence page 33.
6
1
I regard the premier local gas distribution companies in Canada as being Gas Metro, Union Gas, 2
Enbridge Gas Distribution, ATCO Gas and FEI. These are the largest gas distribution companies 3
in Canada and dominate their provincial markets. Traditionally, I have regarded a risk ranking as 4
being ATCO Gas and EGDI as lowest risk with Union and FEI marginally higher and Gaz Metro 5
as the highest risk. This ranking is mainly based on their size, the diversity of the provincial 6
economies and the competitive nature of natural gas in each province. This assessment is broadly 7
in line with that of Ms. McShane who testified on behalf of FEI in the 2012 GCOC hearings. 8
Currently, the Alberta generic ROE is 8.30% and ATCO Gas is allowed 8.30% on 38% common 9
equity where 1% of that is a generic premium for market conditions. EGDI is allowed 9.30%, 10
since its ROE is based on a 2009 formula that the OEB has not recalibrated, and Union Gas 11
8.93%, both with 36% common equity.4 Gaz Metro is allowed 8.90%, but that includes an 12
additional 0.25-0.35% risk premium to reflect Gaz Metro’s risk, while the Regie’s current base 13
4 For 2016 EGDI’s ROE has dropped to 9.19% based on a forecast LTC rate of 2.706% and a 1.83% credit spread. OEB Letter, October 15, 2015. Note the Ontario government is in the process of selling off a significant part of Hydro One Transmission (HONI) which earns the OEB allowed ROE and which may explain the delay in reviewing the OEB ROE methodology. Shortly after the initial sale of HONI shares, they were selling at $21.86 whereas the book value is $15.76, that is, the market to book ratio was 1.39. This implies that the market found the HONI allowed ROE very attractive and excessive for such a low risk pure play utility.
7
allowed ROE for electric distribution and transmission is 8.20%. Gaz Metro’s allowed common 1
equity ratio is 38.5%. 2
My conclusion from this is that since the GCOC hearing the macroeconomic evidence of 3
Concentric, and the judgement of FEI on the two factor ROE adjustment model, together indicate 4
that the allowed ROE for FEI should be lowered. Further, that FEI’s requested financial 5
parameters of 9.5% ROE on 40% common equity exceed that of its peers. The more recent 6
litigated hearings in Quebec, as well as that in Alberta, indicate a generic ROE of 8.20-8.30% 7
for an energy distributor applied to a 35-38% common equity ratio.5 8
I develop my own recommendations in the subsequent sections. First, I consider the general 9
financial and economic outlook, since this is what underpins the current GCOC formula ROE. 10
Second, I consider fair ROE estimates derived from risk premium and discounted cash flow 11
(DCF) models. Here it is important to point out that the recommendation is for 2016 and forecast 12
long Canada rates are not expected to get close to the levels that Mr. Coyne forecasts and uses in 13
his CAPM estimates. Finally I discuss FEI’s business risk and the question of financial access 14
and bond ratings, and make my recommendations. 15
I have relegated most of the more detailed and technical discussion to a series of appendices. A 16
includes my resume; B has a discussion of the market risk premium; C has a relative risk 17
assessment for a benchmark utility; D discusses DCF estimates and E is a new Appendix that 18
discusses alternative risk premium models to the CAPM motivated by the 2012 GCOC decision. 19
I have not repeated the “comparable earnings” appendix since unlike Ms. McShane, Concentric 20
has not produced such evidence. These appendices are written to be free standing references 21
from the main text and are mainly standard updates from 2012. 22
5 Hydro Quebec Distribution is allowed 35% common equity.
8
II FINANCIAL AND ECONOMIC OUTLOOK 1
Q. WHY DO YOU START BY CONSIDERING CAPITAL MARKET 2
CONDITIONS? 3
A. Because the legal standard for a fair rate of return stemmed from “altered conditions in 4
the money market” where we would now understand the money market to mean the capital 5
market. The Supreme Court of Canada determined a fair rate of return in BC Electric Railway 6
Co Ltd., vs. the Public Utilities Commission of BC et al ([1960] S.C.R. 837), where the Supreme 7
Court of Canada had to interpret the following statute: 8
(a) The Commission shall consider all matters which it deems proper as affecting the 9
rate: 10
(b) The Commission shall have due regard, among other things, to the protection of 11
the public interest from rates that are excessive as being more than a fair and 12
reasonable charge for services of the nature and quality furnished by the public 13
utility; and to giving to the public utility a fair and reasonable return upon the 14
appraised value of the property of the public utility used, or prudently and 15
reasonably acquired, to enable the public utility to furnish the service: 16
This statute articulated the "fair and reasonable" standard in terms of rates, and that the 17
regulatory body should consider all matters that determine whether or not the resulting charges 18
are "fair and reasonable." To an economist, "fair and reasonable" means minimum long run 19
average cost, since these are the only costs which satisfy the economic imperative for regulation 20
and by definition do not include unreasonable and unfair cost allocations. The statute also 21
articulated the “prudently and reasonably acquired” test in terms of the assets included in the rate 22
base. 23
Most statutes also allow the regulatory authority to examine all factors that enter into the rates to 24
ensure that the rates are “fair and reasonable.” This includes the firm’s capital structure decision, 25
since this has a very direct and obvious impact on the overall revenue requirement. To allow the 26
regulated utility to freely determine its capital structure will inevitably lead to rates that are 27
9
unfair and unreasonable, otherwise the management of the regulated firm is not fulfilling its 1
fiduciary duties to act in the best interests of its stockholders. 2
In terms of financial charges, in Northwestern Utilities vs. City of Edmonton (1929), it was 3
stated that a utility's rates should be set to take into account ‘altered conditions in the money 4
market.’ A fair rate of return was further confirmed in the BC Electric decision when Mr. Justice 5
Lamont's definition of a fair rate of return, put forward in Northwestern utilities, was adopted:” 6
"that the company will be allowed as large a return on the capital invested in the 7 enterprise as it would receive if it were investing the same amount in other 8 securities possessing an attractiveness, stability and certainty equal to that of the 9 company's enterprise." 10
This definition is referred to as an opportunity cost, in that the fair return is what could be earned 11
by investing in similar securities elsewhere; only if the owners of a utility earn their opportunity 12
cost will the returns accruing to them be fair, i.e., they will neither reward the owners with 13
excessive profits, nor ratepayers by charging prices below cost. 14
To any modern financial economist Mr. Justice Lamont’s definition of a fair rate of return as an 15
opportunity cost means a risk adjusted discount rate or expected rate of return. This is the rate 16
that is determined in the capital market as conditions constantly change. 17
Q. HOW HAS MONETARY EVOLVED IN CANADA? 18
A. Basic macroeconomic data since 1987 is provided as background in Schedule 1. Into 19
2008 we had good economic growth and for a time the unemployment rate was actually below 20
the natural or non-accelerating inflation rate of unemployment (NAIRU) of 6.0%. Consumer 21
spending was strong as low interest rates supported the purchase of consumer durables and new 22
housing as starts exceeded 200,000 for the sixth year in a row. The strong investment position in 23
Canada was partly due to a dramatic improvement in Canada’s terms of trade as commodity 24
prices increased. This created a perception that Canada was a “petro,” or at least a “raw 25
materials” based, economy as commodity prices reached record highs in summer 2008. This 26
perception allied to the continuing strength of the current account surplus resulted in a 27
strengthening Canadian dollar and incipient inflationary pressures. The result was that starting in 28
10
September 2005 the Bank of Canada increased its overnight rate from 2.5% to reduce the 1
stimulus being injected into the economy. 2
The following graph shows the impact of this tighter monetary policy from January 2006, just 3
before the first signs of the financial crisis appeared. Throughout 2006 and up until December 4
2007, the Bank of Canada set the target rate to try and slow down the economy and reduce 5
inflationary pressures. Of importance is that consistent with the Bank of Canada’s 2% inflation 6
target the overnight rate should be at least 3.0%; so 4.5% up until December 2007 was 7
restrictive. The Bank pays interest on deposits that the chartered banks keep with it at 0.25% less 8
than the overnight rate and the banks can borrow at 0.25% more than the overnight rate; a rate 9
that is called the Bank Rate. Bank Prime is then about 2.0% more than the overnight rate. 10
Consequently up until December 2007 the Bank was actively trying to increase borrowing costs 11
to slow interest sensitive demand. This policy stance was reversed due to the impact of the sub-12
prime mortgage crisis coming out of the United States. 13
Canadian Overnight RateCansim 39079
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
1/2/
2006
5/2/
2006
9/2/
2006
1/2/
2007
5/2/
2007
9/2/
2007
1/2/
2008
5/2/
2008
9/2/
2008
1/2/
2009
5/2/
2009
9/2/
2009
1/2/
2010
5/2/
2010
9/2/
2010
1/2/
2011
5/2/
2011
9/2/
2011
1/2/
2012
5/2/
2012
9/2/
2012
1/2/
2013
5/2/
2013
9/2/
2013
1/2/
2014
5/2/
2014
9/2/
2014
1/2/
2015
5/2/
2015
9/2/
2015
14
The above graph shows that the Bank conservatively lowered the overnight rate to 3.0% in May 15
2008 and it kept it there throughout the summer. It was then forced to dramatically and rapidly 16
11
cut the overnight rate to 0.25% in response to the financial crisis triggered by the failure of 1
Lehman Brothers. 0.25% has traditionally been thought to be the lowest the Bank can set the 2
overnight rate, since otherwise it would mean negative deposit rates for the settlement balances 3
the chartered banks keep with it.6 4
Canada recovered quickly from the “Great Recession” coming out of the United States and the 5
Bank of Canada started “normalising” by increasing the overnight rate in June 2010 as there 6
were obvious signs of recovery in the Canadian economy. The Bank of Canada increased the 7
overnight rate on three separate occasions each time by 0.25% to bring it to 1.0% and with it 8
Prime increased to 3.0%. Expectations in 2011 were that the Bank would resume increasing the 9
overnight rate as the economy continued to strengthen as it was still below the “equilibrium” or 10
“normal” rate. 11
Increasingly the Bank of Canada and the Federal Government started to worry that at 1.0% the 12
overnight rate would encourage too much personal borrowing and lead to levels of indebtedness 13
that would have negative implications as interest rates returned to their normal level. They were, 14
and still are, very worried about a housing bubble in Toronto and Vancouver7 where house prices 15
increased strongly in response to both lower interest rates and a stronger economy. In response 16
on July 8, 2012 the Federal Government announced a third round of tightening in the mortgage 17
market for anyone with less than a 20% down payment. The restrictions include limiting the 18
maximum amortisation period to 25 years, capping household debt ratios, limiting CMHC 19
insurance to homes with a purchase price less than $1 million and the maximum on a home 20
equity line of credit to 65%. They also moved responsibility for Canadian Mortgage and Housing 21
Corporation (CMHC) to the Department of Finance, so that it would be subject to OSFI 22
supervision and in June 2015 issued guidelines that CMHC has to follow.8 23
6 The European Central bank (ECB) currently pays negative interest on settlement balances, so there is now a history of this actually happening so the overnight rate could be cut below 0.25%. 7 In September 2015 housing starts increased to an annualized pace of 216,924 indicating a strong housing market in Canada. 8 See the Canadian Bankers Association web site http://www.cba.ca/en/media-room/50-backgrounders-on-banking-issues/657-changes-to-canadas-mortgage-market
12
The conundrum faced by the Bank of Canada and the Federal Government is that while it wants 1
to stimulate the economy by maintaining low interest rates, it does not want a US style debt-2
fuelled housing bubble. The additional problem is that the Canadian economy is not an island 3
and increasingly the Bank of Canada has been concerned about the transfer of events from the 4
Eurozone and the US into Canada as both the Federal Reserve (Fed) and the European Central 5
Bank (ECB) continue to follow “easy” money policies due to weaknesses in their economies. 6
We can see the impact of events outside Canada in the following graph of the capacity utilization 7
levels in both the Canadian manufacturing and non-farm sectors. 8
Capacity Utilisation
65
70
75
80
85
90
1987
-03
1988
-03
1989
-03
1990
-03
1991
-03
1992
-03
1993
-03
1994
-03
1995
-03
1996
-03
1997
-03
1998
-03
1999
-03
2000
-03
2001
-03
2002
-03
2003
-03
2004
-03
2005
-03
2006
-03
2007
-03
2008
-03
2009
-03
2010
-03
2011
-03
2012
-03
2012
-03
2013
-03
2014
-03
2015
-03
Non-Farm Manufacturing
9
The sharp drop in capacity utilization during the recession in the early 1990’s is evident as 10
Canadian firms restructured in response to the Free Trade Agreement as well as a normal 11
cyclical downturn. We can see this again in the slowdown after the financial crisis in 2009-2010. 12
In both cases there followed a normal rapid recovery out of recession and a movement towards 13
stabilisation. However unlike earlier periods, Canada stagnated in 2012-2013 as the recovery did 14
not continue apace. Instead, we were hit with the aftereffects of the Euro crisis and the slow 15
recovery of our major trading partner the United States. Then just as the US recovery started to 16
13
gather speed we were hit with the slowdown in China during 2015 and the dramatic drop in 1
commodity prices.9 2
The following graph shows the Bank of Canada’s commodity price index. 3
Bank of Canada Commodity Price Index
0.00
100.00
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
1000.00
1972
-01
1973
-09
1975
-05
1977
-01
1978
-09
1980
-05
1982
-01
1983
-09
1985
-05
1987
-01
1988
-09
1990
-05
1992
-01
1993
-09
1995
-05
1997
-01
1998
-09
2000
-05
2002
-01
2003
-09
2005
-05
2007
-01
2008
-09
2010
-05
2012
-01
2013
-09
2015
-05
4
We can see the strong increase in commodity prices that started in 2002 as China started to 5
seriously industrialize. The Great Recession in the United States in 2009 caused these prices to 6
collapse, but they quickly recovered until the sharp sell-off in Summer 2014 as fears of a China 7
slowdown proved correct. It is this drop in commodity prices that has severely affected Canada’s 8
resource sector and triggered a technical recession in 2015Q2. We can see this in the weakened 9
capacity utilization at the start of 2015 and the reaction of the Bank of Canada, which surprised 10
markets by cutting the overnight rate twice in early 2015 to the current level of 0.50%. 11
The upshot is that we continue to have spare capacity in the Canadian economy. The following 12
graph shows the overall unemployment rate. Although the current rate of 7.0% is low compared 13
to the average levels of the last thirty years, it is still higher than it was in 2007 immediately 14
before the financial crisis when it was below 6.0%. More to the point there is concern that the 15
9 China’s GDP growth rate has dropped to 6.9% from just over 7%, much of this was targeted by government authorities.
14
quality of many of the new jobs created is low. We also see a pick up in the unemployment rate 1
over the last year in response to the weakened commodity price hit to our resource sector. 2
Canadian Unemployment Rate
4
5
6
7
8
9
10
11
12
13
1987
M01
1988
M07
1990
M01
1991
M07
1993
M01
1994
M07
1996
M01
1997
M07
1999
M01
2000
M07
2002
M01
2003
M07
2005
M01
2006
M07
2008
M01
2009
-07
2011
-01
2012
-07
2014
-01
2015
-07
3
Q. HOW DOES THIS RELATE TO GDP? 4
A. The following graph has the annual change in real GDP since 1961. 5
Real GDP Growth
-4.00
-2.00
0.00
2.00
4.00
6.00
8.00
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
6
The start date reflects the initial iteration of the 1961 GDP accounts. These annual changes 7
clearly show the recessions of the early 1980s, 1990’s and in 2009/10. The graph also shows the 8
15
quick recovery in Canada from that recession, particularly as compared to the FTA recession of 1
the 1990’s. In 2015 the Canadian economy has slowed in response to the decline in commodity 2
prices. The following graphic from the Bank of Canada illustrates the differential impact of the 3
decline in commodity prices.10 4
5
There is a serious retrenchment going on in both the oil and gas sector and non-energy 6
commodity price related sectors of the economy. However, the rest of the Canadian economy is 7
chugging along with 1.4% growth, which the Bank indicates is 83% of GDP. The main reason 8
for this is the decline in the value of the C$. The following graph shows the value of the C$ in 9
terms of US currency back to 2000. 10
10 Monetary Policy Report, October 2015.
16
FX Rate (US:CDN)
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
2000
M01
2000
M07
2001
M01
2001
M07
2002
M01
2002
M07
2003
M01
2003
M07
2004
M01
2004
M07
2005
M01
2005
M07
2006
M01
2006
M07
2007
M1
2007
M7
2008
M01
2008
M07
2009
M1
2009
M7
2010
M1
2010
M7
2011
M01
2011
M07
2012
M1
2012
M7
2013
M1
2013
M7
2014
M01
2014
M07
2015
M01
2015
M07
1
The reason for going back to 2000 is to show the impact of the commodity price increases 2
consequent on China’s industrialisation. Back in 2002 the C$ was worth just 65 cents, but with 3
the dramatic increase in commodity prices it went to a premium to the US$ and peaked at a 4
monthly high of 1.06. However, since the slowdown in China and the US recovery it has 5
dropped to under 71 cents at the start of 2016. In turn this has stimulated the manufacturing 6
industries of central Canada that previously found it very difficult to compete with a C$ at par 7
with the US$. 8
The Bank of Canada expects that this adjustment to low commodity prices will continue to exert 9
a drag on Canadian growth in 2015 and 2016Q1. However, the Bank also expects the growth rate 10
to pick up to reach 2.6% by 2017Q1 and then for both the US and Canadian economies to grow 11
at about that rate. 12
Q. WHAT IS YOUR OUTLOOK FOR INFLATION? 13
A. The Bank of Canada’s 2.0% target rate of inflation was renewed with the Government of 14
Canada in the Fall of 2011 as part of a five year pact. The Bank of Canada increases the 15
overnight rate when it judges the forecast core inflation rate11 to be above this target and likely to 16
11 Core inflation is the all items CPI minus energy and food.
17
go to the top of its 1.0-3.0% operating band. Conversely, it drops the overnight rate when it fears 1
that inflation will drop to the bottom of its range and as a result it needs to stimulate the 2
economy. The inflation rate data in Schedule 1 clearly shows the inflationary pressures in 2008 3
prior to the recession as well as the dramatic drop in 2009 and recovery in 2010. 4
Since 1991, the Federal Government has been issuing two types of bonds: a nominal bond where 5
the interest rate is fixed and a real return bond, which guarantees the investor protection from 6
inflation. The difference between the nominal yield and the yield on the real-return bond is called 7
the break-even inflation rate (BEIR), since if actual inflation is higher than this after the fact you 8
would have been better off in the real bond and vice versa. Consequently the BEIR is often taken 9
as one measure of the market’s inflationary expectations. The following graphs the BEIR (as a 10
%) since 1991. 11
Break-Even Inflation Rate (BEIR)
0
1
2
3
4
5
6
1991
M11
1992
M05
1992
M11
1993
M05
1993
M11
1994
M05
1994
M11
1995
M05
1995
M11
1996
M05
1996
M11
1997
M05
1997
M11
1998
M05
1998
M11
1999
M05
1999
M11
2000
M05
2000
M11
2001
M05
2001
M11
2002
M05
2002
M11
2003
M05
2003
M11
2004
M05
2004
M11
2005
M05
2005
M11
2006
M05
2006
M11
2007
M5
2007
M11
2008
M5
2008
M11
2009
-05
2009
-11
2010
-05
2010
-11
2011
-05
2011
-11
2012
-05
2012
-11
2013
-05
2013
-11
2014
-05
2014
-11
2015
-05
2015
-11
12
We can clearly see the collapse in inflationary expectations in the late 1990’s as the market 13
finally believed the Federal Government’s intentions not to inflate its way out of its deficit 14
problems. Since then the BEIR has generally been slightly above the middle of the Bank of 15
Canada’s operating range for inflation of 2.0%, but never above the 3.0% upper limit set by the 16
Bank. We can also see the impact of the traumatic events of 2008Q3 when the BEIR dropped 17
from its “normal” level of just above 2.0% to 1.26% in November 2008. During this period the 18
fears of a deep recession and deflation were so strong that the BEIR essentially halved in the 19
18
space of a few months. These deflationary fears subsided once economic growth got back on 1
track, but have dropped recently as the decrease in commodity prices has rippled through the 2
economy. The Bank of Canada estimates that through 2015 the actual rate of inflation will be 3
just above 1.0%, while core inflation that strips out the impact of these volatile items will be 4
2.1% or slightly above its 2.0% target. Going forward the Bank estimates inflation both core and 5
actual at 2.0%, whereas the BEIR is more bearish. 6
Q. WHAT HAS BEEN THE RECENT HISTORY OF THE LONG CANADA BOND 7
YIELD? 8
A. Schedule 2 provides data on the full range of interest rates across the broad maturity 9
spectrum as of January 8, 2016. The interest rate on the 30 year Government of Canada bond at 10
2.07% is 1.62% higher than the yield on 91 day Treasury Bills. This is referred to as a normal 11
yield curve as typically this yield difference or yield spread is about 1.25%. However, currently 12
it is marginally higher indicating that the market expects short term interest rates to increase in 13
the future as the Canadian economy improves and with it the demand for loans so that the Bank 14
of Canada no longer follows a stimulative monetary policy. 15
Normally yields on long term Canada (LTC) bonds are not as affected by current monetary 16
policy as short term interest rates, since monetary policy works at the short end of the yield curve 17
via the overnight rate; its influence then weakens as the maturity of the bond increases. The 18
following graph shows that the LTC yield stayed at about 4.5% from 2005 until December 2007, 19
when the Bank of Canada started to cut interest rates after which it stayed at around 4.0% until 20
November 2008 when it dropped by 0.50%, as the market began to understand the severity of the 21
recession. However, as these fears receded the LTC yield recovered to the 4.0% level it was at 22
immediately prior to the financial crisis. The expectation through 2011 was that long Canada 23
bond yields would increase as the economy recovered, consistent with the Bank of Canada 24
increasing the overnight rate. 25
19
Canadian Interest Rates
0
1
2
3
4
5
6
7
2000
M01
2000
M05
2000
M09
2001
M01
2001
M05
2001
M09
2002
M01
2002
M05
2002
M09
2003
M01
2003
M05
2003
M09
2004
M01
2004
M05
2004
M09
2005
M01
2005
M05
2005
M09
2006
M01
2006
M05
2006
M09
2007
M1
2007
M5
2007
M9
2008
-01
2008
-05
2008
-09
2009
-01
2009
-05
2009
-09
2010
-01
2010
-05
2010
-09
2011
-01
2011
-05
2011
-09
2012
-01
2012
-05
2012
-09
2013
-01
2013
-05
2013
-09
2014
-01
2014
-05
2014
-09
2015
-01
2015
-05
2015
-09
LTC T. Bill (91 d)
1
Note that when we look at interest rates, although the normal yield spread is about 1.25%, it 2
narrows when the central bank is trying to slow down the economy by increasing the overnight 3
rate and increases when it is trying to stimulate the economy. We can see, for example, the very 4
narrow yield spread at the end of 2007 and the dramatic increase during the 2008/9 financial 5
crisis. However, in 2011/12 as we headed into the last hearing short term rates had already 6
increased and the yield spread had decreased. For example, the Royal Bank of Canada (Financial 7
Markets Monthly June 3, 2011) had the following forecast in June 2011. 8
9
20
At that time RBC was forecasting that long term Canadian interest rates would increase from the 1
then 3.75% level to a “normal” level of 4.55% and US rates from 4.55% to 5.05%. This did not 2
happen! 3
In 2011Q4 the US Federal Reserve embarked on the most dramatic third round of “quantitative 4
easing” (QE3), which was simply the central bank buying bonds with freshly minted money: the 5
more bonds they buy the higher the bond prices and the lower the interest rate. QE3 amounted to 6
an open-ended commitment by the US Federal Reserve to buy $85 billion of US government 7
bonds and Federal Agency backed mortgages every month while maintaining the Federal funds 8
rate at 0%. As we can see from the previous graph of interest rates the effect was a significant 9
drop in long term Canada yields as investors hunted for higher yields in other markets than the 10
US government bond market. 11
QE3 started to run down in 2013Q3 and again the market expected interest rates to recover. 12
However, although QE3 ended in the US in October 2014 the recovery in interest rates has still 13
not occurred. Instead long term interest rates have essentially been flat through 2015. There are 14
two main reasons why interest rates have still not recovered in Canada. The first is that the US 15
Federal Reserve has accumulated at least $3.5 trillion of securities. The following graphic from 16
the Federal Reserve Bank of Cleveland shows the growth in the Fed’s balance sheet. 17
18
21
As it buys securities it pays for them with US money and by the end of 2014 it had bought about 1
$3.5 trillion of mainly US government and agency (mortgage) bonds. This is $3.5 trillion of 2
securities that have been issued, but are not in the public markets: if they were the huge increase 3
in supply would have depressed prices and increased interest rates. Although this bond purchase 4
program has now ended there is still this $3.5 trillion that is not in the market. 5
This is like filling a bath (liquidity): the US Fed may have turned off the taps so the bath is no 6
longer filling, but the bath is already incredibly full. What this means is that although QE3 has 7
stopped, its impact will last as long as the Fed does not sell its $3.5 trillion hoard of securities. 8
Further as the US government has got closer to balancing its books, and reducing its deficit and 9
need to issue bonds, the supply of US government bonds has also dropped. The problem now is 10
how to drain the bath and here the actions of the Federal Reserve is critical. The bath will be 11
drained as the Fed sells government securities, thereby depressing their price and increasing 12
interest rates. In December 2015 the Federal Reserve did increase its target rate by 0.25% and 13
also indicated that there will be four more similar increases in 2016 to move the Federal Funds 14
rate to 1.50% by the end of 2016. However, this policy change is directed at the short end of the 15
yield curve, not the long end. The Fed seems to be in no hurry to run down its balance sheet by 16
selling its longer term bond holdings, if it did, long term interest rates would increase 17
dramatically. 18
The second reason is that the Fed is not the only central bank to inject money into the global 19
financial system or buy long dated government securities. The Bank of England is in no hurry to 20
increase its target rate and even though it stopped its QE program in July 2012, and like the Fed 21
it is in no hurry to sell its holdings of UK government bonds. Further the European Central Bank 22
did not start serious quantitative easing until January 2015 and currently has a negative target for 23
its policy rate. Finally in October 2014 Japan announced its own expanded QE program. 24
The result is that while the US and UK baths have stopped filling up, but are almost full, there 25
are baths filling up in Europe and Japan. The result is twofold: the supply of liquidity (money) 26
used to buy the bonds has enormously increased, while the supply of bonds has decreased, since 27
trillions have been taken off the market by central banks. Canada has been a by-stander in all this 28
as the Bank of Canada never had to initiate a bond buying program and after it recovered from its 29
own 2008/9 recession started to increase the overnight rate in 2010. Since then Canada has been 30
22
waiting for the rest of the world to sort out its problems. However, with open capital markets 1
Canada is affected by what has happened in the US, UK, Japanese and European financial 2
markets. 3
As global interest rates have dropped the “search for yield” has become very important and 4
investors have started to take note of Canada. Of importance is that Canada remains one of a 5
small number of AAA rate countries so is a particularly attractive location to invest government 6
reserves. The following graphic comes from the Bank of Canada and indicates that almost 30% 7
of the Canadian government bond market is now owned by non-residents. As non-residents have 8
invested in the Canadian government bond market they have driven up market prices and driven 9
down government bond yields far below where they would have been but for the massive bond 10
buying programs elsewhere in the world. 11
12
Q. WHAT IS YOUR FORECAST FOR THE LONG CANADA BOND YIELD FOR 13
2016-7? 14
A. The current (January 8, 2015) RBC forecast is below. 15
23
1
What is clear is that the enormous amount of liquidity in global markets has caused RBC to 2
dramatically lower its interest rate forecast. Whereas in June 2011 it was forecasting a quick 3
rebound to the 4.55-5.05% level for long Canada and US Treasury bonds this has now been 4
scaled back to 3.65%-4.25%. These forecasts are largely keyed off the forecast increases in the 5
over-night and Federal Funds rates where RBC takes the Fed at their word and sees a 1.0% 6
increase in 2016 and then, quite aggressively, further increases to 3.5% by the end of 2017. In 7
Canada RBC sees the Bank of Canada standing pat for 2016 and then increasing the overnight 8
rate in 2017 to end at 1.75%. 9
The RBC forecast is aggressive compared to the general consensus, however, and is reported 10
mainly because I usually report it. In their December 2015 report Consensus Economics places 11
the average (from 18 forecasters) for the ten year bond forecast yield three months out at 1.80% 12
and for one year out at 2.2%. If the current spread from Schedule 2 between the 30 year (2.07%) 13
and ten year (1.32%) bond is added to the end of 2016 forecast this implies that the 30 year bond 14
yield at the end of 2016 will only be 2.95%. The average of the March and December Consensus 15
forecasts which is a proxy for the average for the year as whole, and consistent with the 16
application to an average forward test year rate base, is 2.75%.12 17
12 RBC is the second highest in their forecast; the highest is from my colleagues at the University of Toronto with a 3.1% forecast yield, which translates into 3.22% for the 30 year bond.
24
There has been much discussion into whether there has been a permanent decline in “neutral” 1
interest rates, that is, rates that are neither designed to stimulate nor slow down the economy. 2
Recently Reza and Sarker13 refer to Ms. Yellen’s judgement that the “neutral” or normal interest 3
rate is now a real rate of 1.75%, whereas the neutral rate in Canada has dropped from the 2.5-4
3.5% level of the mid 2000’s to a current level of 1-2%. If correct, with a 2% inflation target 5
these estimates imply a neutral, nominal interest rate of 3.0-4.0% in Canada and a slightly higher 6
3.75% in the US or slightly below RBC’s longer run forecast. However, what happens to long 7
term interest rates depends critically on when quantitative easing ends in Japan and Europe and 8
when, if ever, the central banks start seriously selling their stocks of long term government 9
bonds.14 10
What is important to note is that interest rates are not, and probably will not for the foreseeable 11
future, be set by private investors. Instead, they are being set by what has been termed the 12
“global policy maker.” As a result, forecasting interest rates for a small country like Canada in a 13
global financial system depends critically on central bank decisions elsewhere. 14
Q. WHAT HAS HAPPENED IN THE CORPORATE FIXED INCOME MARKET? 15
A. The following graphs the spread between the yield on 91 day Treasury Bills (TB) and 16
those on three month Bankers Acceptances (BA) and Commercial paper (CP). 17
13 “Is slower growth the new normal in advanced economies,” Bank of Canada Review, Autumn 2015. 14 Note central banks are in no hurry to sell their holdings of long term bonds, since they are financed by issuing cash. As a result, they generate huge profits. The economic reason to sell them will be when there is a pickup up in inflation and the central bank needs to drain the system of liquidity to avoid rampant inflation. Currently, there is no likelihood of this in the foreseeable future.
25
Money Market Spreads
0
50
100
150
200
250
300
1/4/
2006
4/4/
2006
7/4/
2006
10/4
/200
6
1/4/
2007
4/4/
2007
7/4/
2007
10/4
/200
7
1/4/
2008
4/4/
2008
7/4/
2008
10/4
/200
8
1/4/
2009
4/4/
2009
7/4/
2009
10/4
/200
9
1/4/
2010
4/4/
2010
7/4/
2010
10/4
/201
0
1/4/
2011
4/4/
2011
7/4/
2011
10/4
/201
1
1/4/
2012
4/4/
2012
7/4/
2012
10/4
/201
2
1/4/
2013
4/4/
2013
7/4/
2013
10/4
/201
3
1/4/
2014
4/4/
2014
7/4/
2014
10/4
/201
4
1/4/
2015
4/4/
2015
7/4/
2015
10/4
/201
5
1/4/
2016
CP BA
1
Treasury Bill yields are close to the rate that the chartered banks get from their deposits at the 2
Bank of Canada when they have excess cash and are effectively default free, since they are 3
obligations of the Treasury of the Government of Canada. In contrast, the Bankers’ Acceptance 4
rate is the rate the market requires on short term investments in the main chartered banks, 5
whereas the Commercial Paper rate is the rate that large Canadian companies with the best credit 6
rating can get by issuing notes in the money market. As a result, the spreads between these two 7
private rates and that on Treasury Bills is indicative of the state of the short term lending 8
market15 and the willingness of large investors to lend to the banks and very low risk, stable, 9
Canadian companies 10
Before discussing these spreads, it is important to note that investors in the money market are 11
mainly “parking” their money, rather than investing, since their main concern is security of 12
principal, that is, getting their money back. Consequently, with any hint of default the market 13
seizes up. This happens periodically in the CP market as seemingly low risk institutions default 14
and investors panic and refuse to roll over or reinvest the money invested in CP for fear of 15
further losses due to an inability to distinguish between good and bad risks. 16
For example, for the last 20 years the money market has been very quiet with spreads at 10-20 17
basis points. This changed in July 2007 with the US sub-prime problems spilling over into 18
15 The main banks are generally rated R-1 (Mid) equivalent to an AA bond rating while CP is a mixture of R-1 (Mid) and R-1 (low), so generally these spreads are very similar.
26
Canada, where we can see the large spike and again with the Bear Stearns bailout in March 2008. 1
This got much worse in September 2008 as Lehman Brothers failed and contagion hit the 2
world’s financial markets and spreads in the Canadian money market went close to 3.0%. 3
However, of importance is that the measures taken by central banks to stabilise the financial 4
system worked. The BA and CP spreads have been at normal levels for the last six years. 5
At the current point in time with the low overnight rate, 91 day Treasury Bills are at 0.45% and 6
high grade commercial paper is at 0.82% or a premium of 0.37%. In contrast, at the time of my 7
GCOC evidence in 2012 the cost of commercial paper was 0.34% higher at 1.16%. There is no 8
question that top quality credits can now access funds in the short term money market more 9
cheaply than in 2012. 10
The following graphs the generic credit or default, spreads between corporate and government 11
long term bonds using the AA, A, and BBB indexes maintained originally by Scotia Capital 12
markets.16 13
Default Spreads Since Dec 1979
0
50
100
150
200
250
300
350
400
450
500
12/3
1/19
79
12/3
1/19
80
12/3
1/19
81
12/3
1/19
82
12/3
1/19
83
12/3
1/19
84
12/3
1/19
85
12/3
1/19
86
12/3
1/19
87
12/3
1/19
88
12/3
1/19
89
12/3
1/19
90
12/3
1/19
91
12/3
1/19
92
12/3
1/19
93
12/3
1/19
94
12/3
1/19
95
12/3
1/19
96
12/3
1/19
97
12/3
1/19
98
12/3
1/19
99
12/3
1/20
00
12/3
1/20
01
12/3
1/20
02
12/3
1/20
03
12/3
1/20
04
12/3
1/20
05
12/3
1/20
06
12/3
1/20
07
12/3
1/20
08
12/3
1/20
09
12/3
1/20
10
12/3
1/20
11
12/3
1/20
12
12/3
1/20
13
12/3
1/20
14
12/3
1/20
15
AA A BBB 14
Corporate bonds have default risk, since companies can run into financial difficulty whereas 15
governments borrowing in their own currency like Canada cannot.17 These yield spreads usually 16
16 The most recent data is from Datastream, which updates original data from Scotia Capital’s Handbook of Debt Market Indices.
27
behave in a predictable manner. In a recession as the risk of bankruptcy increases, investors sell 1
off default-risky corporate debt and their liquidity drops. As a result their bond prices fall and 2
their yields increase relative to the long Canada bond yield causing a wider spread. Conversely 3
as the economy recovers and this risk recedes the spread narrows. We can see this clearly in the 4
high spreads during the long recession of the early 1990s, the panic of the Asian crisis and the 5
bursting of the Internet Bubble and in particular the financial crisis of 2008-9. Note also that 6
usually the spread increases most for the BBB bond which is the riskiest. 7
At the current point in time generic A spreads are at 1.90% and have marginally increased since 8
the summer when they were just under 1.60%. In contrast, at the time of my 2012 GCOC 9
evidence these spreads were a tad lower at 1.75%. I don’t regard the increase as that significant 10
mainly because they are quite volatile and currently reflect a minor flight to quality as the stock 11
market has swooned since the summer on China fears. More importantly firms do not “pay” the 12
spread they pay the full interest cost. At the time of my GCOC evidence the long Canada rate 13
was 2.41% and the generic A yield was 4.16%, whereas currently with a long Canada rate of 14
2.01% the yield on generic A bonds is 3.91% or 0.25% lower. 15
It is also interesting to compare Canadian interest rates with the US. The following graph is the 16
yield on the generic BBB bond in Canada and the yield on Moody’s BBB rated utility bond18 17
since January 2012 to encompass the period since the last generic hearing. What is clear is that 18
typically the Canadian BBB yield has been below that in the US, particularly since the start of 19
2013. Currently, there is a yield difference of 0.81%, that is, the generic BBB yield in Canada is 20
at 4.70% (January 1) versus the utility BBB yield in the US of 5.51%. This is further 21
confirmation that the financing cost (cost of capital) to Canadian utilities is significantly lower 22
than for US utilities. 23
17 This assumes they simply print more money to pay off their debts. The US can do this, but it has been the behaviour of Tea Party Republicans in Congress arguing that the US should default that has frightened global investors since 2011. 18 The US yield data is from the Federal Reserve Economic Data bank (FRED) and the BBB (actually Moody’s BAA) is the only corporate series available, probably because the typical US utility is rated BBB.
28
US and Canadian BBB rated debt
3.5
4
4.5
5
5.5
6
1/6/
2012
4/6/
2012
7/6/
2012
10/6
/201
2
1/6/
2013
4/6/
2013
7/6/
2013
10/6
/201
3
1/6/
2014
4/6/
2014
7/6/
2014
10/6
/201
4
1/6/
2015
4/6/
2015
7/6/
2015
10/6
/201
5
CAN BBB
US BBB
1
In the past I have also stressed that in Canada the yield on utility debt rated A or BBB has traded 2
on lower yields than equivalent generic debt. This tendency has been particularly true during 3
recessions when the rating agencies seem to be slow to downgrade non-utility debt whereas the 4
market for non-utility debt seems to sell off. I used to use two series from the Canadian Bond 5
Rating service (CBRS) but this was discontinued after it was taken over by S&P. Fortunately, 6
Bloomberg has created a synthetic A rated utility yield series (C29530Y). The following graphs 7
the yields from the Scotia Capital generic A bond index and the Bloomberg utility synthetic 8
series.19 9
Note that utility yields were consistently lower than generic A yields as the financial crisis 10
started to emerge and remained so until two years or so ago when they merged. Currently the 11
market seems to be valuing similarly rated utility and non-utility debt the same. This is further 12
support for a return to normality in the bond market as the extra “safety” implicit in A rated 13
utility bonds is not valued to the same degree that it is during the financial crisis. 14
19 Note the Scotia A index includes a large number of utility bonds so the distinction, by construction, is not as clear as for the two old CBRS series.
29
Utility A vs Generic A yields
3.00
3.50
4.00
4.50
5.00
5.50
6.00
6.50
7.00
7.50
8.00
3/1/
2006
6/1/
2006
9/1/
2006
12/1
/200
6
3/1/
2007
6/1/
2007
9/1/
2007
12/1
/200
7
3/1/
2008
6/1/
2008
9/1/
2008
12/1
/200
8
3/1/
2009
6/1/
2009
9/1/
2009
12/1
/200
9
3/1/
2010
6/1/
2010
9/1/
2010
12/1
/201
0
3/1/
2011
6/1/
2011
9/1/
2011
12/1
/201
1
3/1/
2012
6/1/
2012
9/1/
2012
12/1
/201
2
3/1/
2013
6/1/
2013
9/1/
2013
12/1
/201
3
3/1/
2014
6/1/
2014
9/1/
2014
12/1
/201
4
3/1/
2015
6/1/
2015
9/1/
2015
Utility Scotia A
1
Q. WHAT HAS BEEN THE GENERAL STATE OF CAPITAL MARKETS? 2
A. As indicated previously the bond market has been heavily influenced by the actions of 3
central bankers, it is interesting therefore to look at broader measures of the stress in the financial 4
system. In the US the Federal Reserve Bank of Kansas City has developed the Kansas City 5
“Financial Stress” Index (KCFSI) which is graphed below. 6
30
KCFSI"+" is tough and "-" is loose
-2
-1
0
1
2
3
4
5
6
7
Feb
-90
Feb
-91
Feb
-92
Feb
-93
Feb
-94
Feb
-95
Feb
-96
Feb
-97
Feb
-98
Feb
-99
Feb
-00
Feb
-01
Feb
-02
Feb
-03
Feb
-04
Feb
-05
Feb
-06
Feb
-07
Feb
-08
Feb
-09
Feb
-10
Feb
-11
Feb
-12
Feb
-13
Feb
-14
Feb
-15
1
This index is designed to capture a variety of financial indicators in addition to the two which I 2
have traditionally focussed on, which are the short term spreads in the money market and longer 3
term spreads in the bond market. The additional indicators include the stock market volatility 4
index, the state of bank share prices, and the behaviour of stock and bond returns. When the 5
KCFSI is above 0 it indicates that capital markets are under stress or that access to markets is 6
“tough” similarly when it is below 0 it indicates relatively easy or “stress-free” capital market 7
conditions. The value of the KCFSI is simply that it captures in one number the impact of a 8
variety of capital market indicators.20 The major insight of the KCFSI is that it emphasises the 9
enormous pressure in the US financial system during the financial crisis. Unlike the internet 10
bubble crash in 2001 the crisis in 2008/9 struck at the very core of the US financial system, 11
which is the banking system, where liquidity, that is, the ability to trade securities at close to 12
their true market value, dried up in many parts of the capital market and the US government had 13
to intervene on a massive scale. Since the 2012 GCOC market conditions have been easy, but 14
with the ending of QE3 conditions in the US tightened somewhat as markets returned to average 15
or “normal” conditions. 16
20 Technically it captures the common element in all these indicators by using principal components analysis.
31
The work by the Kansas City Fed follows pioneering work done by researchers at the Bank of 1
Canada who developed a simpler financial conditions stress index,21 which is graphed below. 2
Canadian Financial Conditions Index"-" is loose and "+" is tough"
-3
-2
-1
0
1
2
3
4
5
10/1/199
9
4/1/20
00
10/1/200
0
4/1/20
01
10/1/200
1
4/1/20
02
10/1/200
2
4/1/20
03
10/1/200
3
4/1/20
04
10/1/200
4
4/1/20
05
10/1/200
5
4/1/20
06
10/1/200
6
4/1/20
07
10/1/200
7
4/1/20
08
10/1/200
8
4/1/20
09
10/1/200
9
4/1/20
10
10/1/201
0
4/1/20
11
10/1/201
1
4/1/20
12
10/1/201
2
4/1/20
13
10/1/201
3
4/1/20
14
10/1/201
4
4/1/20
15
3
The Bank of Canada indicator similarly tracks the enormous stress in the financial markets 4
during the financial crisis and similar to the KCFSI, the index reflects the recent easy or stress 5
free financial market conditions. The Bank has now discontinued the stress index and emphasises 6
a broader array of stress indicators. One of the most important indicators is the Bank’s survey 7
results from senior lending officers, which reflects their assessment of the lending conditions to 8
Canadian non-financial firms. The latest survey results from the bank are below and similarly we 9
see the tightening credit market conditions during the financial crisis and the loose conditions 10
thereafter with a gradual return to normal pricing and credit conditions in 2015, similar to the 11
indications from the KCFSI. 12
21 The Bank of Canada index is actually the inverse of this, I multiplied it through by -1 to get the same interpretation as for the KCFSI
32
1
Q. HOW DOES THE STATE OF THE ECONOMY AFFECT PROFITS AND THE 2
CAPITAL MARKET? 3
A. The following graphs the average annual ROE against the spread between the yield on 4
BBB debt and long Canada bonds since 1980. The data for 2015 is for the first half of the year 5
since full year data is not yet available. The graph illustrates the basic inverse relationship that 6
spreads normally increase during recessions when corporate profitability drops and then reduce 7
when the economy recovers and corporate profits increase. We can clearly see this in the 8
recession of the early 1990s and again to a lesser degree during the financial crisis. However, 9
even during the financial crisis basic corporate profitability in Canada remained healthy due to 10
strong commodity prices, indicating that much of the pressure was coming from events in the US 11
and elsewhere. 12
33
Corporate ROE and BBB Spread
0
50
100
150
200
250
300
350
400
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
bas
is p
oin
ts
0.00
2.00
4.00
6.00
8.00
10.00
12.00
14.00
16.00
Per
cen
t
BBB Spread ROE
1
Corporate profitability in Canada has recently dropped from its pre-recession highs as 2
commodity prices have weakened and this has kept spreads marginally high. We can see in the 3
first half of 2015, for example, that the drop in profitability has been matched by a minor 4
increase in spreads over the level for 2014. 5
Q. WHAT ARE YOUR CONCLUSIONS ABOUT CONDITIONS IN THE “MONEY 6
MARKET”? 7
A. The Canadian economy is suffering from two effects. On the one hand it is adjusting to 8
the drop in commodity prices, which is hurting the materials and energy sector. On the other 9
hand, it is being helped by growth in the U.S and a strengthening of traditional manufacturing, 10
particularly in central Canada. Three years ago the Bank of Canada anticipated that the 11
remaining spare capacity in the economy would be used up by mid-2015, while the financial 12
system was firing on all cylinders. The slowdown in China has deferred this growth forecast and 13
caused serious problems in the resource sector. As a result, we are still a couple of years away 14
from the peak of the business cycle with plenty of growth to come. Whether we can use up these 15
34
resources depends on continued growth in the U.S22 and whether the slowdown in China causes 1
more problems. Of particular concern is whether there are significant non-performing loans to 2
the Canadian resource sector resulting from the lower commodity prices. 3
In terms of the equity markets, the following graphs the performance of the TSX composite and 4
the S&P 500 since December 31 2011. In both cases the level is standardised with the December 5
31, 2011 value so we can see the changes since then. What is clear is the outperformance of the 6
US market, since it is not as resource sensitive as the Canadian market, with a 60-70% increase 7
in value over the four year period. In contrast, the Canadian market has only increased by 13% 8
and peaked in Summer 2014 at 15,625, after which it has drifted lower with the impact of 9
commodity prices on the resource sector. This has continued in the first week of 2016 where data 10
is not yet available in the data tapes. 11
Stock market IndexesSince December 2011
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
2012
M01
2012
M03
2012
M05
2012
M07
2012
M09
2012
M11
2013
M01
2013
M03
2013
M05
2013
M07
2013
M09
2013
M11
2014
-01
2014
-03
2014
-05
2014
-07
2014
-09
2014
-11
2015
-01
2015
-03
2015
-05
2015
-07
2015
-09
2015
-11
TSX SP500 12
Despite the recent equity market weakness, general Canadian stock prices are still the same as 13
they were in 2012/2013 and mainly reflect the impact of the resource sector. 14
22 The latest US job market growth statistics indicated an additional 292,000 in December 2015, which is more than enough to soak up new entrants into the workforce and significantly exceeded expectations.
35
Overall the stock market is valuing utilities favourably, credit is still easy and utilities continue to 1
have access to the 40 and 50 year debt market at very low interest rates. Market conditions 2
remain much as they were in 2012 very receptive to good credits like Canadian utilities. 3
Q. THESE COMMENTS ARE FOR CANADA IS THERE ANYTHING DIFFERENT 4
ABOUT BC? 5
A. While almost all the capital market data is relevant for Canada as a whole there are 6
important differences in the BC provincial economic performance. One of the key “risks” for the 7
Bank of Canada is the state of the housing market and the tendency for borrowers to get too deep 8
into debt due to the current low level of interest rates. Although this seems to have kept central 9
bankers awake at night for several years it has been a boon to the BC economy. Currently, retail 10
sales have been exceptional strong as homeowners “feel good” about the equity in their houses. 11
Further the high house prices have stimulated housing starts as developers see strong profit 12
potential in residential real estate. The result has been housing starts running at over 30,000 13
units. The combination of strong retail sales and housing starts has given BC the highest GDP 14
growth rate in Canada with growth over 3.0% forecast for 2016. 15
In terms of industry performance in BC there are mixed messages. On the one hand, the 16
weakness in commodity prices, such as copper and coal, has led to temporary shut downs and 17
severely affected major employers like Teck. On the other hand, the strong US growth and the 18
weak C$ have combined for strong demand for lumber exports to the US. Overall, the impact on 19
BC exports has been minimal. The result is that strong consumer demand gives BC the third 20
lowest unemployment rate in Canada at 6.3% ahead of Manitoba and Saskatchewan, but much 21
lower than the overall Canadian rate of 7.0%. This good provincial economic performance 22
provides the backdrop to FEI’s business risk assessment. 23
36
III RISK PREMIUM ESTIMATES OF THE FAIR ROE 1
Q. WHAT IS THE MOST COMMON WAY OF ESTIMATING THE FAIR ROE? 2
A. The capital asset pricing model or CAPM remains the most common way of estimating 3
the fair rate of return. It is a special form of risk premium model which simply says, 4
*MRPRK F 5
In words, the investor’s required or fair rate of return (K) is equal to the risk free rate (RF) plus a 6
risk premium. Where the CAPM differs from other risk premium models is that it specifies that 7
the risk premium is comprised of the market risk premium (MRP) times the security’s relative 8
risk or beta coefficient (β). In this regard any fair ROE can always be decomposed into a risk 9
free rate and a risk premium, so the CAPM is perfectly general: its contribution is simply to 10
relate an individual risk premium to the overall market risk premium and its relative risk 11
coefficient. 12
Why the CAPM is so widely used is because it is intuitively correct. It captures two of the major 13
“laws’ of finance: the time value of money and the risk value of money. I will discuss the third 14
law of finance the tax value of money later, but the time value of money is captured in the long 15
Canada bond yield as the risk free rate. The risk value of money is captured in the market risk 16
premium, which anchors an individual firm’s risk. As long as the market risk premium is 17
approximately correct the estimate will be in the right “ball-park.” Where the CAPM normally 18
gets controversial is in the beta coefficient; since risk is constantly changing so too are beta 19
coefficients. This sometimes casts doubt on the model as people find it difficult to understand 20
why betas change. Further it also makes testing the model incredibly difficult. However, the 21
CAPM measures the right thing: which is how much does a security add to the risk of a 22
diversified portfolio, which is the central idea of modern portfolio theory. It also reflects the fact 23
that modern capital markets are dominated by large institutions that hold diversified portfolios. 24
Currently, the CAPM remains the most important model used by a company in estimating their 25
cost of equity capital. The following table comes from a survey of 392 US Chief Financial 26
officers by Graham and Harvey in the Journal of Financial Economics 2001: 27
37
1
70% of US CFOs use the CAPM and a further 30% use a multi-beta approach similar to the two 2
factor model I often use. Note in this respect that multi-beta models, while slightly more general, 3
don’t alter the average rate of return which is anchored by the risk free rate and market risk 4
premium: all they do is generate slightly different estimates for individual firms. 5
In response to comments on the Decision for the 2012 GCOC I have developed a new appendix 6
(E) that discusses alternatives to the CAPM. At that time I noted that I was not enthusiastic about 7
such models since they remain extremely controversial. I also don’t think they add much except 8
to add another area that experts can argue over. In particular, as I just mentioned they do not 9
generally have a significant effect on the equity cost estimates. To illustrate this, the current 10
standard in multi-factor models is to extend the one factor CAPM to include two additional 11
factors: one for the size premium (the return difference between small firms minus big or SMB) 12
and the other for the value premium (the return difference between value versus growth stocks). 13
This is the Fama-French three factor model which was raised in the GCOC in 2012. The Fama-14
French equations states, 15
16 HMLSMBMKTRRE jjjFj ,3,2,1)(
38
In this case as well as the market risk premium (MKT) an investor requires a premium for small 1
firms and value stocks. Why the FF3 factor model is controversial is that some believe that small 2
value stocks are riskier and thus deserve a premium, whereas others believe that the market 3
consistently miss-prices large growth stocks, since they are faddish and sexier for financial 4
advisors to sell (think Apple, Facebook etc) and are thus over-valued and consistently earn lower 5
rates of return. 6
I tend to believe the faddish argument, but regardless for individual stocks using the FF3 factor 7
model versus the CAPM rarely makes much of a difference. For example, Estrada (2011)23 8
estimated the equity cost for the Dow 30 firms using both models where the average equity cost 9
using the CAPM was 9.70% versus 9.50% from using the FF3 factor model. The complete 10
estimates are in Schedule 3, but the general point is that we are just allocating the stock’s return 11
to different components, so that the sum of the parts should always (approximately) add up to the 12
same number. The Dow 30 stocks have a beta close to 1.0 since they are a portfolio of stocks and 13
this should be close to average; as we would expect they have negative exposure to the size 14
premium, since they are all large stocks and positive exposure to the value premium, since they 15
are generally value stocks. In this respect they are similar to utilities that tend to be relatively 16
large value stocks. 17
In terms of the “error” in using one model versus another, the difference ranges from +1.5% to -18
1.6% or a range of 3.0%. This is not an insignificant difference but it stems from the confluence 19
of the size and value premiums.24 The +1.5% difference is for American Express, which has a 20
17.7% FF3 Factor equity cost estimate versus the 16.2% for the CAPM. This difference stems 21
from the observation that AmEx is a relatively small value stock and generates a premium for 22
both these factors, which is greater than the lower beta estimate. In contrast, Merck is a large 23
growth stock and its much higher FF3 factor beta coefficient is not enough to offset the negative 24
size and growth premiums. As a result, its CAPM equity cost at 9.1% is higher than its FF3 25
factor cost at 7.5%. The closest to a regulated utility would be AT&T, where the CAPM equity 26
23 Estrada, Journal of Applied Corporate Finance (Spring 2011). Estrada’s estimates are for illustration only I do not recommend them or the process he used to get them. 24 Note also that the range of equity cost estimates is from 4.80% to 17.7% for the FF3 factor model and a slightly smaller 5.3% to 17.5% for the CAPM.
39
cost is 7.80% versus a FF3 factor estimate of 7.30%; again its higher beta is more than offset by 1
the impact of the size and value premiums. 2
Q. IS THERE ANY OTHER SUPPORT FOR THE CAPM? 3
A. Yes. Levy and Roll have recently revisited the question of the empirical support for the 4
CAPM. Richard Roll in a ground-breaking paper25 pointed out that as long as the market 5
portfolio is ex post efficient then by definition the CAPM will work in empirical tests and all 6
securities will lie along a straight line relating returns to betas. In the Levy and Roll paper26 they 7
reverse engineer the tests and consider how much the sample parameters can vary to make sure 8
the normal proxies for the market portfolio are efficient. Levy and Roll find that even slight 9
variations, within the normal estimation bounds, make the proxies efficient. As they conclude: 10
11 “This article shows that a small variation of the sample parameters, well within their 12 estimation error bounds, can make a typical market proxy efficient. Thus, the empirically 13 measured return parameters and the market portfolio weights are perfectly consistent with 14 the CAPM using a typical proxy..... Hence, minor changes in estimation error reverse 15 previous negative and disappointing findings for the CAPM.” 16
17 The point is that the normal criticism of the CAPM is usually based on very old empirical tests 18
that did not recognize estimation error while the Levy and Roll results show that even slight 19
changes in the parameters will nullify those results. 20
Q. WHAT IS YOUR SIMPLE CAPM ESTIMATE FOR A BENCHMARK UTILITY? 21
A. In Appendix B I estimate the market risk premium of common equities over long term 22
Canada bonds at 5.0-6.0%. This estimate is drawn from the Canadian capital market history 23
going back to 1924 so encompasses periods very similar to today, such as the bleak 1930s of 24
slow growth and falling prices, as well as booms and serious inflation problems such as the 25
1970’s. While the Canadian data points to a market risk premium of under 5.0%, I give weight to 26
25 Richard Roll, “A Critique of the Asset Pricing Theory’s Tests: Part 1: On Past and Potential Testability of the Theory”, Journal of Financial Economics 4: 129-76, 1977. 26 Moshe Levy and Richard Roll, “The Market Portfolio May be Mean Variance Efficient After All,” Review of Financial Studies, 2010.
40
the US evidence for three main reasons. First, most of the restrictions on keeping Canadian 1
capital within Canada have been removed resulting in significant capital outflows and higher 2
expected returns on Canadian investments. Second, Canadian governments have moved to a 3
primary surplus on their budgets. The primary surplus is the actual surplus after stimulus 4
expenditures and the impact of an economic slow-down have been removed. Third, the Canadian 5
bond market has received significant foreign capital inflows depressing yields below where they 6
would have been historically with a segmented or closed capital market. The result has been 7
lower interest rates in Canada than the United States for the last five plus years, which has 8
removed the historic bias of a smaller Canadian market risk premium over a higher and riskier 9
Canadian government bond yield. Finally, I give significant weight to survey results by Professor 10
Fernandez, who now annually surveys thousands of academics, financial analysts and corporate 11
executives making investment decisions. This acts as confirmation that my market risk premium 12
range is well within the normal range of other finance professionals. 13
My Appendix C discusses relative risk adjustments or betas. The recent history of Canadian 14
utilities is of beta coefficients in an approximate range of 0.30-0.45. The higher estimate 15
represents their beta estimate post the financial crisis and internet bubble. These estimates are 16
consistent with the price behaviour of Canadian regulated utilities and estimates by external 17
outlets such as the Royal Bank of Canada, Yahoo and Google, so I regard them as reasonable. It 18
is indisputable that as low risk investments the relative risk of Canadian utilities has been 19
significantly lower than for an average Canadian equity security. 20
As robustness checks I also look at beta estimates for a sample of US utilities, where the 21
estimates are somewhat higher, but this could be due to unique effects such as merger activity. I 22
also look at Canadian beta estimates against a US market index to check whether the “low” 23
estimates are due to the hollowing out of the Canadian stock market through acquisitions and 24
find no such evidence. I also look at what would happen if I adjusted my beta estimates using the 25
Blume adjustment methodology toward their grand mean of 0.50 and find it makes very little 26
difference. However, any estimates reflect the time period over which they are estimated and 27
once a unique event falls out of the estimation window it no longer affects the estimate. 28
Consequently on a going forward basis I judge the relative risk of Canadian utilities to move 29
41
back to their historic range reflecting normal market risk. This is why I continue to judge the 1
relative risk of a Canadian utility to be 45-55% of that of the market as a whole. 2
I would therefore judge the going forward utility risk premium to be 2.25% to 3.30% 3
representing the combination of the low end of the relative risk adjustment and the low end of 4
the market risk premium (.45 and 5%) combined with the top end of both (.55 and 6%). If this is 5
added to a 2.75% consensus forecast for the average long term Canadian bond yield for 2016 and 6
a 0.50% flotation cost allowance, I would judge a “simple” CAPM fair return estimate for 2016 7
to be as follows: 8
Simple CAPM Estimates 2016 9
Low end 5.50% 10 High end 6.55% 11
The average estimate of 6.03% is 0.25% lower than the simple average of 6.28% I used in 2012 12
simply because the forecast long Canada bond yield is that much lower. 13
Of importance is that some forecasters, such as RBC, have been predicting a rapid increase in 14
long Canada interest rates for some years, but the forecast has to bear some relationship with the 15
actions of investors in the bond market. If the current bond market yield of about 2.0% does 16
increase to the 2.85% as forecast by RBC over the next year, then a bond market investor would 17
lose on their investment.27 18
Q. DO YOU USE THIS SIMPLE CAPM ESTIMATE? 19
A. No. The CAPM estimate is appropriate under “normal” or average markets, since it uses 20
a normal or average market risk premium and assumes that conditions in the bond market 21
affecting the long Canada bond yield are also driving conditions in the equity market, that is, that 22
the correct “opportunity cost” for an equity investor is the bond market plus a risk premium. 23
However, at the current point in time conditions in the Canadian bond market are largely being 24
27 If RBC’s forecast was generally accepted then current yields would be a lot higher, since the one year Treasury Bill yield is only 0.50%. By the unbiased expectations theorem the loss from investing in one year as opposed to 30 year Canadas is about 1.5% which has to be captured in a higher bond yield next year. This means that the yield has to increase by approximately 0.05%, not 0.85%. This is more consistent with the Consensus forecast.
42
driven by external factors and are still not “average” market conditions. To adjust for this I make 1
two adjustments: the first is simply to make the CAPM estimate conditional on the state of the 2
markets thereby converting it into a conditional CAPM or CCAPM, the second is to adjust for 3
the abnormally low Canada bond yields resulting from rampant bond buying programs by central 4
banks. 5
In terms of the CCAPM there are a variety of financial variables to use to condition on the state 6
of the markets. For example, I have previously discussed the Kansas City financial stress index 7
which attempts to do this by looking at a composite stress index, which should be correlated with 8
investor risk aversion. Another popular index is the Chicago Board of Exchange’s (CBOE) 9
volatility index (VIX). This calculates the expected standard deviation of the overall stock 10
market return from at the money call options.28 The standard deviation, as explained in my 11
Appendix B, is a measure of the overall risk or volatility of the stock market and has averaged 12
18-20% since 1926. Below is a graph of the VIX since 2004. 13
CBOE Volatility Index (VIX)
0
10
20
30
40
50
60
70
80
90
1/2/
2004
7/2/
2004
1/2/
2005
7/2/
2005
1/2/
2006
7/2/
2006
1/2/
2007
7/2/
2007
1/2/
2008
7/2/
2008
1/2/
2009
7/2/
2009
1/2/
2010
7/2/
2010
1/2/
2011
7/2/
2011
1/2/
2012
7/2/
2012
1/2/
2013
7/2/
2013
1/2/
2014
7/2/
2014
1/2/
2015
7/2/
2015
1/2/
2016
14
28 Call options pay off only if a certain event (usually the stock price) reaches a specific value and the only “unknown” to price them is the variability or standard deviation of the price. As a result, we can work back from the market price to estimate the implied standard deviation or volatility.
43
The average value for the VIX over the period 2004-2015 was 19.4%, close to the average from 1
stock market data going back to 1926.29 However, consistent with the KCFSI we can see the 2
huge increase in uncertainty during the financial crisis as the VIX hit a peak value of 80% or 4X 3
the average value. The VIX reflected the huge panic during the financial crisis, but as always the 4
panic subsides and since January 2011 (the last five years) has been below average at 16.2% 5
despite periodic attacks of nerves in August 2015 as the first China fears hit the stock market and 6
the first week of 2016 when they returned. 7
The problem with the VIX and the KCSFI is they are not direct estimates of a rate of return and 8
cannot be used to reliably alter a CAPM expected return estimate. An alternative is to use the 9
default or credit spread between risky corporate debt and default free Government of Canada 10
bonds. This is the corporate spread discussed earlier and similar to the VIX and KCSFI shows 11
the huge impact of the financial crisis when corporate A rated spreads peaked at 3.70% versus a 12
pre-crisis average of 0.94% (1979-2007). 13
In a series of hearings into the fair rate of return in the period after the financial crisis several 14
utility tribunals used the credit spread to adjust the CAPM estimate. For example, the AUC 15
stated in its generic decision (Decision 2009-216, November 12, 2009) 16
17
In arriving at their reasonable CAPM range of 7.13%-8.62% the AUC used a market risk 18
premium range of 5.00-5.75% a relative risk (beta) coefficient of 0.50-0.63 and a forecast long 19
29 It may surprise some but there is no evidence of any structural increase in uncertainty in the stock market: it has always been highly volatile.
44
Canada yield of 4.13-4.50%. Together with a 0.50% flotation allowance these result in a bottom-1
to-top range of 7.13-8.62%. The AUC then adjusted this estimate by a credit spread adjustment. 2
Q. DO YOU AGREE WITH A CREDIT SPREAD ADJUSTMENT? 3
A. Yes. Before several boards in 2009 I stated that much of the increase in credit spreads 4
was caused by liquidity problems in the market making function of investment banks. Since then 5
research at the Bank of Canada has helped to disentangle the liquidity from the pure credit risk 6
components in the corporate spread.30 This research concludes (page 29): 7
“our results show that for investment grade firms, the majority of the spread corresponds to 8 liquidity: on average, the liquidity component accounts for 63% of the spread.” 9
Although it remains very difficult to disentangle the liquidity from the pure credit component in 10
credit spreads, since 2010 as Canadian capital market conditions were returning to normal I have 11
been recommending a 50% adjustment to changes in credit spreads. Over a normal business 12
cycle this adjustment should average out to zero as capital market conditions fluctuate around 13
average levels. This adjustment was incorporated into ROE adjustment models adopted by the 14
OEB, Regie and the BCUC. 15
At the current point in time “A” spreads are at 1.91% or 91bps more than the typical average for 16
the business cycle, which I take to be 1.0%. Consequently I add 0.45% for this credit market 17
effect. I regard this sort of adjustment as converting the CAPM into a conditional CAPM where 18
the CAPM holds conditional upon the state of the financial markets. My estimate for the 19
CCAPM is as follows: 20
CCAPM Estimates 2016 21
Low end 5.95% 22 High end 7.00% 23
30 A. Garcia and J. Yang, “Understanding Corporate Bond Spreads Using Credit Default Swaps,” Bank of Canada Review, Autumn 2009
45
The average of these two is 6.48% and would be appropriate as an estimate for a fair ROE at this 1
stage of the business cycle in a capital market that is also typical for this stage in the business 2
cycle. However, I still regard the resulting ROE as an under estimate at the current point in time. 3
Q. WHY IS THIS CCAPM AN UNDER-ESTIMATE AT THE MOMENT? 4
A. In Appendix B Schedule 6, I develop a model to explain the behaviour of the real yield 5
on long Canada bonds, defined as the nominal yield minus the average of past, current and future 6
CPI inflation. Ignoring the indicator or dummy variables for WW2 and the 1970s, when there 7
was huge liquidity during the petro dollar recycling period, the model essentially says that the 8
real LTC bond yield is 1.34% plus a premium based on bond market uncertainty and a premium 9
based on the size of the government deficit. The model does well in explaining the very high 10
yields when there was huge volatility in the bond market and Canada was running deficits 11
approaching 10% of GDP. 12
However, since the financial crisis while we have seen bond market uncertainty go down, the 13
aggregate deficit in Canada has gone from a surplus to just below 2% of GDP. Normally this 14
would cause an increase in government debt pushing down prices and pushing up yields. 15
Plugging numbers into the regression model would predict a real long Canada yield of about 16
3.88% for 2014, rather than the actual 1.35% (2.60% average 2014 Canada yield minus 1.25% 17
inflation). However, the flood of government debt is being bought in part by non-residents due to 18
the impact of loose monetary policy elsewhere in the world. Once an indicator (dummy) variable 19
is added for the years since 2010, the 2014 real yield estimate is reduced by 2.53%. What this 20
indicates is that the current and forecast long Canada bond yields are severely depressed below 21
where they would have been had Canada remained partially segmented from external events. 22
The model also indicates that current bond yields are not being determined by ordinary investors 23
trading off risk versus return as assumed in standard risk premium models. 24
An additional insight is from looking at preferred yields. In old testimony I (with my late 25
colleague Dr. Berkowitz) presented four ROE estimation methods. One of them estimated the 26
fair ROE by looking at the premium of the earned ROE over the yield on traditional fixed rate 27
preferred shares and how this premium was valued by investors in terms of the market to book 28
ratio. The reason for doing this was that preferred shares are an equity instrument taxed at the 29
46
same rate as dividend income from ordinary shares. As a result, the tax bias from comparing the 1
fair ROE for a regulated utility with the yield on long Canada bonds is removed, since interest 2
income is fully taxed whereas dividend income via the dividend tax credit is not. 3
This tax effect is well known in capital markets. BMO-Nesbitt-Burns used to produce a Preferred 4
Share Quarterly that tracked the performance of the preferred share market. Prior to the 5
distortions introduced by the financial crisis, Nesbitt Burns provided the following yields as of 6
June 2004 when the stage in the business cycle was close to neutral: 7
June 2004 8 9 Retractable Preferreds (%) 10 Dividend yield 4.01 11 Mid Canada yield 4.09 12 After tax spread (corp) 1.77 13 After tax spread (indiv) 0.63 14 15 Straight Preferreds (%) 16 Dividend yield 5.48 17 Long Canada yield 5.34 18 After tax spread (corp) 2.54 19 After tax spread (indiv) 0.98 20 21 Floating Rate Preferreds (%) 22 Dividend yield 3.42 23 BA (3 month) 2.12 24 After-tax spread (corp) 2.25 25 After-tax spread (indiv) 1.22 26 27
The retractable preferreds are compared to mid Canada bonds, since the retraction feature 28
shortens their maturity as compared to a long bond. The traditional straight preferreds are 29
compared to long Canada bonds, while the floating rate preferreds are compared to 90-day 30
Bankers acceptances (BAs), since their dividends are usually reset quarterly. 31
The important point about the comparison is that what we observe in the capital market is a pre-32
tax yield. This is determined by both risk and taxes. Take the straight preferreds, for example, in 33
June 2004 the long Canada bond had a yield of 5.34%, while straight preferreds had a yield of 34
5.48%. Clearly, the preferreds would be regarded as riskier than the long Canada bond, since the 35
47
corporate issuer can default. However, the yield on the preferred shares was only 0.14% higher. 1
The reason is that the dividend income gets more favourable tax treatment than interest income. 2
The correct comparison is the after-tax yield difference, which BMO-Nesbitt-Burns gives as 3
2.54% in favour of the preferred shares for corporate investors and 0.98% for individuals. The 4
after tax yield difference gives the correct result: that on an after tax basis the riskier preferreds 5
give a higher yield. 6
In June 2004 the long Canada bond yield is given by BMO as 5.34% and the preferred share 7
yield at 5.48%. At the end of June 2004 the Scotia Capital “A” yield was 6.34% for a spread of 8
100 basis points over the long Canada bond yield, which is about “average” for a complete 9
business cycle. This is partly why I the 2004 data is useful as a starting point, since in most 10
respects it was “normal”. The Nesbitt-Burns data is no longer available but Standard and 11
Poors/TSX have published a preferred share index and the spread of the yield on this index along 12
with that on the Scotia Capital “A” bonds over equivalent maturity long Canada bonds is 13
graphed below.31 14
Preferred and A Spreads
100
150
200
250
300
350
400
450
1/1/
2010
2/1/
2010
3/1/
2010
4/1/
2010
5/1/
2010
6/1/
2010
7/1/
2010
8/1/
2010
9/1/
2010
10/1
/201
011
/1/2
010
12/1
/201
01/
1/20
112/
1/20
113/
1/20
114/
1/20
115/
1/20
116/
1/20
117/
1/20
118/
1/20
119/
1/20
1110
/1/2
011
11/1
/201
112
/1/2
011
1/1/
2012
2/1/
2012
3/1/
2012
4/1/
2012
5/1/
2012
6/1/
2012
7/1/
2012
8/1/
2012
9/1/
2012
10/1
/201
211
/1/2
012
12/1
/201
21/
1/20
132/
1/20
133/
1/20
134/
1/20
135/
1/20
136/
1/20
137/
1/20
138/
1/20
139/
1/20
1310
/1/2
013
11/1
/201
312
/1/2
013
1/1/
2014
2/1/
2014
3/1/
2014
4/1/
2014
5/1/
2014
6/1/
2014
7/1/
2014
8/1/
2014
9/1/
2014
10/1
/201
411
/1/2
014
12/1
/201
41/
1/20
152/
1/20
153/
1/20
154/
1/20
155/
1/20
156/
1/20
157/
1/20
158/
1/20
159/
1/20
1510
/1/2
015
11/1
/201
512
/1/2
015
1/1/
2016
Pref A
15
On January 1, 2010 long Canada bonds yielded 4.14%, “A” bonds 5.86% and TSX’s preferred 16
share series 5.44%. So the spreads were 130 bps for the preferreds, and 172 bps for the generic 17
31 This index is not comparable to Nesbitt-Burns index, which separated out the difference types of preferred shares.
48
“A” bonds. Compared to the June 2004 Nesbitt-Burns data these spreads had increased. The 1
preferred share spread from 14 bps to 130 bps and the “A” spread from 100 bps to 172 bps but 2
some of the change in the preferred share yield is due to the different index composition. Overall 3
these spread increases were due to the after effects of the financial crisis, but the graph then 4
indicates two things. First, the generic “A” yields moved broadly in tandem with the long 5
Canada bond yield without any trend. This is the change that the corporate credit spread 6
adjustment picks up. Second, while the preferred yield spread moved in tandem with the bond 7
spreads until August 2011, afterwards the spread increased dramatically. Another way of saying 8
this is that by December 2011 the preferred share yield had dropped 0.44% from 5.44% to 9
5.00%, while the A bond yield had dropped by 1.53% from 5.86% to 4.33%. 10
I judge the change in yields from 2010 to 2012 as reflecting the impact of the Federal Reserve’s 11
third round of quantitative easing (QE3) or what I have called Operation Twist.32 It appears that 12
there was a direct impact in Canada as foreign capital slipped into Canada in search of higher 13
yields. This affected both the government and to a lesser extent the corporate bond market, since 14
yields on both came down after September 2011. However, yields in the preferred share market 15
did not come down to the same degree causing the preferred share yield spread to widen. This is 16
probably because preferred shares are unattractive to foreign investors, since the dividends 17
attract with-holding taxes and non-residents do not get the dividend tax credit. 18
I placed the “Operation Twist” impact on the Canadian bond market as approximately 80 bps 19
from August 2011 through to May 2013, which was my judgment in the last generic cost of 20
capital hearing. However, in May 2013 Governor Bernanke spooked the markets by indicating a 21
withdrawal plan for the Federal Reserve’s bond buying program (QE3) and interest rates started 22
to back up in the expectation that the withdrawal would depress prices and increase yields. From 23
May 2013 long term government interest rates increased from 2.20% to 3.07% by October 2013 24
or by 0.87% and my forecast long Canada bond yield increased almost proportionately from 25
3.0% to 3.60%. Of more importance the spread of the preferred share yield over long Canada 26
bonds dropped from 2.54% at the end of April 2013 to 2.01% and was at times even lower. In 27
32 Since the objective was to “twist” the US yield curve and lower the yield on long term bonds.
49
comparison the spread of the yield on A bonds over long Canadas only dropped from 1.68% to 1
1.66%, so was basically constant. Consequently, I reduced my “Operation Twist” adjustment. 2
However, although the US bond buying program finished in 2014, new programs were started or 3
expanded elsewhere and a huge amount of government debt has been taken off bond markets 4
around the world. The result is that long Canada bond yields have backtracked and decreased 5
from 3.0% in October 2013 to the current level of barely over 2.0%, while the A spread has 6
increased from 1.66% to the current level of 1.90% and the preferred spread has increased from 7
2.01% to 3.78% and been highly volatile 8
An easier way of looking at the change in spreads is the following graph which shows the change 9
in the spread between the preferred share spread and the A spread. If all securities reacted in an 10
identical way to the changes in the long Canada bond market then this spread would average out 11
to zero. However, the change in the spreads indicates that this is not true. 12
Pref-A Spread
-100
-50
0
50
100
150
200
250
1/1/
2010
4/1/
2010
7/1/
2010
10/1
/201
0
1/1/
2011
4/1/
2011
7/1/
2011
10/1
/201
1
1/1/
2012
4/1/
2012
7/1/
2012
10/1
/201
2
1/1/
2013
4/1/
2013
7/1/
2013
10/1
/201
3
1/1/
2014
4/1/
2014
7/1/
2014
10/1
/201
4
1/1/
2015
4/1/
2015
7/1/
2015
10/1
/201
5
1/1/
2016
13
The increase in the preferred share yield spread from 0.80% in 2012 to the current 2015 average 14
of 1.3% or an increase of 0.50% reinforces my 2012 judgement that the long term Canada bond 15
50
yield is not as useful as it has been in the past for estimating the cost of equity capital and the fair 1
ROE and that the simple CAPM estimate underestimates the fair return.33 2
If the 1.30% spread of the preferred share yield over the A bond yield is added to the CCAPM 3
estimate as a current “Operation Twist” adjustment similar to 2012 we get the following: 4
Risk Premium Estimate 2014 5
Low end 7.25% 6 High end 8.30% 7 8
This would indicate a 2016 fair ROE of 7.78% for a benchmark utility, which is the mid-point 9
between the high and low estimates. Taking into account the volatility of current yields on utility 10
preferred shares, which are part of shareholder’s equity, but rank ahead of the common 11
shareholders I would tend to be conservative and recommend the same 7.5% as in 2012. 12
However, I note that consistent with Concentric’s review of allowed ROE’s prepared for the 13
Canadian Gas Association, the statistical evidence of lower bond yields is for lower allowed 14
ROEs than in 2012 and that allowed ROEs in both Alberta and Quebec have subsequently been 15
lowered. 16
IV DCF ESTIMATES OF THE FAIR ROE 17
Q. WHAT ARE YOUR DCF ESTIMATES? 18
A. In appendix D I review the DCF model and apply the model to the market as a whole and 19
highlight the problems in applying it to individual stocks. For the market as a whole I would 20
estimate the fair return as being 8.50-9.00% in Canada and slightly higher in the US. With 21
forecast long Canada bond yields at 2.12% for 2016, this implies a current market risk premium 22
of over 6.38-6.88%, which is in excess of both the historic evidence in Canada and the judgment 23
of the respondents to Fernandez’ survey. For the S&P gas and electric index the historic utility 24
risk premium is about 3.4%, which slightly exceeds the top of my Canadian utility risk premium 25
33 On August 7, 2015 Canadian Utilities issued a straight perpetual preferred share at a 5.0% coupon when the TSX preferred share index had a yield of 5.18%. However, many of the issues are from banks and their recent yields seem to reflect the fear of bank loan losses to oil and gas companies.
51
range, but is based on lower ten year US bond yields and not the 30 year bond. Subtracting the 1
spread between these two would indicate it is consistent. 2
I have traditionally viewed my DCF estimates as “checks” on my CAPM estimates, since in my 3
view CAPM estimates have usually been in the right “ball-park.” However, the recent very low 4
long Canada bond yields have forced me to re-evaluate this and look at what drives the 5
difference between the DCF and simple CAPM estimates. This is because they should be 6
consistent. The CAPM equation is as follows: 7
*MRPRK F 8
In words, the required (fair) return is the risk free rate (RF) plus the risk premium comprised of 9
the market risk premium (MRP) times the beta coefficient (β). For the market as a whole we can 10
simply set the beta to one since this is definitional. 11
The risk free rate is directly observable since the practise in Canada is to use the long Canada 12
bond yield as the risk free rate, while the market risk premium is reasonably objective, 13
particularly now that we have Fernandez’ survey data from thousands of professionals in the 14
area. Consequently, the major area of dispute is the relative risk or beta coefficient, and even 15
here there is not much doubt that utilities are lower risk than the market. Hence the big advantage 16
of the CAPM is that it is difficult to make big mistakes. The CAPM also avoids one of the big 17
problems with DCF estimates in that the forecast inflation rate is automatically incorporated into 18
the long Canada bond yield, since we use the nominal rather than the real yield. This is currently 19
not a significant problem, since inflation is so low, but part of the reason the DCF model fell out 20
of favour was that it was giving bad signals when applied mechanically in the 1990s, when there 21
was a structural break in the forecast inflation rate. 22
52
The classic Gordon growth model,34 often referred to as the DCF model in most testimony before 1
regulatory bodies, is as follows: 2
gP
dK 1 3
In words, the required rate of return is the forecast dividend yield plus the long run growth rate. 4
Conceptually the DCF model and CAPM should give exactly the same values but, of course, 5
since they approach it from a different perspective there is always estimation error. For the 6
market as a whole the forecast dividend yield can be estimated with very little error, so the 7
estimation error is with the forecast long run growth rate, which also is easier to estimate for the 8
market than for an individual stock. As a result, if the CAPM and DCF estimates differ 9
significantly, then it is mainly due to the difficulty in estimating the growth rate in the DCF 10
model and the risk premium in the CAPM.35 11
We can assess the relative value of the DCF and CAPM by graphing the “known” parts of both 12
models for the overall market, which are the long Canada bond yield and the TSX dividend 13
yield. Note in the following graph that the current period is quite unusual since dividend yields 14
on the TSX are higher than the long Canada bond yield. This implies that either the long Canada 15
bond yield is unusual or the pricing in the equity market is unusual. 16
34 The model is named after the late Professor Myron Gordon of the University of Toronto. 35 Note since for the CAPM we are dealing with the market return the following analysis is general for any risk premium model
53
Dividend and Long Canada Bond Yields
0
2
4
6
8
10
12
14
16
18
1956M01 1961M01 1966M01 1971M01 1976M01 1981M01 1986M01 1991M01 1996M01 2001M01 2006M01 2011M01
TSXYield Canadas
1
Since both the DCF model and CAPM should give the same answer, we can set them equal to 2
each other, which indicates that for the market as a whole 3
gMRPRPdCAPMDCF F 1 4
Or in words the directly observable spread between the long Canada bond yield and the TSX 5
dividend yield is equal to the long run dividend growth rate minus the market risk premium. 6
From the above graph we can see that there is generally a very large difference between the two 7
indicating that the expected growth rate in the stock market was much higher than the market 8
risk premium but that this growth rate has varied over time. The reason for this difference is 9
mainly the gradual increase and then decrease in the CPI inflation rate over the period since 10
1956. This inflation rate is directly captured in the long Canada bond yield since it is a nominal, 11
and not a real yield, and yet is in the “unobserved” growth rate in the DCF model. Note for 12
example, that the increasing and high rates of inflation in the 1960-1980 period (next graph) 13
coincides with the increasingly large difference between the LTC and TSX dividend yields. 14
54
Average Annual CPI Inflation
0
2
4
6
8
10
12
14
1956
1958
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
1
It is possible to come up with a simple or naïve estimate of the market return by adjusting for this 2
inflation/real yield bias to the estimates. For example, we can assume that for the DCF model the 3
forecast growth rate is the actual CPI inflation rate at the time, based on year over year changes, 4
and then add a 3.50% real growth rate. This gives a simple growth rate forecast to add to the 5
dividend yield and thus a simple or naïve DCF estimate for the market as whole. Similarly, we 6
can add a long run market risk premium of 3.5% to the long Canada yield for a simple CAPM 7
estimate. For the entire period 1956-2014 the average naïve DCF estimate is 10.58%, while the 8
average naïve CAPM estimate is 10.52%, or a difference of only 0.06%between the two, so “on 9
average” these assumptions seem to make sense. 10
To see how robust this simple procedure is, the following graphs the difference between the two 11
estimates for every month since 1956 until October 2015. The graph indicates that the difference 12
was very large from the mid 1970’s until the late 1990’s. The reason for this difference is 13
twofold. First, in the 1970s inflation was increasing and bond yields captured the expected 14
inflation component directly. However, investors did not believe that the Bank of Canada and the 15
Government would allow these high levels of inflation to continue. This resulted in very low real 16
yields on LTC bonds leading to a positive difference between the DCF and CAPM estimates. 17
55
Difference between Naive DCF and Risk Premium estimates for the Market
-8.00
-6.00
-4.00
-2.00
0.00
2.00
4.00
6.00
8.00
10.00
12.0019
56M
01
1957
M10
1959
M07
1961
M04
1963
M01
1964
M10
1966
M07
1968
M04
1970
M01
1971
M10
1973
M07
1975
M04
1977
M01
1978
M10
1980
M07
1982
M04
1984
M01
1985
M10
1987
M07
1989
M04
1991
M01
1992
M10
1994
M07
1996
M04
1998
M01
1999
M10
2001
M07
2003
M04
2005
M01
2006
M10
2008
M07
2010
M04
2012
M01
2013
M10
2015
M07
1
Once investors caught up to the impact of high inflation the reverse set in, as the budget deficits 2
at the Federal level convinced the market that the government would inflate its way out of its 3
deficit problems, rather than bring down inflation. As a result, while the year over year inflation 4
rate dropped dramatically, LTC bond yields did not at first similarly drop, leading to very high 5
real yields and simple CAPM estimates exceeding their DCF equivalents. It is this phenomenon 6
of low real bond yields in the 1970s and 1980s and high real bond yields in the 1990s that is the 7
major reason for the positive deviations from 1970-1982, and the negative deviations afterwards. 8
The second reason is simply that the real GDP growth rate and the market risk premium have not 9
remained constant since 1956. I testified extensively in the 1990s to the effect that the market 10
risk premium was very low due to the high real interest rates and risks attached to government 11
bonds. Subsequently, I have increased my estimates of the MRP as this risk has been removed 12
and currently use 5.0-6.0%, rather than 3.5% as my market risk premium estimate. Similarly, the 13
long run real growth rate may have dropped and might be lower than the 3.5% used in the simple 14
model. 15
However, the point is that we can “ballpark” the broad range for the DCF estimate for the market 16
just as we can for risk premium models like the CAPM. At the end of October 2015 the TSX 17
dividend yield was 3.17% and the year over year inflation rate 1.36%, so with the 3.5% real 18
56
growth rate the simple DCF estimate is 8.18%.36 Similarly with the October 2015 long Canada 1
yield of 2.12% and a 3.5% market risk premium the simple CAPM estimate is 5.62%. As a 2
result, there is currently a positive 2.56% difference when we subtract the CAPM estimate from 3
the DCF estimate. Further note from the graph that this difference between the two has gradually 4
gone from negative to positive over the last 15 years as long Canada bond yields have gone 5
down. The reason for this is simply the fact that the real yield on the long Canada bond has 6
dropped, which has penalised the risk-premium estimate relative to the DCF estimate. 7
Of course current DCF and risk premium estimates are not the naïve ones graphed above, but 8
instead allow for differences in the market risk premium and growth rates, but this analysis 9
confirms the implications of the current problems in the bond market on the CAPM estimates. As 10
a result it supports my adjustments to the CAPM estimates and the value of currently looking at 11
DCF estimates. 12
Q. WOULD YOU USE THESE ESTIMATES? 13
A. No. These are very simple or naive estimates that use average numbers for a very long 14
period of time. They are presented simply to show that while the DCF and CAPM estimates are 15
consistent over long periods of time, they both have problems when used mechanically during 16
periods of very high or low real Canada bond yields. The analysis also helps explain why DCF 17
estimates fell out of favour in the 1990s while the validity of recent CAPM estimates has 18
recently been questioned. 19
Q. IS THERE ANY OTHER EVIDENCE ON THE VALIDITY OF THESE 20
ESTIMATES? 21
A. Yes. What is important is that there is another side to estimating the fair ROE and cost of 22
equity capital. This is that the required rate of return on the part of the investor (cost of equity 23
capital) is also the expected rate of return. Defined benefit pension funds need this expected rate 24
of return to determine whether a fund is in deficit or surplus. On October 19, 2012 TD 25
36 This is 1.0317*(1.0136+.035)-1
57
Economics produced its own analysis of the long run returns of the type needed in defined 1
benefit pension plans.37 2
3
The important point about the TD Economics forecast is that the going forward risk premium for 4
equities minus bonds is 4.00%. This is not the market risk premium, since adjustments need to be 5
made but it is certainly in the right ballpark. 6
Q. WHAT ADJUSTMENTS NEED TO BE MADE? 7
A. As TD Economics notes its return forecast is for ten year geometric returns so they have to 8
be converted to arithmetic returns. To make this adjustment for very long returns we add half the 9
variance of the arithmetic return as explained in my Appendix B, with data in Schedule 8. 10
Historically the standard deviation of equity returns has been about 20% (0.20) so the variance is 11
0.04 and half this is 0.02 or 2.0%. Similarly, the volatility of the long Canada bond return has 12
been about 9% (0.09). I would suspect that this overstates the future volatility, since it is unlikely 13
we will see LTC yields at almost 20% again, but this means a variance of 0.0081 and half this is 14
0.4%. 15
37 TD Economics, An Economic Perspective on Long-term Financial Returns.
58
So converting these long run returns to short run arithmetic returns means market risk premium 1
of 5.60% as follows: 2
Long run 1/2 the variance Arithmetic 3
Equities 7.0% 2.0% 9.0% 4
Bonds: 3.00% 0.40% 3.40% 5
However, the TD Economics forecast is over the yield on the DEX universe bond index and not 6
over long Canada bonds. The universe of bonds would have lower duration or maturity than the 7
30 year long Canada bond, but can be expected to earn more since they have default risk. Given 8
the prior long Canada forecast of 2.12% for 2016, this would increase the market risk premium 9
estimate closer to 6.00%, but is within my 5.0-6.0% range. 10
Q. WHAT IS YOUR FAIR ROE FOR A BENCHMARK UTILITY? 11
A. I would judge a fair ROE for 2016 to be in a range 7.25-8.30% for 2016. My estimates 12
are based on the following: 13
Risk Premium 14
Base LTC forecast: 2.75% 15 Normal utility risk premium: 2.25%-3.30% 16 Issue costs: 0.50% 17 Normal Fair ROE 6.03% 18 Credit Spread Adjustment 0.45% 19 Operation Twist Adjustment 1.30% 20 Fair ROE: 7.78% 21 22
DCF: 23
Canadian equity market return: 8.50-9.00% 24 US SP500 Electric Utility risk premiums: 3.40% 25 Low risk US sample Median DCF: 7.02% 26
Given my increasing skepticism surrounding the interest rate forecasts of some of the main 27
economic forecasters I would judge the fair ROE to be less than the 7.50% I recommended in 28
2012. There is no question that there is objective evidence of a decline in overall interest rates 29
59
since 2012. However, since the long Canada yield has yet to hit my 3.80% trigger I would keep 1
my recommendation at 7.50%. 2
Q. IS THERE ANY OTHER EVIDENCE SUPPORTING 7.50%? 3
A. Yes. Ultimately stock market returns are driven by the returns earned by companies, that 4
is, what is sometimes referred to as “comparable earnings.” In 1925 John Maynard Keynes 5
pointed out38 that there were two sources of returns from investing in the stock market. The first 6
is called the investment return which Keynes defined as “forecasting the prospective yield of an 7
asset over its entire life.”39 In modern terminology this would be the internal rate of return on the 8
firm’s cash flows, or an approximate average return on equity. The second component he called 9
the speculative return, which involved forecasting the psychology of the market and what 10
Keynes referred to as the change in the basis of valuation. In modern terminology this would be a 11
change in the price earnings ratio. Keynes discussed this speculative return as being generated by 12
the “state of confidence” and “animal spirits” but he also pointed out it is affected by the level of 13
interest rates.40 14
Keynes’ point would be that a firm may earn 10%, but if the valuation of that firm changes by 15
10% then the investor would earn both a speculative return as well as an investment return. This 16
total return is then what we look at when we look at the returns over long periods of time on the 17
TSX Composite or the SP500. However, in aggregate the change in the basis of valuation cannot 18
go on forever. We cannot continue to have a state of high confidence any more than interest rates 19
can continue to increase or decrease: both of them will tend to revert back to some long run 20
average. However, professional investors according to Keynes are mainly concerned with 21
speculative returns or forecasting the change in the basis of valuation six months out. In contrast 22
buy and hold or fundamental investors are mainly concerned with the investment return: finding 23
good companies and holding them regardless. 24
38 Quoted in John Bogle, The Lessons of History, September 12, 2011, John Maynard Keynes, 1925, Review of Common Stocks as Long Term Investments, Edgar Lawrence Smith 39 This definition comes from chapter 12 of the General Theory of Employment Interest and Money, Macmillan London, 1936 40 Page 149 of the General Theory
60
Warren Buffet is probably the most successful fundamental investor of the last fifty years. He 1
repeated Keynes’ argument by stating:41 2
“The most the owners in aggregate can earn between now and judgment day is what their 3 businesses in aggregate earn.(italics in original) True by buying and selling that is clever 4 or lucky, investor A may take more than his share of the pie at the expense of investor B. 5 And yes, all investors feel richer when stocks soar. But an owner can exit only by having 6 someone take his place. If one investor sells high, another must buy high. For owners as a 7 whole, there is simply no magic - no shower of money from outer space – that will enable 8 them to extract wealth from their companies beyond that created by the companies 9 themselves.” 10
Buffet’s main criticism was for the financial professionals who help individuals to trade so that 11
in aggregate investors lose part of the pie in fees. However, Keynes, Bogle and Buffet all point 12
out the basic fact that short run returns can deviate from the returns generated by the economy 13
and earned by firms, the investment return, but in the long run this is all there is. 14
This discussion of what generates stock market returns is provided since in the long run the 15
average stock market return should approximate the average investment return,42 that is the 16
speculative return should average out to zero. There are two ways in which we can look at the 17
investment return; the first is to look at average rates of return on equity and the second to look 18
at a Gordon growth model for the economy as a whole. 19
In Schedule 3 is the average ROE for Corporate Canada since 1987 as reported by Statistics 20
Canada (Table # 1800003). Over this period the average ROE has been 9.56%. The third column 21
reports the annual return on the TSX Composite which over the same period has been 10.39% or 22
0.83% more. However, the rough equality over this period hides the significant year to year 23
variation where speculative returns have been significantly high or low. For example, in 1987 24
Corporate Canada earned 11.19% but the TSX Composite only 5.88%, so there was a short term 25
speculative loss of 5.31%. It wasn’t until 1989 than the TSX returned 21.37%, compared to 26
Corporate Canada’s ROE of 11.79%, that the speculative return turned positive. In each year we 27
41 Warren Buffet’s comments in Berkshire Hathaway’s 2006 Annual Report as reported in Fortune, March 20, 2006. 42 It is an approximation since it depends on the market to book ratio at the start of the period.
61
can see that the speculative return is highly volatile and on average 5-6 times more volatile than 1
the investment return. 2
The second way of looking at the investment return is that used by Jack Bogle, the founder of 3
Vanguard Mutual funds. He estimated the investment return using the Gordon model, where at 4
the start of each year he added the subsequent five year earnings growth to the dividend yield. He 5
then took this analysis back to 1900 and provided the graph in Schedule 4. This marginally 6
understates the investment return since he should have used the forecast dividend yield, but as he 7
noted it did not materially affect the results. He estimated this investment return at 8.8% or 8
slightly less than the average US stock market return of 9.1%. However, since he underestimated 9
the investment return the difference in reality is de minimis. Just like Keynes, Bogle also noted 10
the persistent tendency for reversion towards the mean, which is another way of saying that high 11
or low stock markets and PE multiples do not last. As Bogle noted (page 11) 12
“Over the long run it is the durable economics of enterprise – enterprise – that has 13 determined total return: the evanescent emotions of investing – speculation –so important 14 over the short run, has ultimately proven to be meaningless.” 15
The approach of Keynes, Buffet and Bogle is a standard approach used by fundamental investors 16
who look at individual stocks, rather trying to time the market. However, it is sometimes used to 17
time the market over a long horizon. In a US Equity Strategy Report (July 18, 2012) RBC 18
pointed out that historic long run equity returns in the US had been 9.4% nominal or 6.2% real 19
since 1900. However, going forward they used what they termed a “Grinold-Kroner-Siegel” 20
supply side model, which is actually just the modified DCF model, where the long run equity 21
return (R.) was equal to: 22
23
62
This equation says that the long run return is equal to the dividend yield minus share dilution, 1
caused by stock issues, plus inflation and real earnings growth, plus the change in the PE 2
multiple. 3
RBC placed the dividend yield at 2.1% and while they judged average share dilution from new 4
issues to be 2.0% going forward they expect increased share buybacks to cause this to drop to -5
0.50% so the adjusted dividend yield is 1.60%. They anticipated inflation in the US at 2.1% and 6
real growth of 3.8% for a nominal growth forecast of 5.9% almost identical to my estimate for 7
the Canadian market. RBC then forecast that the normalised Shiller PE ratio will move back to 8
16-18X partly due to changing demographics in the US as the proportion of peak savers 9
(demanders of equities) drops as the baby boomers age. They forecast that this would cause a 10
change in the basis of valuation reducing the long run (ten year) equity market return by 1.0%. 11
RBC’s long run forecast is therefore for a 4.9% US equity market return. The following graphic 12
pulls together the RBC forecast: 13
14
I am not as pessimistic as RBC and do not put much faith in the Shiller PE model,43 but the share 15
dilution adjustment is one that many make since increasingly senior executives are taking some 16
of the price appreciation through the exercise of executive stock options. However, it indicates 17
that my own estimate of 8.50-9.0% for the Canadian market as a whole is not “low”. 18
With the stock market as a whole earning about 9.0% and Canadian regulated utilities 19
unambiguously lower risk a 7.5% ROE looks reasonable. 20
43 The problem is the Shiller PE uses a ten year backward average of real earnings, which in the US includes the huge losses from the Great Recession and before that the Internet Bubble. I don’t think these scale of losses will happen until we forget the lessons of the Stock Market crash, probably in another 30 years.
63
Q. WOULD YOU RECOMMEND THE USE OF AN ROE ADJUSTMENT 1
MECHANISM? 2
A. Yes. I agree with the BCUC’s decision on an ROE adjustment model. Adjustment models 3
offer the potential to keep the ROE current without extensive hearings. However, as indicated at 4
that time, and confirmed by my current analysis, keying the fair ROE off the long Canada bond 5
yield forecast currently causes significant problems due to the distortions in that forecast. My 6
judgment is that the fair ROE has not decreased to the same extent as the decrease in the long 7
Canada bond yield, since they are currently not being set by investors but “global policy makers” 8
that is, central banks. In this respect, it is not the normal application of monetary policy that is 9
the problem, but the bond buying programs adopted by the most important central banks. As a 10
result, adjusting the ROE by 50-75% of the decrease in long Canada bond yields I judge to 11
potentially under-estimate the fair ROE. In 2012 I stated: 12
I would recommend an ROE adjustment model where the ROE adjusts by 75% of the 13 forecast change in the long Canada bond yield and 50% of the change in the credit 14 spread. This would be subject to a minimum forecast long Canada bond yield of 3.80% 15 and my going in ROE recommendation. Similar models are in use by the Regie and OEB. 16
In the hearing I also indicated that I currently had no problem with a 50% adjustment to changes 17
in the long Canada bond yield since the inflation risk premium has dissipated. However, at that 18
time I expected that the end of QE3 in the US would lead to some normalisation in the long bond 19
yield, but the expansion/introduction of similar programs elsewhere, as well as the sheer scale of 20
the bonds purchased by the US Federal Reserve, has made me more pessimistic. 21
I still regard 3.8% as a minimum long Canada bond yield consistent with investors trading off 22
risk and return, since this equates to a negligible real after tax rate of return for a taxable 23
investor. RBC and other forecasters expect this minimum to be reached in 2018, but I am not as 24
optimistic and much depends on when, if ever, the Federal Reserve starts to sell its stock-pile of 25
government debt. Consequently, I am also happy for the BCUC to set a fixed rate for the period 26
2016-2018 and if this is considered would recommend a fixed rate of 7.50%, which is generous, 27
since I don’t think one year ahead forecast long Canada bond yields will increase to the 3.80% 28
trigger in the immediate future (next three years). 29
64
V CAPITAL STRUCTURE & BUSINESS RISK 1
Q. WHAT IS YOUR OVERALL JUDGMENT ON FEI’s CAPITAL STRUCTURE? 2
A. As indicated in my full analysis three years ago44 I judge the best way to handle capital 3
structure as the approach adopted by the National Energy Board, the Alberta Utilities 4
Commission, the Regie and the Ontario Energy Board, which is to determine the capital structure 5
based on the business risk of the utility. Utilities with higher business risk should then have more 6
common equity, so that less financial risk offsets higher business risk to equalise total risk. A 7
board can then award the same allowed ROE from a generic cost of capital hearing. This 8
normally works, but sometimes an adjustment to both the common equity ratio and the ROE is 9
necessary, particularly if it results in an inefficient capital structure. For FEI I see no material 10
changes since 2012 and make the same recommendation for 35% common equity as I made then, 11
but will address three issues: changes in FEI’s business risk; 2) the implications of amalgamation 12
and 3) financing issues and the capital attraction standard. 13
Q HOW WOULD YOU ASSESS BUSINESS RISK? 14
A. I agree with the NEB where in RH-4-2001 they differentiated between short run and long 15
run risk. Short run risk is the ability to earn the allowed ROE and reflects the return on capital. 16
Long run risk is the return of capital and reflects the ability of the utility to recover its investment 17
in plant and equipment, that is, capital recovery risk. 18
44 I have not repeated pages 6-26 of that evidence which is largely generic, but my current recommendation is based on the same principles.
65
The NEB stated that for the TransCanada Mainline (Page 24 of the Decision) 1
2
At that time the NEB pointed out that the Mainline had been able to earn its revenue requirement 3
(and allowed ROE) but that the possibility that it may not be able to do this in the future had 4
increased. The NEB has subsequently increased the Mainline’s common equity ratio from 30% 5
to 40% in several hearings to reflect its increased capital recovery risk 6
However, long term risks must eventually become short run risk to have any impact. As I stated 7
before the NEB in RH-4-2004, 8
“If problems occur, then firms bring these problems to the regulator and frequently 9 “compromises” are worked out. This is part of the regulatory bargain and only regulated 10 firms have this capability. For example if a competitive firm suffers a supply shock then 11 the stockholders are directly affected, but in contrast a regulated firm can have losses put 12 in a deferral account and allocated to future customers or apply to the regulator for other 13 means of protecting the stockholders from loss. Consequently it is unreasonable to expect 14 no action on the part of the regulator to the increased risk after year 11 in the above 15 example. 16
The increased risk after year 11 is referring to the present value of the cash flows beyond year 17
10, which I arbitrarily referred to as long run risk. The point is that when serious risks do arise it 18
is extremely rare for a Canadian utility not to come before the regulator to ask for some 19
reallocation of costs to keep the shareholder whole. This is what happened with the Mainline in 20
the RH-03-2011 hearing before the NEB, where TransCanada asked for costs to be reallocated 21
from the Mainline to customers of NGTL; a reallocation of depreciation across different zones to 22
avoid stranded costs; and significant changes in its rate design. The NEB did not allow all of the 23
66
changes asked for, but the fact is there was a hearing and policy measures to deal with the 1
Mainline’s problems. 2
Q. IS THE BCUC AN UNRESPONSIVE REGULATOR? 3
A. No, not in my judgement, in fact exactly the exactly the opposite. In 2002 the BCUC had 4
to decide how to protect Pacific Northern Gas (PNG) from significant load losses. In response it 5
allowed a special Industrial Customers Deliveries Deferral Account to capture the difference 6
between forecast and actual sales to Methanex and some other big industrial customers. It also 7
approved a longer term contract with Methanex at reduced tolls to keep it on PNG’s system. This 8
is the normal regulatory response in Canada, which is to set up a deferral account to capture hard 9
to estimate items to make sure that they are passed on to customers and not born by the utility’s 10
shareholders. No regulator can protect a utility from basic underlying economic forces, such as 11
the disappearance of demand, but it can alleviate them. I would assume that this is also why the 12
BCUC largely denied PNG’s request for more common equity and a higher allowed ROE: 13
increasing these and with them the revenue requirement and tolls does not make a utility more 14
competitive. 15
In a current hearing before the Newfoundland and Labrador Board of Commissioners of Public 16
Utilities, in Appendix A to their October 16, 2015 report Concentric Energy rates the BCUC as 17
the “best” Canadian regulator based on a point system allocation of their DBRS ratings on the 18
following factors: 19
1) Deemed 20 2) Allowed ROE 21 3) Energy cost recovery 22 4) Cost of service vs incentive rate making 23 5) Capital cost recovery 24 6) Political interference 25 7) Retail rate 26 8) Stranded costs 27 9) Rate freeze 28 10) Market structure 29
DBRS is Canada’s premier rating agency and the rating seems to be based on their credit rating 30
support, not protecting the public. Each factor is rated on a scale from 1-5 with 5 being the best 31
and then seemingly added up. Although, this seems to treat all factors equally, BC with 40 is 32
67
regarded as Excellent and is the highest rated regulator in Canada. What this implies is that 1
Concentric sees the BCUC as the most supportive regulator in terms of its DBRS bond rating. 2
The following graphic reproduces Concentric’s full list of scores: 3
4
I don’t necessarily agree with Concentric’s analysis, since there is significant overlap in some of 5
the categories and the ones that count are the ability of the utility to earn their allowed ROE and 6
the treatment of stranded costs. However, I would agree with the general assessment that the 7
BCUC is a very supportive regulator. 8
Why this assessment is important is that Moody’s reported on how it determined its bond ratings 9
in August 2009.45 10
45 Infrastructure Finance; Regulated Electric and Gas Utilities, August 2009.
68
The report refines their assessment into four major areas as follows: 1
Regulatory framework: 25% 2 Ability to recover costs and earn profits: 25% 3 Diversification: 10% 4 Financial strength and liquidity: 40% 5
The weights in the last column are the percentage allocation to each of the four categories. 6
Moody’s states very clearly “for a regulated utility the predictability and supportiveness of the 7
regulatory framework in which it operates is a key credit consideration and the one that 8
differentiates the industry from most other corporate sectors.” A quick glance at Moody’s 9
weights indicates that fully 50% of the weighting is based on the first two criteria which both 10
reflect the supportiveness of the regulatory environment. 11
My judgement would be that the BCUC is a very supportive regulator and that this is the most 12
important aspect of its risk: certainly from the bond rating agencies perspective. I would judge 13
that FEI would be treated fairly if any of its risks ever materialise and that the company and 14
utility analysts make a fundamental mistake when they ascribe risks to the utility shareholders 15
since if these risks ever materialise as far as possible I would judge that the BCUC would allow a 16
new deferral account and the costs will be allocated to rate payers. At Schedule 6 is a list of 17
FEI’s deferral accounts. 18
Q. WHAT HAS BEEN THE HISTORY OF FEI’S ABILITY TO EARN ITS ROE? 19
A. The following graphs FEI’s allowed versus actual ROE, this is essentially where the 20
“rubber meets the road.” In a dictionary sense, risk is defined to be “the probability of incurring 21
harm.” In financial terms, harm is generally meant to mean a financial loss, so that we can 22
examine the ability of FEI to earn its ROE as an objective measure of risk.46 23
46 FEI did not provide full data for 2002 I think because it changed its year end.
69
FEI ROE
8.00%
8.50%
9.00%
9.50%
10.00%
10.50%
11.00%
11.50%
12.00%
12.50%
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
Allowed ACTUAL 1
Over the period since 1994 the difference between FEI’s allowed and earned ROE, prior to any 2
sharing from performance based regulation, has been 0.86%. This average is slightly high due to 3
a 0.92% under earning in 1994. The more recent performance since 2003 when there is 4
uninterrupted data is for a 1.17% over earning. To all intents and purposes FEI’s shareholders 5
have not suffered any losses or experienced any risk. 6
I first testified in an ROE hearing before the BCUC with my late colleague Dr. Berkowitz in 7
1993 and since then have never seen FEI (TGI, BC Gas), as the graph of actual versus allowed 8
ROEs indicates, suffer any material losses. This is despite the fact that at every hearing I listen to 9
FEI’s expert witnesses as to how risky the company is, or how that risk is invariably increasing. 10
Sooner or later FEI should experience some significant losses to offset the fact that it continues 11
to earn a significant risk premium. You have to go back to 1998 to see any loss that could be 12
regarded as material and even the 0.59% loss at that time has to be compared with the minimum 13
2.25% risk premium I recommend. 14
Q. FUTURE RISK 15
A. The answer always given to the historic absence of Canadian utilities actually 16
experiencing any material risk is that risk is a forward concept and it could happen. This is 17
correct and in the case of the TransCanada Mainline some of those risks have materialised in 18
terms of load losses and pipe on pipe competition. In the case of FEI the experience of the 19
Mainline is useful since the major concern is by pass of the utility’s assets leaving them stranded 20
70
and taken out of the rate base, since they are no longer used and useful. Whether this happens 1
depends on alternatives available to FEI’s customers and intervention by the government. 2
In terms of the basic supply and demand for natural gas, the Canadian Gas Association produces 3
an annual report on the industry.47 In their latest 2015 analysis the CGA indicated that Canada 4
has over 200 years of supply coverage at current rates of production. As with any natural 5
resource how much is commercially available depends on market prices, so low prices reduce 6
supply even though it is a boon to consumers. As a result, the 200 year estimate underestimates 7
actual supply since as the resource is used up we might expect prices to rise and what are 8
currently marginal supplies to come on stream with more drilling. Currently drilling is well 9
below the peaks in 2005. The CGA indicates for example that current low prices have retarded 10
some drilling in traditional “dry” gas areas. Regardless, the CGA data from the National Energy 11
Board does not indicate any supply problems. As the following graphic indicates supply has 12
shifted to shale gas and tight gas. 13
14
With such a plentiful cheap resource to distribute we should expect an expansion in supply 15
which is what has happened as there has been significant expenditure on the distribution system. 16
47 http://www.cga.ca/wp-content/uploads/2015/05/CGA-Annual-Natural-Gas-Report-2015_EN-Final.pdf
71
The following graphic indicates the total capital investment in the natural gas industry in 2014. 1
2
As the CGA explains, upstream extraction invested $19 billion, pipeline expansion another $1 3
billion and there was a further $2.6 billion in distribution spending. The significant amount spent 4
on distribution indicates that the industry does not see any long run market problems. 5
72
The reason for the significant expenditures in the natural gas industry is that it is plentiful and 1
cheap as the following CGA graphics indicates, 2
3
The graphic indicates that natural gas has a significant cost advantage over propane, electricity 4
and heating oil in the residential space heating and water heating markets. Currently the CGA 5
estimates that the “typical” user can save upwards of $2,000 by using natural gas rather than 6
competing fuels and as a result natural gas residential users have increased by an average of 7
100,000 users a year for the past ten years. 8
Q. DOES THIS GENERAL ASSESSMENT APPLY TO FEI? 9
A. Yes. The cost competitiveness of natural gas for residential users is particularly important 10
for FEI since unlike Union Gas and Gaz Metro it has very limited industrial load and revenues 11
from industrial users typically account for only about 7% (Table C3 company evidence) of total 12
revenues.48 This depends on the stage in the business cycle but residential and commercial 13
48 A recent report from Simon Fraser University in Burnaby for the CGA indicates that natural gas is particularly important for the following industries: Food and Beverage, Wood Products, Pulp and Paper, Chemical Products,
73
revenues are very important for FEI. In answer to BC (AMPC) IR 1.2.12 FEI provided the 1
following cost comparison for someone in the Lower Mainland using natural gas for heating. 2
3
In 2012 I pointed out that the cost advantage of natural gas over electricity had turned around 4
from the time of the 2009 hearing. At that time FEI claimed that natural gas had a 28% cost 5
advantage over electricity and its risk had increased. However, by 2012 that cost advantage had 6
increased to 45% and it is now 57%. Whatever “increased” risk existed at the time of the 2009 7
hearing, when FEI’s common equity ratio was substantially increased from 35% to 40%, has 8
since dissipated. In terms of fundamental risk natural gas is highly competitive for residential 9
space heating and water heating and is expected to continue to be so. 10
Q WHAT ABOUT THE HIGHER CAPITAL COST OF NATURAL GAS VS 11
ELECTRICTY INSTALLATIONS? 12
A. I don’t think there is any reason to doubt that there is a higher capital cost attached to a 13
natural gas installation than electricity, but this has always been the case and is recovered from 14
the lower running costs. As I indicated in 2012 this has been a major factor in the low 15
penetration rate of natural gas in the two Maritime, greenfield, natural gas distributors in Nova 16
Scotia and New Brunswick. Both Heritage Gas and Enbridge Gas New Brunswick have failed to 17
Cement and Lime, and the Primary Metals industries. Most of these are also important for the BC economy, which generally ranks third behind Ontario and Quebec.
74
meet their initial penetration targets for this reason. However, the same argument works in 1
reverse: FEI has a huge installed residential customer base and they are not going to rip out their 2
natural gas systems and replace them with electricity given that their systems are a sunk cost and 3
natural gas is so much cheaper than electricity. Moreover the capital cost is essentially the same 4
across Canada, as is the declining customer use as retail customers move to high efficiency 5
furnaces. Despite these long known facts, recently the Ontario Energy Board has confirmed the 6
36% common equity ratios for both Enbridge Gas Distribution (EGDI) and Union Gas. 7
Consequently I do not regard them as either new or material. 8
Q. WHAT IS FEI’S GROWTH PROSPECTS? 9
A. As in 2012 FEI indicates that a risk factor is its capture rate in the residential market. 10
There is no denying that the Greater Vancouver area is condifying, the same as downtown 11
Toronto, and that natural gas capture rates in this market are lower than for traditional single 12
family dwellings. To some extent the amalgamation may reduce some of this impact as it reduces 13
some of the impact of the lower Mainland. 14
75
The following graphic from FEI indicates this. 1
2
However, I do not regard this as a risk factor. FEI continues to capture 77% of the traditional 3
single family and 59% of the semi-detached market, but it is inevitable that multiple dwellings 4
are better served by electricity.49 More to the point it is difficult to see how this affects FEI’s 5
ability to earn its allowed ROE and it would only affect its long run risk if it increases its cost 6
recovery risk. This in turn requires not that it fail to capture more new customers but that existing 7
residential customers drop off the system at an increasing rate and there does not seem to be any 8
evidence for this. 9
49 Normally one would expect single family housing to sue more energy for space heating.
76
The graphic below indicates that FEI continues to add customers. 1
2
Q. HOW DOES FEI COMPARE TO THE OTHER MAJOR CANADIAN LDCS? 3
A. In Table C-12 of the company’s application they provided the following comparison of 4
the cost competitiveness of natural gas versus electricity. 5
77
1
A similar chart was provided in 2012 in answer to an information request from the BCUC. The 2
data is consistent with that provided above for the lower Mainland. In 2012 natural gas had a 3
53% competitive advantage in BC and now for Vancouver it is 59%. The chart provides three 4
natural comparison groups: Alberta, Ontario and Quebec since each has at least one very large 5
major natural gas local distribution company (LDC). As a result I have normally used ATCO 6
Gas in Alberta, Gaz Metro in Quebec and EGDI and Union in Ontario as comparators for FEI. 7
The Regie regards Gaz Metro as above average risk since Quebec Hydro is a formidable 8
competitor as the above data demonstrates. Natural gas at the current low prices has only a 31% 9
competitive advantage over electricity in the home use market and in 2012 that was only 9%. So 10
consistent with the argument of FEI for themselves, the capital cost of switching to natural gas 11
has historically not justified the switch and electricity has been the fuel of choice in Quebec. As 12
a result, Gaz Metro has focussed on the new construction market and industrial load. To 13
compensate for the higher risk of Gaz Metro the Regie allows it a 0.25% to 0.35% premium to its 14
ROE over the low risk company for an 8.90% allowed ROE.50 In addition, the Regie allows a 15
50 By implication the Regie regards the benchmark as an 8.60% ROE. This has not changed for some time as the ROE has been settled through negotiation.
78
38.5% common equity and an additional 7.5% in deemed preferred shares. Like the Regie I have 1
traditionally viewed Gaz Metro as the highest risk of the very large gas LDCs. Notably the Regie 2
allows Hydro Quebec Distribution 35% common equity and an 8.2% ROE. 3
Traditionally the other slightly higher risk gas LDC has been Union Gas which has been allowed 4
a 0.15% premium to the ROE allowed EGDI. This was also justified by its acquisition of Centra 5
Gas Ontario when it assumed a large industrial load, in addition to its existing industrial load. 6
However, in recent negotiations and settlements this premium seems to have disappeared and 7
Union Gas currently is allowed an 8.93% ROE on 36% common equity. 8
The final two gas LDCs I regard as the lowest risk in Canada. ATCO Gas has primarily 9
residential load since most of ATCO’s industrial customers are served directly off ATCO 10
Pipelines. It is currently allowed an ROE of 8.3% on 38% common equity, but that includes a 11
1% premium for the impact of the “financial crisis” and its aftermath. There is a hearing 12
scheduled by the AUC for the Summer and this premium might finally be removed, but the 13
AUC’s underlying judgement is that ATCO Gas has a 37% common equity ratio. Generally 14
EGDI is regarded as Canada’s premier utility since it serves the Greater Toronto area with 5.5 15
million people and has mainly residential load. It had a business risk hearing in 2013 when its 16
common equity ratio was confirmed at 36%, but has not had a litigated ROE hearing since 2004. 17
It currently has an allowed ROE of 9.19%.51 18
Relative to these gas distributors I continue to rate FEI as in the same risk bucket, or perhaps 19
slightly riskier, than EGDI and ATCO Gas and lower risk than either Union Gas or Gaz Metro. 20
This seemed to also be the judgment of Ms. Kathy McShane, FEI’s long-time expert witness in 21
cross examination in 2012. However, to a great extent these risk differences are tiny relative to 22
the fact that they all distribute a very competitive fuel source that is an abundant supply. 23
One final comparator is Nova Scotia Power Inc. (NSPI), which is the provincial electricity 24
provider in Nova Scotia. NSPI has a monopoly on the supply of electricity in Nova Scotia as the 25
former crown corporation and has recently been allowed a comprehensive fuel adjustment 26
51 OEB letter dated October 15, 2015.
79
mechanism that acts much like the RSAM for FEI in removing commodity price exposure. The 1
“competition” from Heritage Gas in Nova Scotia is also limited and not material. NSPI 2
negotiated a settlement on September 14, 2012 where it maintained its 37.5% common equity 3
ratio in return for a 9.0% ROE. 4
Based strictly on these comparators, and not my own judgment, FEI’s common equity ratio 5
should be about 37% in a range from 36% (Union and EGDI) to 38.5% (Gaz Metro). However, I 6
continue to recommend the same 35% common equity I have traditionally recommended for 7
both gas and electricity distribution companies. My normal ranking places distribution 8
companies as slightly higher risk than transmission companies for which I recommend a 30% 9
common equity ratio. The Regie allows Hydro Quebec Distribution 35% common equity and 10
Hydro Quebec Transmission 30% common equity. 11
Q. DOES FEI FACE HIGHER POLITICAL RISK? 12
A. In its own assessment of its business risk FEI indicated that almost all categories used by 13
the BCUC to assess its business risk indicated that there were no changes since 2012. In total 19 14
categories indicated the same risk, one lower and four higher. Two of these I can dismiss since 15
lower commodity prices indicate lower risk, but the volatility of them higher risk. In my 16
judgment there is limited potential for increasing commodity prices, given the previously 17
mentioned assessment of supply, and volatility is only important if it means that natural gas 18
prices can materially affect the competitive margin against electricity, which I doubt.52 I also 19
dismiss aboriginal rights since this is a governmental and not an FEI factor and I judge that the 20
BCUC will continue to protect FEI. Finally, security of supply has not changed for FEI and I will 21
deal with amalgamation issues separately. 22
This only leaves higher political risk. There is no question that the reduction in greenhouse gas 23
(GHG) emissions is something that may affect everyone in BC. However, many parties accept 24
that natural gas is a solution to the GHG problem, not the cause. This is true in transportation, for 25
52 FEI confers with the BCUC every three months to make sure that the cost of gas charged to ratepayers is sufficient to cover its cost and associated pipeline service and storage. The volatility of natural gas prices only affects FEI, if at all, to the extent that a very high price may encourage some to use alternative fuels. This is highly unlikely at the moment and any impact is not material.
80
example, as well as in electricity generation. In terms of the distribution system itself, the 1
Canadian Gas Association provided the following graphic. 2
3
Despite the continued expansion of the natural gas distribution system, green-house gas 4
emissions from the system (GHG) are declining. What this means is that the distribution system 5
itself is getting cleaner. 6
I do not see the provincial government requiring that residential users modify their natural gas 7
space heating in order to replace them with more expensive electricity. It would be easier and 8
more effective to use a cap and trade system, where the province buys credits from the reduction 9
of GHG emissions elsewhere to meet its targets. Further, I do not see slower growth prospects as 10
a risk factor, since it does not affect the value or the risk of assets in place. If FEI does see the 11
provincial government requiring the removal or modification of natural gas heating systems, the 12
correct response is a depreciation study to depreciate the assets more quickly and reduce any 13
stranded asset risk. In this way t keeps FEI whole in terms of its risk exposure. 14
In my judgement FEI is marginally lower risk than in 2012 and I do not see any basis for the 15
conclusion that its risk has increased. 16
81
VI FINANCING AND CONCLUSIONS 1
Q. DO YOUR RECOMMENDATIONS SATISFY THE FAIR RETURN STANDARD? 2
A. Yes. My recommendations are based on the fair return standard. The most basic thing to 3
remember is that my recommendation for an ROE of 7.5% is approximately 3.5% over the 4
company’s long term borrowing cost, 5.50% over the current long Canada bond yield and 5
approximately 2.0% over the preferred share yield should FEI decide to issue preferred shares.53 6
In my judgement it can attract capital at my recommended financial parameters. 7
In terms of its “financial metrics” I am extremely reluctant to benchmark my recommendations 8
against guidelines issued by the rating agencies, such as Moodys, for two reasons. First, DBRS 9
has long maintained the exact same “A” rating on FEI and its predecessor companies through 10
periods when it had a 33% common equity ratio, a 35% common equity ratio, a 40% common 11
equity ratio and most recently a 38.5% common equity ratio. The following graphic from DBRS 12
(March 18, 2014) indicates its credit history back to 2001. 13
14
53 Fortis, FEI’s ultimate parent issues preferred shares on a regular basis and there would be no problem “mirroring” those costs down to FEI as is done with the subsidiaries of Canadian Utilities such as ATCO Gas, having FEI issue preferred shares or the BCUC simply deeming them as is done by the Regie for Gaz Metro.
82
Also as DBRS indicates FEI has had an interest coverage ratio below 2.0 on many occasions in 1
the past and it is a variant of this ratio that FEI focusses on in its evidence. The fact is that DBRS 2
has given FEI an A rating for the last 15 years even during periods when it had 33% common 3
equity and an interest coverage ratio below 2.0. The idea that it would change this rating if the 4
BCUC were to give it the same common equity ratio as ATCO Gas, EGDI and Union, or the 5
33% it used to have, is not reasonable. 6
I would also note that DBRS has maintained the A rating despite the introduction of performance 7
based regulation (PBR) that previously FEI and its witnesses have introduced as a risk factor. In 8
fact, Moody’s explicitly stated in their July 15, 2014 credit report. 9
“TRANSITION TO PBR EXPECTED TO HAVE MINIMAL CREDIT IMPLICATIONS 10 FortisBC utilities have submitted detailed PBR proposals for both FEI and FBC for the 11 period 2014-2018. We have assumed that it does not represent a material change in risk 12 and that the company continues to earn its allowed ROE. The proposed PBR plan is 13 broadly similar to the previous PBR plan and would have both an annual and mid-term 14 review. FEI's proposal would set controllable O&M and non-CPCN (CPCN includes 15 large capital projects that currently require regulatory pre-approval) capex by formula 16 with substantial costs remaining as pass through items. The proposal contains a 17 proposed symmetrical earnings sharing mechanism on up to 200bps and is subject to 18 meeting service quality targets. Performance above or below the allowed ROE by more 19 than 200bps would trigger an automatic review of the PBR plan. There are no proposed 20 changes to key deferral accounts. While we don't expect it, a key risk to the proposal is 21 that the regulator adopts very difficult efficiency targets within the formula. The PBR 22 plan does not propose to modify support for CPCN capex. A final decision on the PBR is 23 expected from the regulator in Q3 or Q4 2014. FEI previously” 24
The second reason is that the guidelines are heavily based on the degree of regulatory protection, 25
where 50% of the weight applied by Moodys is explicitly for this and not the financial metrics. 26
Consequently, the metrics are not the most important issue. As Moody’s also stated in its July 27
15, 2-14 credit note 28
“FEI's investment grade rating is driven by its credit supportive regulatory environment 29 and its monopoly position. Rates are typically set using a cost of service framework and a 30 forward test year that enables the company to recover its costs and earn an allowed 31 return established by the regulator, resulting in stable cash flow. The company has a 32 track record of passing through its commodity costs in rates and has no direct exposure 33 to commodity price risk and limited volume risk. To the extent that these and many other 34 costs differ from forecast values, deferral or true up mechanisms limit exposure to 35
83
forecast error. As a result the company has a long track record of earning the return on 1 equity (ROE) established by the regulator.” 2
And most importantly 3
“The company benefits from a monopoly position. We believe that its customers, who are 4 primarily residential, continue to have the capacity and willingness to pay their bills.” 5
I therefore judge regulation by targeting what the rating agencies say they want in one part of 6
their analysis to the exclusion of the most important issues in other parts as being incorrect. I 7
would also add that recently Moody’s has changed its view of US regulatory protection. In a 8
request for comment on September 23, 2013 Moody’s stated: 9
“Our revised view that the regulatory environment and timely recovery of costs is in most 10 cases more reliable than we previously believed is expected to lead to a one notch 11 upgrade of most regulated utilities in the US, with some exceptions. This evolving view is 12 independent of the proposed changes in the methodology that are highlighted in the 13 Summary section that follows, and would have taken place even if the 2009 methodology 14 were to remain in place without modification.” 15
The comment basically says that since the regulatory protection afforded US utilities seems to 16
have increased it will pretty much apply a one notch upgrade to their credit ratings independent 17
of their credit metrics. To the extent that Moody’s has traditionally viewed Canadian regulation 18
as more protective than that in the US, this comment indicates that we can take the US guidelines 19
and add a notch for Canadian utilities, rather than just reading off from the guidelines. This 20
vitiates their usefulness in indicating what guidelines produce what ratings in a mechanical 21
manner. 54 22
I also agree with the BCUC statement in its 2013 decision when it stated 23
“The Commission Panel is supportive of maintaining an “A” category credit rating but 24 only to the extent that it can be maintained without going beyond what is required by the 25 Fair Return Standard.” 26
The fair return standard trumps the relevance of a particular bond rating since there are other 27
ways of ensuring market access without giving the shareholder an unfair rate of return. 28
54 Note this is not to say that Moody’s views Canadian and US utilities as equivalent in risk. This is a request for comment and reflects Moody’s “evolving” view.
84
Q. ARE SOME FINANCIAL METRICS IMPORTANT? 1
A. Yes. FEI is to some extent constrained by the prospectus of the actual entity that issues its 2
medium term notes (MTNs) (unsecured long term debt). To issue MTNs a 2.0X new issue test 3
based on its interest coverage ratio (ICR) has to be met. FEI analyses in depth (evidence pages 4
27-31) the implications of this restriction as it refinances its purchase money mortgages and the 5
implications of the BCUC reducing its financial parameters to 37% common equity and an 6
8.25% allowed ROE. Note that these financial parameters are similar to what other premium 7
LDCs are allowed in Alberta and Quebec so their choice does not seem to be arbitrary, but 8
instead an acknowledgement that these parameters are reasonable. 9
To understand how the interest coverage restriction (ICR) works I reproduce below the filing 10
with the Ontario Securities Commission that FortisBC Energy made on November 6, 2015, 11
12
Note that for the ICR the consolidated interest was $140 million for 2014 and consolidated 13
earnings $317 million. So the ICR is simply 317/140 or 2.26. This had decreased marginally to 14
2.21 by September 2015. Note that in the ICR neither the interest charge nor the tax charge are 15
the actual FEI expenses. FEI’s total interest in 2014 as in the application was $189 million which 16
includes $136 in interest on the long term debt (MTNs), as well as $55 million in interest on 17
short term debt. However, whereas consolidated earnings includes the total interest expense, the 18
ICR only looks at funded or long term interest. Similarly the ICR includes the notional tax 19
85
charge, since it specifically excludes the impact of FEI’s “tax loss utilisation plan” or TLUP. In 1
2014 the use of TULP effectively lowered FEI’s tax expense from $35 million to $23 million but 2
it is the notional $35 million tax expense that is used to generate the consolidated earnings in the 3
ICR test.55 4
There are some timing differences in the numbers used in the ICR as there are some smoothing 5
options, but the net result is that FEI has considerable financing flexibility and is not currently 6
constrained by the ICR in issuing MTNs. For example, in August 2015 FEI negotiated a 7
syndicated $700 million credit facility of which approximately half is currently unused. On April 8
8, 2015 FEI issued $150 million of 30 year MTNs at 3.375% using the proceeds to repay short 9
term debt. At the time the 30 year long Canada bond (Cansim V39056) was yielding 2.03% for a 10
1.35% credit premium.56 In September 2015 it repaid $75million of the PMMs with short term 11
notes. 12
Q. SHOULDN’T THE COMMON EQUITY RATIO BE INCREASED DUE TO THE 13
AMALGAMATION? 14
A. No. The BCUC approved the amalgamation effective January 1, 2015 and the financial 15
parameters of the amalgamated FEI continued at FEI’s allowed 38.5% common equity and 16
8.75% ROE. The effects of the amalgamation should have been reflected in the interest coverage 17
ratio filing of 2.21 on November 15, 2015, so the amalgamation per se has not had a material 18
effect on FEI’s market access. Further both DBRS and Moody’s reviewed FEI after the 19
amalgamation and did not change their credit rating. FEI did note in its November 6, 2015 20
management and discussion analysis (MD&A) filed with the OSC that the impact of the 21
amalgamated financial parameters was $3 million for the first nine months of 2015 and in answer 22
to BC (AMPC) IR # 1.1 that it would have been about $4 million in 2014. I do not regard such 23
numbers as material in affecting FEI’s financial flexibility for three reasons: 24
55 TULP is a tax loss sharing scheme between controlled affiliates, where one sells preferred shares to another and receives non-taxable dividend income whereas the other buys debt where the interest is taxable. I assume the TULP has been approved by the BCUC and only actual interest expense is included in the revenue requirement. 56 The MD&A does not indicate the price received. RBC reports the spread on FEI seasoned debt on that day at 1.44%.
86
1) FEI has significant financial flexibility under its current financial parameters 1 2) Under PBR I expect FEI to continue to over-earn its allowed ROE, so that its ICR 2 is improved beyond the values implied by the regulated ROE and common equity ratios. 3 3) Other financial strategies such as TULP may generate savings that also enhance 4 the ICR. 5
FEI concludes this analysis by stating (Company evidence page 31) 6
“Under certain circumstances, and in a period of higher debt capital requirements, debt 7 issuance may be constrained. An increase in equity and/or allowed ROE would be 8 reasonable to support issuance capacity going forward.” 9
I disagree with this statement. First, the conclusion is based on highly speculative assumptions 10
and if such circumstances do obtain FEI can come to the BCUC for relief. Second, the analysis 11
does not take into account how the ICR actually works, where on page 25 of the company 12
evidence it specifically only includes $96.31 million in funded debt interest, whereas the ICR 13
uses all interest in calculating consolidated earnings.57 Third and most importantly, the equity 14
holder should not be awarded a higher common equity ratio or ROE due to issue problems in the 15
bond market. Put another way FEI should not be allowed more than the fair return standard due 16
to bond market problems. If any such problems exist, and I don’t think they currently do, they 17
can be addressed with a short term solution, such as issuing term preferred shares. However, 18
since other Canadian gas LDCs also have this interest coverage restriction, as well as similar 19
financial metrics that satisfy the fair return standard, I am not persuaded that FEI’s shareholder 20
should get such special treatment. 21
Q. WHAT ARE YOUR ROE AND COMMON EQUITY RATIO 22
RECOMMENDATIONS? 23
A. I judge a benchmark ROE for 2016 to be 7.78%, but consistent with my judgment in 24
2012 would not change the allowed ROE until the forecast long Canada bond yield exceeds 25
3.80%. I therefore recommend a continuation of both my ROE and ROE adjustment mechanism 26
recommendations from 2012 for 2016. In terms of the common equity ratio my personal 27
recommendation is for 35%, which is the same that the Regie allows a distributor in Quebec and 28
57 I don’t know why there is a different approach between FEI’s statutory filings and its current evidence.
87
would point out that a comparison with other Canadian LDCs would indicate a 37% common 1
equity ratio, which is marginally higher than that allowed EGDI and Union Gas in Ontario. 2
Q. DOES THIS CONCLUDE YOUR TESTIMONY? 3
A. Yes, at this time.4
1
Schedule 1
Macro data (* indicates year to date) Unemployment Real CPI T Bill Canada FX Rate Average
Rate Growth Inflation Yield Yield US$ ROE1987 8.81 4.04 4.42 8.17 9.93 0.75 11.191988 7.77 4.74 3.94 9.42 10.23 0.81 12.971989 7.58 2.38 5.06 12.02 9.92 0.84 11.791990 8.16 0.13 4.81 12.81 10.81 0.86 7.481991 10.32 -2.12 5.61 8.83 9.81 0.87 3.531992 11.24 0.85 1.45 6.51 8.77 0.83 1.561993 11.42 2.61 1.90 4.93 7.88 0.78 3.691994 10.43 4.55 0.12 5.42 8.58 0.73 6.571995 9.54 2.74 2.22 6.98 8.35 0.73 9.551996 9.73 1.68 1.48 4.31 7.54 0.73 10.291997 9.16 4.25 1.69 3.21 6.47 0.72 10.861998 8.35 4.14 1.00 4.74 5.45 0.67 8.831999 7.58 5.00 1.75 4.70 5.68 0.67 9.822000 6.85 5.12 2.69 5.48 5.92 0.67 10.922001 7.23 1.69 2.52 3.85 5.79 0.67 7.412002 7.66 2.80 2.25 2.57 5.67 0.65 5.692003 7.61 1.93 2.80 2.87 5.29 0.72 9.652004 7.18 3.14 1.85 2.27 5.08 0.77 11.622005 6.77 3.16 2.21 2.71 4.41 0.83 12.702006 6.32 2.62 2.00 4.02 4.29 0.88 13.952007 6.03 2.01 2.14 4.17 4.32 0.94 12.872008 6.15 1.18 2.37 2.62 4.06 0.94 9.442009 8.23 -2.71 0.30 0.40 3.85 0.88 8.062010 7.99 3.37 1.78 0.50 3.71 0.97 9.972011 7.46 2.96 2.39 0.94 3.22 1.01 9.922012 7.29 1.92 2.03 0.96 2.35 1.00 10.682013 7.07 2.00 0.94 0.98 2.71 0.97 9.822014 6.90 2.44 1.91 0.91 2.65 0.91 10.372015 6.90 2.03* 1.11* 0.54 2.06 0.78
Cansim V13682111 v62305752 v41690973 V122484 V122501 V37426 V634672/V634628
Schedule 2
CANADA BOND YIELDS
Overnight money market rates 0.52
Benchmark bonds
Canada 91 day Treasury Bill yield 0.45
Canada Six month Treasury Bills 0.44
Canada One year Treasury Bills 0.44
Canada Two year 0.41
Canada Three year 0.43
Canada Five year 0.66
Canada Seven year 0.96
Canada Ten year 1.32
Canada Long term (30 year) 2.07
Canada Real return bonds 0.63
Marketable Bond Average yields
Canada 1-3 year 0.42
Canada 3-5 year 0.55
Canada 5-10 1.06
Canada Over tens 1.95
Source: Bank of Canada’s web site at http://bankofcanada.ca/en/securities.htm, for January 8, 2016.
Schedule 3
Fama-French Application
Schedule 4
Investment and Speculative TSX Returns back to 1987
ROE Spec TSX1987 11.19 -5.31 5.881988 12.97 -1.89 11.081989 11.79 9.58 21.371990 7.48 -22.28 -14.801991 3.53 8.48 12.021992 1.56 -2.99 -1.431993 3.69 28.86 32.551994 6.57 -6.75 -0.181995 9.55 4.98 14.531996 10.29 18.06 28.351997 10.86 4.12 14.981998 8.83 -10.42 -1.581999 9.82 21.90 31.712000 10.92 -3.51 7.412001 7.41 -19.98 -12.572002 5.69 -18.12 -12.442003 9.65 17.08 26.722004 11.62 2.86 14.482005 12.70 11.43 24.132006 13.95 3.31 17.262007 12.87 -3.04 9.832008 9.44 -42.44 -33.002009 8.06 26.99 35.052010 9.97 7.64 17.612011 9.92 -18.63 -8.712012 10.68 -3.49 7.192013 9.82 3.18 13.002014 10.37 0.18 10.55
Schedule 5
Jack Bogle’s Investment and Speculative Returns in the US back to 1900
Schedule 6
1
105 St George Street, Toronto, Ontario M5S 3E6 E-Mail [email protected]; (416)-978-6311
TEACHING AND Main interest is teaching domestic and international corporate RESEARCH finance. Research interests centre on the cost of capital, empirical INTERESTS corporate finance and capital market theory. ACADEMIC D.B.A., Indiana University, (Finance). BACKGROUND M.B.A., Indiana University, (Finance).
M.A., Indiana University, (Economics). B. Sc.(Econ), London School of Economics.
AWARDS & MBA Second Year Instructor of the Year Award, 1996, 1998 (joint) HONOURS & 2000
Best paper in corporate finance, 1999 SFA meetings ASAC Distinguished Professor Address 1990, Director Financial Management Association 1988-90, English Speaking Union Fellow, Fulbright, Elected to Beta Gamma Sigma, First class honours B.Sc.(Econ) CBV (Chartered Business Valuator), National Post Leader in Management Education Award 2003
ACADEMIC CIT Chair in Structured Finance (1999-), Professor of Finance, EMPLOYMENT Rotman School of Management, University of Toronto (1987-
Present). Visiting Professor Nankai University (China) 1989, the Czech Management Centre (1998). Visiting scholar London School of Economics (1985).
Professor Laurence Booth CIT Chair in Structured Finance
Appendix A
2
TEACHING Graduate (MBA) courses on The Economics of Enterprise, the EXPERIENCE Economic Environment of Business, Business Finance, Corporate
Financing, International Financial Management, Mergers & Acquisitions, Financial Management, Capital Markets & Corporate Financing (EMBA), Applied Asset Management, Financial Theory of the Firm (Ph.D), Capital Markets Workshop (Ph.D). Undergraduate courses (B.Comm) in International Business, Business Finance and Introduction to Financial Markets. Executive courses (2-5 days) on Money and Foreign Exchange Markets, Business Valuation, Financial Strategy, Equity Markets, Capital Market Innovations, Mergers & Acquisitions and Finance for Non-Financial Managers.
JOURNAL "Stochastic Demand, Output and the Cost of Capital: A ARTICLES Clarification," Journal of Finance, 35 (June 1980),
"Capital Structure, Taxes and the Cost of Capital," Quarterly Review of Economics and Business, 20 (Autumn 1980),
"Stock Valuation Models Under Inflation," Financial Analysts Journal, (May-June 1981),
"Market Structure, Uncertainty and the Cost of Equity Capital," Journal of Banking and Finance, (May 1981),
"Capital Budgeting Frameworks for the Multinational Corporation," Journal of International Business Studies, (Fall 1982),
"Hedging and Foreign Exchange Exposure," Management International Review, (Spring 1982),
"Correct Procedures for Discounting Risky Cash Outflows," Journal of Financial and Quantitative Analysis, (June 1982),
"Total Price Uncertainty and the Theory of the Competitive Firm," Economica, (May 1983),
"Portfolio Composition and the CAPM," Journal of Economics and Business, (June 1983),
"On the Negative Risk Premium for Risk Adjusted Discount Rates," Journal of Business Finance and Accounting, (Spring 1983),
3
"On the Unanimity Literature and the Security Market Line Criterion," Journal of Business Finance and Accounting (Winter 1983), "Empirical Tests of the Monetary Approach to Exchange Rate Determination," (with R. Vander Kr,aats) Journal of International Money and Finance, (December 1983),
"The Ex-Dividend Day Behaviour of Canadian Stock Prices: Tax Changes and Clientele Effects," Journal of Finance, (June 1984) (with D. J. Johnstone), "On the Relationship Between Time State Preference and Capital Asset Pricing Models," Financial Review (May 1984),
"Bid-Ask Spreads in the Market for Foreign Exchange," Journal of International Money and Finance (August 1984),
"An Economic Analysis of Hedging and The Canadian Accounting Treatment of Revenue Hedges," Canadian Journal of Administrative Sciences, (June 1987),
"The Dividend Tax Credit and Canadian Ownership Objectives," Canadian Journal of Economics (May 1987),
"A Note on the Demand for Labour and the Phillips curve Phenomenon," Journal of Economics and Business (July 1987) (with W. Y. Lee and J. Finkelstein),
"Adjustment to Production Uncertainty and the Theory of the Firm: A Note," Economic Inquiry (1988),
"The Deregulation of Canada's Financial System," Banking and Finance Law Review, (Jan 1989),
"Stock Returns and the Dollar," Canadian Investment Review, (Spring 1990), (With W. Rotenberg),
"Taxes, Funds Positioning and the Cost of Capital,' in R. Aggarwal (ed) Advances in Financial Planning and Forecasting, JAI Press, 1990,
"Assessing Foreign Exchange Exposure: Theory and Application Using Canadian Firms," Journal of International Financial Management and Accounting (Spring 1990) (With W. Rotenberg),
4
"Research in Finance at Canadian Administration and Management Faculties," Canadian Journal of Administrative Studies, (With F. Heath), (December 1990),
"The Influence of Production Technology on Risk and the Cost of Capital," Journal of Financial and Quantitative Analysis (March 1991),
"Evidence on Corporate Preferences For Foreign Currency Accounting Standards", Journal of International Financial Management and Accounting, (with W. Rotenberg) (Summer 1991)), "Peoples Acquisition of Zale: An application of Valuation Principles," in Canadian Investment Banking Review, (R. Rupert, Editor), McGraw-Hill Ryerson, 1992,
"The Cost of Equity Capital of a Non-Traded Unique Entity," Canadian Journal of Administrative Sciences, (June 1993), "Lessons From Canadian Capital Market History," Canadian Investment Review (Spring 1995),
"Making Capital Budgeting Decisions in Multinational Corporations," Managerial Finance 22-1, (1996), "Great Lakes Forest Products" Accounting Education 5 (Winter 1996) (with Professor W. Rotenberg),
"On the Nature of Foreign Exchange Exposure" Journal of Multinational Financial Management" (Spring 1996),
“The Importance of Market to Book Ratios in Regulation,” Quarterly Bulletin, National Regulatory Research Institute, Winter 1997, “A New Model for Estimating Risk Premiums (Along with Evidence of their Decline)” Journal of Applied Corporate Finance, (Spring 1998), “The Case Against Foreign Bonds in Canadian Fixed Income Portfolios,” Canadian Investment Review, (Spring 1998),
“The CAPM, Equity Risk Premiums and the Privately Held Business,” Journal of Business Valuation (1999),
5
“Estimating the Equity Risk Premium and Equity Costs: New Ways of Looking at Old Data,” Journal of Applied Corporate Finance, (Spring 1999),
“Time to Pass the Old Maid,” Canadian Investment Review, (Spring 1999), “Risk and Return in Capital Markets,” Canadian Treasurer 16-2, March 2000,
“What Drives Shareholder value,” Canadian Treasurer 16-3, June 2000.
“Capital Structures in Developing Countries,” Journal of Finance 61-1 (March 2001, pp 87-130) (with V. Aivazian, V. Maxsimovic and A. Demirgic Kunt), (abstracted in the CFA Digest-31 -3 August 2001) “Discounting Expected Values with Parameter Uncertainty,” Journal of Corporate Finance 9- 2 (Spring 2003, pp 505-519) "Equity Risk Premiums in the US and Canada," Canadian Investment Review (Spring 2001), "Financial Planning with Risk," Canadian Journal of Financial Planning (December 2001), "How to Find Value when None Exists: Pitfalls in Using APV and FTE," Journal of Applied Corporate Finance (Spring 2002), "Do Emerging Market Firms Follow Different Dividend Policies than Firms in the US: Evidence From Firms in 8 Emerging Markets," Journal of Financial Research 26-3, (September 2003, pp 371-387) (Abstracted in CFA Digest 34-1, Feb 2004) (With V. Aivazian and S. Cleary), "Dividend Policy and the Organisation of Capital Markets, Journal of Multinational Financial Management, 13-2 (April 2003, pp 101-121 (With V. Aivazian and S. Cleary), “What to do with Executive Stock Options,” Canadian Investment Review 16-2, (Summer 2003, pp 12-18), “Formulating Retirement Targets and the Impact of Time Horizon on Asset Allocation,” Financial Services Review 13-1, (Spring 2004),
6
“Dividend Policy and the Role of the Contracting Environment,” FSR Forum, December 2005, pp 13-22,
“Dividend Smoothing and Debt Ratings,” Journal of Financial and Quantitative Analysis, with V. Aivazian and S. Cleary (June 2006), “Capital Cash Flows, APV and Valuation,” European Financial Management, (Spring 2007). “What Drives Provincial-Canada Yield Spreads” Canadian Journal of Economics, (Summer 2007) with Walid Hejazi and George Georgoplous. “Blast from the Past,” Canadian Investment Review, Summer 2007. “Cash Flow Volatility, Financial Slack and Investment Decisions,” China Finance Review 2-1, (January 2008) with Sean Cleary, pp 63-87. “Collateral Damage,” 2008, Canadian Investment Review 21-4, pp 10-17. “Capital Market Developments in the Post 1987 Period: A Canadian Perspective,” Review of Finance and Accounting 8-2 (with Sean Cleary), 2009. “The Secret of Canadian Banking: Common Sense?” World Economics, September 2009 “Information Asymmetry, Dividend Status and SEO Announcement Day Returns" (with Bin Chang), Journal of Financial Research, (Spring 2011) “Target Date Funds: Good News and Bad News,” (with Bin Chang) Journal of Risk, Spring 2011, pp 1-28. “The Influence of Productivity Growth on Equity Market Performance, Journal of Wealth Management (with Bin Chang, Walid Hejazi and Pauline Shum) (Summer 2011) “Asset Allocation and the Performance of American Target Date Funds,” Rotman International Journal of Pension Management, (With Bin Chang) Fall 2011.
7
“Import Competition and Disappearing Dividends, “Journal of International Business Studies 44-1 (January 2013) with Jun Zhou and Bin Chang. “Debt Rating Initiations: Natural Evolution or Opportunistic Behavior? Journal of Modern Accounting and Auditing (with Lynettte Purda and Sean Cleary), Fall 2013. “Increase in Cash Holdings: Pervasive or Sector Specific” Frontiers in Finance and Economics: 10-2 (October 2013) (with Jun Zhou). “The Choice between Non-Callable and Callable Bonds” Journal of Financial Research, XXXVII-4 (Winter 2014), with Frank Skinner and Dimitrious Gounopoulos. “Dividend Policy and Market Power: A Risk based Approach, Managerial Finance 41-2 with Jun Zhou. “Which Analysts Lead?” Journal of Accounting, Auditing and Finance, (forthcoming) with Jun Zhou and Bin Chang. “Financial Constraints, R&D Investment and the Value of Cash Holdings,” Quarterly Journal of Finance 5-2, May 2015 with Jun Zhou and Chrystos Ntantamis) .
NON-JOURNAL “Financial Considerations for Providing Incentives for Private PUBLICATIONS Industry and their Implications for Employment Level and
Stability,” (with M. J. Gordon) Technical study #2, Labour Market Development Task Force, Ministry of Supply and Services Canada, 1982.
“A Comparison of the Car Insurance Industry in Ontario with The Public Monopolies in Saskatchewan, Manitoba and British Columbia,” 122 pp, in C. Osbourne (ed) Report of the Inquiry into Motor Vehicle Accident Compensation in Ontario, Ontario 1988.
"Securities Market Regulation: Institutional Ownership and Diversification;” “TSE Listing Proposals for Junior Companies,” and “Discount Brokerage and the Entry of Financial Institutions." Reports submitted to the Ontario Securities Commission, July 1982, June 1983 and December 1983.
8
"Bank Profitability, Is It Excessive? (With M. Jensen and S. Klein), Report to the House Standing Committee on Finance, Trade and Economic Affairs, May 1982.
"Survey of Foreign Bank Affiliates," Chapter 8 in Small Business Financing and Non-Bank Financial Intermediaries, Facsym 1981.
"A Methodological Error in the Application of the Capital Asset Pricing Model" Proceedings ASAC, (May 1981).
International Business, (with A. Rugman and D. Lecraw), McGraw Hill, 1985.
"Hedging Foreign Exchange Exposure," in Rugman (ed), International Business in Canada: Strategies for Management, Prentice-Hall, 1988.
"Section 1650 of the CICA Handbook: Interpreting Foreign Results Under a Flexible Accounting Standard," (With W. Rotenberg), CGA Communications, 1989.
"Liability Management in the Public Sector," Report for Ministry of Treasury and Economics, May 1990 (with P. Halpern,)
"The Tax Deductibility of Interest and Hostile Takeovers," John Deutsch Institute, May 1990.
"Regulation of Transmission and Distribution Activities of Ontario Hydro,” in R. Daniels, Editor, Ontario Hydro at the Millenium: Has Monopoly’s Moment Passed? McGill-Queens University Press Fall 1996 (with P. Halpern).
“Competition and Profitability in the Financial Services Industry in Canada,” in J. Mintz & J. Pesando (editors) Putting Consumers First C.D Howe Institute, 1996.
"What Drives Shareholder Value,” Financial Intelligence IV-6, Federated Press , Spring 1999. "Canada's Competitiveness over the last 20 years," Rotman Management, Spring/Summer 1999. “A Walk through Risk and Return,” Advisor’s Guide to Financial Research, 1999.
9
“Picking the Right Stocks,” Advisor's Guide to International Financial Research, 2000.
“The CAPM, Equity Risk Premiums and the Privately Held Business,” reprinted in W. Albo et al, Purchase and Sale of Privately Held Businesses, CA Press, Toronto, Ontario, 2000
"Investments, Alternative Investments and Bubbles," in Advisor's Guide to New Investment Opportunities, 2001. "The Increasing Complexity of Bank Brands," Rotman Management, Spring/Summer 2001. "Asset Allocation in the Long Run," Advisor's Guide to Risk Management, 2002. "The Competitiveness of Corporate Canada," Financial Post, July 2002. "Corporate Responsibility," Rotman Management, Spring/Summer 2003. "The MBA International Finance course: a course whose time has come and gone, in A. Rugman (editor) Research in Global Strategic Management, JAI press, June 2003. "The fundamentals of finance all business professionals should know and remember," Inside the Minds: Textbook Finance, Aspatore Books, June 2003. “Anticipating the Big Boom,” Rotman, the magazine of the Rotman School of Management, Fall 2005. “Asset Allocation: The Long View,” in H. Evensky (Editor) Retirement Income Redesigned: Master Plans for Distribution, Bloomberg Press, Princeton, 2006. “Loyalty in Finance,” Rotman, the magazine of the Rotman School of Management, Fall 2006. Introduction to Corporate Finance, John Wiley and Sons, 2007 (with Sean Cleary) “Blast from the Past”, Canadian Investment Review, Summer 2007.
10
“Saving Capitalism from the Capitalists,” Rotman, the magazine of the Rotman School of Management, Summer 2008. “Sub-prime market meltdown and learning form the past,” in The Finance Crisis and Rescue, University of Toronto press, November 2008. “An Overview of Value Based Management,” in Advanced Corporate Finance, C. Krishnamurti and S.R. Vishwanath Prentice Hall International, 2009. Introduction to Corporate Finance, John Wiley and Sons, (2nd edition) 2010 (with Sean Cleary) “The Cost of Equity Capital and Fair Rate of Return on Equity (ROE) for a Canadian Utility” Canadian Regulation, Gordon Kaiser Editor) 2011. Corporate Finance, John Wiley and Sons 2013 (with Sean Cleary and Pamela Drake Petersen). Introduction to Corporate Finance, John Wiley and Sons, (3rd edition) 2013 (with Sean Cleary) “Estimating Discount Rates” School of Public Policy Research Paper Volume 8, Issue 18, University of Calgary, April 2015.
TESTIMONY Expert financial witness in a variety of civil cases plus (some with the late Professor M.K. Berkowitz) rate hearings for Altalink partners, ATCO Gas (South), ATCO Pipelines (South), ATCO Electric, Bell Canada, Consumers Gas, Teleglobe, Maritime T&T, Island Tel, BC Tel, AGT, Newfoundland Tel, Union Gas, Intragaz, Gazifere, Gaz Metropolitain, Ontario Hydro, Centra Gas Ontario, NB Tel, Northwestel, Pacific Northern Gas, BC Gas, West Kootenay Power, TransCanada Pipelines, TransEnergie, Trans Mountain Pipelines, IPL, Westcoast Energy, Nova Gas Transmission, Foothills Pipeline, TQ&M, ANG, Centra Gas Manitoba, Maritime Link, Newfoundland Power, AOSPL, Alberta Bottlers Depot Association.
Ph.D George Pink, A Dominance Analysis of Canadian Mutual Funds, SUPERVISOR 1988,
Greg Lypny, An Experimental Study of Managerial Pay and Firm Hedging Decisions, 1989,
11
Frank Skinner, Credit Quality Adjustments and Corporate Bond
Yields, 1990, Rui Pan, Probability Analysis of Option Strategies, 1994, Peter Klein, Three Essays on the Capital Gains Lock-in Effect, 1996,
Guy Bellemare, Capital Market Segmentation: US -Canada, 1996,
Kevin Lam, The Pricing of Audit Services, 1997,
Sean Cleary, The Relation Between Firm Investment and Financial Slack, 1998, Xinlei Zhao, Three Essays on Financial Markets, 2002, Lynnette Purda, Elements of Corporate Debt Policy, 2003, Themis Pantos, Investment Distortions in the Presence of a Sovereign Debt Overhang, 2003. Zhao Sun, PEG ratios and Stock Returns, 2004.
Zhaoxia Xu, Dynamic Adjustment of Financial Policy, 2007. Bin Chang, Information in Financial Markets, 2008. Ambrus Kesckes, Three Essays on IPOs, 2008 (Co-chair with Jan Mahrt-Smith) Jun Zhou, Industry Influences on Corporate Financial Policy, 2010.
CASE A fair rate of return for Bell Canada, 1986. WRITING Canvend 1984, A & B, 1988.
Peoples Jewellers, 1988. Great Lakes Forest Products A, 1989. Inco, 1989. Peoples acquisition of Zale, 1990. American Can Canada, 1990. Great Lakes Forest Products A, 1993 (with W. Rotenberg) BC Telephone, 1993 103 Kirsten Avenue, 1994 Great Lakes Forest Products B, 1994 (with W. Rotenberg) Mill Creek Jewellery, 1995 (With E. Kirzner)
12
Chapters, draft 2002. Second Cup Valuation, draft 2002.
SERVICE Executive Committee: 1980-2, 1989-90, 1993-4, 2001-3, 2009-10
Finance Area Co-ordinator 1987-91, 1994-2008 External Advisory Board, Health Administration Faculty, 1985-92. Editorial Board Activities:
Journal of Economics & Business 1982-87. Finance Section Editor, Canadian Journal of Administrative Sciences 1993-2005. Journal of Multinational Financial Management 1989-. Journal of International Business Studies 1992- Associate Editor, Multinational Finance Journal, 1995-2013 Journal of Applied Finance 2003-2007
Director at large Multinational Finance Society 1998- Co-Chair 1991 Northern Finance Association meetings. Chair 1998 Northern Finance Association meetings Chair 2008 MFS annual meetings. President Multinational Finance Society, 2010-11 Programme Committee member FMA meetings, October 1993. Programme Committee member SFA meetings November 2002. Programme Committee member, MFS meetings 2002-14 Programme Committee Member, Global Finance Conference, 2006. Programme Committee Member, European Financial Management 2006-2014 Programme Committee member, NFA meetings 2008-2013 Investment Committee, Trinity College, U of T. Pension Committee, Governing Council University of Toronto, 2011-14 Special committee on the Supplementary Retirement Arrangement (SRA) University of Toronto, 2011 Frequent media commentator. Vice-President Mid-West Finance Association 2012-14. President Elect Mid-West Finance Association 2014-2015 President Mid-West Finance Association 2015-2016.
November 2015.
1
APPENDIX B 1
ESTIMATION OF THE MARKET RISK PREMIUM 2
Introduction 3
In this appendix I estimate the market risk premium by examining realised rates of return on 4
different broad classes of securities over long periods of time.1 The reason for doing this is that if 5
the underlying relationship generating these returns has remained reasonably constant then these 6
realised returns can be used as a forecast of the market's future requirements. The difference 7
between these returns is then commonly used as an estimate of the market risk premium. In 8
analyzing the actual data, however, we first need to be aware of some estimation problems and 9
the impact of changes that have occurred in the markets. 10
Different Risk Premium Estimation Procedures 11
Suppose an investor puts $1,000 into an investment. If the investment doubles, i.e., a 100% 12
return, to $2,000 and then halves, i.e., a -50% return, to $1,000, we can calculate two rates of 13
return. The arithmetic rate of return would be 25% i.e., the average of +100% and -50%. The 14
arithmetic rate of return is the average of the two per period rates of return. However, it would be 15
difficult to convince an investor, who after two years only has the same $1,000 that he started 16
with, that he has earned an average rate of return of 25%. Quite obviously, the investor is no 17
better off at the end of the two periods than he was at the start! To counterbalance this 18
potentially misleading statistic, most mutual funds advertise geometric or compound rates of 19
return. This compound rate of return is often called the true rate of return. It is calculated as the 20
nth root of the terminal value divided by the initial value, minus one. In our case, there are two 21
periods, so that n=2 and the compound rate of return is calculated as (1/1)1/2 which is 1, 22
indicating a zero rate of return. This gives the common sense solution that if you started and 23
finished with $1,000, then your rate of return is zero. 24
Both the arithmetic and compound rates of return are normally calculated when evaluating 25
1 This appendix covers similar material to that covered in Laurence Booth "Equities Over Bonds: But By How Much?" Canadian Investment Review, Spring 1995 and "Equity Risk Premiums in the US and Canada," Canadian Investment Review (Spring 2001). The latter paper is available for download from Professor Booth’s web site http://www.rotman.utoronto.ca/~booth
2
investments. If we need the best estimate of next period's rate of return, this is the arithmetic 1
return. If we need the best estimate of the return over several periods, the arithmetic return 2
becomes less useful and more emphasis is placed on the compound return. If we want the best 3
estimate of the annual rate of return earned over a long period of time, this is the compound rate 4
of return, since this indicates the long run expected change in wealth. Moreover, if we ignore 5
intervening periods, then the arithmetic return over a very long period is the compound rate of 6
return, that is, the difference between the arithmetic and compound returns is essentially the 7
definition of the period over which the investment is held. 8
What causes the two rates of return to differ is the uncertainty in the per period arithmetic rates 9
of return. If the arithmetic rate of return is constant, then both rates of return are identical. 10
However, the more uncertain the arithmetic rate of return, the larger the discrepancy between the 11
two estimates. For instantaneous rates of return the following equation approximately describes 12
their relationship: 13
Compound rate of return = Arithmetic return – (var/2) 14
In the previous example, there is a large amount of uncertainty, that is, high variance (var), so 15
that the difference between the arithmetic return and the geometric return is very large. 16
Moreover, as we estimate over a longer and longer period, the estimated arithmetic rate of return 17
earned on an investment approaches that of the compound return. In estimating the market risk 18
premium, I believe that the correct time period for calculating arithmetic rates of return is a one-19
year holding period. The reason for this is primarily because most regulated firms are regulated 20
on the basis of annual rates of return and rates are almost always expressed as annual 21
percentages. 22
In addition to the arithmetic and compound rates of return I also estimate the rate of return by 23
means of an ordinary least squares regression model. This is a statistical technique that 24
estimates the annual rate of return by minimizing the deviations of the annual values around the 25
estimate. Ordinary least squares (OLS) is the standard technique for estimating economic models 26
and is commonly used for estimating other annual growth rates, such as the growth rate in 27
dividend growth models. 28
3
Market Risk Premium Estimates Going Forward and Backwards 1
In Schedule 1 I graph the market risk premium using Canadian data and these three estimation 2
techniques in two ways.2 In the top graph starting in 1924-1928 the realised market risk premium 3
is estimated using each of the three techniques and is then updated each year with the new data 4
so the second observation is for the period 1924-1929. In this way the graph captures the 5
“learning” that goes on from 1924. The instability in the 1920s is evident: as the estimates are 6
very high, due to the strong equity markets in the 1920’s, and then in the 1930s it declines 7
precipitously as a result of the great stock market crash. However, the market risk premium 8
stabilizes by the late 1950s, and then begins its long gradual decrease. Note that with over ninety 9
years of data, the impact of any one-year is now very small and the market risk premium is 10
"stuck" around 5.0%. However, it is apparent that the realised market risk premium has been 11
declining almost continuously since the mid 1960's as the importance of the prewar period gets 12
smaller and smaller and the impact of the post war bond market uncertainty increases. 13
An alternative to the above procedure is to work backwards, that is, start in the five-year period 14
2010-2014 and then go back in time, which is the lower graph in Schedule 1. In this way we 15
capture what current market participants have experienced. Note that whereas the previous graph 16
always includes the period 1924-1928, this graph always includes the most recent five year 17
period. In this case the last five years includes the post-crash period stock market volatility that 18
mimics in many ways what was observed in the 1930’s after the great stock market crash. 19
However, as we work back through time and add in progressively older data the influence of the 20
recent market volatility recedes and once we get back to the 1950's we finally get a market risk 21
premium about 4.0%. However, this graph illustrates why current market participants generally 22
assess the risk premium of equities over bonds as much lower than 5.0%, since this is what they 23
have experienced over the last 20-30 years. 24
In Schedule 2 is the earned risk premium (using arithmetic returns) for various holding periods. 25
If we look at the last row we have the earned risk premium for various start dates finishing in 26
2014, this is essentially a subset of the data graphed in Schedule 1. Note for example, that the 27
2 The graphs use data from the Canadian Institute of Actuaries, "Report on Canadian Economic Statistics" April 2011 updated with more recent data.
4
most recent 20-year period has an earned risk premium of 1.12%, as this period goes back 1
successively by adding an extra ten years of data each time the earned risk premium drops and 2
then increases until it stabilizes at just below 5.0%. 3
The usefulness of the different holding periods in Schedule 2 is simply to note the variability in 4
the experienced risk premium that results from individuals choosing to base estimates on a subset 5
of the data. A high estimate could, for example, be generated by ending the time period in the 6
1980s or 1990’s by using stale data from old textbooks, since this was the period when interest 7
rates were at their peak and as result realised returns on bonds were much less than anticipated. 8
Equivalently a low market risk premium could be generated by emphasizing the most recent 9
period since 1989 when the very high returns from holding bonds during this declining interest 10
rate period often gives a negative market risk premium. 11
We can illustrate this problem simply by graphing the behaviour of interest rates which is done 12
in Schedule 3. Note for example, that there was very little interest rate variability in the 1930’s 13
but then starting in the 1950’s interest rates started to increase with inflation, thereby causing 14
losses to anyone holding long term bonds, since as interest rates go up the return to holding 15
bonds goes down. This process ended in the period 1981-1989, after which it has gone into 16
reverse until we reach the current period of very low interest rates when long Canada bonds 17
ended 2014 at just 2.22% (V122487). For 2015 the average long Canada bond yield until 18
November was 2.00% and actually lower than the average level for 1936 of 2.97% as massive 19
government intervention in the fixed income markets caused record low interest rates. 20
Changes in the Market Risk Premium 21
The fact that estimates of the market risk premium change over time indicates that some 22
adjustments are in order. In my judgement the riskiness of the equity market is relatively stable. 23
In fact, going back as far as 1871, there is substantial evidence that the real return on US equities 24
has been quite stable at under 9.0%.3 However, there is no support for the assumption that either 25
bond market risk or average bond market returns have been constant. As Schedule 3 shows, from 26
3 See Laurence Booth, “Estimating the Equity Risk Premium and Equity Costs: New Ways of Looking at Old Data”, Journal of Applied Corporate Finance, Spring 1999.
5
1924-1956, there was very little movement in nominal interest rates as monetary policy was 1
subordinate to fiscal policy. As a result, the standard deviation of annual bond market returns 2
was only 5.20%. In contrast from 1956-2014, monetary policy became progressively more 3
important and interest rates much more volatile. As a result, the standard deviation of the returns 4
from holding the long Canada bond increased substantially. Effectively bond market risk 5
dramatically increased while equity market risk was much the same, if not less. 6
This changing bond market risk is illustrated in Schedule 4 which graphs the equity market risk 7
divided by the bond market risk, where each is estimated as the standard deviation of returns 8
over the prior ten year period, so the series start with the first observation for the period 1924-9
1933. We can clearly see the dramatic decrease in equity relative to bond market risk starting in 10
the 1950s, where equities dropped from being six times riskier than long term Government of 11
Canada (GOC) bonds to their low point, prior to the Internet Bubble, of similar risk. Since then 12
the new relative stability in long Canada bond yields has caused equities to revert to being 13
significantly riskier than GOC bonds. 14
However, what is crucial for the investor is whether this risk is diversifiable, that is, is the bond 15
market beta or risk positive? In Schedule 5 I show that the Canadian bond market beta was very 16
large during the period from the mid-1980s until the early 2000’s. This was the period when 17
governments had severe financing problems and flooded the market with government debt. This 18
caused both the bond and equity markets to partly be moved by a common risk factor: interest 19
rates. This is why adding long Canada bonds to an equity portfolio during the 1990's did not 20
reduce portfolio risk to the extent that it did in the 1950's and more recently. However since the 21
Canadian government solved its structural budget problems in the 1990’s we have seen the bond 22
market beta revert to its more typical negative or insignificant relationship 23
In Schedule 6 are the results of a regression analysis of the real Canada bond yield against 24
various independent variables. The real Canada yield is defined as the nominal yield reported by 25
the Canadian Institute of Actuaries minus the average CPI rate of inflation, calculated as the 26
average of the current, past and forward year rates of inflation.4 The regression model explains a 27
4 Before 1991 there was no real return bond.
6
large amount of the variation in real Canada yields, and five variables are highly significant. 1
The two main independent variables capture bond market uncertainty (risk) and the endemic 2
problem of financing government expenditures (deficits). Risk is the standard deviation of the 3
return on the long Canada bond over the preceding ten years. In earlier periods when monetary 4
policy was not used, interest rates barely moved and the returns on long Canada bonds were very 5
stable. As a result the risk of investing in them was very low. The coefficient on the bond market 6
risk variable indicates that for every 1% increase in volatility, real Canada yields increased by 7
about 24 basis points. That is, a 5% increase in the standard deviation of bond market returns 8
before and after 1956 was associated with about a 1.25% increase in real Canada yields between 9
these two periods. 10
The deficit variable is the total amount of government “lending” (from all levels of government) 11
as a percentage of the gross domestic product. Statistics Canada reports this as lending but 12
usually it is negative, that is, deficits and government borrowing. As governments increasingly 13
ran deficits, this figure became a very large negative number. For 1992, the number was about -14
9.1%, a record peacetime high, indicating that government net borrowing was 9.1% of GDP and 15
was flooding the markets with Canada bonds. For 1997, this deficit turned into a surplus, which 16
increased every year until 2000 when the surplus hit almost 3.0% of GDP. The coefficient in the 17
model indicates that for every 1% increase in the aggregate government deficit, real Canada 18
yields have increased by about 25 basis points. That is, increased government borrowing by 19
competing for funds with other borrowers has driven up real interest rates. At the peak of the 20
government's financing problems in 1992 a 9% deficit was adding well over 2.0% to the real 21
Canada yield relative to what would have happened with a balanced budget. 22
When these two effects are added together we can explain the huge increase in real yields in the 23
early 1990s. In 1994, for example, when real yields were over 7%, the deficit added about 1.75% 24
and the bond market uncertainty about another 2.65% or in total close to 4.5 % to the real yield. 25
It is easy to see that with this dramatic increase in real yields there was very little "extra" risk for 26
low risk equities over bonds at this time. 27
The three “dummy” or indicator variables represent unique periods of intervention in the 28
7
financial markets. An indicator variable simply inserts a “1” for the years when this special 1
phenomenon was in effect. Dum1 is for the years from 1940-1951, which were the "war" years, 2
when interest rates were controlled in order to finance the war and recovery. The coefficient 3
indicates that government controls reduced real Canada yields by about 5.0% below where they 4
would otherwise have been. Similarly, Dum2 is for the years 1972-1980, which were the oil 5
crisis years, when huge amounts of "petrodollars" were recycled from the suddenly oil rich 6
OPEC countries back to western capital markets, where they essentially acted as a tax to depress 7
real yields. The sign on Dum2 indicates that, but for this recycling and the oil crisis, real yields 8
would have been about 3.5% higher. 9
Dum3 is for the recent period of quantitative easing since 2010, which indicates that the real 10
yield has recently been about 2.5% below where it would have been without the recent extreme 11
measures taken in the US, UK, Europe and Japan. These dummy variables are included due to 12
known periods of intervention that have prevented the “normal” application of financial 13
principles, that is, interest rates were determined largely by political factors. In previous evidence 14
I have only used the first two indicator variables and the explained variability in the real yield 15
was only “81%” and was dropping each year. The reason for this was that recently the 16
abnormally low yields could not be explained without recognising the impact of quantitative 17
easing in the major capital markets in the world. With the addition of an indicator variable for 18
the years 2011-2014 the adjusted R square jumps from 81% to 86% 19
In Schedule 7 is a graph of the real yield produced directly from the real return bond. 20
Unfortunately this data is not available for earlier periods, since these bonds did not exist. 21
However, we can see directly the huge decline in the real yield over the last ten years or so as 22
governments have got their budgets under control, uncertainty in the bond market has declined 23
and monetary policy has been loose. For the period 1991-2000 the real yield was 4.0-4.5%, 24
whereas in the after math of the financial crisis it has averaged less than 2.0% and is currently at 25
0.63%. 26
If we use the regression model in Schedule 6 the real yield should be about 3.9% or slightly 27
below where it was in 2000. The 2014 government deficit of 1.93% adds about 0.5% to the 28
intercept of 1.34% and the 2014 bond market volatility of 8.6% another 2.03%. At the 1.35% 29
8
inflation rate this implies a long Canada bond yield consistent with 2014 data of 5.25%. If 1
Canada were still insulated from the rest of the world, the deficit problems of Canadian 2
governments would have driven up Canadian bond yields. Instead, the dire shape of the rest of 3
the developed world has made Canada look good with about 1/3 of the government bond market 4
now owned for non-residents. Of course, capital markets have changed, but what is clear is that 5
current long term Government of Canada bond yields are well below what would normally be 6
regarded as “normal” or average and as long as we have loose monetary policy elsewhere in the 7
world this is likely to continue. 8
US Estimates
The Canadian data is one time series of equity and bond market returns and reflects unique 9
events that happened in Canada; looking at US data we can assess whether these estimates are 10
reasonable. Schedule 8 provides US estimates of the market risk premium along with the 11
comparable Canadian estimates for the period 1926-2014. 12
Based on annual holding periods the US realised equity risk premium is higher than the 13
Canadian equivalent. Given the "higher" quality of the US data as well as the volatility of the 14
estimates, many put greater faith in the US estimates, even for the Canadian market. This is also 15
frequently justified by the doubt expressed at the “higher risk”5 Canadian market having a lower 16
realized market risk premium, as well as the increasing integration between the two capital 17
markets, which “presumably” moves Canada closer to the US experience. 18
However, the difference between the US and Canadian AM market risk premium estimates of 19
1.45% (6.07%-4.62%) is split between a difference in the average equity return of 0.87% and a 20
difference in the average government bond return of 0.67%, that is approximately a 50:50 bond 21
market-equity market split. In explaining this, note that: 22
The difference between the equity market returns can partly be explained by the 23 historic efforts of Canadian governments to deliberately segment the Canadian equity 24
5 Note, however, that the standard deviation or variability of the S&P500 equity returns was 20.07% or 1.51% higher than that for the Canadian market. Over this whole period US equities were marginally more risky than Canadian equities.
9
market from that in the US6 as well as by the historically lower risk of the Canadian 1 market. 2
The difference in the returns on Canadian and US government bonds reflects the 3 pivotal role of the US government bond market in the world capital market as the US $ 4 has became the world’s reserve currency. 5
If we take the US equity market return as a better estimate of the “true” Canadian equity market 6
return, now that most of the protectionist policies in Canada have receded, this would increase 7
the Canadian market risk premium estimate to just over 5.0%. 8
Finally we have to bear in mind that Canada is in a favourable position as a AAA rated borrower 9
that has solved most of its structural deficit problems.7 The favourable finances have resulted in 10
low inflation and interest rates, and allowed the removal of the foreign property restriction on tax 11
preferred investments. We can see this in the graph of long term interest rates in Canada and the 12
US in Schedule 9. In the mid 1990s the nominal yield on long Canada bonds was routinely 13
higher than that on equivalent US bonds. However, this started to change as the Government of 14
Canada move into a surplus position and since the mid 2000’s long Canada bonds have usually 15
had lower yields than US treasuries. As of September 2015 20 year US Treasuries were yielding 16
2.62% or 0.62% more than long Canada bonds. This is consistent with lower required rates of 17
return in Canada. It also means that the lower historic market risk premium in Canada estimated 18
over riskier Canadian GOC bonds may no longer reflect expected market risk premiums over the 19
currently lower Canadian GOC bond yields. As a result although my direct estimate of the 20
Canadian market risk premium is under 5.0% I judge a reasonable range to be 5.0-6.0%, since 21
this reflects the recent behaviour of interest rates in Canada and the removal of regulatory 22
protection in the Canadian equity market. 23
Reasonableness of the Estimates 24
The prior statistical work indicates that the Canadian market risk premium has been about 5.0% 25
while that for the US has been about 1.5% higher, but points out that the Canadian market risk 26
6 The dividend tax credit only applies to dividends from Canadian corporations; foreign withholding taxes apply to foreign source income, while portfolio restrictions have existed in tax-preferred plans. 7 We can see this in a Globe and Mail article (June 22, 2015) “investors buying Canadian debt at fastest pace on record”
10
premium has almost certainly increased. These estimates are consistent with the judgment of 1
professionals in the area of capital markets. Professor Fernandez8 and his co-authors survey 2
finance professionals around the world to find out what they actually use for the market risk 3
premium. A key result from his 2015 survey is reproduced bas follows: 4
5
This table indicates that the average US market risk premium was estimated at 5.5% with the 6
typical (median) slightly lower at 5.3%. In contrast, the average market risk premium in Canada 7
was reported at a slightly higher 5.9% with a median of 6.0%. The median Canadian response of 8
6.0% was amongst the highest of the developed countries, but across these countries there is an 9
obvious 5.0-6.0% grouping. 10
A feature of Fernandez’s recent surveys is that they also surveyed the use of the risk free rate in 11
estimating the required rate of return. Textbooks normally use a Treasury Bill yield, rather than 12
the long term bond yield used before regulatory boards. However, as reported below in the US 13
the average risk free rate was given as 2.4% and that in Canada 2.3%, in both cases these look 14
more like long term bond yields than Treasury Bill yields. 15
8 Market risk Premium and Risk-Free Rate Used for 41 countries in 2015,” IESE Business School, 2015.
11
1
Finally with the both the market risk premium and the risk-free rate the survey reports the overall 2
return on the market as below: 3
4
12
Again there is a remarkable commonality that the overall market return is in the 6.0-8.0% range 1
with that for the both the U.S and Canada having a median value of 8.0%. 2
Conclusions 3
Fernandez’s survey work supports my own judgement that the Canadian market risk premium is 4
between 5.0-6.0%, as well as indicating that the overall equity market return is about 8.0%. This 5
is also confirmed by professional opinion in Canada, where on October 19, 2012 TD Economics 6
came out with a report “An Economics Perspective on Canadian Long Term Financial Returns.”9 7
The following table captures the TD Economics analysis: 8
9
The TD analysis placed long run Canadian equity returns at 7.00%, the same as in the US and 10
internationally, whereas bond returns were forecast at 3.0% for the Dex universe bond index, that 11
is, including corporate as well as government bonds. The implication is for a long run market 12
risk premium of 4.00% of equities over bonds and slightly higher over government bonds. This 13
is an increase compared to a similar report in March 2011, where Canadian equity returns were 14
forecast at 7.5% and bond returns at 4.00%. TD Economics is predicting a return to a balanced 15
portfolio of 4.0-6.0%, which with 2% inflation implies a real return at a maximum of 4.0%. This 16
is the same sort of analysis that underlies most defined benefit pension plans. 17
9 http://www.td.com/economics/special/ca0311_long_run_returns.pdf
13
These particular return estimates are for long run (next decade) which would be closer to 1
geometric (compound) returns than arithmetic returns. An adjustment to convert them to 2
arithmetic returns would move the market risk premium closer to 6.0%. 3
To summarise: 4
My own direct estimate of the experienced Canadian market risk premium is about 4.6% 5 since 1926. 6
This historic market risk premium is probably low given the removal of barriers to capital 7 flows and the current very low level of Canadian bond yields. 8
I would judge the market risk premium to currently be in a range of 5.00-6.00%. This 9 estimate is supported by the survey results of Fernandez and the implicit judgment of TD 10 Bank.10 11
The overall market return from Fernandez survey is about 8.0%, which benchmarks the 12 return for low risk Canadian utilities.13
10 The latest issue of Credit Suisse’ “Global Investment Returns Yearbook 2015,” has the equity market risk premium over bonds from 1900-2014 at 3.6% for Canada; 4.5% for the US and 3.7% for the UK.
14
SCHEDULE 1 1
Risk Premium Estimates Forward from 1924
-10
-5
0
5
10
15
20
25
30
1928 1938 1948 1958 1968 1978 1988 1998 2008
AM GM OLS
Market Risk Premium Estimates Back From 2014
-4
-2
0
2
4
6
8
1930 1940 1950 1960 1970 1980 1990 2000 2010
AM GM OLS
15
SCHEDULE 2
Arithmetic Earned Risk Premiums for Different Holding Periods
Start dates on the horizontal and ending dates on the vertical. For example, an investor would have earned a 7.31% arithmetic risk premium investing from 1925-1984.
1925 1935 1945 1955 1965 1975 1985 19951944 4.501954 8.12 9.821964 8.64 9.94 12.771974 7.18 7.79 8.96 5.761984 7.31 7.83 8.72 6.51 4.671994 5.94 6.15 6.52 4.31 2.35 2.862004 5.23 5.30 5.48 3.50 1.83 1.99 -1.022014 4.87 4.89 4.98 3.25 1.86 0.51 -0.01 1.12
16
SCHEDULE 3
Interest Rates and Inflationto Ocotber 2015
-5
0
5
10
15
20
25
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
2014
Canadas CPI T Bills
17
SCHEDULE 4
RELATIVE UNCERTAINTYEquities versus Bonds
0
1
2
3
4
5
6
7
1934 1939 1944 1949 1954 1959 1964 1969 1974 1979 1984 1989 1994 1999 2004 2009 2014
18
SCHEDULE 5
Canadian Bond Beta
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
2014
19
SCHEDULE 6
FACTORS INFLUENCING THE REAL CANADA YIELD
Dependent variable: Long Canada (over 10) yield minus the average CPI inflation rate for the past, current and forward year. Independent variables: Coefficient T-Statistic Constant: 1.341 3.51
Risk: standard deviation of return on 0.24 5.19 long bond index for prior ten years. Deficit: aggregate government lending -0.26 -8.76 as a % of GDP. Dum1: dummy variable for years 1940-51 -5.31 -13.09 Dum2: dummy variable for years 1972-80 -3.63 - 9.01 Dum3: dummy variable for years 2010-2014 -2.53 - 4.85 Adjusted R2 of the regression 85.6% Data 1936-2014
20
SCHEDULE 7
Real Bond Yield
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
1/1/
1997
7/1/
1997
1/1/
1998
7/1/
1998
1/1/
1999
7/1/
1999
1/1/
2000
7/1/
2000
1/1/
2001
7/1/
2001
1/1/
2002
7/1/
2002
1/1/
2003
7/1/
2003
1/1/
2004
7/1/
2004
1/1/
2005
7/1/
2005
1/1/
2006
7/1/
2006
1/1/
2007
7/1/
2007
1/1/
2008
7/1/
2008
1/1/
2009
7/1/
2009
1/1/
2010
7/1/
2010
1/1/
2011
7/1/
2011
1/1/
2012
7/1/
2012
1/1/
2013
7/1/
2013
1/1/
2014
7/1/
2014
1/1/
2015
7/1/
2015
1/1/
2016
21
SCHEDULE 8
Annual Rate of Return Estimates 1926-2014
U.S. CANADA
S&P
Equities
Long US
Treasury
Excess
Return
TSE Equities Long
Canadas
Excess
Return
AM
12.07 6.00 6.07 11.20 6.57 4.62
GM
10.62 5.85 4.77 10.04 6.52 3.52
OLS
10.96 5.37 5.59 10.30 6.07 4.24
Volatility1
20.07 9.70 18.56 8.98
22
SCHEDULE 9
Canadian and US Interest Rates
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
Jan-
94
Jan-
96
Jan-
98
Jan-
00
Jan-
02
Jan-
04
Jan-
06
Jan-
08
Jan-
10
Jan-
12
Jan-
14
US Canada
1
APPENDIX C
RELATIVE RISK ASSESSMENT FOR A BENCHMARK UTILITY
Introduction 1
In risk premium models the relative risk coefficient adjusts the overall market risk premium up 2
or down depending on whether the individual security (company) is more or less risky than the 3
market as a whole. More risky stocks have a relative risk coefficient greater than 1.0 and less 4
risky stocks a relative risk coefficient less than 1.0. All risk premium models have this same risk 5
assessment relative to the market, whether they are the capital asset pricing model (CAPM)1 6
where the only source of risk is the market risk, or models that introduce other sources of risk. 7
However, even within a two factor model, where the risk free rate is often regarded as risky due 8
to interest rate risk,2 or the Fama-French three factor model3 where size and the market to book 9
ratio (in their model termed the book to market ratio) are additional sources of risk, the 10
coefficient on the market is still the main measure of risk. Estrada,4 for example, shows that for 11
the DOW 30 US stocks the simple CAPM expected return at 9.70% is only 0.20% more than that 12
estimated using the three factor Fama-French Model and that the market risk premium is much 13
larger than either the size or book to market premiums. 14
With the CAPM the relative risk assessment is the expected covariance between the security’s 15
return and that on the market scaled by the variance of the return on the market. This is called the 16
security’s beta coefficient (β) and measures the contribution of the security to the risk of a 17
diversified portfolio. We normally estimate actual historic beta estimates by a simple ordinary 18
least squares (OLS) regression of the security’s return on that of the market. In any OLS 19
regression the intercept is called alpha and the slope coefficient is called beta, which is why these 20
terms are used pervasively in finance. However, estimating beta coefficients entails the exact 21
1 William Sharpe, “Capital asset prices: a theory of market equilibrium under conditions of risk,” Journal of Finance 19, 1964.
2 Fisher Black, “capital market equilibrium with restricted borrowing”, Journal of Business, July 1972.
3 Eugene Fama and Ken French, “The cross section of expected stocks returns,” Journal of Finance 59, 1992.
4 “The three factor model a practitioners guide,” Journal of Applied Corporate Finance, Spring 2011.
2
same problem as that involved in estimating the market risk premium, since both use the actual 1
or historic returns. This is, the estimate is very sensitive to what happened during the estimation 2
period. To overcome this problem in estimating the market risk premium we go back over very 3
long periods of time. For estimating beta coefficients we generally can’t do this, since the risk of 4
a firm or industry changes much more than the overall risk of the market. Instead, we tend to use 5
estimates from similar firms and industries as well as more judgment in understanding the 6
economic and financial factors underlying the beta estimates. In this way we can get a better 7
understanding of the expected beta coefficient, which is what is required. 8
Historic Beta Estimates for Canadian utilities 9
In 2002 the Toronto Stock Exchange outsourced its market indexes to Standard and Poors (S&P) 10
and changed the composition of our sub-indexes. These changes roughly coincided with the loss 11
of many traditional Canadian utilities. It was also controversial in transferring Enbridge and 12
TransCanada from pipelines, where they were regarded as similar to utilities, into energy 13
services. 14
Regardless of these changes the great advantage of the sub-indexes is that they include more 15
companies than possible with individual company estimates since companies are constantly 16
being reorganised as business strategy changes. This is particularly important due to the fact that 17
a large number of Canadian regulated firms, like Consumers Gas, Maritime Electric, Bell 18
Canada, Union Gas etc., have disappeared through corporate reorganisation. Although this 19
means that their individual company betas disappeared, it does not mean that their economic 20
impact has also disappeared. Consumers Gas now shows up as part of Enbridge, Terasen Gas as 21
Fortis etc., so their economic impact continues to show up in the sub index betas. However, there 22
is a disadvantage, which is that these are not simple averages but market value weighted 23
averages, since this is the way that stock market indexes are normally calculated. As a result 24
large market value companies have a disproportionate impact on the indexes. 25
It is important to remember that betas are simply a statistical estimate of the extent to which a 26
stock moves with the general market over a particular period of time. By convention, betas are 27
estimated over a five-year period. This means that if a critical event happens during the 28
3
estimation period, then the beta estimate will pick it up. However, once the event “passes out” of 1
the five-year estimation window, the impact of the event will disappear from the beta estimate. 2
In Schedule 1 is a graph of rolling betas on the Canadian utility sub index since 1988. Betas are 3
normally estimated over the prior five years of data so the first observation is from January 1988 4
until December 1992 and then each month as a new return is available the five year estimation 5
window moves forward a year. This process is repeated using two estimation techniques; the first 6
Beta 1 is the simple beta against the Canadian market index, whereas the second Beta 2 also 7
includes the impact of interest rate changes by adding the monthly return on the long Canada 8
bond. However, to all intents and purposes the beta estimates are almost the same, but it does 9
allow an estimate of the sensitivity of utility shares to interest rates, which I discuss later, and 10
refer to as “gamma.” 11
Using this procedure I can pick up the impact of unique events. For example, the utility betas 12
were both in a range of 0.40-0.60 until 1997. The betas then dropped to negative values during 13
2001-2004 before reverting to more “normal” levels. Did this mean that utility shares had no 14
risk during this period? The answer is no, since a special event, the behaviour of Nortel and the 15
Internet bubble, drove the estimates. During the late 1990s, the technology and internet boom 16
were driving North American markets and as the prices of Nortel and JDS Uniphase5 increased, 17
so did the TSE Composite, since at their peak they were about 35% of the index. When this 18
boom turned into a crash and Nortel declined from $1,240 to bankruptcy, Nortel took the 19
Canadian market down with it. 20
It is important to understand that betas measure the risk of a security against the risk of a 21
diversified portfolio, in this case the TSX Composite. Utility betas were pulled down as Nortel 22
and the tech boom affected the Canadian market but not utility shares. As the effect of the 23
internet bubble and crash passed thought the estimation window utility betas reverted to a more 24
normal pattern. By 2008 the beta estimates covering the period 2004-2008 were largely devoid 25
of the effects of the internet bubble. The message was that during this period utility shares added 26
very little risk to a diversified portfolio since that portfolio was dominated by the effect of 27
5 JDS Uniphase resulted from a merger of the Canadian fibre optic company JDS Fitel in 1999.
4
internet stocks. However, without the internet bubble effect utility shares add a normal amount 1
of risk to an investor holding a diversified portfolio. 2
Finally, utilities are clearly interest sensitive stocks as the consistent positive gamma coefficients 3
indicate. It is also clear that this interest rate sensitivity exhibits a negative correlation (-0.44) 4
with the beta estimates, that is, beta coefficients tend to fall as gamma coefficients increase. This 5
is because interest rates tend to increase during good times as the stock market booms and then 6
fall in recessions. This interest rate sensitivity reduces the exposure of utility investors to the 7
market during recessions when interest rates tend to fall. For this reason utility shares are 8
generally regarded as defensive shares. 9
This statistical result echoes the comment of former RBC utility analyst Maureen Howe who 10
commented that Canadian utilities are6 11
“like convertible bonds. When interest rates are low, as they currently are, the companies 12 trade on their bond value and are supported by tax-efficient dividend yields. When the 10-13 year GOC yield rises above 6%-6.5%, the Canadian companies trade on the basis of their 14 underlying earnings and P/E.” 15
Maureen Howe’s observation is confirmed by the relative performance of the PE multiples for 16
the TSX versus the Utilities as indicated in the following graph provided in the 2012 BCUC 17
hearing. The graph indicates that whereas the PE multiple of the TSX was then weaker than 18
before 2005, the very low interest rates supported the valuation of dividend rich utilities, so that 19
their PE ratios had increased. This observation is consistent with Maureen Howe’s observation 20
that with low interest rates utilities trade on their “bond or fixed income” value in line with the 21
observation that their cost of equity capital has declined. 22
6 October, 3 2001 RBC Morning Comment.
5
1
In Schedule 2 are the results of two multiple regression estimates of utility risk. The first panel 2
has the estimates for the entire period from 1988 where the utility beta is 0.26 and the gamma 3
0.43. However as noted previously this period reflects the Internet bubble and crash. In the 4
second panel are the estimates for the last five year period ending in 2014. For this period the 5
beta estimate is 0.45 closer to traditional levels and the gamma 0.44. 6
We can see the same effects in the individual beta estimates, where the average utility betas are 7
graphed in Schedule 3 and the individual values since 1996 are in Schedule 4.The average is 8
both with and without TransAlta, since TransAlta is not strictly a rate of return regulated utility 9
anymore.7 Again we see the Internet Bubble effect and the trend of the betas back towards their 10
normal level, in this case the average has stabilised around 0.30. Professional judgement would 11
indicate that it is unreasonable to use statistical estimates without recognising the underlying 12
events that generated them. This means that some judgment is called for when using the beta as a 13
forward estimate of risk. 14
It is my judgement, and consistent with the data in Schedule 1, the interest sensitivity of the 15
utilities has caused them to trade based on their defensive or income characteristics for the last 16
several years. This is evident from the fact that their betas vary inversely with their interest 17
7 Much of its output is still sold under power purchase contracts.
6
sensitivity. As interest rates increase back to normal levels, I would expect their betas to increase 1
as they trade less on their bond values and more as regular equities. I would therefore expect 2
some tendency for their betas to revert back to their long run average level: for the market as a 3
whole this is 1.0, but for regulated firms this is about 0.45-0.55.8 4
US utility stocks as a comparison 5
I have started looking at the relative risk of a sample of low risk US utilities. The US utilities 6
originally represented the intersection of two samples used previously by Ms. McShane and Dr. 7
Vilbert both of whom have testified before Canadian boards on behalf of utilities. The 8
intersection of their two “sets” might then be regarded as a smaller and unambiguously purer set 9
of low risk US utilities, rather than the bigger more diversified holding companies that are in the 10
S&P500 index. However, two of the six utilities that resulted from this exercise have now been 11
taken over by other utility holding companies, which forced me to widen my search. In 12
Appendix 1 to this Appendix is a list of the 17 US companies included in Google’s US gas 13
industry along with a description of their activities. It is clear that many of these companies are 14
mid-stream pipeline and storage companies that are not heavily rate of return regulated 15
consequently I excluded them and arrived at the following list: 16
Original set of companies 17
WGL Holdings 18 New Jersey Resources 19 Northwest Natural Gas Company 20 Piedmont Natural Gas Company 21 Vectren Corporation 22
Added Companies 23
South West Gas Corporation 24 The Laclede Group 25 ATMOS Energy Corporation 26
This gives a sample of 8 gas companies. 27
8 This is also accepted in the literature. Gombola and Kahl, “Time series properties of utility betas,” Financial Management, 1990, come to the same conclusion.
7
Schedule 6 provides a graph of their average beta estimates based on data back to 1971 with the 1
estimated betas in Schedule 7. The betas are estimated in the same way as for the Canadian betas 2
from monthly holding period returns over a five year time period updated monthly. 3
The estimates from these US utilities behave in a similar manner as for the Canadian utility 4
holding companies. This is clear from the observation that they also exhibit an “internet bubble” 5
effect, although not quite as severe as for the Canadian utility holding companies. However, the 6
recent average level of the betas from these companies is higher than for the Canadian. This is 7
clear from the graph in Schedule 7 that graphs the average for both groups across the same time 8
period since 1988. Noticeably since the passing of the internet bubble (the period 2003-2007) 9
the average beta for the sample of Canadian firms has been 0.29 versus 0.39 for the US firms; a 10
beta difference of 0.10. 11
Adjusted betas 12
Utility witnesses frequently adjust utility betas not toward their grand mean of around 0.50, but 13
the overall market average of 1.0. As low risk businesses this inevitably increases their “adjusted 14
beta.” Such a process is justified by the seminal work of Marshall Blume9 who showed that if 15
there is measurement error, when we estimate a very low beta the chances are the “true” beta is 16
underestimated and vice versa. For the whole universe of stocks he recommended that we adjust 17
betas by taking 2/3 of the estimated beta and adding 0.33, which essentially means weighting 18
them 1/3 with the market mean of 1.0 and 2/3 with the actual beta. This procedure means that 19
low betas are increased and high betas are reduced. However, low estimates for utilities do not 20
mean they are under-estimated, since utility betas are perennially low due to their low risk. 21
Instead as Gombola and Kahl demonstrated utility betas are better mechanically adjusted by 22
weighting with their grand mean. If I were to do this I would get an adjusted beta as follows: 23
Adjusted beta = 0.67 * 0.45 +0.33 * 0.5 = 0.47 for the utility sub index 24
Adjusted beta = 0.67 * 0.3 + 0.33 * 0.5 = 0.37 for the individual large companies 25
9 Marshall Blume, Betas and their regression tendencies, Journal of Finance June 1975.
8
However, rather than a mechanical weighting I generally prefer to use judgment constrained by 1
the actual historic evidence on betas. 2
Canadian “utilities” are generally not inter-listed in the US and apart from the very largest, like 3
TransCanada and Enbridge, trade on the pink sheets if at all, so they mainly trade on the TSX. 4
On November 18, 2015, I recorded basic data for the following Canadian utility holding 5
companies. In particular the following captures their beta estimates as reported by RBC, Yahoo 6
and Google as well as my own estimates up until December 2014.10 7
Canadian Regulated BetasCAP $B RBC Yahoo Booth Google
TransCanada TRP 30.6 0.47 0.96 0.28 0.27Enbridge ENB 43.1 0.21 0.6 0.11 0.11Canadian Utilities CU 9.2 0.09 0.44 0.2 -0.02Emera EMA 6.2 0.01 0.06 0.32 0.07Fortis FTS 10.57 0.11 0.05 0.26 0.08Valener VNR 0.7 0.34 0.13 0.27 0.19Veresen VSN 3.2 0.76 1.67 0.34 0.58Average 14.80 0.28 0.56 0.25 0.18Median 9.20 0.21 0.44 0.27 0.11 8
The average and median beta estimates by the Royal Bank of Canada were 0.28 and 0.21 9
respectively very similar to my own estimates (Booth) of 0.25 and 0.27. Google’s estimates seem 10
to be slightly lower than mine, whereas Yahoo’s are higher. However, there is no sign of the 11
mechanical Blume adjustment favoured by utility witnesses, if there were then no beta would be 12
less than 0.33 which is what you would get with the Blume adjustment and an actual beta of 0. 13
The key insight, however, is that “publicly available” betas, as distinct from my privately 14
estimated betas, all indicate the low risk nature of Canadian utilities. 15
Similarly the following table gives the betas for the US utilities. Again the average beta is 0.53 16
according to RBC and 0.60 for my estimates ending in 2014, 0.74 for Yahoo and 0.50 for Google 17
and the medians are 0.52, 0.60, 0.68 and 0.49 respectively. None of the beta estimates seem 18
10 I use conventional estimation procedures but the data end in December 2014. It is not clear how the public services estimate their betas in terms of time period and estimation window.
9
consistent with any type of Blume adjustment methodology, but some are clearly affected by 1
recent takeovers and speculation.11 2
US Regulated BetasMKT Cap RBC Yahoo Booth Google
5.5Vectren VVC 3.38 0.50 0.94 0.53 0.46WGL WGL 2.89 0.54 0.78 0.71 0.55Piedmont PNY 4.57 0.84 1.30 0.63 0.83Northwest NWN 1.27 0.29 0.46 0.57 0.31New Jersey Resources NJR 2.56 0.56 0.67 0.62 0.51Laclede LG 2.45 0.39 0.48 0.45 0.36ATMOS ATO 6.19 0.45 0.61 0.57 0.41South West Gas SWX 2.6 0.63 0.69 0.73 0.58
Average 3.49 0.53 0.74 0.60 0.50Median 2.89 0.52 0.68 0.60 0.49 3
One important observation is that in comparing the Canadian versus the US samples of utilities 4
the US firms are quite small with average market capitalisation (total equity market value) of 5
US$3.49 billion versus the average for the Canadian companies of $14.8 billion. Even after we 6
adjust for the exchange rate and the impact of outliers, the typical market capitalisation is still 7
much higher for the Canadian sample at $9.20 billion. Why this is important is that one of the 8
criticisms levelled against the CAPM is that beta adjusted small firms earn higher rates of return 9
than large firms, which some attribute to risk.12 So we might expect a higher risk level for these 10
US firms than for the Canadian sample due to their smaller average size. This is what we observe 11
where the average betas are consistently higher in the US than Canada whether we use my own 12
beat estimates or those of RBC, Yahoo or Google. 13
Of more importance is that the way these estimates are derived is consistent with conventional 14
practise. One of the biggest data providers in Canada is the Financial Post, where their Corporate 15
Analyzer data base includes ten year financial data for larger publicly listed Canadian 16
companies. Their definition of beta is: 17
11 Piedmont is subject to a takeover offer from Duke Energy
12 This is the Fama-French model discussed in Appendix G and in my main evidence.
10
Beta (Corporate Profiles) Beta factors are derived from a historical regression of percentage share price changes for the selected company on percentage changes in the TSE 300 price index. The unadjusted slope coefficient from this regression is the beta factor. Beta factors may be computed on a variety of weekly or monthly data. Betas shown in FP Analyzer are for 52 weeks, 36 months, 60 months and 120 months.
1
Again there is no discussion of “adjusting” betas using the Blume procedure, in fact they very 2
specifically state the “unadjusted slope coefficient” which is what the beta estimate is. However, 3
the FP does note that different time horizons can be used other than my conventional use of five 4
years of data. 5
One final comment is that there is often discussion of the “hollowing out” of the Canadian 6
capital market and the fact that the TSX is no longer a diversified index being heavily weighted 7
towards bank stocks and resource firms. This does not affect utility stocks, since they are 8
generally Canadian held and not inter-listed in the US, but as a check I looked at the risk of the 9
TSX utility index against the S&P500 index both hedged and unhedged against foreign exchange 10
rate movements. The results are in the graph is in Schedule 8. Can is the same beta estimate 11
graphed in Schedule 1, while US H is the beta of the Canadian utility index against the S&P500 12
index hedged against changes in the C$:US$ foreign exchange rate, while US UN takes that 13
change into account. The beta estimates against a more diversified US index are a little more 14
stable in that the high values during the 1990s are removed and the lows during the internet 15
bubble are not quite as low, but overall there is no dramatic change. In fact, since 2005 (2009 16
estimates start using data in 2005) the betas estimated against the US index have been 17
persistently lower than those obtained from using a Canadian market index. 18
Conclusion 19
What is clear from the above analysis is that the market recognises that Canadian utilities are 20
lower than average risk. This comes through after: 21
I recognise that the low values during the internet bubble period were an anomaly 22
I analyse the utility sub index versus individual Canadian firms 23
I check estimates against those publicly available and those from Canada’s largest bank 24
I use a US index as a market proxy to account for the “hollowing out” of Canada and 25 greater market integration 26
11
From this analysis I can see no reason that would cause me to deviate from my normal generic 1
risk assessment for a Canadian utility of a beat range of 0.45-0.55. In fact, the persistent low beta 2
estimates after recovery from the recession suggest that the risk assessment may now be 3
permanently lower. 4
12
SCHEDULE 1
Utility Index Beta Estimates
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
12
/31
/19
92
12
/31
/19
93
12
/31
/19
94
12
/31
/19
95
12
/31
/19
96
12
/31
/19
97
12
/31
/19
98
12
/31
/19
99
12
/31
/20
00
12
/31
/20
01
12
/31
/20
02
12
/31
/20
03
12
/31
/20
04
12
/31
/20
05
12
/31
/20
06
12
/31
/20
07
12
/31
/20
08
12
/31
/20
09
12
/31
/20
10
12
/31
/20
11
12
/31
/20
12
12
/31
/20
13
12
/31
/20
14
Gamma Beta 1 Beta 2
13
SCHEDULE 2
SUMMARY OUTPUT
Regression StatisticsMultiple R 0.43238257R Square 0.18695468Adjusted R Squ 0.18188898 Utilities Against Market and Bond ReturnsStandard Error 3.34947633 1988-2014Observations 324
ANOVAdf SS MS F Significance F
Regression 2 828.0955571 414.0478 36.90597 3.745E-15Residual 321 3601.296323 11.21899Total 323 4429.39188
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%ower 95.0%Upper 95.0%Intercept 0.20090919 0.189501627 1.060198 0.289852 -0.171913 0.573731 -0.17191 0.573731TSX 0.26262321 0.045513657 5.770207 1.86E-08 0.1730805 0.352166 0.17308 0.352166GOC 0.43213464 0.079294008 5.449777 1.01E-07 0.2761331 0.588136 0.276133 0.588136
SUMMARY OUTPUT
Regression StatisticsMultiple R 0.49139917R Square 0.24147314 Utilities Against Market and Bond ReturnsAdjusted R Squ 0.21485817 2010-2014Standard Error 2.46801123Observations 60
ANOVAdf SS MS F Significance F
Regression 2 110.5266457 55.26332 9.072829 0.0003795Residual 57 347.1915266 6.091079Total 59 457.7181723
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%ower 95.0%Upper 95.0%Intercept 0.07934724 0.348243018 0.22785 0.820578 -0.617997 0.776692 -0.618 0.776692TSX 0.45089255 0.115232023 3.91291 0.000246 0.2201444 0.681641 0.220144 0.681641GOC 0.44228936 0.1531105 2.888694 0.00546 0.1356909 0.748888 0.135691 0.748888
14
SCHEDULE 3
Utility Company Betas
-0.200
-0.100
0.000
0.100
0.200
0.300
0.400
0.500
0.600
Feb
-93
Feb
-94
Feb
-95
Feb
-96
Feb
-97
Feb
-98
Feb
-99
Feb
-00
Feb
-01
Feb
-02
Feb
-03
Feb
-04
Feb
-05
Feb
-06
Feb
-07
Feb
-08
Feb
-09
Feb
-10
Feb
-11
Feb
-12
Feb
-13
Feb
-14
Average Average minus TAU
15
SCHEDULE 4
CUL Emera Enbridge Fortis GMI PNG TRP VERESEN TAU
12/29/00 0.36 0.28 0.05 0.22 0.18 0.47 0.17 0.05
12/31/01 0.25 0.21 -0.13 0.13 0.10 0.47 -0.07 0.06
12/31/02 0.18 0.16 -0.20 0.13 0.07 0.47 -0.08 0.08
12/31/03 0.05 -0.05 -0.40 -0.05 0.02 0.36 -0.40 0.02 -0.06
12/31/04 0.03 -0.02 -0.32 0.03 0.16 0.48 -0.19 0.10 0.14
12/30/05 0.21 0.05 -0.18 0.23 0.19 0.51 -0.19 0.19 0.42
12/29/06 0.33 0.09 0.22 0.48 0.42 0.47 0.30 0.33 0.42
12/31/07 0.53 0.21 0.52 0.61 0.75 0.23 0.48 0.33 0.51
12/31/08 0.18 0.14 0.32 0.20 0.51 0.25 0.37 0.51 0.86
12/31/09 0.09 0.16 0.32 0.20 0.38 0.45 0.40 0.44 0.74
12/31/10 0.09 0.22 0.34 0.16 0.35 0.41 0.40 0.37 0.76
12/31/11 0.06 0.21 0.32 0.15 0.36 0.50 0.37 0.35 0.73
12/31/12 0.01 0.23 0.22 0.13 0.32 0.33 0.40 0.63
12/31/13 0.03 0.25 0.19 0.28 0.18 0.33 0.22 0.27
12/31/14 0.20 0.32 0.11 0.26 0.27 0.28 0.34 0.20
16
SCHEDULE 5
US Utility Betas
0.000
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
0.900
1.000
"197
5-11
-28"
"197
7-04
-29"
"197
8-09
-29"
"198
0-02
-29"
"198
1-07
-31"
"198
2-12
-31"
"198
4-05
-31"
"198
5-10
-31"
"198
7-03
-31"
"198
8-08
-31"
"199
0-01
-31"
"199
1-06
-28"
"199
2-11
-30"
"199
4-04
-29"
"199
5-09
-29"
"199
7-02
-28"
"199
8-07
-31"
"199
9-12
-31"
"200
1-05
-31"
"200
2-10
-31"
"200
4-03
-31"
"200
5-08
-31"
"200
7-01
-31"
"200
8-06
-30"
"200
9-11
-30"
"201
1-04
-29"
"201
2-09
-28"
"201
4-02
-28"
17
SCHEDULE 6
VVC WGL PNY NWN NJR LG ATO SWX Average
"1975-12-31" 0.81 0.42 0.61 0.73 0.92 0.70
"1976-12-31" 0.98 0.44 0.65 0.71 0.80 0.72
"1977-12-30" 0.99 0.48 0.65 0.73 0.67 0.71
"1978-12-29" 0.86 0.51 0.61 1.01 0.60 0.72 0.60 0.70
"1979-12-31" 0.98 0.90 0.73 1.19 1.03 0.66 0.80 0.90
"1980-12-31" 0.82 0.36 0.47 1.14 0.63 0.64 0.78 0.69
"1981-12-31" 0.60 0.33 0.39 1.08 0.64 0.76 0.66 0.64
"1982-12-31" 0.60 0.35 0.43 0.91 0.61 0.66 0.67 0.61
"1983-12-30" 0.53 0.30 0.42 0.93 0.56 0.69 0.58 0.57
"1984-12-31" 0.27 0.24 0.43 0.74 0.39 0.60 0.51 0.45
"1985-12-31" 0.13 0.40 0.42 0.51 0.40 0.36 0.46 0.38
"1986-12-31" 0.18 0.57 0.47 0.53 0.36 0.25 0.48 0.41
"1987-12-31" 0.19 0.46 0.61 0.52 0.61 0.37 0.72 0.50
"1988-12-30" 0.30 0.48 0.66 0.51 0.57 0.38 0.76 0.53
"1989-12-29" 0.47 0.50 0.59 0.58 0.63 0.40 0.51 0.71 0.55
"1990-12-31" 0.46 0.47 0.57 0.47 0.66 0.31 0.40 0.78 0.51
"1991-12-31" 0.37 0.36 0.54 0.45 0.60 0.26 0.34 0.71 0.45
"1992-12-31" 0.27 0.31 0.40 0.24 0.31 0.06 0.41 0.20 0.28
"1993-12-31" 0.23 0.30 0.33 0.22 0.32 0.03 0.34 0.14 0.24
"1994-12-30" 0.23 0.38 0.42 0.18 0.37 0.07 0.36 0.39 0.30
"1995-12-29" 0.23 0.45 0.36 0.21 0.34 0.11 0.49 0.00 0.27
"1996-12-31" 0.41 0.59 0.33 -0.04 0.48 0.15 0.80 0.18 0.36
"1997-12-31" 0.57 0.63 0.29 0.39 0.43 0.29 0.08 0.49 0.40
"1998-12-31" 0.30 0.47 0.51 0.52 0.46 0.34 0.14 0.96 0.46
"1999-12-31" 0.17 0.30 0.33 0.25 0.34 0.23 0.22 0.70 0.32
"2000-12-29" 0.22 0.26 0.17 0.12 0.36 0.21 -0.02 0.61 0.24
"2001-12-31" 0.17 0.21 0.17 0.08 0.24 0.05 -0.18 0.54 0.16
"2002-10-31" 0.22 0.21 0.20 0.01 0.16 0.04 -0.01 0.57 0.17
"2003-12-31" 0.36 0.14 -0.05 -0.21 0.03 0.01 -0.01 0.19 0.06
"2004-12-31" 0.40 0.21 0.10 -0.04 0.09 0.13 0.01 0.28 0.14
"2005-12-30" 0.34 0.22 0.26 0.06 -0.04 0.15 0.19 0.26 0.18
"2006-12-29" 0.52 0.27 0.34 0.14 0.03 0.49 0.45 0.23 0.31
"2007-12-31" 0.49 0.57 0.46 0.60 0.44 0.79 0.72 0.42 0.56
"2008-12-31" 0.27 0.26 0.10 0.36 0.14 0.10 0.50 0.63 0.30
"2009-12-31" 0.37 0.16 0.18 0.24 0.12 0.01 0.49 0.70 0.28
"2010-12-31" 0.43 0.27 0.27 0.35 0.22 0.08 0.51 0.73 0.36
"2011-12-30" 0.39 0.29 0.31 0.32 0.25 0.06 0.50 0.72 0.36
"2012-12-31" 0.35 0.22 0.30 0.26 0.23 0.07 0.44 0.69 0.32
"2013-12-31" 0.53 0.43 0.56 0.39 0.44 0.32 0.54 0.73 0.49
"2014-12-31" 0.53 0.71 0.63 0.57 0.62 0.45 0.57 0.73 0.60
18
SCHEDULE 7
US and Canadian Regulated Betas
-0.200
-0.100
0.000
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
Feb
-93
Feb
-94
Feb
-95
Feb
-96
Feb
-97
Feb
-98
Feb
-99
Feb
-00
Feb
-01
Feb
-02
Feb
-03
Feb
-04
Feb
-05
Feb
-06
Feb
-07
Feb
-08
Feb
-09
Feb
-10
Feb
-11
Feb
-12
Feb
-13
Feb
-14
Canada US
19
SCHEDULE 8
Canadian Betas with US Market Index
-0.400
-0.200
0.000
0.200
0.400
0.600
0.800
12/3
1/19
92
12/3
1/19
93
12/3
1/19
94
12/3
1/19
95
12/3
1/19
96
12/3
1/19
97
12/3
1/19
98
12/3
1/19
99
12/3
1/20
00
12/3
1/20
01
12/3
1/20
02
12/3
1/20
03
12/3
1/20
04
12/3
1/20
05
12/3
1/20
06
12/3
1/20
07
12/3
1/20
08
12/3
1/20
09
12/3
1/20
10
12/3
1/20
11
12/3
1/20
12
12/3
1/20
13
12/3
1/20
14
Can US H US UN
20
Appendix 1 US Gas Industry
Description Targa Resources Corp. (Traga), through its general and limited partner interests in Targa Resources Partners LP (the Partnership), is a provider of midstream natural gas and natural gas liquid (NGL) services in the United States. The Partnership is engaged in the business of gathering, compressing, treating, processing and selling natural gas and storing, fractionating, treating, transporting, terminaling and selling NGLs, NGL products, and gathering, storing and terminaling crude oil and refined petroleum products. The Company operates in two divisions Gathering and Processing and Logistics and Marketing. Gathering and Processing includes Field Gathering and Processing and Coastal Gathering and Processing. Logistics and Marketing includes Logistics Assets and Marketing and Distribution.
Description ONEOK, Inc. is the sole general partner of ONEOK Partners, L.P. (ONEOK Partners), a master limited partnership engaged in the gathering, processing, storage and transportation of natural gas in the United States. The Company operates through three segments: Natural Gas Gathering and Processing, Natural Gas Liquids, and Natural Gas Pipelines. The Natural Gas Gathering and Processing segment provides nondiscretionary services to producers, including gathering and processing of natural gas produced from crude oil and natural gas wells. The Natural Gas Liquids segment owns and operates facilities that gather, fractionate, treat and distribute natural gas liquids (NGLs), and store NGL products, primarily in Oklahoma, Kansas, Texas, New Mexico and the Rocky Mountain region. The Natural Gas Pipelines segment owns and operates regulated natural gas transmission pipelines and natural gas storage facilities
Description ONEOK Partners, L.P. is engaged in gathering, processing, storage and transportation of natural gas in the United States. In addition, the Company owns natural gas liquids (NGL) systems, connecting NGL supply in the Mid-Continent and Rocky Mountain regions. The Company operates in three segments: Natural Gas Gathering and Processing, Natural Gas Liquids and Natural Gas Pipelines. The Company’s Natural Gas Gathering and Processing segment provides nondiscretionary services to producers, which include gathering and processing of natural gas produced from crude oil and natural gas wells. The Company’s Natural Gas Liquids segment owns and operates facilities that gather, fractionate, treat and distribute NGLs and store NGL products in Oklahoma, Kansas, Texas, New Mexico and the Rocky
21
Mountain region. The Company’s Natural Gas Pipelines segment owns and operates regulated natural gas transmission pipelines and natural gas storage facilities
Description EnLink Midstream Partners, LP, formerly Crosstex Energy, L.P., focuses on providing midstream energy services, including gathering, transmission, processing, fractionation and marketing, to producers of natural gas, natural gas liquids (NGL), crude oil and condensate. The Company also provides crude oil, condensate and brine disposal services to producers. The Company owns and operates around 360 miles of gas gathering pipelines, approximately 300 Million Cubic Feet per Day (MMcf/d) of processing capacity, as well as crude trucking and logistics services extending through seven counties in the core of the Midland Basin. EnLink Midstream operates approximately 9,100 miles of gathering and transportation pipelines, 16 processing plants, seven fractionators, as well as barge and rail terminals, product storage facilities, brine disposal wells, truck terminals and a fleet of approximately 100 trucks.
Description Energy Transfer Partners, L.P. is a master limited partnership. The Company's operating segments include Intrastate Transportation and Storage segment; Interstate Transportation and Storage segment; Midstream segment; Liquids Transportation and Services segment; Investment in Sunoco Logistics segment; Retail Marketing segment and All Other segment. It is engaged in natural gas operations, including natural gas midstream and intrastate transportation and storage, and interstate natural gas transportation and storage; Liquids operations, including NGL transportation, storage and fractionation services; product and crude oil operations, including product and crude oil transportation, terminalling services and acquisition and marketing activities and retail marketing of gasoline and middle distillates. It is managed by its general partner Energy Transfer Partners GP, L.P. Its brands include Sunoco, Stripes, Aplus, Aloha Island Mart, Exxon, Valero, Mobil, Shell and Chevron, among others.
Description Enable Midstream Partners LP, is a large-scale, growth-oriented limited partnership formed to own, operate and develop strategically located natural gas and crude oil infrastructure assets. It serves current and emerging production areas in the United States, including several unconventional shale resource plays and local and regional end-user markets in the United States. The Company operates in two business segments: Gathering and Processing and Transportation and Storage. Gathering and processing provides natural gas gathering, processing and fractionation services and crude oil gathering for producer customers. The Company’s natural gas gathering and processing assets are located in four states and serve natural gas production from shale developments in the Anadarko, Arkoma and Ark-La-Tex basins. Transportation and storage provides interstate and intrastate natural gas pipeline transportation and storage service to natural gas producers, utilities and industrial customers.
Description National Fuel Gas Company is a holding company. The Company is engaged in the production, gathering, transportation, distribution and marketing of natural gas. It also develops and produces oil reserves, primarily in California. The Company operates in five segments. The Exploration and Production segment seeks to discover and produce raw materials (natural gas, oil and hydrocarbon liquids), and its operations are carried out by Seneca Resources Corporation. The Pipeline and Storage segment operations are carried out by National Fuel Gas Supply Corporation, which provides interstate natural gas transportation and storage services for affiliated and nonaffiliated companies through an integrated gas pipeline system. The Gathering Segment consists of National Fuel Gas Midstream Corporation’s operations. The Utility segment’s operations are carried out by National Fuel Gas Distribution Corporation. The Energy Marketing segment’s operations are carried out by National Fuel Resources, Inc.
22
Description JP Energy Partners LP (JP Energy Partners) owns, operates, develops and acquires a portfolio of midstream energy assets in the United States. The Company operates through four segments: Crude Oil Pipelines and Storage, Crude Oil Supply and Logistics, Refined Product Terminals and Storage, and NGL Distribution and Sales. The Company's crude oil businesses are situated in areas, including the Permian Basin, Mid-Continent and Eagle Ford shale. The Crude Oil Supply and Logistics segment manages the physical movement of crude oil origination to final destination. The Refined Product Terminals and Storage segment consists of two refined products terminals located in North Little Rock, Arkansas and Caddo Mills, Texas. The NGL Distribution and Sales segment includes its propane cylinder exchange business and the retail, commercial and wholesale sale of NGLs and other refined products. The Company markets its propane cylinder exchange business under the Pinnacle Propane Express.
Description Atmos Energy Corporation (Atmos Energy) is engaged primarily in the regulated natural gas distribution and transmission and storage businesses as well as other nonregulated natural gas businesses. The Company delivers natural gas through regulated sales and transportation arrangements to over three million residential, commercial, public authority and industrial customers in nine states located primarily in the South. The Company also operates intrastate pipelines in Texas based on miles of pipe. The Company operates through the following three segments: The regulated distribution segment, which includes regulated distribution and related sales operations; the regulated pipeline segment, which includes the pipeline and storage operations of its Atmos Pipeline Texas Division and the nonregulated segment, which includes nonregulated natural gas management, nonregulated natural gas transmission, storage and other services.
Description Vectren Corporation (Vectren) is an energy holding Company. The Company segregates its operations into groups, including Utility Group and Nonutility Group. The Company's wholly owned subsidiary, Vectren Utility Holdings, Inc. (Utility Holdings), serves as the intermediate holding company for three public utilities: Indiana Gas Company, Inc. (Indiana Gas), Southern Indiana Gas and Electric Company (SIGECO) and Vectren Energy Delivery of Ohio, Inc. (VEDO). The Company, through Vectren Enterprises Inc. (Enterprises), is involved in non-utility activities in two primary business areas: Infrastructure Services and Energy Services. Infrastructure Services provides underground pipeline construction and repair services. Energy Services provides energy performance contracting and sustainable infrastructure
Description Piedmont Natural Gas Company, Inc. is an energy services company whose principal business is the distribution of natural gas to over one million residential, commercial, industrial and power generation customers in portions of North Carolina, South Carolina and Tennessee, including customers served by municipalities who are the Company's wholesale customers. The Company is invested in joint venture, energy-related businesses, including unregulated retail natural gas marketing, and regulated interstate natural gas storage and intrastate natural gas transportation. The Company operates in two segments: regulated utility and non-utility activities. The regulated utility segment is the gas distribution business, where it includes the operations of merchandising and its related service work and home warranty programs, with activities conducted by the utility. The non-utility activities segment consists of the Company's equity method investments in joint venture, energy-related businesses.
Description American Midstream Partners, LP owns, operates, develops and acquires a diversified portfolio of midstream energy assets. The Company gathers, treats, processes and transports natural gas,
23
fractionates natural gas liquids (NGLs) and stores specialty chemical products through its ownership and operation of 12 gathering systems, five processing facilities, three fractionation facilities, three interstate pipelines, five intrastate pipelines and four marine terminal sites. The Company operates through three segments: Gathering and Processing, Transmission and Terminals. The Gathering and Processing segment gathers, processes, transports and treats natural gas. The Transmission segment transports natural gas. The Terminals segment provides above-ground storage services at the Company's marine terminals that support commodity brokers, refiners and chemical manufacturers, to store a range of products, including petroleum products, distillates, chemicals and agricultural products.
Description Southwest Gas Corporation is engaged in the business of purchasing, distributing and transporting natural gas in Arizona, Nevada and California. The Company has two business segments: natural gas operations and construction services. The Company operates two pipeline transmission systems: a system, including a liquefied natural gas (LNG) storage facility owned by Paiute extending from the Idaho-Nevada border to the Reno, Sparks, and Carson City areas and communities in the Lake Tahoe area in both California and Nevada and other communities in northern and western Nevada, and a system extending from the Colorado River at the southern tip of Nevada to the Las Vegas distribution area. The natural gas operations segment includes acquiring and arranging delivery of natural gas to its system. The construction services segment consists of Centuri, which includes NPL Construction Co., Link-Line Contractors, W.S. Nicholls Construction and Brigadier Pipelines.
Description The Laclede Group, Inc. (Laclede Group) is a public utility holding company. The Company, through its subsidiaries, operates in two business segments: Gas Utility and Gas Marketing. The Gas Utility segment includes the regulated operations of Laclede Gas Company (Laclede Gas, the Missouri Utility) and Alabama Gas Corporation (Alagasco) (together with Laclede Gas, the Utilities). Laclede Gas is a public utility engaged in the purchase, retail distribution and sale of natural gas, serving residential, commercial and industrial customers. Alagasco is a public utility engaged in the purchase, retail distribution and sale of natural gas principally in central and north Alabama, serving over 0.4 million residential, commercial and industrial customers. The Gas Marketing segment includes Laclede Energy Resources, Inc. (LER), a wholly owned subsidiary engaged in the marketing of natural gas and related activities on a non-regulated basis
Description WGL Holdings, Inc. (WGL Holdings), is a United States of America-based public utility holding. Through its wholly owned subsidiaries, it sells and delivers natural gas and provides energy-related products and services to customers primarily in the District of Columbia and the surrounding metropolitan areas in Maryland and Virginia. Promotes the efficient use of clean natural gas and renewable energy to improve the environment for the benefit of customers, investors, employees, and the communities it serves. The Company’s business segments include Regulated Utility, Retail Energy-Marketing, Commercial Energy Systems and Midstream Energy Services. WGL Holdings owns all of the shares of common stock of Washington Gas, Washington Gas Resources, Hampshire Gas Company (Hampshire) and Crab Run Gas Company (Crab Run). Washington Gas Resources owns four unregulated subsidiaries that include WGEServices, WGESystems, and WGL Midstream.
Description Questar Corporation (Questar) is an integrated natural gas holding company that develops, produces and delivers clean energy in the Rockies, North America. The company, through its subsidiaries is principally engaged in three line of business: Questar Gas Company (Questar Gas), which provides retail natural gas distribution in Utah, Wyoming and Idaho; Wexpro Company (Wexpro) that develops and produces natural gas from cost-of-service reserves for Questar Gas customers, and Questar Pipeline Company, which operates interstate natural gas pipelines and storage facilities in the western United States and
24
provides other energy services. The Company also operates through Questar Fueling Company (Questar Fueling), which builds, owns and operates compressed natural gas (CNG) fueling stations for fleet operators with medium- and heavy-duty trucks and tractors.
Description Northwest Natural Gas Company is engaged in the distribution of natural gas. The Company operates in two business segments: Local Gas Distribution and Gas Storage. The Company maintains operations in Oregon, Washington and California and conducts businesses through NW Natural and its subsidiaries. The local gas distribution business, referred to as the utility segment, serves residential, commercial, and industrial customers in Oregon and southwest Washington; and the gas storage business, referred to as the gas storage segment, provides storage services for utilities, gas marketers, electric generators and industrial users from storage facilities located in Oregon and California. As of December 31, 2014, the Company had 700,000 customers with around 89% located in Oregon and 11% located in Washington. It provides natural gas service to over 100 cities in around 18 counties.
1
APPENDIX D 1
DISCOUNTED CASH FLOW ESTIMATES 2
The DCF Model 3
The standard alternative to risk premium models is the discounted cash flow model. This model 4
infers the required rate of return by replicating the actions of an investor in valuing the firm's 5
securities. To do this we need to define the costs and benefits attached to an investment. The cost is 6
simply the price of the security (P0, price at time zero) and the benefits the stream of cash inflows 7
expected at time t in the future (Ct). However, since the investor can always invest in alternative 8
investments, future expected cash flows are not of equal value. As a result future cash flows are 9
"discounted," or reduced in value, to reflect this "opportunity cost." This is the basic idea behind 10
using the discounted cash flow model, 11
10 )1(t
tt
K
CP 12
where K is the discount rate or investor's required rate of return. 13
Once we estimate the stream of future cash inflows, we can equate them to the current price and 14
solve for the investor's required rate of return. For example, this is the standard way of valuing 15
bonds. At the end of every business day investment banks simply take the coupon payments on a 16
bond and its terminal value, and use the last trading value for the bond to solve the above equation 17
for the bond's "yield to maturity." This yield to maturity is then published in the newspaper as an 18
objective measure of the investors' required rate of return for a default free security. I already use 19
this DCF estimate as part of my risk premium estimates. However, we can take this a stage further 20
and estimate the DCF required return on equity directly using this same procedure. 21
The expected equity cash flows are the future expected dividends. Unlike the stream of cash flows 22
on a bond the dividends are not contractual and are more difficult to forecast, particularly for 23
individual stocks. Consequently the DCF model is only used for low risk dividend paying stocks or 24
the market as a whole, where the expected dividends can be assumed to grow at some long run 25
average growth rate g. 26
2
In this case, each dividend is expected to grow at the rate g, so we can substitute d1 = d0 * (1+g) into 1
the valuation equation to get: 2
gK
dP
1
0 3
where the stock price is equal to the expected dividend per share, divided by the investor's required 4
rate of return, minus the dividend growth expectation, g. The advantage of this formulation of the 5
problem is that we can easily rearrange the equation to obtain, 6
gP
dK
0
1 7
which states that the investor's required rate of return can be estimated as the expected dividend 8
yield plus the expected growth rate in dividends. This is the direct analogy with the yield to maturity 9
on a bond. This formulation of the model is often called the Gordon (or dividend discount) model 10
after my late colleague Professor Myron Gordon of the University of Toronto. 11
Further it is straightforward to show that increased dividends primarily come from increased future 12
earnings, which are generated by the firm retaining some of its current earnings for re-investment. If 13
we set X as the earnings per share and denote b as the fraction of earnings retained within the firm, 14
then (1-b)X is the dividend and bX, the retained earnings.1 Provided the assumptions of the DCF 15
model hold, it is straightforward to show that dividends and earnings will then grow at a long run 16
growth rate estimated as the product of the firm's retention rate (b) and its return on common equity 17
(r), which is referred to as its sustainable growth rate. Note that while K is the return that investor's 18
require, r is the actual return on equity (ROE) the firm is expected to earn.2 19
An example may help to make these assumptions clear. Suppose, as in Schedule 1, the firm's 20
book value per share is $20 and its return on equity expected to be 12%. In this case, its 21
earnings per share are expected to be $2.40 and with a 50% dividend payout rate, its dividends 22
1 This assumes that the only change in shareholder’s equity comes from retentions, that is, everything flows through the income statement. 2 There is an additional term if the firm repeatedly sells shares at a premium to its book value, but this term is small and usually dwarfed by estimation problems.
3
per share and retained earnings are both expected to be $1.20. Moreover, since $1.20 has been 1
retained and reinvested within the firm, next period's book value per share increases to $21.20. 2
As a result, the firm is expected to earn $2.544 in the following year, i.e., 14.4 cents more. This 3
additional 14.4 cents comes from earning the 12% return on equity on the $1.20 of retained 4
earnings. The increase in earnings per share, dividend per share and retained earnings is 6% each 5
year and is calculated directly as the product of the firm's return on equity of 12% and its 6
retention rate of 50%. Moreover, the value of the firm's common stock can be calculated from 7
equation (1), which also increases at this 6% rate, since only the dividend per share is expected 8
to change. 9
The importance of Schedule 1 is in showing some of the implications of the dividend growth 10
model. First, note that if the investor's fair rate of return is 10%, the stock price in Schedule 1 is 11
$30, determined as the expected dividend of $1.20 divided by the discount rate minus the growth 12
rate (or 0.04). This price exceeds the book value of $20 by 50%. This is because the firm's 13
return on equity (r ) is 12% and the investor's required or fair rate of return (K) is only 10%. This 14
is the reason why economists look at market-to-book ratios to infer the investor's opportunity 15
cost. If market-to-book ratios exceed one for a regulated company, most economists immediately 16
assume that the firm's return on equity exceeds the return required by stock holders, implying 17
that the regulator should lower the firm's allowed rate of return. In our example the ROE exceeds 18
the required rate of return by 2% which results in a market to book ratio of 150%. 19
Second, it is the return on equity that drives the growth in both dividends per share and earnings 20
per share, provided that the dividend payout is constant. If the dividend payout is gradually 21
increased over time, then it is possible to manufacture a faster growth rate in dividends than 22
earnings per share, from the same underlying level of profitability. 23
For example, in Schedule 2 the same data is used as in Schedule 1 except that the dividend 24
payout starts at 50% and then increases by 2% per year. By the end of year 5 earnings per share 25
have only risen to $2.99 instead of the $3.03 in Schedule 1, because less money has been 26
reinvested within the firm. As a result, there is less capital to generate earnings. Thus the 27
earnings in Schedule 2 only grow at a 5.6% compound growth rate, down from the 6% of 28
4
Schedule 1. Conversely, since more of the earnings are being paid out as dividends, dividends 1
per share are up to $1.73 instead of $1.52. This is a 9.6% compound growth rate, rather than the 2
6% in Schedule 1. 3
In the short-run, Schedule 2 demonstrates that the growth in dividends per share can be 4
artificially manipulated by increasing the dividend payout. This is not sustainable in the long 5
run, since the dividend payout cannot be increased indefinitely. Moreover, the manipulation can 6
be detected by performing the basic 'diagnostic' check of tracking the behaviour of the firm's 7
dividend payout over time, and the firm's return on equity. However, if the analyst is not aware 8
of the change in the dividend payout, estimating the fair rate of return by adding this 9
manipulated dividend growth rate to the expected dividend yield will overstate the investor's 10
required rate of return. It is important in this case to base the estimate of the investor's required 11
rate of return on a long run sustainable growth rate, estimated from the underlying growth in 12
earnings and dividends and the two components of growth. 13
The third implication of Schedule 1 is that the DCF estimate using the historic growth rate is 14
appropriate only when the assumptions of the model hold. This means that non-dividend paying 15
firms, firms with highly fluctuating earnings and dividends, and firms with non-constant 16
expected growth cannot be valued accurately using the formula. Usually these assumptions hold 17
for regulated utilities, so the DCF estimate is particularly appropriate for use in determining the 18
fair rate of return for a regulated utility. However, for non-regulated firms and utility holding 19
companies (UHCs), these assumptions are frequently violated. As a result, estimating the 20
investor's required rate of return by using the formula K=d1/P0 + g, is tenuous and subject to 21
significant measurement error. 22
Circularity 23
When we apply the DCF model to estimate a fair return we estimate the dividend yield and 24
future growth rate. In the example in Schedule 1 the dividend is forecast to be $1.20 which with 25
a $30 stock price means a 4% dividend yield. When this is added to the sustainable growth rate 26
of 6% we get back the investor’s fair rate of return of 10.0%. However, it is sometimes alleged 27
that this DCF estimate is circular, since the ROE used to forecast the future growth rate of 12% 28
5
differs from the investor’s required or fair rate of return estimated at 10%. The allegation is that 1
if a regulatory body were to accept the 10% estimate and reduce the allowed ROE then future 2
growth will drop and with it the stock price. As a result, there is an inconsistency between the 3
forecast ROE and the DCF fair return estimate. However, this inconsistency or circularity is 4
false. 5
Note that there will always be a difference between the forecast ROE and the investor’s fair 6
return, whenever the market to book ratio differs materially from 1.0.3 However, this does not 7
affect the estimate produced by the DCF model. Suppose for example the ROE was decreased to 8
10%, after the fair return is correctly estimated at 10% using the DCF model, what happens? In 9
this case the forecast earnings per share drop to $2 from $2.40 and with the same 50% payout the 10
dividend is cut to $1.0 and the forecast growth rate drops to 5% (50% retention times the 10% 11
ROE). The stock price will then also drop, which is what critics of the DCF model worry about! 12
However, using the same DCF equation the market price will fall back to its book value of $20. 13
20$05.010.0
1$0
P 14
At the new price the dividend yield increases to 5% ($1/$20), so that with the new lower forecast 15
growth rate of 5%, we again estimate the investor’s fair return accurately at 10%. 16
Investors will be far from happy that the allowed ROE has been cut from 12% to 10%, but that 17
does not invalidate the use of the DCF model to estimate their fair, or required, rate of return of 18
10%. Similarly, if the regulator for some reason increases the allowed ROE to 14% then the 19
dividend would increase to $1.40 and the forecast growth to 7%. In this case the stock price 20
would increase to $46.67 and the dividend yield drops to 3.0%, so again the dividend yield plus 21
growth correctly estimates the investor’s fair rate of return of 10.0%. 22
The fact is that the DCF model simply reverse engineers the forecast cash flows to extract the 23
investor’s fair rate of return; it says nothing about whether or not the investor would be happy if 24
3 We see this every day in the bond market where a bond selling above (below) par has a stated coupon interest rate higher (lower) than the current market interest rate.
6
the firm earned that rate of return on its book value. Further proponents of this circularity 1
argument often apply the DCF model based on analyst growth estimates, yet these same analysts 2
have to get their forecast growth rates from somewhere and invariably they are based on future 3
profitability, that is, forecast ROEs. Moreover, even if they are not explicitly based on a forecast 4
ROE, one is always implicit in any growth forecast. For example, if an analyst’s growth forecast 5
of 7% is used in a DCF model, then with a 50% dividend payout this means by definition the 6
analyst is forecasting an ROE of 14%. It is impossible to ignore the result that any forecast 7
growth rate carries with it a forecast ROE which will almost certainly deviate from the investor’s 8
required rate of return. 9
DCF Estimates for the “Market” as a whole 10
In terms of DCF estimates we can go from the broad to the specific. By broad I mean the market 11
as a whole, since by holding a diversified portfolio an investor reduces the possibility of gains 12
from one firm being the result of losses by another. In Schedule 3 is a graph of the dividend yield 13
on the TSX Composite (Cansim V122628) along with the yield to maturity on the long Canada 14
(LTC) bond (Cansim V122501). At the end of October 2015 the TSX yield was 3.17%, while the 15
Long Canada yield was 2.12%. This is an unusual situation that has prevailed since the end of 16
2011. It is unusual since equities are a claim on real resources and should grow in line with the 17
growth rate in profits and GDP. In contrast, the yield on the long Canada bond is fixed and is all 18
the investor can earn if the bond is held to maturity. 19
In Schedule 5 is a graph of the after tax profits and dividends earned and paid in Canada by 20
Canadian corporations. The data is from the GDP accounts and goes back to 1956 and in both 21
cases is scaled by dividing by GDP. The after tax profits are those reported for tax purposes and 22
do not reflect the accounting “games” that are often used to inflate accounting or GAAP profits 23
to “please” investors. For example, non-cash items like capital gains are removed and 24
inventories are adjusted to remove inflationary gains. As is to be expected, aggregate dividends 25
(right side axis) are more stable than aggregate after tax profits. After-tax profits plummeted, for 26
example, during the recessions in 1981, the early 1990s, marginally in the early 2000s and 27
during the recent financial crisis. Overall dividends on average have been 3.0% of GDP and after 28
7
tax corporate profits 6.7%, but much more variable. Until recently after tax profits have been 1
above these long run averages and reached over 10.0% in 2008 before the financial crisis as high 2
resource prices benefitted Corporate Canada. 3
Note that dividends are more stable than earnings as firms don’t like to cut their dividends. This 4
is important since some utility analysts “key” dividend growth forecasts off earnings forecasts. 5
This is suspect since the greater variability in earnings means that their average growth rate 6
always exceeds that of dividends in the same way that the arithmetic return always exceeds that 7
of the geometric (compound) growth rate.4 However, with this caveat it is hard not to conclude 8
that in the long-run dividends and after tax profits grow at about the same rate as the overall 9
economy, but are more variable. The average real Canadian growth rate since 1961 has been 10
3.25%, while the Bank of Canada’s operating band for inflation centres on 2.0%, this implies a 11
long-run growth rate in dividends and earnings at about 5.35% (1.02*1.034). This is probably a 12
low estimate for two reasons; first the GDP accounts have become less reliable as the economy 13
has shifted to a knowledge based economy, since it has become more difficult to estimate the 14
value of productivity changes; second the arithmetic vs compound growth rate problem also 15
affects the GDP accounts, which are less variable than similar accounts for companies. 16
An alternative estimate of future growth for the market as a whole is to use the “br” or 17
sustainable growth rate. In Schedule 6 is the aggregate dividend payout from the GDP accounts. 18
We can see very clearly the jump in the payout during the severe recessions in the early 1980s 19
and 1990s when Corporate Canada had serious profitability problems. The median payout is 20
41%. In Schedule 7 is the dividend payout based on the earnings and dividends of the TSX 21
Composite. We can see the impact of the recessionary periods even more clearly, but this time 22
the payout is truncated for the over 100% payout periods. The TSX data is based on GAAP 23
profits and reflect “big bath” accounting, that when times are bad and the stock market expects 24
bad news, firms tend to exaggerate their losses and build reserves that allow them to smooth 25
profits in the future. The median payout for the TSX is higher at 51%. Overall I judge a 26
reasonable dividend pay-out to be in a range 41-51% so that the retention rate is (b) of 49-59%. 27
4 The standard deviation of after tax profits as a % of GDP has been about twice that of dividends.
8
From Schedule 1 of my main testimony the average ROE of corporate Canada back to 1987 has 1
been 9.26% and the median 9.87%. Multiplying these ROEs by the retention rates gives a 2
sustainable growth rate range of 4.72% (0.51*9.26) – 5.77% (.59*9.78) which brackets the 3
estimate of 5.35% from the long run GDP growth rate. If this range is combined with the recent 4
(end of September 2015) TSX dividend yield of 3.17%, the DCF estimate for the market as a 5
whole is 8.1% (3.17% and 4.72%) to 9.2% (3.17% and 5.77%). Similarly with the 5.35% growth 6
rate the DCF estimate is 8.75%. These would seem to be reasonable estimates if the market were 7
at the mid-point of the business cycle, rather than just leaving the “recession or slowdown” 8
phase. 9
At the current point in time Canadian business has recovered from recession and is running close 10
to average or normal capacity. The median capacity utilisation levels since 1987 have been 82.8 11
& 81.9% for non-farm and manufacturing respectively, but currently they are at 82.7-82.9%, so 12
right about average. This observation is confirmed by the December 2015 unemployment rate of 13
7.1% graphed in Schedule 9, which would place us similar to 2005. The peak of the business 14
cycle is when inflationary pressures set in, where the non-accelerating inflation rate of 15
unemployment (NAIRU) is regarded as 6.0%, which is where we were just before the financial 16
crisis when the Bank of Canada stated to tighten to ward of incipient inflation. We should be 17
further along the growth path than we are, but Canada has been stuck at around 7.0% 18
unemployment since 2012 as a result of weaker commodity prices. So there is still spare 19
capacity in the economy. 20
The following Bank of Canada graphic illustrates the divergent growth paths of the oil and gas 21
sector versus the rest of the economy. As a result, the Bank of Canada5 anticipates that a weak 22
2015 will be followed by sub-par growth of 2.0% in 2016 and 2.5% in 2017. 23
5 Monetary Policy Report, October 2015.
9
1
What this means is that a simple application of the Gordon model using the current dividend 2
yield plus growth estimates in a range of 8.1-9.2% probably marginally understates the market’s 3
fair return. Overall I would judge the fair rate of return on the Canadian market to be 8.50-4
9.00% which means a real rate of about 6.50-8.0% with 2.0% long run inflation. This is right 5
about average for the Canadian market. 6
In Schedule 9 is a graph of the dividend yield on the S&P500 index and in Schedule 10 a graph 7
of the dividend payout rate on the S&P500 firms. The average dividend payout since 1956 is 8
47.7% while the median payout is 43% meaning that typically 52.3% of the earnings for S&P500 9
firms are reinvested to generate future growth in earnings. However, note from the graph that 10
the S&P500 firms suffered significant problems in 2007-2009 during the financial crisis, which 11
was not as evident in the Canadian data, particularly the tax data. In contrast, there is no 12
evidence of the serious problems suffered by Corporate Canada in the recessions in the early 13
1980s and 1990s. 14
S&P has profitability data for the S&P500 firms since 1987 where the average ROE was 13.43% 15
and the median ROE 14.07%. These are higher than the Canadian average ROE since the data is 16
for the largest firms in the US economy and includes a large proportion of foreign earnings, 17
whereas that for Canada is for all firms. Over this same period the average and median retention 18
10
rates were 51.5% and 57%. If I pair the median payout and ROE the “br” growth rate is 7.99% 1
and if I pair the averages the growth rate is 6.91% reflecting both the higher average payout and 2
lower average ROE. Combining these with the current dividend yield on the S&P500 index of 3
1.99% gives a fair return on the US market of 9.04%-10.14%. I would judge a fair return on the 4
US market to be very similar to that in Canada at 9.0-10.0%. Note the higher US growth forecast 5
is offset by a lower US dividend yield. 6
Of note is that in a recent US strategy report (July 18, 2012) RBC estimated the historic equity 7
market return in the US at 9.4%, which would be consistent with the above analysis. However, 8
RBC then estimated a “supply side” model to get a 4.9% long run (ten year) forecast for average 9
US equity market returns. Their approach was a variant of the above calculations. They started 10
with a 2.1% dividend yield on the SP500, from which they subtracted 0.50% for dilution caused 11
by firms issuing shares and thus reducing the investor’s proportionate interest. This gives a cash 12
yield of 1.60%. RBC then used a 2.1% forecast inflation rate and estimated earnings growth at 13
3.8% or slightly higher than the US economy’s average growth rate of 3.50%. However, RBC 14
pointed out that margins in the US were at an all-time high and they forecast that if the US 15
economy grows at 3.5% profits will only grow at 2.2%. Finally, RBC judged the US market to 16
be over-valued based on the long run “Shiller” PE ratio and predicted a 1.0% valuation drag on 17
expected returns. 18
This analysis resulted in a 4.9% expected return on the US stock market as follows: 19
Dividend Yields: 2.1% 20
Net Share issuance: -0.50% 21
Inflation: 2.1% 22
Real Earnings growth: 2.2% 23
Change in PE: -1.0% 24
Total equity return: 4.90% 25
I am nowhere near as pessimistic as RBC about the US and judge there to be excellent short run 26
growth prospects leading to top line revenue growth to match the improved margins that have 27
already generated bottom line growth in a weak economy. However, what it points out is that 28
professionals in the investment business often generate much lower expected rates of return than 29
those that I have generated. 30
11
S&P US Utility DCF cost estimates 1
As well as the data for the S&P500 as a whole, Standard and Poors also publishes data on the 2
utilities that meet the requirements to be included in the S&P500 index. In Schedule 11 is the 3
summary data for the traditional utilities which include the standard electric and gas utilities. 4
Note that the S&P data includes the firms that at the time were classified into these groups so 5
whereas there was only 1 utility included as a gas utility in 2014, there were 13 in 1993. The data 6
for each year reflects the S&P value weighted average of the firms in that industry that year. 7
The schedules provide the basic data needed for a DCF analysis. The data includes dividends, 8
earnings, book value per share, average market values and the return on equity. From this it is 9
possible to calculate several pieces of useful information. First, is the average payout, which is 10
in the fourth column and its inverse, which is the retention ratio. Clearly, utilities as low risk and 11
low growth investments have relatively high payouts: in not one of the 22 years is the payout 12
less than 50% for the electric utilities and the average and median payouts are both 71%. This 13
may be marginally biased high by the large payout in 2000. For the gas utilities the payout is 14
more variable as there are fewer gas utilities in the SP500 index. In this case there are 7 years in 15
which the payout is under 50% and three years when the payout was over 100%. However, the 16
median payout for the gas utilities is still is very high at 67% and only slightly less than for the 17
electric utilities. 18
The very high dividend payout means that the growth potential for these utilities is low, which 19
reduces the error in using the DCF model. It also means that utilities are quintessentially 20
dividend or income stocks. The average 2014 dividend yield for the electric utilities was 3.82% 21
and for the gas utilities 3.91% both of which were significantly higher than the S&P500 dividend 22
yield in December 2014 of 1.92%. Again we would expect lower long run growth for these 23
utilities than the US market as a whole, since they are mature industries. 24
To estimate the future growth rate I can assume that each year the utility is expected to earn its 25
current ROE, so that its earnings will grow by the retention rate times this ROE. For example, in 26
1993 the retention rate was 10.57% and the ROE 11.25% for the electric utilities implying future 27
earnings growth of 1.19%, which is the g (b*ROE) in the next column. For 1993 the dividend 28
12
yield for the S&P Electric utilities was 5.73% (column 8), so that the DCF equity cost estimate 1
was 6.99%, which is in column 10. In 1993 the average long term (ten year) US Treasury yield 2
was 5.87% implying that the electric utility risk premium was only 1.12%. Column 11 gives the 3
market to book ratio for these utilities, which in 1993 was 1.59, implying correctly that the ROE 4
of these utilities of 11.25% exceeded their equity cost. 5
The above calculation is a mechanical exercise and obviously includes estimation error in both 6
the earned ROE which affects both the forecast ROE and the retention rate. To reduce individual 7
estimation errors the exercise is repeated for each year from 1993 until 2014. This gives the 8
average and median electric utility risk premium of 3.42% and 3.43% with 2.75% and 3.09% for 9
the gas utilities. However, the br growth rate is sensitive to the actual earnings which affect the 10
retention rate and may not capture the full amount of growth expectations. To check for this the 11
last two columns estimate the utility risk premium with two alternative growth expectations. 12
URP2 assumes that the expected ROE is the median ROE for the whole period 1993-2014 which 13
avoids the problem of fluctuating earned returns. URP3 also assumes that the retention rate is the 14
constant median growth rates for the whole period. This avoids the problem of declining 15
retention rates as earnings are squeezed and the dividend maintained. The average and median 16
URP2 is 3.32% and 3.70% for the electrics and 1.75% and 3.09% for the gas utilities and for 17
URP3 the values are 3.35% and 2.98% for the electrics and then 2.25% and 2.09% for the gas 18
utilities. 19
As a final check I looked at the growth in the dividend per share for the electric utilities relative 20
to GDP.6 Over the period from January 1995 to 2014 the median nominal US GDP growth rate 21
was 4.7%, whereas the median dividend per share growth rate of the S&P500 Electrics was 22
1.75% or barely 40% of the growth in the US economy. In Schedule 12 is a graph of the ten year 23
US Treasury yield against the dividend yield of the S&P500 electrics. The correlation is over 24
0.50 and clearly divided yields have dropped with the drop in US market interest rates. A naïve 25
forecast that these utilities would have dividend growth of 40% of US GDP or about 2% would 26
produce a DCF estimate of about 6.0%. 27
6 The dramatic reduction in the number of gas companies makes their use problematic and even for the electrics where the number of firms is not constant there are problems.
13
From the data in Schedule 11, I derive the following conclusions: 1
Risk premiums of the order of 2.7-3.70% for a typical US utility over ten year US 2
government bond yields is reasonable as reflecting typical experience over the last 22 3
years. 4
For the more stable US electric utilities the risk premium for the period 1993-2014 is 5
about 3.42% with evidence that it is higher at lower interest rates. 6
Although 2015 data is not yet available the impact of lower US treasury yields has 7
shown up in the more recent US data as increasing utility risk premiums, since the 8
most recent risk premiums are higher than average. 9
Individual company estimates 10
The DCF estimates for the market as a whole and the S&P utility indexes are more reliable than 11
for individual companies due to the significant measurement error attached to forecasting future 12
growth rates. For example, the forecast growth rate for the economy is more accurate since the 13
growth rate in profits for the market as a whole is constrained in the long run by the growth rate 14
in the economy. However, the growth rates are mechanically estimated and do not reflect market 15
estimates. Consequently some use analyst forecast of earnings growth as a proxy for the 16
sustainable growth rates in the former estimates. However, in my judgment these are no more 17
reliable as can be illustrated by looking at the sample of US utilities that I analysed in Appendix 18
C. These utilities tend to be smaller gas utilities that are not in the SP500 and tend to be less 19
diversified. As a result, one might expect their DCF estimates to be more reliable, which turns 20
out to be not the case. 21
The following table has data I extracted on November 18, 2015 22
Yield Past G Future G #AnalystsDIV EPS Retention ROE Sust G MB K5 year K SG
Vectren 3.92 18.8 5 7 1.6 2.28 0.30 11.68 3.48 2.05 9.12 7.54WGL 3.18 -17.09 7 4 1.85 2.62 0.29 10.31 3.03 2.33 10.40 6.31Piedmont 2.29 2.88 5 6 1.32 1.8 0.27 10.37 2.77 3.25 7.40 5.12Northwest 4.02 -11.31 4 1 1.87 1.92 0.03 6.94 0.18 1.68 8.18 4.21New Jersey Resources 3.22 10 6 3 0.96 1.77 0.46 14.26 6.53 2.27 9.41 9.96Laclede 3.31 5.16 4.44 7 1.84 3.24 0.43 8.96 3.87 1.52 7.90 7.31ATMOS 2.79 29.23 7 5 1.68 3.09 0.46 10.03 4.58 1.92 9.99 7.49South West Gas 2.97 -13.7 4 2 1.62 2.78 0.42 8.74 3.65 1.67 7.09 6.73
Average 3.21 3.00 5.31 4.38 1.59 2.44 0.33 10.16 3.51 2.09 8.69 6.83Median 3.2 4.02 5.00 4.50 1.65 2.45 0.36 10.17 3.57 1.99 8.65 7.02 23
14
Note that the current dividend yields range from 2.29% to 4.02% due to the particular 1
circumstances of each utility, but the median dividend yield of 3.2% is basically the same as the 2
the average of 3.21%. Moreover, these are much higher than the current (November 18, 2015) 3
dividend yield on the S&P 500 index of 1.98% consistent with the lower growth prospects of the 4
utility sector as a mature industry. However, the earnings of these utilities are not stable; their 5
past five year growth rates range from -17.09% for WGL to +29.23% for ATMOS Energy. It is 6
this growth rate instability that makes the use of the DCF model suspect even for “low risk” 7
“utilities.” Going forward the current analyst growth forecasts are in a tighter range of 4%-7%. If 8
these growth rates are mechanically included in a constant growth model the fair equity cost (K 5 9
year) ranges from 7.09% to 10.40% for an average of 8.69% and a median of 8.65%. 10
These estimates may look superficially appealing to some, but they present formidable problems. 11
The first is that the DCF model assumes growth forever at the average forecast (five year) 12
growth rate of 5.31%, which with the forecast US inflation at a long run rate of 2.0% means 13
about 3.3% long run real growth, whereas at least in the shorter run the US economy is forecast 14
at best to grow at less than 3.0%. It is impossible that these utilities will grow faster than the 15
growth rate of the US economy forever and there is no evidence that they have at least grown at 16
the US GDP growth rate consistently in the past. It is this problem that causes some analysts to 17
taper their forecast growth rates to the future growth rate in the economy. However, historically 18
as noted above US utilities have only been able to grow their dividends at about 40% of the 19
growth rate in the US economy and I have never seen any evidence produced to justify such an 20
optimistic forecast. Moreover, whereas the average (median) five year growth rate over the 21
previous five years was 3.0% (4.02%), the forecast growth rate averages 5.31% (5.0%) or 22
significantly more. This indicates that these DCF estimates have to be reduced to reflect these 23
optimistic forecasts. 24
At Schedules 13 and 147 are reprints of a Globe and Mail article that reports on an update of a 25
study by the consulting firm, McKinsey. They report that analysts start out optimistic when 26
making their five year forecast, but gradually as they get more information (generally from the 27
7 Schedule 13 has a clearer repeat of the content of the article.
15
company) they hone in on the correct number. This is a result that has been in the academic 1
literature for some time and is not necessarily driven by any conflict of interest, as was evident in 2
the global settlement,8 but simply reflects an attachment effect, where analysts tend to become 3
attached to a stock and see good in it until proven otherwise. 4
Easton and Sommers9, for example, have documented the optimism bias at 2.84% and in their 5
conclusions (page 1012) state: 6
7
8
Easton and Sommers also state (page 986) 9
10
Easton and Sommers estimates are in line with my own estimate of the expected return on the US 11
market being in a range of 9.0-10.0%, even though it was based on data several years ago. More 12
importantly there is no reason to believe that analyst optimism has suddenly disappeared. In fact, 13
this optimism bias persists in current studies to the extent that authors refer to it as “well 14
documented”10 that is, researchers are so used to the optimism bias that they automatically take it in 15
8 This was the 2003 US$1.4 billion settlement between US Attorney General for New York Elliot Spitzer and a series of major US investment banks, where the investment banks admitted that security analyst compensation was tied to investment banking income and that analyst reports were in some instances fraudulent and lacked objectivity. 9 “Effect of analyst’s optimism on estimates of the expected rate of return implied by earnings forecasts, Journal of Accounting Research, 45-5, December 2007. 10 See Huang and Tan, for example, “Analyst target price optimism around the world,” November 2013.
16
to account. In addition to the Globe and Mail article referred to previously the Financial Times also 1
noted that analyst optimism exists in Europe, where they quote Goldman Sachs that “going back 25 2
years analysts have been too optimistic about earnings growth in 20 years out of the 25 and by 8 3
percentage points on average over the whole period.”11 A Google search on analyst optimism 4
produced 645,000 hits! 5
A final problem with the use of analyst forecasts is that they are based on earnings not dividends, 6
whereas the DCF model values dividends and not earnings. As Schedule 5 showed earnings are more 7
volatile than dividends even after we aggregate over all firms. What this means is that the short term 8
growth forecast for earnings is higher than for dividends even if their long run, or compound, growth 9
rates are the same. This is due to the common practise of smoothing dividend payments or said 10
another way firms only increase their dividend after their fundamental earnings have increased and 11
not due to temporary factors. For example, suppose the compound nominal (real plus inflation) 12
growth rate for the economy is 5% and profit uncertainty is 20%, then the short term earnings 13
growth rate would be 5% + .5* (0.2^2) or 7% whereas if its dividend uncertainty were only 5%, due 14
to this smoothing, its dividend growth rate would be 5% + .5 * (.05^2) or 5.12%.12 What this means 15
is that even if analyst growth forecasts were unbiased their earnings growth rates would be biased 16
high estimates of the future growth rate in dividends, which is what is needed for cost of capital 17
estimation.13 18
A standard way of alleviating the effects of analyst growth optimism is to use the sustainable growth 19
rate, which indicates that growth in earnings and dividends generally comes from reinvestment of 20
earnings and their earning a positive rate of return. From the data on these US utilities their retention 21
rate of earnings averages 33%, that is, 67% of earnings are paid out as dividends similar to the data 22
for the large S&P utilities. As we would expect, these mature utilities reinvest less of their earnings 23
than do companies on average, so we would expect them to grow at less than the rate of companies 24
11 Sarah Gordon, “European corporates thwart analyst’s optimism,” Financial Times, April 27, 2014. 12 This uses the standard result that the arithmetic growth rate is the compound growth rate plus half the variance in the growth rate. 13 The standard deviation of the past 5 year growth rates for these 8 US utilities was 16.4% indicating that the average growth rate of 3.0% over estimated the compound growth rate by 1.3%.The analysts do not reveal the range of their forecasts, so the extent of the bias is difficult to assess.
17
in general or the economy. With the recent average ROE for each utility the sustainable growth rate 1
is 3.51% and the median slightly higher at 3.57%. This forecast seems more consistent with a US 2
economy growing at about 5.0% and the mature nature of their operations. This produces a median 3
fair ROE of 7.02% consistent with their median market to book (MB) ratio of 2.0, indicating that 4
investors are very happy with their level of profitability. 5
One final comment is that we can always back out from analyst growth forecasts an implicit rate of 6
return. For example, with a median growth forecast of 5.0% and a retention rate of 36%, the implied 7
rate of return on new investment is ROE = .05/.36 or 13.9%14 which exceeds their current median 8
ROE of 10.17% by 3.73%. This indicates that these analysts are projecting a large increase in the 9
profitability of these utilities. The fact that this is a persistent phenomenon indicates analyst 10
optimism, in this case that their growth forecasts are 30-40% too high. 11
Conclusion 12
From the forgoing DCF estimates I draw the following conclusions: 13
The Overall equity market return in Canada is in a range 8.50%-9.50%; 14
The overall US equity market return is in a range 9.0-10.0%; 15
The US S&P electric utility risk premium is the most stable at about 3.4% which 16
based on an RBC forecast 2016 ten year US Treasury Yield of 2.60% would put a fair return 17
at about 6.00% before flotation costs; 18
The individual DCF estimates for 8 US utilities would put their equity cost at 8.65% 19
based on median analyst five year growth forecasts, which suffers both the optimism bias the 20
inconsistent growth forecast bias. When based on sustainable growth forecasts, the median 21
DCF equity cost for these 8 US utilities is 7.02%. 22
Overall I would estimate the DCF long run expected return on the equity market in a range 8.0-23
10.0% and the equity cost for a low risk US utility at about 7.0%. 24
14 This just reverses g = b*ROE.
18
SCHEDULE 1
BEGINNING BOOK VALUE EARNINGS DIVIDEND RETENTIONS YEAR PER SHARE PER SHARE PER SHARE PER SHARE 1 20.00 2.40 1.20 1.20 2 21.20 2.54 1.27 1.27 3 22.47 2.70 1.35 1.35 4 23.80 2.86 1.43 1.43 5 25.24 3.03 1.52 1.52 ASSUMPTIONS: Return on Equity = 12% Dividend Payout = 50% Cost of Equity = 10%
19
SCHEDULE 2
YEAR BEGINNING EARNINGS DIVIDENDS RETENTIONS BOOK VALUE PER SHARE PER SHARE PER SHARE PER SHARE 1 20.00 2.40 1.20 1.20 2 21.20 2.54 1.32 1.22 3 22.40 2.69 1.45 1.24 4 23.70 2.83 1.59 1.25 5 24.90 2.99 1.73 1.26 ASSUMPTIONS: Return on Equity = 12% Dividend Payout = 50% + 2% p.a. Required Return = 10%
20
SCHEDULE 3
Dividend and Long Canada Bond Yields
0
2
4
6
8
10
12
14
16
18
1956M01 1961M01 1966M01 1971M01 1976M01 1981M01 1986M01 1991M01 1996M01 2001M01 2006M01 2011M01
TSXYield Canadas
21
SCHEDULE 4
Canadian CPI Inflation back to 1914(all Items)
-20.00
-15.00
-10.00
-5.00
0.00
5.00
10.00
15.00
20.00
25.00
1915
-01
1916
-10
1918
-07
1920
-04
1922
-01
1923
-10
1925
-07
1927
-04
1929
-01
1930
-10
1932
-07
1934
-04
1936
-01
1937
-10
1939
-07
1941
-04
1943
-01
1944
-10
1946
-07
1948
-04
1950
-01
1951
-10
1953
-07
1955
-04
1957
-01
1958
-10
1960
-07
1962
-04
1964
-01
1965
-10
1967
-07
1969
-04
1971
-01
1972
-10
1974
-07
1976
-04
1978
-01
1979
-10
1981
-07
1983
-04
1985
-01
1986
-10
1988
-07
1990
-04
1992
-01
1993
-10
1995
-07
1997
-04
1999
-01
2000
-10
2002
-07
2004
-04
2006
-01
2007
-10
2009
-07
2011
-04
2013
-01
2014
-10
22
SCHEDULE 5
After Tax Profits % GDP
0.000
0.020
0.040
0.060
0.080
0.100
0.120
1926
1930
1934
1938
1942
1946
1950
1954
1958
1962
1966
1970
1974
1978
1982
1986
1990
1994
1998
2002
2006
2010
2014
Note: Statistics Canada issued a new revision of the GDP accounts starting in June 2012 where there was substantial revision to profits and dividends.
23
SCHEDULE 6
Aggregate Canadian Dividend PayoutMedian payout 41%, there were no aggreagte profits in 1932.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
1926
1930
1934
1938
1942
1946
1950
1954
1958
1962
1966
1970
1974
1978
1982
1986
1990
1994
1998
2002
2006
2010
2014
24
SCHEDULE 7
TSX Based Payout(Median 51.2%)
0
0.2
0.4
0.6
0.8
1
1.2
1956
M01
1957
M09
1959
M05
1961
M01
1962
M09
1964
M05
1966
M01
1967
M09
1969
M05
1971
M01
1972
M09
1974
M05
1976
M01
1977
M09
1979
M05
1981
M01
1982
M09
1984
M05
1986
M01
1987
M09
1989
M05
1991
M01
1992
M09
1994
M05
1996
M01
1997
M09
1999
M05
2001
M01
2002
M09
2004
M05
2006
M01
2007
M09
2009
M05
2011
M01
2012
M9
2014
-05
25
SCHEDULE 8
Capacity Utilisation
65
70
75
80
85
90
1987
-03
1988
-03
1989
-03
1990
-03
1991
-03
1992
-03
1993
-03
1994
-03
1995
-03
1996
-03
1997
-03
1998
-03
1999
-03
2000
-03
2001
-03
2002
-03
2003
-03
2004
-03
2005
-03
2006
-03
2007
-03
2008
-03
2009
-03
2010
-03
2011
-03
2012
-03
2012
-03
2013
-03
2014
-03
2015
-03
Non-Farm Manufacturing
26
SCHEDULE 9
Canadian Unemployment Rate
4
5
6
7
8
9
10
11
12
13
1987
M01
1988
M07
1990
M01
1991
M07
1993
M01
1994
M07
1996
M01
1997
M07
1999
M01
2000
M07
2002
M01
2003
M07
2005
M01
2006
M07
2008
M01
2009
-07
2011
-01
2012
-07
2014
-01
2015
-07
27
SCHEDULE 10
S&P500 Dividend Yield
0
0.01
0.02
0.03
0.04
0.05
0.06
0.0719
56
1958
1961
1963
1965
1967
1969
1971
1974
1976
1978
1980
1982
1984
1987
1989
1991
1993
1995
1997
2000
2002
2004
2006
2008
2010
2013
2015
28
SCHEDULE 11
S&P500 Dividend PayoutMedian 47.7%
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
1956
1958
1960
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2013
2015
29
SCHEDULE 12 S&P Electric UHC Data
EPS DPS PAYOUT RETAIN ROE g (B*ROE) YIELD US TSY K MB URP URP2 URP31993 7.95 7.11 89.43 10.57 11.25 1.19 5.73 5.87 6.99 1.59 1.11 1.10 3.251994 8.45 7.05 83.43 16.57 11.71 1.94 6.55 7.08 8.62 1.37 1.54 1.44 2.891995 9.23 6.97 75.51 24.49 12.36 3.03 6.23 6.58 9.45 1.39 2.87 2.55 3.061996 9.07 6.96 76.74 23.26 11.64 2.71 5.86 6.44 8.73 1.43 2.29 2.17 2.821997 7.63 6.64 87.02 12.98 10.16 1.32 5.49 6.35 6.88 1.49 0.53 0.67 2.531998 8.52 6.5 76.20 23.80 11.05 2.63 4.45 5.26 7.19 1.82 1.93 1.95 2.531999 9.31 6.24 67.02 32.98 12.36 4.08 4.60 5.64 8.87 1.69 3.23 2.81 2.322000 6.06 6.36 104.95 -4.95 7.04 -0.35 4.40 6.03 4.04 1.80 -1.99 -2.20 1.722001 10.58 5.42 51.23 48.77 13.63 6.65 3.41 5.00 10.28 1.88 5.28 4.03 1.722002 7.31 5.93 81.12 18.88 10.18 1.92 4.82 4.53 6.83 1.63 2.30 2.49 3.652003 8.44 5.29 62.68 37.32 10.61 3.96 4.31 4.02 8.44 1.51 4.42 4.63 3.642004 11.12 5.77 51.89 48.11 12.37 5.95 3.74 4.28 9.91 1.68 5.63 5.02 2.792005 10.22 6.85 67.03 32.97 11.86 3.91 3.69 4.31 7.75 2.04 3.44 3.20 2.722006 12.35 6.99 56.60 43.40 12.68 5.50 3.37 4.82 9.06 2.13 4.24 3.55 1.862007 14.82 7.85 52.97 47.03 12.81 6.02 3.09 4.54 9.30 2.20 4.76 3.95 1.862008 15.27 8.57 56.12 43.88 12.83 5.63 3.75 3.57 9.59 1.92 6.03 5.26 3.522009 13.37 8.8 65.82 34.18 10.53 3.60 5.01 3.27 8.79 1.38 5.52 5.74 5.112010 14.56 9.06 62.23 37.77 10.96 4.14 4.96 3.28 9.30 1.38 6.02 6.10 5.042011 13.94 9.49 68.08 31.92 10.1 3.22 4.70 3.16 8.07 1.47 4.91 5.26 4.892012 12.46 9.78 78.49 21.51 8.38 1.80 4.53 1.85 6.41 1.45 4.56 5.19 6.032013 11.52 9.52 82.64 17.36 7.47 1.30 4.18 2.35 5.54 1.48 3.18 3.85 5.182014 12.68 9.43 74.37 25.63 8.04 2.06 3.82 2.54 5.96 1.57 3.41 4.24 4.61
average 71.44 28.56 10.91 3.28 4.58 4.58 8.00 1.65 3.42 3.32 3.35Median 71.22 28.78 11.15 3.13 4.49 4.54 8.53 1.58 3.43 3.70 2.98
S&P Gas UHC Data1993 6.11 3.43 56.14 43.86 11.55 5.07 3.15 5.87 8.37 1.93 2.50 1.86 0.741994 7.21 3.82 52.98 47.02 12.29 5.78 3.57 7.08 9.56 1.78 2.48 1.43 -0.031995 5.25 4.02 76.57 23.43 8.28 1.94 3.45 6.58 5.45 1.75 -1.13 -0.67 0.341996 9.75 4.36 44.72 55.28 13.75 7.60 2.78 6.44 10.59 2.14 4.15 2.10 -0.211997 6.25 5.01 80.16 19.84 8.19 1.62 2.74 6.35 4.41 2.15 -1.94 -1.54 -0.161998 5.89 5.36 91.00 9.00 7.85 0.71 2.69 5.26 3.41 2.32 -1.85 -1.64 0.871999 7.4 9.34 126.22 -26.22 6.57 -1.72 3.84 5.64 2.05 1.99 -3.59 -4.56 1.692000 18.7 8.43 45.08 54.92 12.96 7.12 2.61 6.03 9.91 2.18 3.88 2.30 0.032001 9.87 8.16 82.67 17.33 7.33 1.27 2.47 5.00 3.77 2.38 -1.23 -0.73 0.912002 13.45 8.58 63.79 36.21 13.69 4.96 4.01 4.53 9.17 2.15 4.64 3.31 2.982003 14.77 7.23 48.95 51.05 13.82 7.06 4.24 4.02 11.59 1.57 7.57 5.61 3.722004 13.37 9.92 74.20 25.80 9.84 2.54 4.99 4.28 7.66 1.43 3.38 3.46 4.242005 10.42 19.06 182.92 -82.92 10.14 -8.41 9.05 4.31 -0.12 2.03 -4.42 -4.43 8.412006 8.26 8.89 107.63 -7.63 9.59 -0.73 3.94 4.82 3.18 2.62 -1.64 -1.69 2.612007 16.54 4.39 26.54 73.46 17.95 13.19 1.63 4.54 15.03 2.92 10.48 4.66 0.502008 19.61 4.21 21.47 78.53 18.46 14.50 1.60 3.57 16.33 2.48 12.76 6.13 1.452009 11.17 4.73 42.35 57.65 10.15 5.85 2.32 3.27 8.31 1.85 5.03 5.03 2.492010 12.04 7.48 62.13 37.87 9.7 3.67 2.91 3.28 6.69 2.07 3.41 3.58 3.092011 15.48 10.83 69.96 30.04 9.3 2.79 3.08 3.16 5.96 2.11 2.80 3.06 3.382012 18.22 13.92 76.40 23.60 10.96 2.59 3.39 1.85 6.06 2.47 4.21 4.02 5.012013 17.11 15.72 91.88 8.12 9.64 0.78 3.33 2.85 4.14 2.66 1.29 1.33 3.952014 54.24 24.16 44.54 55.46 13.3 7.38 3.91 3.85 11.58 1.58 7.73 5.91 3.56
average 71.29 28.71 11.15 3.89 3.44 4.66 7.41 2.12 2.75 1.75 2.25Median 66.88 33.12 10.15 3.23 3.24 4.54 7.17 2.13 3.09 2.20 2.09
URP assumes actual br growth, URP2 assumes that the expected ROE is the median value and URP3 also assumes a median retention rate. Source data is Standard & Poors Analyst's Handbook 2015.
30
SCHEDULE 13
US Electric Utility Dividend Yields and Treasury Yields(correlation +0.53)
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
Electrics US TSY
31
SCHEDULE 14
32
Wall St.'s woeful forecasting not getting better David Parkinson The Globe and Mail
PublishedFriday, May. 21 2010, 6:00 PM EDT
http://www.theglobeandmail.com/globe-investor/investment-ideas/wall-sts-woeful-forecasting-not-getting-better/article4353202/
Nearly a decade ago - about the time the bursting tech bubble had raised serious questions about conflicts of interest in Wall Street equity research - consulting firm McKinsey & Co. did a study on the accuracy of analysts' company earnings forecasts. The results were discouraging: Analysts were routinely over-optimistic about earnings growth, too slow to revise forecasts when economic conditions changed, and prone to increasingly inaccurate forecasts when the economy slowed.
Since then, major scandals involving tainted research have come to light, Wall Street's biggest firms have paid $1.4-billion (U.S.) in penalties for those practices, and regulators have put rules in place aimed at creating equity research with more independence and distance from the investment-banking side of the business. Unfortunately, McKinsey reports, the changes have had little effect on the accuracy of analysts' projections.
Downturn reveals same old habits In an update of the 2001 study, McKinsey researchers found that from 2003 to 2006, analysts' earnings projections actually did look less unrealistically rosy. In each of those years, analysts, on average, actually underestimated S&P 500 annual earnings for significant portions of the year - and undershot through the entire year in 2005 and 2006.
But lest we think this was evidence of a new kind of thinking within Wall Street research departments, the Street's wide-eyed optimism came back with a vengeance starting in 2007.
Going back over the past 25 years, McKinsey found that, on average, analysts' earnings-growth forecasts "have been nearly 100-per-cent too high." Annual S&P 500 consensus growth forecasts have typically been in the 10- to 12-per-cent range, while actual earnings growth has averaged 6 per cent.
Broken-clock accuracy Looking at five-year rolling average growth estimates, there have only been two periods in the past 25 years when the earnings met or exceeded analysts' forecasts. Both were in recovery periods after the U.S. recessions of the early 1990s and the early 2000s.
"This pattern confirms our earlier findings that analysts typically lag behind events in revising their forecasts to reflect new economic conditions," McKinsey researchers wrote. "When economic growth accelerates, the size of the forecast error declines; when economic growth slows, it increases."
This pattern means that when the analysts are accurate with their forecasts, it's sort of the same way a broken clock is accurate - twice a day.
"As economic growth cycles up and down, the actual earnings S&P 500 companies report occasionally coincide with the analysts' forecasts."
1
APPENDIX E 1
ALTERNATIVE MODELS TO THE CAPM 2
1. Introduction 3
In the 2012 GCOC the decision stated (page 64) that: 4
A number of experts raised concerns about the validity of the simple ‘single factor’ 5 CAPM. The term single factor refers to the reliance of the prediction of the model on 6 only the market portfolio, the factor implied by the theory. Ms. Ahern introduced in 7 evidence a summary article that shows how other factors such as firm market 8 capitalization (size) and market-to-book ratios can provide added explanatory power to 9 the single factor CAPM (Exhibit B2-7, Evidence of Ms. Ahern, pp. 13-14; Exhibit B2-7, 10 Attachment PMA-1, Exhibit PMA-9) 11
and concludes 12
“At this time the Panel is not persuaded that the specific model extensions that are 13 presented are valid and hence places no weight on them. However, the Panel 14 recommends that in the future improvements in the model can also be brought into 15 evidence, but the evidence should then include both the model extensions as well as a 16 basis on which to judge the validity of the extensions.” 17
This appendix is motivated by these comments and will discuss some basic principles in what 18
finance academics refer to as “asset pricing.” However, it should be mentioned that all of the 19
following is highly contentious for reasons that will be developed. Further it is my judgement 20
that the panel should place no reliance on them. Instead, I recommend continued reliance on the 21
CAPM conditional on the state of the economy or what is termed the conditional CAPM 22
(CCAPM). 23
2. A Brief History of the Capital Asset Pricing Model 24
A.D. Roy provided what is generally acknowledged as the first rigorous approach to pricing 25
capital assets.1 Roy envisaged an investor faced with the problem of investing in a set of assets 26
where they know the expected value and standard deviation of the return on those assets. Their 27
objective is to minimize the risk of a disaster scenario happening such that the actual rate of 28
return drops below some critical level (d). Roy derived what we would now call an efficient 29
1 A.D Roy, “Safety First and the Holding of Assets,” Econometrica, (July 1952)
2
frontier and showed how the investment varied depending on the value d. A key result from Roy 1
is not to evaluate investments based on their expected rate of return, but instead to take into 2
account their risk. Roy is generally regarded as the first to show that what matters is the 3
incremental expected return over this disaster level return divided by the level of risk or what has 4
become known, with slight changes, as the Sharpe ratio. 5
At the same time as Roy developed his safety first criterion, Markowitz2 also developed the 6
efficient set, but derived the optimal portfolio for an investor interested in expected return and 7
variance, since it implied the superiority of diversification. Markowitz’s approach is somewhat 8
more general than Roy’s since a “utility function” can encompass a broader range of behaviour 9
than Roy’s focus on a disaster level of return. Markowitz’s contribution was also more accessible 10
since it was graphical, whereas Roy was a Professor of Statistics at Cambridge and his paper is 11
“dense.” 12
Roy and Markowiz are usually regarded as the “fathers” of portfolio theory, but it was left to 13
William Sharpe as Markowitz’ student to add “general equilibrium” content.3 What I mean by 14
this is that both Roy and Markowitz were concerned with the actions of an individual investor, 15
whereas Sharpe derived the relationship between risk and return for an individual asset within a 16
general equilibrium, where all assets have to be held by someone. Sharpe showed that when 17
investors diversify the expected return on a portfolio can be defined by the Capital Market Line 18
or 19
20
where E(.) defines the expected value, σ the standard deviation and the subscripts on the rates of 21
return (R) denote the market, risk free rate or the portfolio. 22
2 H. Markowitz, “Portfolio Selection”, Journal of Finance, (March 1952). 3 William Sharpe “Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk,” Journal of Finance, (September 1964)
3
The CML states that the expected return on a portfolio is equal to the risk free rate plus a risk 1
premium comprising the market price of risk times the standard deviation of the portfolio’s rate 2
of return. In the CML the price of risk is the expected excess return on the market portfolio over 3
the risk-free rate divided by the standard deviation of the return on the market portfolio. The 4
market price of risk thus measures the incremental expected return per unit of incremental risk 5
similar to Roy, but instead using market not individual specific values. 6
In this equilibrium Sharpe showed that the expected return on any individual asset is determined 7
as 8
9
This is Sharpe’s celebrated ‘Capital Asset Pricing Model” or CAPM. It states that the expected 10
return on an individual asset is the risk free rate plus the expected excess of the market return 11
over the risk free rate, or what is now commonly referred to as the market risk premium, times 12
the security’s beta coefficient. The beta coefficient in turn is the correlation coefficient of the 13
return on the security with that on the market (ρ) times the ratio of the security’s standard 14
deviation to that of the market. 15
16
Low risk securities thus have low correlation with the market and/or low volatility because they 17
add very little risk to the portfolio in the sense that what variability risk they have is 18
diversifiable. 19
Both Sharpe and Markowitz were awarded the Nobel Prize in economics, but Roy was not and 20
neither were Jan Mossin4 nor John Lintner5 who developed almost identical models to Sharpe’s. 21
Probably this was because in some sense the Markowitz-Sharpe approach was more accessible 22
4 Jan Mossin “Equilibrium in a Capital Asset Market,” Econometrica, (October 1966) 5 John Lintner, “The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets,” Review of Economics and Statistics, (February 1965).
4
and had immediate implications that were quickly assimilated into both the academic literature 1
and the practice of finance.6 Two very important implications were: 2
1) The CML allows the evaluation of mutual fund performance. Rearranging the CML 3 allows us to judge the actual performance of any mutual fund against the market price of 4 risk, which has come to be called the Sharpe ratio. Although this is almost identical to 5 Roy’s result it is tied in with the capital market and is directly observable, whereas Roy’s 6 disaster rate of return is not. 7
2) As Sharpe noted the measure of risk is amenable to estimation ex ante (before the fact) as 8 a linear regression line and it is “the part of an asset’s risk which is due to its correlation 9 with the return on a combination” (ie., a diversified portfolio).7 10
In deriving the CAPM the rates of return have to be defined over some time horizon, since 11
conceptually the CAPM is a single period model of indeterminate length. However, in estimating 12
whether or not the model “works” the statistician has to choose a time period and normally a one 13
month period is used. For this reason the risk-free rate used in the tests is normally the yield on a 14
one month Treasury Bill. This means that in practice the statistical tests of the CAPM amount to 15
seeing whether the one month return on a security can be predicted in part by its beta coefficient. 16
A huge literature exists in formalizing tests of the CAPM that have involved some of the best 17
minds in statistics, since there is voluminous data on monthly rates of return. These huge data 18
sets excite statisticians! The standard procedure is to first estimate the betas on the universe of 19
securities by a regression of the actual return in excess of the risk free rate against the same for 20
the market return, normally over the preceding five year period. The securities are then usually 21
grouped into ten beta portfolios to minimize measurement error.822
6 Arguably Sharpe was first, but he derived the CAPM in a footnote, rather than being the focus of his research as it was with Lintner and Mossin. 7 Regression is a basic statistical technique that minimises the error around a straight line, so it estimates a line of best fit. The CAPM is one of the few economic models that don’t need to be linearized for estimation purposes. 8 I have testified on many occasions as to the instability of actual beta estimates, since they simply reflect what did or did not happen over the estimation period and may not as a result reflect the risk going forward. For this reason I use longer run averages and also look at industry betas.
5
After this first “pass” exercise, the betas on the portfolios are then measured against the excess 1
returns in a subsequent period in a second pass as follows: 2
3
The absolutely critical assumption here is that the actual return on a security is that which was 4
expected plus a random error term or 5
6
where the second term is this random error term. If this error term is completely random adding a 7
large number of securities into one of these beta portfolios should remove the portfolio’s error 8
term, such that, as best we can, we are estimating the relationship between the true beta and the 9
true expected rate of return. However, in practise both these assumptions are problematic as 10
betas are estimated with significant error and one month security returns are so volatile that 11
grouping them into beta portfolios rarely removes all the estimation error.9 12
To partly adjust for these problems this procedure, of estimating security betas, grouping them 13
into beta portfolios and estimating the coefficients in the second pass regression, is then repeated 14
each month to create a time series of coefficient estimates. In this way the volatility of any 15
individual estimate is reduced and standard significance tests can be performed on the estimates 16
from the second pass regression procedure. 17
The early tests of the CAPM were generally positive as tests showed that the market portfolio 18
was generally efficient, which is to say the coefficient on beta was positive and plausible. 19
However, there was early evidence that the estimated coefficients were not what some expected. 20
If the CAPM holds then the first coefficient, which is the intercept (γ1) should be equal to zero, 21
while the second coefficient (γ2) should equal the market risk premium. In fact, there was 22
9 If there is time variation in betas that is correlated with the market risk premium then a conditional CAPM better explains returns. For example, stocks that have lower betas when risk (utilities?) is high deserve a lower risk premium than predicted by the CAPM. It is well known that the conditional CAPM can hold when an unconditional or static CAPM does not.
6
evidence that the intercept was positive and the slope coefficient, (γ2), was lower than the historic 1
value for the market risk premium.10 In combination these observations implied that the true 2
risk-free rate was higher than the Treasury Bill yield, which led to a major contribution by Fisher 3
Black. 4
Black examined what would happen if there were restrictions on borrowing and lending or no 5
risk free asset at all.11 Black’s result was that the CAPM would still hold in terms of the critical 6
result that beta is a sufficient statistic to determine risk and expected rates of return. However, 7
the correct value for the risk free rate is the expected return on a minimum variance portfolio 8
with zero correlation with the market portfolio. He dubbed this the minimum variance, zero beta, 9
portfolio. Black’s model is then 10
11
where E(RZ) is the expected return on this minimum variance, zero beta, portfolio. To the extent 12
that borrowing rates exceed lending rates, so that the financial intermediary earns the spread, the 13
actual risk free rate must be higher than the Treasury Bill yield, which only the government can 14
borrow at and in which normally individuals do not directly invest in. The higher return on this 15
zero beta portfolio then causes a significant intercept in the second pass regressions and a lower 16
market risk premium than the historic values estimated using short term Treasury bill yields. 17
In rate of return evidence I have never used the Treasury Bill yield as the risk free rate in my 18
CAPM estimates. It is only useful as a risk free rate over a very short time horizon, which is not 19
consistent with the fact that equities have an infinite horizon and allowed ROEs are not set over 20
one month horizons. I regard the correct risk free rate to use for long term investing as that 21
embedded in the risk return tradeoff for investing in long term investments, so I use the yield on 22
the long Canada bond. It also has the following three advantages: it is free of default risk; it 23
10 This spawned a reconsideration of what is a reasonable value for the market risk premium, where R. Mehra and E. Prescott, “The Equity Premium a Puzzle” Journal of Monetary Economics (1985) regarded the historic values as being too high and inconsistent with simple capital market models. 11 F. Black, “Capital Market Equilibrium with Restricted Borrowing,” Journal of Business, (July 1972).
7
normally has a zero beta with respect to equity markets and has minimal variance risk.12 So I 1
regard the long Canada bond yield as satisfying the principles underlying Black’s minimum 2
variance, zero beta, portfolio. 3
At Schedule 1 is a graph of the beta of the return on the long Canada bond against the return on 4
the Canadian market. As the graph shows this is normally around zero if not slightly negative, 5
but was significantly positive in the 1990’s as Canada ran huge fiscal deficits and investing in the 6
long bond was regarded as extremely risky. For this reason in the 1990’s I used a low market risk 7
premium as measured over long Canada bond yields, since with a beta of about 0.50 the long 8
Canada bond was half as risky as the Canadian equity market and similar in risk to low risk 9
Canadian utilities. Normally the yield on the Long Canada bond is about 1.25% higher than the 10
yield on short term treasury bills, but varies with monetary policy and interest rate forecasts. 11
3. Anomalies 12
With the development of the CAPM, in the 1970’s financial researchers started testing whether 13
or not markets were efficient and whether the CAPM held. Note that these tests were joint tests 14
that both financial markets are efficient and the CAPM holds. For example, if there are 15
persistent deviations from market efficiency the CAPM may still hold but tests might find 16
irregularities,13 while similarly the CAPM may not hold but the market may still be efficient 17
based on another pricing model. With this caveat, however, starting in the late 1970s persistent 18
anomalies emerged. The following reflect the major types of anomalies, but there is a vast 19
literature documenting anomalies that many refer to simply as empirical regularities or as a 20
different way of viewing the same anomaly.14 21
The first set of anomalies arose as a result of forming portfolios based on price. Basu (1977)15 22
12 Note the spot yield on a long Canada bond might be marginally better, since it removes reinvestment rate risk, but in most cases it makes little difference. 13 For example, many believe the market is efficient and good at pricing on a relative basis but not on an absolute basis, which is why we have periodic crashes. Also there may be restrictions on trading such that prices are not efficiently formed. 14 Many of these anomalies are clearly related. 15 S. Basu, “The relationship between earnings Yield, Market Value and Return for NYSE Common Stocks,” Journal of Financial Economics, 1981.
8
found that portfolios of firms with low earnings yield (inverse of the price earnings ratio) 1
outperformed the market. Similarly Naranjo et al16 have found that dividend yields have a similar 2
effect while Fama-French use the ratio of book value to market value (inverse of the market to 3
book ratio), The common feature in all of these (and many more studies) is that low stock price 4
firms tend to have higher future returns, whether this stock price is used as the denominator for 5
earnings per share, dividend per share or book value per share. 6
A second set of anomalies revolved around size. Banz (1981)17 found that portfolios of small 7
firms in terms of total market value tend to earn more than predicted by the CAPM. This result 8
has been found by many others and is also reflected in the so called January effect documented 9
by Keim (1983)18 where the bulk of the small firm effect comes in January returns from very 10
small stocks. 11
The third major anomaly relates to momentum and over-reaction which together explain 12
bubbles. Jagedish and Titman19 were among the first to recognize in the academic literature that 13
creating a portfolio of winners based on past returns tends to earn excess rates of return in 14
subsequent 3-12 month periods. Although momentum was well known among technical 15
analysts20 using such rules as cross over points for short and long run moving price averages, it 16
was a surprise to those schooled on early efficient market tests.21 17
The existence of size, price and momentum based anomalies leads to the question of whether 18
there is something other than the market portfolio against which risk must be measured, that is, 19
whether there are other “factors” that drive security returns. In 1976 Ross provided the answer 20
16 A. Naranjo, M. Nimalendran and M. Ryngaert “Stock returns Dividend Yields and taxes”, Journal of Finance, (December 1998). 17 R. Banz, The Relationship between Return and Market Value of Common Stocks, Journal of Financial Economics, (March 1981) 18 D. Keim, “Size Related Anomalies and Stock Return Seasonality’, Journal of Financial Economics (June 1983). 19 N. Jegadeesh and S, Titman, "Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency". Journal of Finance 1993. 20 A famous phrase in finance is that “the trend is your friend” which essentially means momentum and is a clear violation of efficient markets 21 Early tests used auto-correlation tests and “runs” tests to see whether there was momentum and found nothing significant.
9
by deriving the arbitrage pricing model or APT.22 Ross’ model is extremely elegant, since rather 1
than deriving an asset pricing model from the optimal behavior on the part of investors and 2
equilibrium considerations, he assumes the absence of arbitrage opportunities and that investors 3
diversify. Ross starts from the proposition that there are multiple factors (δ) driving security 4
returns, so that the excess return over the zero beta return is driven from the surprises in these 5
factors, that is, deviations of the actual from the expected value for these factors and a security’s 6
exposure to that factor. That is, 7
8
It is a well-known fact that by holding a portfolio of randomly picked stocks the volatility of the 9
portfolio’s rate of return decreases as the number of stocks is increased, such that their returns 10
are highly correlated.23 What this means is that as long as there are more than 30-40 securities in 11
a portfolio all the risk in a portfolio stems from these common market wide factors and we can 12
ignore the random error term. Further since there are thousands of securities with different betas 13
we can target a particular exposure in a myriad of ways. However, the arbitrage assumption, that 14
there are no risk-free profits, means that all these strategies should generate the same rate of 15
return. 16
For example, suppose we have a factor we label as A, we can write the portfolio’s exposure as 17
AAAAA WWWW 44332211 18
where the W’s are simply the portfolio weights for four securities. If there are also four factors 19
we can solve directly for any desired exposure or beta to these four factors.24 However, in 20
practice, we have thousands of securities, so we can arbitrarily fix the weights on all of them 21
except for our holdings in four securities and then directly solve for the weights for these four, 22
since we are back to four equations and four unknown portfolio weights. Ross then invoked the 23
22 S. Ross, “The Arbitrage Theory of Capital Asset pricing” Journal of Economic Theory, 1976. 23 This can be derived from a simple mathematical model of naïve diversification as well as from simulation results using actual security rates of return. 24 Technically we have another equation that the portfolio weights sum to one.
10
no arbitrage assumption so that these different ways of forming portfolios with a particular beta 1
weight should give the same rate of return. 2
Since we can target any exposure we like, it is possible to create portfolio weights that are 3
unitary with respect to each factor and have zero exposure to the others. In this way we can 4
generate a risk premium for each factor and get the following APT equation: 5
6
where the lambdas (λ) are the risk premiums with respect to each factor. 7
Ross’ APT is the theoretical foundation for multi-factor models and has had both positive and 8
negative effects. The positive effect is that it liberates the CAPM from many of the assumptions 9
required to generate it. For example, if there is only one factor the APT degenerates to Black’s 10
model and if there is a risk-free rate to Sharpe’s CAPM. Consequently, the justification for the 11
CAPM relies on the importance of the market portfolio, the observation that investors diversify25 12
and the absence of risk-free profits, that is, arbitrage opportunities. This means the CAPM is 13
more general than originally assumed with its more formal derivation. 14
The negative effect is due to the fact that it is so general; in particular, the factors are not 15
specified. All that APT says is that quite generally returns are generated by a set of factors each 16
of which attracts a risk premium, but there are no economic implications. The result is that the 17
APT has generated a huge literature dealing with factors, many of which result from mining the 18
data tapes over and over to find commonalities, which must exist at some time.26 Further it has 19
been a boost to industry professionals that can now sell proprietary factor models to their clients, 20
rather than actively managed portfolios that have proven difficult to beat the market based on 21
their Sharpe ratios.27 22
23
25 More formally it is the marginal investor that counts and the assumption is this is a large institution. 26 For example certain economic factors move the stock market at some time but not others, such as trade deficits or government fiscal deficits, but they are only relevant if investors require a risk premium for bearing those risks. 27 In Malkiel’s classic work, A Random Walk Down Wall Street,1973, he suggested that what was needed was a low cost index fund; a challenge that Jack Bogel took up in founding Vanguard.
...)( 2,21,1 jjZj RRE
11
4. Multi-Factor Models 1
The APT has spawned three literatures. The first is that there are statistical techniques called 2
factor analysis and principal components that can be used to “mine” all the security returns to see 3
whether there is common movement. A principal components analysis reduces all the correlated 4
variables to a smaller sub-set of independent “composite” variables.28 If the CAPM holds there 5
would only be one composite variable, but with a multi-factor model there would be more than 6
one. The empirical results are somewhat mixed, some researchers have found only one 7
meaningful factor, whereas others find up to three or four. However, the importance of each 8
component usually drops off quite quickly. The problem is that since it is data driven, much 9
depends on the time period that is used and what happened during that period. Also the literature 10
has little positive value, since it does not explain what investors should do with the results, that 11
is, knowing a few factors drives the market does not help unless we know what they are! 12
The second literature involves pre-specifying the economic variables that drive the market. We 13
know for example that equities are valued based on the discounted value of their stream of cash 14
flows, which in turn are affected by interest rates, the state of the economy, central bank policy 15
and risk aversion (sentiment) etc. So it makes sense to test APT by using pre-specified economic 16
variables. 17
The first such model by Chen, Roll and Ross29 used four pre-specified economic factors: 18
Unexpected inflation 19
Unexpected changes in GDP (production index) 20
Changes in default spreads 21
Changes in the spread between long and short term bond yields 22
In their view the fundamental determinants of security rates of return were unexpected changes 23
in inflation, economic growth, default risk and the shape of the yield curve. However, their 24
results were mixed, even though the default spread between BBB and AAA rated corporate debt 25
28 Technically factor analysis is more general than principal components, since it estimates an underlying structural model, whereas principal components is a mathematical transformation. 29 N Chen, S. Ross and R. Roll, “An Empirical Investigation of the Arbitrage Pricing Theory,” Journal of Finance, 1980.
12
was generally significant as was the change in GDP. Of importance was that when the return on 1
the market was added to these four economic variables it was not highly significant. However, as 2
Sharpe later showed the CAPM can be given a multi-beta or multi-factor representation, since 3
the market rate of return, as the sum of all the individual securities, must be driven by the same 4
pre-specified economic factors. 5
The pre-specified economic factor approach has spawned a huge professional literature. The 6
main reason is that it ties in with basic macro-economic forecasting. One group will analyse the 7
state of the economy and forecast the pre-specified economic factors and then a portfolio will be 8
structured or recommended to take into account this forecast. One of the best money managers, 9
Blackrock, uses a multi-factor approach. At Schedule 2 is a summary of Blackrock’s pre-10
specified economic factors and their performance analysis as of the Summer of 2014. In their 11
analysis they use six factors: 12
Real interest rates 13
Inflation 14
Credit 15
Economic 16
Emerging markets 17
Liquidity 18
Of these six factors, four are clearly in the spirit of Chen Roll and Ross. Moreover, some 19
investment professionals tout what has come to be known as “risk parity” that a portfolio should 20
be constructed to have equal risk (parity) regarding these basic economic factors without ever 21
explaining why that composition is optimal. 22
The final strand of this literature involves a combination of the ideas behind principal 23
components analysis and the anomalies literature and is due to the path breaking work of Fama 24
and French.30 Principal components analysis indicates that the factors are different clusters of 25
stock market returns and there are relatively few of them, whereas the anomalies literature 26
indicates that there are securities with particular characteristics that earn excess returns. When 27
equities outperform bonds we don’t think of it as an anomaly, but a return for bearing greater risk 28
30 E. Fama and K. French, “Common Risk Factors in the Returns on Stocks and Bonds,” Journal of Financial Economics (1993).
13
and earning the equity risk premium. Fama-French used the same logic for the anomalies: if the 1
markets are efficient the excess return from investing in stocks with low earnings yields, low 2
market capitalization or good momentum must be due to their greater risk. 3
Consequently Fama-French formed factor “mimicking” portfolios based on securities issued by 4
firms with these different characteristics. A factor mimicking portfolio is simply a portfolio of 5
long positions in securities with a high value for this characteristic and a short position in 6
securities of firms with low values. Similar to principal components analysis, they then held this 7
portfolio constant for the next twelve months to create a time series of rates of return for that 8
particular factor. 9
Fama-French used a variety of sorting mechanisms to build portfolios of securities based on 10
different exposures to these various factors, much like the original portfolios constructed with 11
different average betas, and then did second pass regression tests on the market pricing of these 12
factors. Fama-French estimated the following three factor model: 13
14
where MKT is the normal excess return on the market over the risk free rate, SMB is the return 15
on a portfolio of small minus large NYSE stocks and HML is the return on a portfolio of high 16
minus low stocks based on the book to market, that is, the inverse of the market to book ratio. In 17
the second pass results Fama-French found that the intercept was insignificant and the coefficient 18
on the market was consistent with a 5% market risk premium. On their own these results would 19
have been broadly consistent with the CAPM. However, they found that the coefficient on SMB 20
was consistent with a risk premium of just over 3%, while that on HML was a very large 4.8%. 21
Consistent with the anomalies literature Fama-French documented the excess return earned on 22
small value stocks, which they characterised as a risk premium. 23
Fama-French’s results remain controversial. Most people would accept that historically small 24
stocks might be riskier, since they have less market power than large ones, tend not to be covered 25
in depth by security analysts and are less liquid so it is difficult for large institutions to hold 26
HMLSMBMKTRRE jjjFj ,3,2,1)(
14
positions in them.31 However, firms with low book to market ratios tend to be value stocks and it 1
is difficult to rationalise value stocks as being riskier than glamour or growth stocks, particularly 2
when there is more to lose on glamour stocks with higher market to book ratios simply because 3
their stock prices are so high.32 Instead, many believe that the Fama-French results are consistent 4
with the persistent market mis-pricing of glamour stocks, since there is no significant difference 5
in their stock market behaviour, that is, the variability of glamour stocks looks much the same as 6
for value stocks.33 7
At Schedule 3 is the Fama-French taxonomy used by Morningstar to evaluate mutual funds 8
based on their style. As is clear from the taxonomy Morningstar regards large glamour/growth 9
stocks as the least risky and small value stocks as the most risky. Consequently the Fama-French 10
analysis has had a significant effect in the financial markets. Whether it makes a significant 11
difference in cost of capital estimation is another question. 12
At Schedule 4 is a table from a Fama-French versus CAPM cost of equity capital analysis of the 13
Dow-Jones 30 in the US.34 The average CAPM equity cost is 9.7% versus 9.5% for the Fama-14
French three factor model. Of interest, is that the average beta for these stocks is 0.95, and the 15
exposures to size (SMB) and value (HML) -0.22 and +0.19 indicating that the DJIA30 are 16
generally large value stocks with a bit less than average exposure to the stock market. For the 17
low risk stocks Wal-Mart has a three factor cost of equity of 4.8% and a CAPM estimate of 18
5.5%, Kraft Foods 7.1% and 7.2%, J. P Morgan 10.2% and 10.0%, and Home Depot 9.1% and 19
8.1%. Only J.P Morgan’s values look strange, since its Fama-French market beta is half the size 20
of its simple beta.35 21
31 Some claim the small firm effect is no longer obvious, since these causes have receded, so the “outlook” for the size effect is bleak. M. Dijk “Is Size Dead? A Review of the Size Effect in Equity Returns” Journal of Banking and Finance, 2011. It is also mostly driven by penny stocks and may be driven by market micro-structure concerns.. 32 As Canadians we can think of our two big glamour stocks as Nortel and RIM, both of which generated low ex post rates of return as they flirted with bankruptcy, but neither would be regarded as low risk! 33 K. Daniel and S. Titman, “Evidence on the Characteristics of Cross-Sectional Variation in Common Stock Returns, Journal of Finance, 1997. 34 Estrada, Journal of Applied Corporate Finance (Spring 2011). 35 Looking at any US bank stocks at this time is difficult due to the strain the US financial system was under during the estimation period.
15
The fact that the Fama-French three factor estimates on average are similar to the CAPM 1
estimates should not be a surprise. If the CAPM holds, the factor risk premiums are simply 2
3
or that the factor’s risk premium is simply the exposure of the factor to the market times the 4
market risk premium, in this case using the Black model. If we then group this expression for 5
two factors we get 6
7
which is the Black version of the CAPM, except that the beta has been broken out to show its 8
factor determinants. In this case the beta is the sum of the security’s exposure to each factor and 9
each factor’s exposure to the market. This is the approach of Sharpe and others; that multi-factor 10
models can be regarded as “better beta” models. 11
Of interest is that long before Fama-French there were services offering better beta estimates to 12
solve the problem of beta instability over particular time horizons. Beaver, Kettler and Scholes36 13
for example estimated the following, 14
15
where the X’s are the independent variables used to explain beta to make better forecast 16
estimates. Examples of these predictive values include dividend payout; growth; leverage; 17
liquidity; size, earnings variability, industry dummy variables and book to market. These are the 18
same types of variables now being used to create factor mimicking variables similar to the Fama-19
French factors. To the extent that these accounting variables are more stable than beta estimates, 20
they may overcome the serious measurement problems evident in asset pricing tests. 21
One way of estimating the market risk in the Fama-French factors is simply to estimate their 22
factor betas in the same way as for normal securities. This is a straightforward exercise since the 23
36 W. Beaver, P. Kettler and M. Scholes “The Association between Market Determined and Accounting Determined Risk Measures,” Accounting Review, (October 1970)
])([11 ZM RRE
])()[()( 2211 ZmZj RREbbRRE
.......2211 XaXaaj
16
data is on Ken French’s web site and readily available.37 At Schedule 5 are these factor beta 1
estimates for the two Fama-French factors using five years of monthly data updated each month 2
with a new observation. The SMB factor clearly has persistent market risk at about 0.3 whereas 3
the HML factor has not been stable. At times it has been growth stocks that have significant 4
market risk and at other times value stocks. However, it seems that the HML factor beta started 5
out positive in the 1930’s and then went into a secular decline until the time of the internet 6
bubble when it changed from negative to positive. This observation that the market is sometimes 7
value and sometimes growth oriented seems to support the Daniel and Titman argument that it is 8
characteristics, not risk, that is the driving force behind the HML factor.38 9
After Fama-French the search for factors became a major academic exercise. One of their 10
students (Carhart) augmented the Fama-French three factor model to include a momentum 11
factor. Others have included liquidity, short-term reversal, long-term reversal, accounting 12
accruals, net share issues, total variance, idiosyncratic risk, investment and the latest “flavour of 13
the month” which is “quality” as estimated by operating profitability. Mimicking portfolios for 14
all these factors are on Ken French’s web page. 15
However, what has to be emphasised is the following: 16
Finding exposures does not necessarily mean there is a factor risk premium. Currently 17 the market is dominated by the actions of central banks, but this is not normal it 18 simply reflects the current economic environment; 19
The one constant is the historic market risk premium. When we add up all the factor 20 risk premiums they have to sum to the overall historic market return. It is self-evident 21 that summing over all the securities in the market gives the market return and over 22 time the market risk premium.39 23
Whether multiple factors changes the return estimate of different securities, that is 24 for example, whether low beta securities deserve higher than their CAPM risk 25 premium depends in practise on how their beta is estimated and their exposure to the 26 size and value premiums. For example, these asset pricing tests use the most recent 27 unadjusted beta estimate, but I am not aware of anyone that actually does this! In the 28
37 http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html 38 Fama-French (page 176) “Industry Equity Costs, Journal of Financial Economics, (1997 ) recognised that the risk premiums attached to the SMB and HML factors were not as predictable, that is, stable as the market risk premium and (page 183) that “industry slopes on HML wander through time”. 39 We have to do this since any consistent model has to recognise that all securities have to be held.
17
past, for example, I have used a beta variant based on the work of Beaver, Kettler and 1 Scholes and others.40 2
The CAPM is still the only useful equilibrium model, since all securities have to be 3 held and when we sum over all securities the market portfolio is still pivotal. 4
In terms of mutual fund performance as far as I am aware the Sharpe Ratio is still the 5 dominant statistic used in the profession. I am not aware, for example, of a residual 6 from a Fama-French model being widely used to judge fund performance. 7
5. Example Application for a Canadian Utility 8
To consider whether multi-factor models are useful in a Canadian regulatory context I will repeat 9
the above analysis and historical development of multi-factor models using the main sub-indexes 10
of the TSX. At Schedule 6 is a factor analysis of these sub-indexes using monthly returns since 11
1988. Using a principal axis routine that allows a residual error the model estimated three main 12
factors. One of the nuances with factor analysis is choosing how many factors to use in the 13
analysis. This is done by looking at a “scree plot” which measures the amount of the variance or 14
uncertainty that is explained by estimating an additional factor, that is, it suggests where to stop 15
the analysis. In this case the scree plot is as follows: 16
eValue % Cum %3.802101 61.08% 61.08%1.084623 17.42% 78.50%0.609839 9.80% 88.30%0.32146 5.16% 93.47%
0.199134 3.20% 96.66%0.1205 1.94% 98.60%
0.054773 0.88% 99.48%0.038605 0.62% 100.10%-0.00014 0.00% 100.10%-0.00608 -0.10% 100.00% 17
The first estimated factor explains 61.08%; the second explains another 17.42% and the third 18
another 9.80%, so together the first four factors explain almost 90%. Normally the 19
recommendation is to stop adding factors when the “eigenvalue” drops below 1.0, so in this case 20
the third and certainly the fourth factor become questionable and once we are down to the fifth it 21
40 At times both I and my late colleague Professor Berkowitz have used such an “instrumental variables” approach to estimate the equity cost for a unique non-traded utility. See for example M. Berkowitz “Estimating the Market Risk for Non-Traded Securities: An Application to Canadian Public Utilities, International Review of Financial Analysis, Spring 1998.
18
is essentially explaining random noise. This confirms the general belief in finance that there are 1
only a few factors that drive security returns and there is one major factor (the market) as in our 2
case. 3
If we look now at Schedule 6, we can see how the first factor is composed. These weights were 4
estimated using regression and other techniques would give slightly different results, but it 5
indicates that the first factor is based on the Industrial (IND), Consumer Discretionary 6
(CONSD), Health (Health), Finance (FIN), Information Technology (IT) and 7
Telecommunications (TELCO) sub-indexes, that is, six out of the ten sub-indexes. The negative 8
weights are for energy (ENGY, Materials (MAT), Consumer Staples (CONSS) and Utilities 9
(UTE). If the weights had all been positive we would say immediately that this is the market 10
factor, since all securities have to be held. The second factor has very big negative weights on 11
energy and materials, so it looks like a “commodity price” factor which is what mostly affects 12
energy and materials stocks. The third weights Consumer Staples, and Utilities very heavily, so 13
the factor “looks” like a “defensive” or “interest sensitive” factor. 14
The return series for each of these three factors is then creating by multiplying the weights in 15
Schedule 6, for example, -10005 for energy in the first factor times the return for that sub index, 16
the weight of -.08714 for materials, times the return on the materials sub index etc. In this way 17
we can create the rates of return on the three factors that explain a large proportion of the return 18
on each of the ten sub-indexes. This reduces the dimensionality of the problem from ten sub-19
indexes to three factors. However, the drawback is the need to now forecast the returns and risk 20
premiums on these three factors when we don’t know what they represent! 21
I looked at three factors that might determine these rates of return: the excess return on the TSX, 22
the return on the Government of Canada long bond (GOC) and the return on the Bank of 23
Canada’s commodity price index (COMM). These are factors that might affect the Canadian 24
market, but were chosen simply for illustrative purposes. 25
19
The following table provides the results of a regression analysis of these pre-specified factors to 1
see whether they are explained by the three factors. Under each coefficient is the T statistic that 2
tells us how significant the coefficient is. 3
Table 1 4
Factor 1 Factor 2 Factor 3 Adjusted R Square 5 Market 0.42 -.62 0.21 81.2% 6 36.6 -26.2 10.9 7 Commodity Prices -0.19 0.08 -0.09 1.5% 8 -1.5 2.4 -2.4 9 GOC bond 0.13 -0.07 0.07 4.1% 10 0.86 -2.5 2.87 11
So for the return on the TSX (Market) there is a very large T statistic on all three factors, 12
indicating that the return on the TSX is being driven by these three factors. We would expect this 13
since the factors were in turn derived from the sub-indexes of the TSX! For commodity prices 14
and the return on the government bond, we see that both the second and third factors are 15
relevant, but not the first. For commodity prices there is a positive weight on the second factor 16
and for the return on the long Canada bond the opposite. So the initial suspicion is partially 17
supported that the first factor is mainly related to the overall market, the second exposure to 18
commodity prices and the third to interest rates.41 However, the statistical results are not exactly 19
strong, which is the main drawback of both these types of pre-specified variables and factor 20
analysis generally. 21
In Schedule 7 are the results of a more conventional Canadian model explaining utility returns. 22
The first panel is a simple regression of the excess return on the utilities sub index against the 23
Canadian excess return on the market (TSX Composite) for the period 1988-2014. The adjusted 24
R square indicates that the Canadian market explains just less than 11% of the utilities’ return. 25
This low value for the R square is normal for low risk securities as the adjusted R square is 26
simply the explained uncertainty divided by the total uncertainty. As the return becomes more 27
certain, the R square of course goes down until in the limit it is zero for the risk-free security. 28
41 We would not expect a high explanatory power for the interest rate and commodity price estimates since they are affected by other variables. The estimates simply show there is a relationship.
20
The second panel has the return on the long Canada bond added as another explanatory variable. 1
As expected the adjusted R square goes up to 18% as the utility return is interest sensitive similar 2
to the long Canada bond. Note the stock market risk (beta) of the utility index over this long 3
period has been 0.26-0.29, but it is a very long period and includes the Nortel years. Schedule 8 4
has the beta estimated every month over a five year estimation period with the beta re-estimated 5
every month. This estimate is against three market proxies: the Canadian market index, and the 6
US market index both hedged and unhedged against changes in the foreign exchange rate.42 7
The point of the analysis in schedules 7 and 8 is to set up a CAPM benchmark so that we can 8
then analyse the usefulness of a Fama-French three factor model for Canadian utilities. In 9
Schedule 9, are the coefficient estimates for the US excess market return, SMB and HML factors 10
both adjusted and unadjusted for foreign exchange risk. The adjusted R squares are 16% for the 11
hedged model and 12% for the unhedged model as the extra foreign exchange uncertainty 12
increases the volatility of the US parameters. Looking at the hedged model, the estimated beta on 13
the US market return is 0.31, on the size factor -0.13 and the HML factor 0.29. This would 14
indicate slightly higher market risk than against the Canadian market index over the same period 15
and that these utilities as expected exhibit the characteristics of large, value, stocks, since the 16
coefficient on size is negative and on HML positive. This would accord with the intuition that 17
utilities generally are capital intensive and are relatively large, but regulated to have high book to 18
market or low market to book ratios.43 19
As a final check I estimated the models over the most recent five year period in Schedule 10. The 20
Fama-French model has an adjusted R Square of 14%, and the coefficient on the US market is 21
0.36 and on the size factor -0.42, while that on HML is insignificant. This again indicates low 22
market risk for a relatively large firm. In the second panel in Schedule 9 are the results for the 23
two-factor model where the adjusted R Square is 21% and the betas on the market and long 24
Canada bond 0.45 and 0.44 respectively. 25
In comparing the “two factor” Canadian market estimates with the Fama-French estimates my 26
42 Investing in the US involves a Canadian investor bearing the uncertainty attached to the C$:US$ exchange rate. 43 Note that whereas the beta on the market return averages out to 1.0 over all securities, there are no such restrictions on the betas for SMB and HML
21
overall conclusions are: 1
The Canadian and US markets are more highly integrated now than they have been 2 for a long time, so the coefficient on the US market of 0.31-0.35 is similar to the beta 3 coefficient against the Canadian market. It is actually a commonplace observation 4 that global capital markets over the last 5 plus years have all been driven by the same 5 post Financial Crisis factors. 6
Utilities are relatively large value stocks as confirmed by the negative beta against 7 size and positive beta against value in typical Fama-French estimates. This means that 8 they get a negative premium for size and a positive premium for value relative to the 9 simple CAPM. 10
The two factor model with the Canadian market return and the return on the long 11 Canada bond confirms that Canadian utilities have low market exposure and are 12 interest sensitive. This accords with the intuition that utilities behave like convertible 13 bonds and have defensive characteristics in down markets. 14
If we take the latest Fama-French risk premiums I would regard the US beta of 0.36 15 against the Fama-French estimate of the US market risk premium of 5% as 16 warranting a 1.8% risk premium. The size beta of -0.42 and the size premium of 3.0% 17 then warrants a subtraction of 1.26%. Since the HML coefficient is insignificant and 18 negative together these two estimates imply a risk premium of 0.54%, which I regard 19 as unreasonably low and unreliable. 20
In contrast, the two factor model is for a beta of 0.45 against a Canadian market risk 21 premium of 5-6.0% or 2.5% plus an interest rate or term premium of 0.44 against the 22 current term premium of about 1.75% or 0.77%. Combining these two gives a more 23 reasonable risk premium of 3.27%.44 24
Personally, I don’t support the use of the Fama-French model. I tend to be in the camp that 25
believes it is the result of data mining with no firm theoretical basis. So without compelling 26
empirical evidence to the contrary, which I don’t see, I would not recommend its use in a 27
Canadian context for a regulated utility.45 28
44 This would be slightly higher using Treasury Bills as the risk free rate in the market risk premium. 45 Note the following analysis is illustrative and is not mean to meet journal standards of empirical research.
22
Schedule 1
Canadian Bond Beta
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
2014
23
Schedule 2
Industry Pre-specified Factors
Source: Blackrock Market Advantage
24
Schedule 3
25
Schedule 4
Fama-French Comparisons for the DJ 30
26
Schedule 5
Market BetasRolling 5 year monthly betas
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1931
0
1933
0
1935
1
1938
0
1940
0
1942
0
1944
1
1947
0
1949
0
1951
0
1953
1
1956
0
1958
0
1960
0
1962
1
1965
0
1967
0
1969
0
1971
1
1974
0
1976
0
1978
0
1980
1
1983
0
1985
0
1987
0
1989
1
1992
0
1994
0
1996
0
1998
1
2001
0
2003
0
2005
0
2007
1
2010
0
2012
0
SMB HML
27
Schedule 6
Composition of the Three Canadian factors
1 2 3ENGY -0.10005 -0.71502 -0.20046MAT -0.08714 -0.69138 -0.20488IND 0.239009 -0.1349 -0.03352CONSD 0.28004 0.144925 0.148573CONSS -0.07432 0.253175 0.608837HEALTH 0.251445 0.014204 -0.10763FIN 0.102725 0.014789 0.250277IT 0.454502 -0.04964 -0.40871TELCO 0.272657 0.197583 0.046317UTE -0.26262 0.015156 0.62311
28
Schedule 7
SUMMARY OUTPUT
Regression StatisticsMultiple R 0.3293648R Square 0.10848117Adjusted R Squ 0.10571248Standard Error 3.48760527Observations 324
ANOVAdf SS MS F Significance F
Regression 1 476.5784043 476.5784 39.18138 1.232E-09Residual 322 3916.611739 12.16339Total 323 4393.190144
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%ower 95.0%Upper 95.0%Intercept 0.7241534 0.194738372 3.718596 0.000236 0.3410332 1.107274 0.341033 1.107274TSX 0.2934313 0.046877732 6.259503 1.23E-09 0.201206 0.385657 0.201206 0.385657
SUMMARY OUTPUT
Regression StatisticsMultiple R 0.43098078R Square 0.18574443Adjusted R Squ 0.18067119Standard Error 3.33824227Observations 324
ANOVAdf SS MS F Significance F
Regression 2 816.0106179 408.0053 36.61256 4.754E-15Residual 321 3577.179526 11.14386Total 323 4393.190144
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%ower 95.0%Upper 95.0%Intercept 0.55620611 0.188866043 2.944977 0.003466 0.1846345 0.927778 0.184635 0.927778TSX 0.25670035 0.045361006 5.659053 3.37E-08 0.1674579 0.345943 0.167458 0.345943GOC Bond 0.43615412 0.079028058 5.518978 7.04E-08 0.2806758 0.591632 0.280676 0.591632
29
Schedule 8
Can Utility Betas
-0.40
-0.20
0.00
0.20
0.40
0.60
0.80
12/3
1/19
92
12/3
1/19
93
12/3
1/19
94
12/3
1/19
95
12/3
1/19
96
12/3
1/19
97
12/3
1/19
98
12/3
1/19
99
12/3
1/20
00
12/3
1/20
01
12/3
1/20
02
12/3
1/20
03
12/3
1/20
04
12/3
1/20
05
12/3
1/20
06
12/3
1/20
07
12/3
1/20
08
12/3
1/20
09
12/3
1/20
10
12/3
1/20
11
12/3
1/20
12
12/3
1/20
13
12/3
1/20
14
US Hedge Can Index US No hedge
30
Schedule 9
Fama-French Hedged and Unhedged for Canadian Utilities
Regression StatisticsMultiple R 0.3990953
R Square 0.1592771
Adjusted R 0.1513953
Standard Er 3.4113283 Hedged
Observation 324
ANOVA
df SS MS F ignificance FRegression 3 705.5005 235.1668 20.20827 5.11E-12
Residual 320 3723.891 11.63716
Total 323 4429.392
Coefficientstandard Erro t Stat P-value Lower 95%Upper 95%ower 95.0%Upper 95.0%Intercept 0.2349951 0.193361 1.215318 0.225141 -0.14542 0.615414 -0.14542 0.615414
MKT-RF 0.3138317 0.046555 6.74113 7.34E-11 0.22224 0.405424 0.22224 0.405424
SMB -0.139855 0.062622 -2.23331 0.02622 -0.26306 -0.01665 -0.26306 -0.01665
HML 0.2932924 0.067045 4.374554 1.65E-05 0.161388 0.425197 0.161388 0.425197
Regression StatisticsMultiple R 0.3537087
R Square 0.1251098 Unhedged
Adjusted R 0.1169077
Standard Er 3.4799566
Observation 324
ANOVA
df SS MS F ignificance FRegression 3 554.1605 184.7202 15.2534 2.67E-09
Residual 320 3875.231 12.1101
Total 323 4429.392
Coefficientstandard Erro t Stat P-value Lower 95%Upper 95%ower 95.0%Upper 95.0%Intercept 0.3260425 0.195736 1.665723 0.096747 -0.05905 0.711135 -0.05905 0.711135
MKT-RF 0.2636696 0.045961 5.736816 2.24E-08 0.173246 0.354093 0.173246 0.354093
SMB -0.249773 0.057853 -4.31739 2.11E-05 -0.36359 -0.13595 -0.36359 -0.13595
HML 0.1435248 0.056607 2.535458 0.011706 0.032156 0.254894 0.032156 0.254894
31
Schedule 10
Two factor vs Fama-French for Utilities (last 5 years)
Regression StatisticsMultiple R 0.433347R Square 0.1877896Adjusted R 0.1442784Standard Er 2.5765545Observation 60
ANOVAdf SS MS F ignificance F
Regression 3 85.95473 28.65158 4.315885 0.008294Residual 56 371.7634 6.638633Total 59 457.7182
Coefficientstandard Erro t Stat P-value Lower 95%Upper 95%ower 95.0%Upper 95.0%Intercept 0.2381774 0.355324 0.670311 0.505415 -0.47362 0.949977 -0.47362 0.949977MKT-RF 0.3554792 0.101244 3.511129 0.000889 0.152664 0.558294 0.152664 0.558294SMB -0.416617 0.180191 -2.31208 0.024474 -0.77758 -0.05565 -0.77758 -0.05565HML -0.04054 0.188204 -0.21541 0.830233 -0.41756 0.336477 -0.41756 0.336477
Regression StatisticsMultiple R 0.4913992R Square 0.2414731Adjusted R 0.2148582Standard Er 2.4680112Observation 60
ANOVAdf SS MS F ignificance F
Regression 2 110.5266 55.26332 9.072829 0.000379Residual 57 347.1915 6.091079Total 59 457.7182
Coefficientstandard Erro t Stat P-value Lower 95%Upper 95%ower 95.0%Upper 95.0%Intercept 0.0793472 0.348243 0.22785 0.820578 -0.618 0.776692 -0.618 0.776692TSX 0.4508925 0.115232 3.91291 0.000246 0.220144 0.681641 0.220144 0.681641GOC Bond 0.4422894 0.15311 2.888694 0.00546 0.135691 0.748888 0.135691 0.748888
![BOOTH’S ALGORITHM DESIGN USING FIELD ... ALGORITHM DESIGN USING FIELD...multiplication [3]. This project is to modified booth’s algorithm in order to improve the performance which](https://img.dokumen.tips/doc/110x75/5aa19ecc7f8b9a80378be9ad/booths-algorithm-design-using-field-algorithm-design-using-fieldmultiplication.jpg)
