Upload
sckchris
View
215
Download
0
Embed Size (px)
Citation preview
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
1/36
Equilibrium ExchangeRates and Exchange Rate
ForecastingRonald MacDonald
University of Strathclyde
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
2/36
Introduction
Main theme of lecture: macroeconomicfundamentals useful for determination ofexchange rates at long (equilibrium) and
medium-run (forecasting) horizons.
Controversial as seems to go againstcurrent conventional wisdom (CCW).
What is CCW?
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
3/36
Introduction
CCW argues for the abandonment ofmacroeconomic fundamentals foranalysing exchange rate movements.
CCW takes as its starting point dailyvolume of global foreign exchangetransactions = $1.2 trillion
Given on daily basis macro-fundamentalsdont change much - How explain $1.2T?Move to Market Microstrucure (MM).
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
4/36
Introduction
MM focuses on institutional features of theForex inter-bank behaviour; inter-bank / broker.Two key MM variables are bid-ask measure of
transaction costs - and order flow measure ofinformation flow.
Theoretical MM literature focuses on det. of B-Aand influence of order flow on volatility.
Highlights Heterogeneity vs. Homogeneity ofexpectations Important.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
5/36
Introduction
Empirical literature provides support forinfluence of order flow on volatility-volumeand B-A. So perhaps MM helpful for
explaining high frequency volatility. But time frame day so not helpful for
predictability or equilibrium.
The new CCW arose because of theexistence of the so-called Exchange RateDisconnect. This has three aspects:
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
6/36
Introduction
1. Volatility: Exchange rates when flexible areexcessively volatile. Intra- and inter-regimeaspects.
2. Level: Exchange rates unpredictable athorizons of < 3 years - relates to forecasting andMeese and Rogoff Random walk result.
3. PPP Puzzle: If PPP taken as measure ofequilibrium then equilibrium is ill-defined meanreversion too slow i.e. life too large.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
7/36
Introduction
I intend focussing on 2 aspects of theexchange rate disconnect Level andPPP puzzle.
Will argue: can forecast currencies usingmacro fundamentals as short as 2 monthhorizons; produce sensible measure of
equilibrium if abandon PPP and focus on areal exchange rate relationship.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
8/36
Introduction
The issue of an equilibrium exchange rate, andthe related concept of misalignment, is ofinterest to policy makers/ central banks/ financial
institutions. Forecastability of currencies of interest to
financial institutions such as hedge funds.Recently returns from international portfolios
have often come from exchange ratemovements so getting these right important.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
9/36
Equilibrium Exchange Rates
Why is the concept of equilibrium of interest?
Issues of misalignment in managed floatuseful to have indicator for on-going policy
debate. Issues of joining a currency union or locking
currency to a monetary standard. History repletewith example of countries getting this wrong (UKin 1926 and again in 1992)
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
10/36
Equilibrium Exchange Rates
How Measure Equilibrium?
Purchasing Power Parity (PPP) firstmeasure Economists reach for.
Is it useful? To answer think of derivation.
*
t t tS P P
*
0t t t t Q S P P
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
11/36
Equilibrium Exchange Rates
Relies on Law of One Price (LOOP)
LOOP Arbitrage + Substitutability
Arbitrage: Individual and Wholesale issues- Are Traded goods PerfectlySubstitutable?
What Does the Empirical Evidence say?
LOOP violated for all but generic goods(i.e. commodities).
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
12/36
PPP Puzzle therefore not surprising How explain?
Straightforward to decompose real exchange rate:
i. QTTrading frictions: Impart neutral band Qfollows non-linear process. Quite a bit of evidence infavour of this view but issues of arbitrage andsubstitutability.
Equilibrium Exchange Rates
,T T NT
t t tQ Q Q
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
13/36
Equilibrium Exchange Rates
ii. Real Factors a) QT,NTBest knownBalassa-Samuelson. TFP shocks in tradedsector generates appreciation of CPI-
based Qimpart systematic trend.
Evidence: 1. Indirect: Studies using pricedata suggest it is movements in QTexplain
violations of PPP.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
14/36
Equilibrium Exchange Rates
2. Direct: build TFP from OECD sectoral database CPI-based qand qT,NT. Studies reportsignificant evidence of importance of Balassa-Samuelson effect + adjustment speeds/ life
faster with BS. Real Factors b) QTFailure of LOOP means
driven by NFA or TB.A lot of evidencesupportive of this + life faster.
iiiQT
Market structure i.e. PTM and passthrough. Empirical evidence shows meanreversion speed significantly related to marketstructure.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
15/36
Equilibrium: in search of a suitableacronym.
In thinking about equilibrium issues otherapproaches take explicitly real perspective.
Internal External Balance Best knownFundamental Equilibrium Exchange Rate or
FEER. FEER is Q consistent with both internal and
external balance in medium run (5 years hence)- not stock-flow equilibrium.
IB = high employment + low inflation. EB:sustainable desired net flow of resourcesbetween countries when in internal balance.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
16/36
Equilibrium: in search of a suitableacronym.
Issues with FEER.
1. CAP sustainability controversial - Williamson usesvariety of factors such as Investment needs, effect ofdemographics on savings.
2. Method of calculation: does not actually say how Qconverges to FEER,
3. Depends on underlying trade elasticities veryimprecise
4. not clear how good implicit ex rate relationship is. 5. Normative approach. Useful to have approach which
separates the behavioural from normative.
6. Micro foundations?
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
17/36
Equilibrium: in search of a suitableacronym.
BEER Approach of Clark and MacDonald(1999,2000). Sequence: 1. Take standardbehavioural real exchange rate model.
2. Use best practice econometrics to estimatemodel.
3. Use this for assessment purposes - therebyseparating positive from normative. General
framework for assessment issues. 4. Show real factors important since mean
reversion speeds fast: life 1 year or less.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
18/36
Equilibrium: in search of a suitableacronym.
Permanent Equilibrium Exchange Rates:PEERs
Rely on decomposing Q into permanent and
transitory components. Interpret the former as ameasure of equilibrium:
where QPpermanent component; QT is
transitory. Advantages straightforward. Disadvantages
lack of theoretical underpinnings.
p T
t t tQ Q Q
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
19/36
Equilibrium: in search of a suitableacronym.
A capital enhanced equilibrium exchangerate, or CHEER [UIP/PPP].
Focuses on x=[s,p,p*,Il, il*]
Produces sensible measures ofequilibrium in sense that homogeneityrestrictions can be imposed on prices and
sensible coefficients on idiff. Useful in presence of limited data, but
limited in terms of structure.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
20/36
Equilibrium: in search of a suitableacronym.
New Open Economy Macroeconomic Approach(NOEM) to Assessment Issues.
Basic idea: optimising behaviour of consumers hasimplications for CA which, in turn, has implications forexchange rates.
Optimising rule of consumers suggests elasticity ofsubstitution, , is key determines how relative price ofT to NT affects rel quantaties.
Given and required in consumption of traded goodsshow how much ofQ needed to restore currentbalance.
Advantages theoretically rigorous. Disadvantages: whatis ? and as in FEER normative.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
21/36
Summary of Equilibrium exchangerate Issues
1. Measure useful for assessmentpurposes/ locking currencies together.
2. PPP not a suitable vehicle. Lesson of
mean reversion important. 3. Use a measure which makes the
normative/ positive split transparent.
BEER PEER? 4. Perhaps use a range of indicators and
produce weighted average?
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
22/36
Exchange Rate Forecasting
Since Meese and Rogoff (1984) Forecastingexercises usually defined w.r.t. Random Walkthe acid test. Can the profession beat it?
A lot of eyeball evidence in favour ofapproximate random walk. But issue of timedimension 1, 2 ,3 months? See Figures.
Econometric evidence gauges forecastabilityusing RMSE criterion:
1t t tS S
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
23/36
Exchange Rate Forecasting
The RMSE criterion is:
How useful? Direction perhaps more so.
But this is benchmark in academic literature. How good are the professionals see the
distributions for professionals.
From Consensus Economics, period -.
RMSEF A
n
t t ( )
2
RMSERMSE
RMSE
rm
rw
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
24/36
Exchange Rate Forecasting
Econometric work has as starting point study ofMeese and Rogoff (1984) who take variants ofthe monetary model i.e.:
For USD of DM, Yen and , M+R unable tooutperform a random walk at horizons of
between one and 12 months ahead Since M+R gave models an unfair advantage
random walk very strong.
* * *[( ),( ),( )]s f m m y y i i
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
25/36
Exchange Rate Forecasting
Update surveying post M+R:
Frankel and Rose (1995) ...the Meeseand Rogoff analysis of short horizons [less
than 36 months] has never beenconvincingly overturned or explained. Itcontinues to exert a pessimistic effect onthe field of empirical exchange ratemodeling in particular and internationalfinance in general
Rogoff (1999) reaffirms this.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
26/36
Exchange Rate Forecasting
Although Frankel and Rose quote typifiesview in profession, now numerous papers(over 30) using monetary model which
have overturned this result:
Why discounted?
Seems to be view in profession that we
need new conventional wisdom mentionedat start of lecture.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
27/36
Exchange Rate Forecasting
Because of RW result, Obstfeld and Rogoff andFlood and Rose suggest moving from macrofundamentals to market microstrucure.
But does this thrown baby out with bath water? Why the random walk result? M+R and all those
who are unable to outperform use simple staticmodels.
But exchange market, and underlying, marketsinherently dynamic. So should recognise this inany estimation.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
28/36
Exchange Rate Forecasting
Take MacDonald and Marsh Model UIP/CIP
x=[s,p,p*,Il, il*]
Dynamic:
V=[xt,xt-1, x]
Key advantage of strategy:
1. gives full system of equations, for all variables, ratherthan a single reduced form.
2. Facilitates a stringent test of forecast ability sincepredicted values of all terms (exchange rates, prices andinterest rates) are used rather than actual data values.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
29/36
Exchange Rate Forecasting
Currencies : yen, DM and against USD, Jan1974 to December 1992, last 24 obs held backfor forecasting purposes.
The forecasts fully simultaneous and dynamic
and could therefore have been used by apotential forecaster. Also significance levels offorecasting performance are provided.
Criteria:
1. RMSErrelative to a random walk 2. In terms ofdirectional ability:
Dn
(1 if forecast direction = actual,else 0)
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
30/36
Exchange Rate Forecasting
Pure Chance: D = 0.5
3. RMSErRelative to a panel of 150professional forecasters, located in G7 financialcentres, as collected by Consensus Economicsof London.
Summary of Findings: 1. Able to beat a randomwalk at horizons as short as 2 months aheadand this continues to longer horizons.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
31/36
Exchange Rate Forecasting
2. Models have excellent directional ability on average between 0.6 and 0.7 overthe different horizons
3. No forecaster ranks as consistentlyhighly across currencies and/ or horizons.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
32/36
Exchange Rate Forecasting
The essential point of this modeling: an Smodel which has sensible long-runequilibrium and dynamic properties, rich
enough to capture the underlying marketdynamics, will do better than a staticmodel or one with very simple dynamics.
Modelling approach has been usedextensively by financial institutions andforecastability seems robust over time.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
33/36
Summary and Conclusions
Main theme of todays lecture has been to have
argued against the current conventional wisdomin the exchange rate economics literature.
Tried to argue that standard macrofundamentals are useful for an analysis ofexchange rate behaviour.
Specifically we have argued that it is possible toaddress the Exchange Rate Disconnect.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
34/36
Summary and Conclusions
Specifically, I argued that PPP on its own is nota sufficient measure for thinking aboutequilibrium issues but there are alternatives.
These alternatives explicitly recognises that realexchange rates have real determinants.
They have an exotic array of mnemonics fromBEERs to FEERs to Cheers. Useful forassessment issues and for currency locks.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
35/36
Summary and Conclusions
We have also argued that the reason sofew researchers have been able to beat arandom walk model is because of the very
simplistic empirical models used. Realistic dynamic models are able to
outperform a random walk and offer gooddirectional performance.
Such models should be of use topractitioners.
7/31/2019 Presentation Slides by 31 October by Prof MacDonald
36/36
Summary and Conclusions
In sum, the current trend in the professiontowards market microstructure issues isworthwhile and interesting.
However, to abandon a role formacroeconomic fundamentals istantamount to ignoring key information in
the process of exchange ratedetermination.