Upload
dogan
View
212
Download
0
Embed Size (px)
Citation preview
Applied Economics, 2008, 40, 2671–2696
The inflation-hedging properties of
Turkish REITs
Isil Erola,* and Dogan Tirtiroglub
aDepartment of Economics, Middle East Technical University,
06531 Ankara, TurkeybDepartment of Finance, Concordia University, Montreal, QC H4V 1Z9,
Canada
This study empirically tests the inflation-hedging abilities of Turkish
REITs in comparison to the indices of common stocks listed on the
Istanbul Stock Exchange (ISE) over the period December 1999 to
December 2004. Two main factors motivate this study. First, compared
to their counterparts in developed capital markets, Turkish REITs have
some important tax incentives as well as flexibility in managing their
portfolios. Second, the Turkish economy provides a rare and good
opportunity to test the hedging behaviour of real estate stocks in periods of
both high- and moderate-inflation rates. The empirical results show that
Turkish REITs, in general, provide a better hedge against both actual and
expected inflation than do the ISE common stock indices. Dividing the
entire sample period into the high- and moderate-inflation sub-periods, we
find that the hedging ability of REITs is better under high inflation than
under moderate inflation.
I. Introduction
One basic objective of investors holding investments
is to protect their wealth against inflation. There is a
rich literature, mainly with results from the developed
economies, on the inflation hedging ability of both
financial and real estate assets. There is a consensus
that assets differ in their hedging behaviour against
inflation and that while real estate investments
provide a good hedge against both the expected and
unexpected components of inflation, common stock
investments act as a perverse hedge against both
components in most of these markets.Real estate has traditionally been a popular
investment tool, especially in high-inflation econo-
mies, as it is a hedge against inflation. But, one must
keep in mind that bias, which is likely to originate
from appraisal smoothing in the data of several
published empirical papers, is a point of criticism for
the conclusions in these studies. As stated by
Yobaccio et al. (1995), a logical succession to the
data that is potentially contaminated with appraisal
smoothing is to use market data from publicly traded
real estate investment trusts (REITs). Publicly traded
REIT returns reflect the value of the income and
price appreciation components of the underlying
assets, determined in a market that adjusts overall
rapidly to changes in information or expectations.
Moreover, REITs are a means to address the
illiquidity problem of real estate assets and offer an
efficient integration of real estate markets to the
capital markets.Turkey established its REIT structure in 1998.
Turkish REITs can and do invest in buildings,
land, development projects, real estate-backed
securities, and to a limited extent other capital
*Corresponding author. E-mail: [email protected]
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online � 2008 Taylor & Francis 2671http://www.informaworld.com
DOI: 10.1080/00036840600970237
market instruments. At the beginning of 2005, therewere nine REITs listed on the Istanbul StockExchange (ISE)(http://www.imkb.gov.tr) with atotal portfolio of US$1.5 billion. Real estate invest-ment trusts make up approximately 1% of the totalTurkish stock market capitalization.
The present study is the first attempt to evaluateand provide empirical evidence on the inflationhedging properties of Turkish REITs. The study isimportant for at least two reasons. First, the Turkishmarket provides a rare and good opportunity to testcomparatively the hedging behaviour of real estatestocks under high and moderate inflation sub-periods. Turkey has experienced a persistent hyperin-flation between late 1970s and until the early years ofthe 21st century, when the inflation has started tochange its course for a rather fast and sudden decline.Second, compared to their counterparts in developedcapital markets, Turkish REITs have some importanttax incentives. Specifically, REITs are exempt fromboth corporation and income taxes. Turkish REITshave the freedom to choose their dividend policies.This is another interesting and important differenceof Turkish REITs from their counterparts in othercountries. The law, which governs Turkish REITs,does not impose any dividend payout requirementson them. The freedom to choose their dividend policyallows the Turkish REITs to enjoy the financialflexibility to accumulate dividends, if needed, forfurther investments.
In this study, we use two different models; namely,the Fama-Schwert and the Fisherian Direct Causalitymodels, to examine comparatively the inflationhedging properties of Turkish REITs and the indicesof common stocks listed on ISE in relation to actualinflation and its expected and unexpected compo-nents between December 1999 and December 2004.Overall, our findings show that Turkish REITs, ingeneral, provide a better hedge against both actualand expected inflation than do the ISE common stockindices. Dividing the entire sample period into thehigh and moderate inflation sub-periods, we also findthat the hedging ability of REITs as well as ISEcommon stock indices is better under high inflationthan that under moderate inflation.
The remainder of this article proceeds as follows.Section II reviews the existing empirical studies oninflation hedging characteristics of real estate andcommon stocks in different markets. Section IIIprovides an overview of both real estate market inTurkey and the characteristics and the developmentof Turkish REITs. Section IV presents the empiricalmethodology and the data. Section V provides ananalysis of the empirical results, and the final sectionconcludes the article.
II. Review of the Literature
Real estate has long been regarded as a good hedge
against inflation. Empirical findings from country-
specific and cross-country studies have been suppor-
tive of the hypothesis that real estate returns move in
one-to-one correspondence with inflation rates.Evidence from the US indicates consistently that
real estate offers a hedge against inflation, in
particular, against the expected component of infla-
tion. Hartzell et al. (1987) documented that a
portfolio of commercial real estate provided an
effective hedge against both components of inflation
between 1973 and1983. An earlier study by Fama and
Schwert (1977) suggested that residential real estate
was a complete hedge against both components of
inflation. Rubens et al. (1989) found that residential,
commercial and farmland real estate provided at least
partial hedging against inflation. These authors also
found that including real estate assets into their
portfolios improved their hedging effectiveness.
Brueggeman et al. (1984), Ibbotson and Siegel
(1984) and Miles and Mahoney (1997) also reported
results, which indicated hedging capabilities of real
estate assets. Overall, we note that these papers’
findings on the unexpected component of inflation
were less conclusive. Similarly, Wurtzebach et al.
(1991) found that office and industrial real estate
returns offered no significant hedging possibility
against unexpected inflation. The low inflation in
the US could be an explanation of the insignificant
real estate performance against unexpected inflation.In the UK, reported evidence indicated that real
estate provided good protection against inflation (see
Limmack and Ward, 1988; Hoesli et al., 1997; Miles,
1996). Limmack and Ward (1988) found that office
and shop offered no significant hedge against
unexpected inflation. Meanwhile, Hoesli et al.
(1997) showed that while the capital appreciation
component of real estate returns offered a hedge
against unexpected inflation, the income component
did not.The results were similar for Switzerland (Hoesli,
1994; Liu et al., 1997), Canada (Newell, 1995),
New Zealand (Newell and Boyd, 1995), Australia
(Newell, 1996) and Hong Kong (Ganesan and
Chiang, 1998). In a recent study, Sing and Low
(2000) examined the inflation hedging characteristics
of real estate in Singapore. The authors found that
real estate provided a better hedge against inflation
than did stock and securitized real estate. Industrial
property was the most effective hedge against both
components of inflation, whereas shop offered a
significant hedge only against expected inflation.
2672 I. Erol and D. Tirtiroglu
For the equity markets, findings from the US and
the UK, in general, suggested that common stock
returns offered no significant hedge against inflation
(Fama and Schwert, 1977; Gultekin, 1983; Hartzell
et al., 1987; Liu et al., 1997; Coleman et al., 1994).1
Many studies also noted the existence of a perverse
hedging property of domestic common stock returns.
For example, both Hasbrouck (1983) and Gultekin
(1983) examined the relationship between inflation
rates and common stock returns from 1950s to 1980
and concluded that unexpected inflation, but not
expected inflation, explained the negative relationship
between the two measures. Spyrou (2005) finds a
statistically significant positive (negative) hedging
property of common stocks in three (one) emerging
markets out of the sample 10 emerging markets in
the 1990s.Evidence on the inflation-hedging properties of
REITs or similar liquid real estate assets is not as
clear as that of other real estate assets. On the one
hand, a number of studies indicated that REIT
returns did not provide a significant hedge against
inflation and that they might act as a perverse hedge
against either actual inflation or one or sometimes
both components of inflation.2 These researchers
attributed their findings to REITs’ tendency to
behave like other common stocks. On the other
hand, a number of studies indicated that REITs
possessed some inflation-hedging properties.3 This
presence of mixed evidence is a source of motivation
for the current study.So far, Onder’s (2000) paper on the inflation
hedging properties of residential real estate
investments in the capital city of Ankara, Turkey isthe only empirical study with Turkish data. Sheexamined the hedging behaviour of residential realestate investment from 13 neighbourhoods againstexpected and unexpected components of inflationunder the high-inflationary conditions between 1977and 1996. Her results suggested that, unlike thebehaviour of real estate assets in other countries,residential real estate investments in Ankara did notprovide a hedge against inflation.
Based on the literature review, we can generalizethat the studies involving the use of real estate assetsas an inflation hedge are encouraging. Real estateassets were found to offer a significant hedge againstinflation, particularly the expected component ofinflation in most of the countries. But, the hedgeagainst unexpected inflation is not statistically well-established. In fact, existing evidence shows thatREITs could hedge perversely against the unexpectedcomponent of inflation.
III. Real Estate Market in Turkey
Turkey is a dynamic, emerging market economycountry. It has a young population of over 70 millionpeople, of which 73% lives in urban areas. Real estatedevelopment and construction are two of the leadingsectors of the Turkish economy. Both sectors havebeen playing a crucial role by accounting forapproximately 10% of the country’s gross nationalproduct over the years. During the last decade, the
1 Brooks and Tsolacos (2001) show within a UK context that short- and long-term interest rates and interest rate spreads maybe leading indicators for real estate markets.2Murphy and Kleiman (1989) examined REIT returns and found that, over short periods of time, equity REITs acted as aperverse hedge against both expected and unexpected inflation. Chan et al. (1990) noted the perverse relationship only againstunexpected inflation. Park et al. (1990) found that REITs generally exhibited the same perverse hedging characteristics ascommon stocks, although REITs appeared to be a partial hedge against expected inflation when one used the Livingston pricedata. Yobaccio et al. (1995) examined the inflation-hedging abilities of REITs over the period February 1972 to December1992. They expanded Park et al.’s (1990) model by allowing the real return to the market portfolio to vary. They argued thatfailure to account for potential variability in the real return on risky assets might account for the apparent perverse hedgingproperties of real estate. The results indicated that REITs acted as poor hedges against any measure of inflation (actual,expected or unexpected) with the poorest performance relative to unexpected inflation. In this respect, REIT returns wouldseem not to be proxies for direct investment in real estate. The authors concluded that evidence on REITs indicated that realestate, at best, acted as a partial hedge against expected inflation and a perverse hedge against unexpected inflation.3Gyourko and Linneman (1988) documented that REIT returns provided a partial hedge against actual inflation – adjustedby subtracting out the Home Purchase Price component of the CPI – and expected inflation, but a perverse hedge againstunexpected inflation. Chen and Tzang (1988) found that both equity and mortgage REITs showed some hedging abilityagainst expected inflation. Examining the Hong Kong real estate market, Ganesan and Chiang (1998) showed that real estatestocks had a partial hedge against expected inflation, but a perverse hedge against unexpected inflation. Chatrath and Liang(1998) examined the inflation-hedging characteristics of REITs between 1972 and 1995. They also investigated whether theapparent lack of a positive relationship between general prices and REIT returns in prior studies was an outcome ofthe impact that stock market movements might have had on REITs. Under Johansen co-integration tests, the authors foundsome evidence that REITs provided a long-run inflation hedge. They also ruled out the possibility that a stock market-induced proxy effect was one of the causes for the apparent lack of a relationship between REIT returns and inflation rates.
The inflation-hedging properties of Turkish REITs 2673
real estate market has made spectacular advances.
One reason for this is the high demand, both in
residential and commercial real estate, by foreign
investors. The size of the total real estate market in
Turkey is about US$ 82 billion.4
In spite of the high demand for real estate assets, a
well-organized and deep enough mortgage market did
not exist in Turkey until quite recently. The absence
of an efficient mortgage market was mainly due to a
long-running process of persistently high inflation.5
In the last 3 years, as to be seen in Fig. 1, however,
inflation has steadily dropped to the lowest levels of
the past decades.6 Under the stable economic condi-
tions of the recent years, the current government has
recently prepared a legal framework not only for a
proper mortgage system, designed especially for
financing residential real estate for the middle-
income households, but also for the eventual
securitization of these mortgages. The government
recently proposed this framework to the Turkish
Parliament for its passage as a law.Easing restrictions on foreigners’ property pur-
chases in July 2003 was one of the most important
recent developments in the Turkish real estate
markets. As of the end of 2004, 44 740 foreigners
from 68 countries owned properties in Turkey; the
total value of their investments was approximately
US$2 billion. Another important and essential
improvement in the Turkish real estate markets is
the ongoing growth of REITs. This is a major step
towards bringing international standards, providing
reliable and quality information and fostering foreign
investments, especially at the institutional scale.
Essentially, in parallel to her application to join the
European Union, Turkey has tightened its fiscal
policy and pushed for major improvements in its
−2.0
0.0
2.0
4.0
6.0
8.0
10.0
1995
/01
1995
/07
1996
/01
1996
/07
1997
/01
1997
/07
1998
/01
1998
/07
1999
/01
1999
/07
2000
/01
2000
/07
2001
/01
2001
/07
2002
/01
2002
/07
2003
/01
2003
/07
2004
/01
2004
/07
2005
/01
2005
/07
Moderate inflation period
High-inflation period
Fig. 1. The monthly inflation Rate (% Changes in CPI) in Turkey between 1990 and 2005
Notes: Turkey experienced a persistent hyperinflation between late 1970s and until the early years of the 21st century, whenthe inflation started to change its course for a rather fast and sudden declining one. This figure shows monthly inflation ratesbetween 1990 and 2005. As seen from this figure, high-(moderate)inflation characterizes the period from January 2000 toJanuary 2002 (from February 2002 to December 2004). We use these sub-periods to further examine inflation-hedgingproperties of returns on (1) various Istanbul Stock Exchange (ISE) common stock indices (2) ISE REIT Price Index and(3) ISE listed individual REITs.Source: The State Institute of Statistics of Turkey, Ankara, Turkey.
4NAREIT, Real Estate Portfolio, Special Issue, 2005.5 The absence of commercial banks from the mortgage markets in the 1980s and 1990s (and earlier too) led government-subsidized Housing Development Administration (HDA) and a few state-owned banks to try to fill, in a very limited way, thisvoid by creating some inflation-sensitive alternative mortgage instruments (see Erol and Patel, 2005 for details). But, highinflation eventually shut down these instruments, too.6 See Us (2004) and Gunay et al. (2005) for a detailed examination of the inflation issues in Turkey. In particular, Us (2004)finds that Monetary Conditions Index, which is a combination of the changes in the short-term real interest rate and in thereal effective exchange rate, could have policy implications in curbing inflation in Turkey.
2674 I. Erol and D. Tirtiroglu
economic environment within the last few years. Theimproving economic conditions have revivedTurkey’s real estate markets. Turkish REITs standto be the highest beneficiaries of this resurgence,especially attracting the attention and the interest ofinstitutional investors.
Turkish REITs
Turkey established its REIT structure, namedGayrimenkul Yat., in 1998. The law allows TurkishREITs to invest in buildings, land, developmentprojects, real estate-backed securities, and to a limitedextent, other capital market instruments, such asgovernment bonds and stocks.
Compared to REITs in developed capital markets,Turkish REITs have a remarkable growth potentialin the near future due to the high expectations placedon them to bring transparency and professionalism tothe broader real estate industry. In order to create afavourable growth environment for the fledglingindustry, authorities have provided REITs withsome important tax incentives as well as flexibilityin managing their portfolios. An amendment by theCapital Markets Board (CMB), introduced in 1998,requires REITs to float at least 49% of their shares tothe public. This modified legal framework also hasexempted REITs from both corporation and incometaxes to promote the formation and growth of theindustry (see Aydinoglu, 2004).
Following the adoption of the modified legalframework, a number of large and well-establishedTurkish financial entities set up their REITs andretain a controlling interest in them. At the beginningof 2005, there were nine REITs listed on ISE with atotal portfolio of US$ 1.5 billion. REITs make upapproximately 1% of the total Turkish stock marketcapitalization. These were Alarko, Atakule, EGS,Garanti, Ihlas, Is, Nurol, Vakif and Yapi KrediREITs.7
A key and important difference of Turkish REITsfrom those of developed economies is that TurkishREITs do not have to pay out dividends to theshareholders on an annual basis.8 This makes their
dividend withholding tax rate 0%. This incentiveallows Turkish REITs to finance relatively cheaply
their growth through 100% plow back of profits intonew investments. This is an important advantageirrespective of inflationary conditions.
Turkish REITs can engage in a variety of activities.For the purposes of generating capital gains orearning rental income, REITs can (1) purchase andsell real estate, (2) lease real estate from third partiesand rent them in return to generate rental income, (3)
purchase and sell capital market tools and do reverserepo transactions with such tools, (4) buy land inorder to carry out real estate development projectsand (5) purchase foreign real estate on the conditionof obtaining ownership and investment in real estate-
backed foreign marketable securities.9
According to 1998 Communique,10 REITs may be
found (i) for a specific period to realize a certain
project, (ii) for a specific or unlimited period to invest
in specific areas, (iii) for a specific or unlimited period
without any limitation of objectives.11 All Turkish
REITs are of the third type, so they are not limited by
a certain product type or geographic location but are
still bound by the general principles set by the CMB.Currently, REITs must invest at a minimum 50%
of their portfolios in real estate and real estate-backed
securities. Earlier, this ratio was 75%. This reduction
has given them further flexibility to construct a more
diversified portfolio with short-term and long-term
fixed income securities and/or equity.We examine the market value (MV) and the net
asset value (NAV) of each of the nine listed REITs asof 31 December 2004. Specifically, the MV is thenumber of shares outstanding multiplied bythe closing price per share on 31 December 2004.
The NAV for a REIT is equal to the sum of its totalportfolio value and its nonportfolio liquid assets lessits total debt. Is-REIT stands out as the industryleader both in MV and NAV terms. Its share of theTurkish REIT industry is 54% in NAV and 43% inMV. Is-Bank, which is one of the largest, oldest and
privately-owned commercial banks in Turkey, is thefounder and main backer of Is-REIT. Atakule and
7The tenth Turkish REIT, Akmerkez, was listed on ISE in April 2005. Since our study examines the inflation-hedgingcharacteristics of REITs from December 1999 to December 2004, Akmerkez is not included in our study. Akmerkez has oneasset in its portfolio, the Akmerkez complex. It is a shopping centre with 246 separate shops and office spaces as well.8 REITs in the US, UK, Canada, Japan, Hong-Kong and Singapore have a minimum payout ratio, ranging from 90% to100% of their net income after taxes.9However, the total value of such investments is not to exceed 10% of their portfolios.10 The operations of REITs are governed by the 1998 Communique on the REITs, issued by the CMB. At least one of theinitiator shareholders (founders) should be ‘leader entrepreneur’, who should hold at least 25% of this REIT’s capital.11 1998 Communique, Article 5.
The inflation-hedging properties of Turkish REITs 2675
Alarko, the two big financial groups in Turkey, haveMV shares of 33% and 6% of the REIT industry,respectively. In terms of the NAV, Alarko, Atakule,Yapi Kredi, Garanti and EGS have 6–12% shares ofthe industry’s total net assets.
Table 1 summarizes the portfolio structure of eachof the nine REITs as of 31 December 2004. With theexception of Alarko REIT, all REITs had at least75% of their asset portfolios invested in real estate.More specifically, Atakule, EGS, Ihlas, Is and NurolREITs had 93–100% shares of their portfolios in realassets. Investments in buildings and land (real estateprojects) comprised approximately 75% (14%) of theaggregated or total REIT asset portfolio, respectively.The remaining 11% was a combination of liquidassets, such as short-term government securities andreverse repurchased bond contracts (repo).Examining the aggregated sector-based real estateportfolio structure of Turkish REITs reveals thatinvestments in office and shop had the largest sharesof the total portfolio; that is 46% and 22%,respectively. Residential investments had only 12%share of the total real estate portfolio.
Figure 2 exhibits the ISE’s REIT Price Index, theISE All-share Index and the ISE sector-based indicesbetween December 1999 and March 2005. The ISEREIT Price Index is value-weighted and comprises ofthe nine individual REITs listed on ISE. Since REITsare newly developing in Turkey, there is still not anysector-based division, such as hotel REITs orapartment REITs. The ISE All-share Index is value-weighted and represents the price movements of allfirms listed on the ISE. The ISE has also compiledsub-indices, categorized by industrial type. The mainsub-indices are industrial, financial, service, andtechnology indices (Fig. 3).
As seen from Fig. 2, the ISE REIT Price Indexdominated the ISE All-share Index from December1999 to the early 2002 and more recently in 2004.Although REITs generally outperformed both theindustrial and service sectors, financial firms listed onthe ISE performed better than REITs did. We discussin detail some of the performance of these indices andtheir properties in Section ‘Descriptive statistics forasset returns and inflation rate’.
IV. Methodology and Data
Broadly speaking, we undertake two inter-relatedempirical investigations. The first considers thequestion of whether the Turkish REITs’ returns area hedge against the actual inflation rates. The secondextends the same question to include the expected and
unexpected rates of inflation. Before we proceed anyfurther, let us also note that we use monthly datasince Consumer Price Index (CPI) data to measureactual inflation rates during the sample period areavailable on a monthly basis.
The following model provides empirical evidenceon the first question for each sample REIT, the ISEREIT Price Index and each of various ISE indicesconsidered in this study:
Rit ¼ �þ ��t þ "it ð1Þ
where:
Rit ¼ lnPit
Pit�1
� �and is the nominal rate ofreturn on the ith TurkishREIT or the ISE REITPrice Index or one of theconsidered ISE commonstock index at time t,
Pit¼ the end-of-month closingprice on the ith TurkishREIT or the ISE REITPrice Index or one of theconsidered ISE commonstock index at time t,
�t ¼CPIt � CPIt�1
CPIt�1¼ the actual rate of inflation at
time t,CPIt¼ the Consumer Price Index at
time t,� and � are coefficients to be
estimated," is the error term of theempirical model.
We examine the inflation-hedging effectiveness ofREITs against the expected and unexpected compo-nents of inflation under two different models. Thefirst one is the Fama and Schwert (1977) model,which provides an effective definition that is widelyused in the empirical tests of an inflation-hedginghypothesis. According to the definition, an asset issaid to be a complete hedge against inflation, if andonly if the nominal return of the asset varies in a one-to-one relationship with both expected and unex-pected inflation. According to Fama and Schwert’smodel, the expected nominal rate of return of an assetcan be expressed as the sum of the expected real rateof return, expected inflation and unexpected infla-tion. The model can be written as follows:
E Ritj�t�1ð Þ ¼E �itj�t�1ð ÞþE �tj�t�1ð Þþ �t�E �tj�t�1ð Þ½ �
ð2Þ
where:
EðRitj�t�1Þ ¼ expected nominal rate of returngiven the information � at t� 1,
2676 I. Erol and D. Tirtiroglu
Table
1.PortfoliostructuresofindividualREIT
s(31Decem
ber
2004)
Alarko
REIT
(%)
Atakule
REIT
(%)
EGS
REIT
(%)
Garanti
REIT
(%)
Ihlas
REIT
(%)
IsREIT
(%)
Nurol
REIT
(%)
Vakif
REIT
(%)
YapiKredi
REIT
(%)
REIT
Total(%
)
Asset
portfolio
Financialassets
42.2
4.3
0.4
25.5
07.1
0.6
17.2
12.0
11
Realassets
57.8
95.7
99.6
74.5
100
92.9
99.4
82.8
88.0
89
Buildings
33.1
95.7
65.0
21.6
66.0
81.3
66.3
73.4
36.5
68
Land
24.3
–0.3
–2.0
0.4
32.5
9.4
34.5
7RealEstate
Projects
0.4
–34.3
52.9
32.0
11.2
0.6
–17.0
14
Sector-basedrealestate
portfolio
Shop
30.0
50.0
91.9
––
4.0
3.1
36.3
37.2
22
Office
3.7
22.5
8.1
29.0
35.8
67.0
19.4
52.3
46.2
46
Residential
18.6
––
71.0
64.2
4.0
77.5
–16.6
12
Tourism
47.7
10.5
––
–19.0
––
–15
Service
–17.0
––
–6.0
–11.4
–5
Notes:Thistable
presents
both
theasset
portfoliosandthesector-aggregatedrealestate
portfoliosofninesample
individualREIT
slisted
ontheIstanbulStock
Exchange.
Source:
Thewebsite
oftheIstanbulStock
Exchangehttp://w
ww.imkb.gov.tr
The inflation-hedging properties of Turkish REITs 2677
Eð�itj�t�1Þ ¼ expected real rate of
return given the information � at
t� 1,
Eð�tj�t�1Þ ¼ expected inflation rate given the
information � at t� 1,�t ¼ actual or observed inflation rate at
time t,
½�t � Eð�tj�t�1Þ� ¼unexpected inflation rate, which issimply the difference between theactual and expected inflation attime t.
As stated by Fisher’s well-known hypothesis, theexpected nominal rate of return is the sum of
Price indices
0
5000
10000
15000
20000
25000
30000
35000
4.1.
2000
4.5.
2000
4.9.
2000
4.1.
2001
4.5.
2001
4.9.
2001
4.1.
2002
4.5.
2002
4.9.
2002
4.1.
2003
4.5.
2003
4.9.
2003
4.1.
2004
4.5.
2004
4.9.
2004
4.1.
2005
ISE_All shares
REITs
Fig. 2. Price performance of Istanbul stock exchange all share index and REIT price index between 2000 and 2005Notes: This figure shows for comparative purposes the levels of the ISE All Share Index and REIT Price Index between 2000and 2005. There were nine REITs listed on ISE. They make up the ISE REIT price index. ISE all share index does not includeany REITs.
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
04.0
1.20
00
04.0
4.20
00
04.0
7.20
00
04.1
0.20
00
04.0
1.20
01
04.0
4.20
01
04.0
7.20
01
04.1
0.20
01
04.0
1.20
02
04.0
4.20
02
04.0
7.20
02
04.1
0.20
02
04.0
1.20
03
04.0
4.20
03
04.0
7.20
03
04.1
0.20
03
04.0
1.20
04
04.0
4.20
04
04.0
7.20
04
04.1
0.20
04
04.0
1.20
05
REITs
ISE_Industrial
ISE_National_Service
ISE_Financial
Fig. 3. Price performance of REITs and industrial, national service and financial companies listed on the istanbul stock
exchangeNotes: This figure shows for comparative purposes the levels of various ISE sector-based common stock indices and REITPrice Index between 2000 and 2005. There were nine REITs listed on ISE. They make up the ISE REIT price index. Theconsidered ISE common stock indices do not include any REITs.
2678 I. Erol and D. Tirtiroglu
the expected real rate of return and expectedinflation. If the market is efficient, the expected realreturn should be independent of the expected infla-tion. Based upon this assumption, Fama and Schwert(1977) develop the following empirically testablemodel specification:
Rit ¼ �þ �E �tj�t�1ð Þ þ � �t � E �tj�t�1ð Þð Þ þ eit ð3Þ
where � and g are coefficients to be estimated and e isthe error term.
Fama and Schwert argue that, given financialtheory, both � and g should be positive ex ante. Morespecifically, an asset is said to hedge against expected(unexpected) inflation when �¼ 1 (g¼ 1) and offers acomplete hedge against both components when�¼ g¼ 1.
We also consider the Fisherian Direct Causality(FDC) model in Solnik’s (1983) work as a supple-mentary alternative model. It assumes that only realfactors, such as productivity of capital and investors’time and risk preferences, influence expected realreturns. Therefore, expected real returns should beindependent of both the levels and variations inexpected inflation. In order to test the hypothesis thatinflationary expectations do not influence realreturns, we subscribe to the following regression
model:
�it ¼ �þ �1E �tj�t�1ð Þ þ �2½E �tþ1j�tð Þ
� E �tj�t�1ð Þ� þ "it ð4Þ
where
�it¼ (Rit��t) and is the ex post real rate ofreturn on the ith asset at time t,
�1¼ the coefficient for expected inflation level,�2¼ the coefficient for the changes in expecta-
tions of inflation.
Under the null hypothesis, both coefficients should be
zero since real returns are independent of any
expected inflation level or expected changes in
inflation.We use Ordinary least square (OLS) to estimate all
these models and consider two alternative measures
of expected and unexpected components of inflation.
Following Fama and Schwert (1977), the first
measure sets the current expected inflation rate to
previous period’s 3-month Treasury bill rate. The
second measure, following Fama and Gibbons
(1982), estimates the expected inflation rate afterregressing the actual inflation rate on the 3-month T-
bills yield.12 The following model ascertains each
proxy’s effectiveness as a predictor of inflation:
�t¼ ’þ E �tð Þ þ t ð5Þ
where ’ and are coefficients to be estimated and is
the error term.In an efficient market, we expect and ’ to be
equal to unity and zero, respectively. Table 2 showsthat the estimated coefficient is equal to unity under
the first model and that both models have the same
R2. However, the estimated constant is significantly
different from zero under the first model. Thus, our
data exhibit better harmony with the Fama–Schwert
(1977) model.This study uses monthly data between December
1999 and December 2004 period.13 These data come
from a number of sources. The websites of the State
Institute of Statistics of Turkey (http://www.die.-
gov.tr) and the Central Bank of Turkey (http://
www.tcmb.gov.tr) are the sources of data on monthly
inflation rates and the 3-month Treasury bill rates,respectively. The holding period return on the
Treasury bills is our measure of short-term interest
Table 2. Results to evaluate the expected inflation rates
Parameter estimate t-Statistic R2
Model 1: Eð�tÞ ¼ T� Billt�1’ �0.008 �2.045** 0.564 0.870 8.664***
Model 2: Eð�tÞ ¼ ð�0:0081Þ þ ð0:8703ÞT� Billt�1’& �0.000 �0.005 0.564 1.000 8.664***
Notes: This table presents the estimation results for theeffectiveness of two alternative measures of expectedinflation rate. The first measure, following Fama andSchwert (1977), sets the current expected inflation rate to aone-period lagged 3-month Treasury bill rate (Model 1).The second measure, following Fama and Gibbons (1982),estimates the expected inflation rate by regressing the actualinflation rate on the 3-month T-bill yield (Model 2).&The estimated value for ’ is �0.00001686.** and *** Refer to the 5 and 1% significance levels,respectively.
12We note that there are other different approaches to obtain the expected and unexpected components of inflation(Fama and Schwert, 1977; Fama and Gibson, 1982; Gultekin, 1983).13A proxy that is widely used in the literature for measuring return on real estate is the appreciation rate in the real estate(housing) purchase price component of the CPI (see, for example, Fama and Schwert, 1977). Unfortunately, only rents, theprices of utilities and the maintenance costs are considered as housing expenses in the Turkish CPI calculations. Hence, it isnot possible to use this component as a return on real estate investment. Therefore, in parallel to the literature, the return onREITs is used as a measure of the inflation-hedging properties of real estate.
The inflation-hedging properties of Turkish REITs 2679
rates. The ISE provides the data on stock marketreturns; namely, ISE All-share Index, the related sub-indices (ISE-100, ISE-30, ISE-Industrial, ISE-Financial), and the ISE REIT Price Index and allsample individual REITs. The ISE All-share Indexrepresents the price movements of all companieslisted on the ISE. The ISE also compiles sub-indices, categorized by industrial type. The mainsub-indices are industrial, financial, service, andtechnology indices. Consistency in the time periodof data series requires a subscription to themonthly series of the Treasury bill rates and thestock indices.
The number of REITs in Turkey currently is few.Therefore, it is not feasible to consider the perfor-mance of REIT sub-sectors, such as hotel REITs orapartment REITs. There is only an overall REITPrice Index in place. Therefore, we are able toexamine price movements of both the overall REITPrice Index and all individual REITs over thespecified time period.14
Descriptive statistics for asset returns andinflation rate
Table 3 reports the descriptive statistics of assetreturns and inflation rate for the sample period. TheISE REIT Price Index was constructed from onlythe nine individual REITs listed on ISE as ofDecember 2004. Therefore, we caution that theidiosyncratic risk is likely not diversified away totallyin this index. Thus, one must keep this caveat in mindin interpreting the results in Table 3 (and in Fig. 2,too). The results show that both the ISE REIT PriceIndex and all individual REITs did better than theISE stock indices in terms of their average monthlyreturns (footnote 14). This is not surprising, given ourreservations in the previous sentences about theidiosyncratic risk inherent in individual REITs andthe ISE REIT Price Index. In fact, the SDs ofindividual REITs and the ISE REIT Price Index arevisibly higher than those of the ISE indices, providingsome substance for the presence of the idiosyncraticrisk in REIT returns.
Garanti REIT, Is-REIT, and Alarko REIT werethe best performers with the two highest mean
monthly returns of 5.28% and 3.90%, whereas the
ISE’s diversified indices realized the lowest mean
returns of 0.6–1.02%. Over the period December
1999 to December 2004, except for Atakule REIT, all
individual REITs outperformed the mean rate of
actual inflation. In comparison, the monthly mean
return on all ISE related indices were lower than the
mean rate of actual inflation, as measured by CPI.
The mean and SD of Housing Price Index (HPI) are
close to those of CPI.15
Table 3. Monthly asset returns and inflation rate between
December 1999 and December 2004
VariablesMean(%)
SD(%) Correlation
Asset return: Rt ¼ lnðPt=Pt�1Þ
ISE Stock
ISE-All Index 0.70 17.4 0.139ISE-100 Index 0.60 17.5 0.134ISE-30 Index 0.60 17.4 0.138ISE-Industrial Index 1.02 12.7 0.181ISE-Financial Index 0.62 16.0 0.121House price index (HPI) 2.20 0.07 0.841ISE REIT Price Index 3.68 19.7 0.232
Individual REITsAlarko REIT 3.90 18.7 0.209Atakule REIT* 2.10 17.4 0.096Ihlas REIT 3.70 27.4 0.115Is REIT 3.90 25.7 0.146Nurol REIT 3.40 19.9 0.185Yap| Kredi REIT 3.73 22.3 0.261Vak|f REIT 3.71 20.5 0.232Garanti REIT 5.28 24.1 0.185EGS REIT 2.73 29.2 0.173Consumer priceindex (CPI)
2.30 1.87 –
Notes: This table presents descriptive statistics (mean, SD)for the sample asset returns along with Pearson correlationcoefficients between returns on different sample assets andactual inflation rates, as measured by CPI. We acknowl-edge that the mean return of 1.37% on Istanbul StockExchange (ISE) REIT Price Index is inconsistent with themean returns on sample individual REITs. We do not haveany information about the reasons and sources of thisinconsistency. The returns are from natural logarithms ofprice relatives between time t� 1 and t.Source: Price data for Atakule REIT are available fromFebruary 2002.
14 In Table 3, the mean return on REIT Price Index is 1.37% while the smallest mean return on any one of the nine REITs,which comprise of the REIT Price Index, is 2.10%. This is puzzling. Our data come from the Istanbul Stock Exchange and wedo not have any information about the source of this puzzle.
Given this situation, we constructed our own equally-weighted REIT Price Index, using the monthly returns on the nineREITs to offer a more informative point of comparison for the results in Table 3. The mean return and SD of our REIT PriceIndex are 3.68% and 19.7%, respectively.15 This is likely because HPI is also a component of CPI. Unfortunately, CPI calculations in Turkey consider only rents, utilityprices and maintenance costs as housing expenses. Hence, it is not possible to use HPI as a return on real estate investmentsince it captures only the income component and leaves out capital gains of real estate assets.
2680 I. Erol and D. Tirtiroglu
The last column of Table 3 reports the Pearsoncorrelation coefficients between the returns ondifferent assets and actual inflation rates. These areall positive, but not high. REIT returns in generalmove closely with the changes in inflation rates. Thecorrelation coefficient between returns on the ISEREIT Price Index and actual inflation rates is 19.2%.Among the individual REITs, Yapi Kredi REITreturns have the highest correlation coefficient at26.1%, followed by Vakif REIT returns (23.2%),Alarko REIT returns (20.9%), and Garanti REITreturns (18.5%). The correlation coefficient betweenstock returns and inflation rates is ranging from12.1% to 18.1%. Thus, multicollinearity is not anissue for our article.
V. Analysis of Results
Below, we report empirical results first for the entiresample period and then for sub-periods of high andmoderate inflation.
Analysis of results for the entire sample period
Table 4 reports estimation results for Equation 1 forthe sample individual REITs, the ISE REIT PriceIndex and various ISE common stock indices. Allcoefficient estimates are positive, but not all of them
attain statistical significance. Among individualREITs, coefficient estimates for Vakif REIT andYapi Kredi REIT attain statistical significance at the5% level while those for the ISE REIT Price Index,Alarko REIT, EGS REIT, Garanti REIT and NurolREIT are statistically significant at 10% level. In thestock market, the only statistically significant (atthe 10% level) coefficient estimate is for the ISE-Industrial Index. These estimates are all greater thanone. So, they indicate that return on (1) someindividual REITs, (2) the ISE REIT Price Index,and (3) a diversified portfolio of industrial firmshedge more than one-to-one against the actualinflation rate and that the ISE REIT Price Indexoffers a better hedge against inflation than the ISEIndustrial Index does. Moreover, the results suggestthat REITs’ inclusion in a portfolio with less thanadequate protection against inflation is an appealingidea. We also note that Durbin–Watson (DW)statistics show no evidence of serial correlation inthe residuals.
Table 5 presents empirical results for Equation 3,under both the Fama–Schwert (1977) and the Fama–Gibbons (1982) definitions of expected inflation, forsample REIT returns, returns on the ISE REIT PriceIndex and returns on various ISE common stockindices. Empirical results are the same qualitativelyand statistically significantly under both definitionsof expected inflation. While there is statistical andeconomic significance for coefficient estimates in
Table 4. Empirical results for hedging against actual inflation: Rit ¼ aþ dpt þ eit for the entire sample period
Dependent variable � � t-Statistic (�) R2 DW statistic
ISE-Stock
Rt (ISE_ALL) �0.024 1.299 1.073 0.020 2.587Rt (ISE_100) �0.023 1.255 1.032 0.018 2.586Rt (ISE_30) �0.024 1.282 1.059 0.019 2.512Rt (ISE_Industrial) �0.018 1.231 1.401* 0.033 2.524Rt (ISE_Financial) �0.018 1.032 0.927 0.015 2.317Rt (REIT_ALL) �0.025 1.671 1.488* 0.037 2.469
Individual REITsRt (Alarko) �0.015 2.017 1.635* 0.044 2.463Rt (Atakule) �0.003 1.726 0.577 0.010 2.370Rt (EGS) �0.045 2.554 1.334* 0.030 2.339Rt (Garanti) �0.011 2.169 1.439* 0.034 2.191Rt (Ihlas) �0.012 1.536 0.866 0.013 2.234Rt (Is) �0.016 1.316 1.114 0.021 2.440Rt (Nurol) �0.019 1.813 1.423* 0.034 2.281Rt (Vakif) �0.028 2.278 1.808** 0.053 2.504Rt (Yapi Kredi) �0.041 2.866 2.051** 0.068 2.357
Notes: This table presents the OLS regression results, using monthly observations for each of nine sample individual TurkishREITs, Istanbul Stock Exchange (ISE) REIT Price Index and each of various ISE common stock indices over the sampleperiod from January 2000 to December 2004, on the hedging properties of nominal sample asset returns against actualinflation rate. When �i¼ 1, asset i offers a complete hedge against inflation.* and ** Indicate statistical significance at the 10 and 5% levels, respectively.DW denotes Durbin–Watson statistics. For �¼ 1%, DL¼ 1.38 and DU¼ 1.45.
The inflation-hedging properties of Turkish REITs 2681
Table
5.EmpiricalresultsfortheFama–SchwertModel:R
it^
aþ�Eðp
tj/t�
1Þþ
cðpt�Eðp
tj/t�
1ÞÞþe itunder
twoalternative
measuresofexpectedinflationfortheentire
sample
period
ExpectedinflationModel
1:�e t¼ðT�BillÞt�
1ExpectedinflationModel
2:�e t¼�þ�ðT�BillÞt�
1
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
ISE-stock
Rt(ISE_All)
��0.068
�1.175
1.052
2.493
0.036
�0.048
�1.095
1.048
2.494
0.036
�2.029
1.426*
2.336
1.448*
g�0.048
�0.026
�0.043
�0.023
Rt(ISE_100)
��0.067
�1.158
1.004
2.490
0.034
�0.047
�1.076
1.000
2.491
0.034
�1.984
1.388*
2.289
1.413*
g�0.088
�0.048
�0.083
�0.045
Rt(ISE_30)
��0.069
�1.194
1.064
2.412
0.036
�0.049
�1.109
1.061
2.413
0.036
�2.031
1.428*
2.346
1.456*
g�0.101
�0.055
�0.096
�0.052
Rt(ISE_Industrial)
��0.042
�0.998
1.231
2.461
0.041
�0.031
�0.981
1.228
2.462
0.041
�1.622
1.564*
1.786
1.520*
g0.510
0.382
0.512
0.383
Rt(ISE_Financial)
��0.038
�0.716
0.544
2.263
0.019
�0.029
�0.715
0.543
2.264
0.019
�1.371
1.041
1.513
1.014
g�0.408
�0.240
�0.410
�0.242
Rt(R
EIT
_Index)
��0.027
�0.504
1.289
2.463
0.047
�0.028
�0.389
1.289
2.463
0.437
�2.703
1.679**
1.733
1.683**
g1.613
0.940
1.613
0.940
IndividualREIT
sRt(A
larko)
��0.026
�0.437
1.340
2.439
0.045
�0.021
�0.467
1.340
2.439
0.045
�2.191
1.699**
2.263
1.726**
g1.695
0.899
1.699
0.902
Rt(A
takule)
�0.104
1.178
1.230
2.635
0.071
0.061
0.922
1.237
2.636
0.072
2682 I. Erol and D. Tirtiroglu
��0.574
�0.172
�1.287
�0.358
g4.152
1.228
4.161
1.231
Rt(EGS)
��0.124
�1.363*
1.518
2.218
0.051
�0.089
�1.278*
1.513
2.219
0.050
�3.873
1.724**
4.426
1.739**
g0.123
0.043
0.132
0.046
Rt(G
aranti)
��0.001
�0.016
1.034
2.206
0.035
�0.006
�0.100
1.034
2.206
0.035
�2.003
1.121
1.931
0.954
g2.475
1.075
2.477
1.076
Rt(Ihlas)
��0.215
�0.253
0.379
2.232
0.013
�0.017
�0.268
0.379
2.232
0.013
�1.691
0.805
1.758
0.739
g1.249
0.461
1.249
0.461
Rt(Is)
�0.009
0.153
0.775
2.519
0.027
�0.002
�0.053
0.778
2.519
0.027
�0.899
0.644
0.717
0.453
g2.083
1.158
2.090
1.161
Rt(N
urol)
��0.007
�0.119
1.026
2.300
0.035
�0.012
�0.263
1.027
2.300
0.035
�1.620
1.074
1.535
0.898
g2.169
1.115
2.173
1.117
Rt(V
akif)
��0.052
�0.862
1.745
2.456
0.058
�0.041
�0.895
1.742
2.456
0.058
�2.679
1.798**
2.845
1.684**
g1.539
0.801
1.545
0.803
Rt(Y
apiKredi)
��0.057
�0.846
2.117
2.324
0.069
�0.049
�0.968
2.115
2.325
0.069
�3.129
1.891**
3.236
1.725**
g2.382
1.116
2.388
1.119
Notes:
TheFama–Schwert(1977)model
decomposesactualinflation
into
itsexpected
and
unexpected
components.Weuse
both
theFama–Schwert(1977)and
Fama–Gibbons(1983)modelsto
obtain
theexpectedinflationrate
anduse
monthly
observationsforeach
ofninesample
individualTurkishREIT
s,IstanbulStock
Exchange(ISE)REIT
Price
Index
andeach
ofvariousISEcommonstock
indices
over
thesampleperiodfrom
January
2000to
Decem
ber
2004.OLSregressionsprovidethe
tabulatedresults.FamaandSchwertarguethat,given
financialtheory,both�andgshould
bepositiveex
ante.More
specifically,anasset
issaid
tohedgeagainstexpected
(unexpected)inflationwhen�¼1(g¼1).It
offersacomplete
hedgeagainst
both
components
when�¼g¼1.
*,**,and***Indicate
statisticalsignificance
at10,5and1%
levels,respectively.
DW
denotesDurbin–Watsonstatistics.For�¼1%
,D
L¼1.35andD
U¼1.48.
The inflation-hedging properties of Turkish REITs 2683
many cases against expected inflation, there is noneagainst unexpected inflation. Once again, as per DWstatistics, there is no evidence of serial correlation inthe residuals.
All ISE common stock indices, except for the ISE-Financial Index, offer a statistically significant hedge,with estimates greater than one, against expectedinflation. Both ISE-All and ISE-30 indices attain acoefficient estimate slightly greater than 2 while ISE-100’s estimate is not too far from 2. There is noevidence from any of the estimations for the ISEcommon stock indices of a hedge against unexpectedinflation. Overall, these findings indicate ISE-stocks’strong hedging performance against expected infla-tion. The coefficient estimate for the ISE REIT PriceIndex is 2.7. It is both greater and more significantthan the coefficient estimates for the statisticallysignificant estimates for the ISE common stockindices. This result suggests a strong inflation hedgingperformance of the REITs as a group. But, oneshould keep in mind the influence of the idiosyncraticrisk on the ISE REIT Price Index.
It is useful to contrast the results for the ISEcommon stock indices and the ISE REIT Price Indexin Table 5 against those in Table 4 before we proceedwith the results for the sample individual REITs inTable 5. If the ISE common stock indices hedgeagainst expected inflation, but have almost noprotection against actual inflation, then one canconclude that inflation-hedging losses in unexpectedinflation are likely to offset sufficiently the expectedinflation-hedging gains of the ISE common stockindices. Evidence for this is hidden in the negative,but not significant coefficient estimates for unex-pected inflation in Table 5. Meanwhile, the ISEIndustrial Index and the ISE REIT Price Index arethe only two indices in Table 4 with statisticallysignificant estimates. In Table 5, these estimates are,unlike those for other indices, positive, but once againnot significant. We suggest that both the ISEIndustrial Index and the ISE REIT Price Indexhave the potential to offer a rather weak protectionagainst the unexpected component of actualinflation.16
Among the sample individual REITs, AlarkoREIT, EGS REIT, Vakif REIT, and Yapi KrediREIT provide a statistically significant hedge againstexpected inflation at the 5% level. While the smallestsignificant estimate is 2.191 (for Alarko REIT), the
largest is 3.873 (for EGS). All offer strong and muchmore than a one-to-one hedge against expectedinflation. Furthermore, all coefficient estimates forunexpected inflation for the nine sample REITs arepositive, unlike those for the ISE common stockindices, but not statistically significant. These find-ings, along with those for individual REITs underactual inflation in Table 4, suggest that the sampleREITs have the potential to offer a weak protectionagainst the unexpected component of actualinflation.17
Table 1 indicates that sample REITs with coverageagainst both actual and expected inflation had jointly28% share of the total net assets and 15% share ofthe total market capitalization as of the end of oursample period. Moreover, those with coverage forexpected inflation had a focus mainly in office andshops in constructing their real estate portfolios.More specifically, EGS REIT invested 91.9% of itsreal assets in shop and 8.1% in office, Vakif REITinvested 52.3% of its real assets in office and 36.3%in shop, Yapi Kredi REIT invested 46.2% of its realassets in office and 37.2% in shop, and Alarko REITinvested 47.7% of its real assets in tourism and 30%in shop. The results suggest that REIT managersform fairly accurate inflation expectations and writelease contracts for their office and shop buildingswith clauses for rapid rent adjustments under the highand moderate inflationary conditions in Turkey.
Table 6 presents empirical results on the FDCModel in Equation 4. Under this model, the mainhypothesis is that only real factors should influenceexpected real returns and that expected real returnsare independent of inflationary expectations. Thus,the estimates for both �1 and �2 in Equation 4 shouldbe zero. Once again, empirical results are the samequalitatively and statistically significantly under boththe Fama–Schwert (1977) and the Fama–Gibbons(1982) definitions of expected inflation. None of theestimates for �1, with the exception of that forAtakule REIT, attain statistical significance.Meanwhile, the estimates for �2 for many sampleassets, including the ISE common stock indices, arehighly statistically significant, rather large andnegative. Once again, as per DW statistics, there isno evidence of serial correlation in the residuals.
The dependent variable of Equation 4 is thedifference between the nominal rate of return on theith asset and actual inflation rate at time t. Results in
16 These results differ from those in Hasbrouck (1983) and Gultekin (1983). Both show that unexpected inflation, not expectedinflation, explains the negative relationship between actual inflation and common stock returns.17Yapi Kredi REIT and Vakif REIT offer hedge against both actual and expected inflation at 5% significance level. AlarkoREIT and EGS REIT returns are statistically significant at 5% against expected inflation, but statistically significant at 10%level against actual inflation. These findings may suggest that Alarko REIT and EGS REIT have a weaker hedging capacityagainst unexpected inflation than Yapi Kredi REIT and Vakif REIT do.
2684 I. Erol and D. Tirtiroglu
Table
6.EmpiricalresultsfortheFisheriandirectcausality
model:qit¼
aþ�1pteþ�2½p
e�pte�þ
e it
tþ1
under
twoalternative
measuresofexpectedinflationfortheentire
sampleperiod
Expectedinflationmodel
1:�e t¼ðT�BillÞt�
1ExpectedinflationModel
2:�e t¼�þ�ðT�BillÞt�
1
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
ISE_Stock
�t(ISE_All)
��0.023
�0.41
2.771
2.639
0.089
�0.022
�0.499
2.776
2.639
0.089
�1
0.097
0.067
0.107
0.065
�2
�4.528
�2.191**
�5.217
2.194**
�t(ISE_100)
��0.021
�0.374
2.858
2.633
0.091
�0.021
�0.468
2.863
2.634
0.091
�1
0.029
0.02
0.029
0.018
�2
�4.645
�2.242***
�5.352
�2.245***
�t(ISE_30)
��0.022
�0.388
3.082
2.550
0.098
�0.021
�0.482
3.088
2.550
0.098
�1
0.047
0.033
0.050
0.031
�2
�4.778
�2.324**
�5.505
�2.328**
�t(ISE_Industrial)
��0.161
�0.391
1.830
2.524
0.060
�0.016
�0.484
1.832
2.524
0.060
�1
0.044
0.041
0.048
0.039
�2
�2.718
�1.785**
�3.131
�1.787**
�t(ISE_Financial)
��0.001
�0.027
2.237
2.347
0.073
�0.006
�0.146
2.246
2.348
0.073
�1
�0.495
�0.372
�0.571
�0.373
�2
�3.994
�2.085**
�4.605
�2.089**
�t(R
EIT
_Index)
�0.003
0.054
2.705
2.522
0.087
0.007
0.111
2.706
2.523
0.087
�1
�0.409
�0.305
�0.542
�0.304
�2
�4.374
�2.272**
�5.806
�2.273**
IndividualREIT
s�t(A
larko)
��0.015
�0.258
0.723
2.439
0.025
�0.009
�0.202
0.723
2.439
0.025
�1
0.623
0.409
0.714
0.408
�2
�2.025
�0.925
�2.333
�0.923
�t(A
takule)
�0.133
1.624
4.476
2.471
0.219
0.086
1.459
4.467
2.470
0.218
�1
�5.002
�1.879**
�5.756
�1.881**
�2
�24.62
�3.718***
�28.325
�3.719***
�t(EGS)
��0.03
�0.361
7.364
2.304
0.205
�0.026
�0.398
7.368
2.304
0.205
�1
0.425
0.199
0.485
0.198
�2
�10.86
�3.537***
�12.497
�3.540***
(continued
)
The inflation-hedging properties of Turkish REITs 2685
Table
6.Continued
ExpectedinflationModel
1:�e t¼ðT�BillÞt�
1ExpectedinflationModel
2:�e t¼�þ�ðT�BillÞt�
1
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
�t(G
aranti)
�0.012
0.171
0.828
2.239
0.028
0.013
0.228
0.833
2.239
0.028
�1
0.055
0.029
0.061
0.029
�2
�3.203
�1.200
�3.698
�1.204
�t(Ihlas)
��0.004
�0.047
0.352
2.270
0.012
�0.003
�0.048
0.354
2.270.
0.012
�1
0.082
0.037
0.093
0.037
�2
�2.446
�0.776
�2.824
�0.779
�t(Is)
�0.035
0.641
2.416
2.509
0.078
0.023
0.540
2.417
2.510
0.078
�1
�1.289
�0.915
�1.485
�0.917
�2
�4.441
�2.191**
�5.111
�2.191**
�t(N
urol)
��0.479
�0.008
0.455
2.330
0.016
�0.542
�0.011
0.455
2.330
0.016
�1
�0.013
�0.008
�0.017
�0.009
�2
�2.040
�0.90
�2.349
�0.901
�t(V
akif)
��0.012
�0.335
2.903
2.560
0.092
�0.015
�0.318
2.905
2.560
0.092
�1
0.525
0.349
0.599
0.346
�2
�4.599
�2.124**
�5.298
�2.127**
�t(Y
apiKredi)
��0.032
�0.490
2.420
2.394
0.078
�0.024
�0.460
2.421
2.394
0.078
�1
0.900
0.532
1.031
0.530
�2
�4.448
�1.827**
�5.125
�1.829**
Notes:
This
table
reportsresultsfrom
theFisheriandirectcausality
(FDC)model
asin
Solnik’s
(1983)work.TheFDC
model
assumes
thatonly
realfactors,such
as
productivityofcapitalandinvestors’timeandrisk
preferences,influence
expectedrealreturns.Therefore,expectedrealreturnsshould
beindependentofboth
thelevelsand
variationsin
expectedinflation,indicatingthatboth
coefficientestimatesshould
bezero.Weuse
both
theFama–Schwert(1977)andFama–Gibbons(1983)modelsto
obtain
theexpectedinflationratesanduse
monthly
observationsforeach
ofninesampleindividualTurkishREIT
s,IstanbulStock
Exchange(ISE)REIT
Price
Index
andeach
of
variousISEcommonstock
indices
over
thesample
periodfrom
January
2000to
Decem
ber
2004.OLSregressionsprovidethetabulatedresults.
*,**and***indicate
statisticalsignificance
atthe10,5and1%
levels,respectively.
DW
denotesDurbin–Watsonstatistics.For�¼1%
,D
L¼1.35andD
U¼1.48.
2686 I. Erol and D. Tirtiroglu
Tables 4 and 5 demonstrate the hedging ability of
several sample assets against actual inflation and its
expected component. So, without a necessity to
invoke the underlying theory and from a practical
empirical viewpoint, deducting actual inflation from
nominal returns should remove the significance
effects for �1, found earlier for actual inflation and
its expected component in Table 5. Our findings in
Table 6 both conform to this empirical viewpoint and
are, moreover, consistent with the first prediction of
the FDC model and positive estimates for the
expected component of inflation in Table 5.Overall, we have strong evidence that the level of
expected inflation does not associate with real returns
on both the ISE common stock indices and sample
REITs. Thus, results from the FDC model solidify
our earlier results from the Fama–Schwert (1977)
model that ISE common stock indices, the ISE REIT
Price Index and sample REITs provide a positive
hedge against expected inflation.All estimates for �2 for the ISE common stock
indices and the ISE REIT Price Index are negative
and statistically significant at the 5% level (1% level
for the ISE 100 Index). These estimates are not
consistent with the second prediction of the FDC
model and indicate that these assets’ current ex-post
real returns move strongly in the opposite direction to
the difference between the next and current time
periods’ inflationary expectations. These findings are
consistent with those in Nelson (1976) and Stulz
(1986), who documented a perverse hedging property
for common stocks.Alarko REIT, Garanti REIT, Nurol REIT, and
Ihlas REIT have real returns that exhibit indepen-
dence from both the level and the changes in the
expected inflation rate. According to the FDC model,
these REITs provide a complete inflation hedge.
When we analyse these particular REITs’ portfolios
in Table 1, we observe that, except for Alarko REIT,
they invest mainly in residential property, projects
and office property to some extent. Meanwhile, the
estimates for �2 of the remaining sample REITs are
negative, relatively large and statistically significant,
either at the 1% or 5% levels. Once again, these
estimates are not consistent with the second predic-
tion of the FDC model.
Overall, the evidence on the effect of the changes inexpected inflation on our sample assets’ real returnsindicates, except for some sample individual REITs, astrong negative association between the two vari-ables. Thus, changes in inflation expectations are oneof the determinants of real returns on some financialassets in Turkey.
In summary, empirical results in Tables 4, 5 and 6demonstrate that REITs differ considerably in theirinflation hedging abilities from the ISE commonstock indices. REITs provide a more comprehensivepositive hedging against actual inflation than do ISEcommon stock indices. Moreover, REITs’ hedgingperformance against the expected component ofinflation is stronger than that of ISE common stockindices. Also, empirical results under the Fama–Schwert model suggest, for REITs, a potentially weakpositive hedging ability against the unexpectedcomponent of inflation.18 Meanwhile, the corre-sponding evidence for the ISE common stock indicesindicates a potentially weak perverse hedging ability.
Analysis of results for the high and moderateinflation periods
Turkey has experienced a persistent hyperinflationbetween late 1970s and until the early years of the21st century, when the inflation has started to changeits course for a rather fast and sudden declining one.19
The earlier part of our sample period is years withhyperinflation while the last 3 years, from 2002 to2004, constitute a period of increasingly and rapidlydeclining inflation. So, in order to capture the suddenswitch in inflation’s course in the present work, weanalyse the inflation hedging characteristics of all ISEcommon stock indices and sample individual REITsby dividing the last 5 years into high and moderateinflation periods. As seen from Fig. 1, there is anotable difference at the inflation level before andafter 2002. During the period January 2000 toJanuary 2002, monthly average inflation rate was3.7%, while it was only 1.3% between February 2002and February 2004. So, we estimate Equations 1, 3and 4 for each sub-period to examine whetherthe inflation-hedging properties of our sample assetsare time-dependent. We subscribe only to the
18REITs, which invest mostly in shops, office buildings, and residential properties, offer hedging against inflation. Bothforeigners’ increasing investments in residential and commercial properties and the peaking prices of residential properties inTurkey over the last 2 years are likely contributors to the sample REITs’ hedging performance. For example, foreigninstitutional investors are investing and seeking to purchase shops and office buildings with long-term rent leases by credibletenants In Turkey (Colliers Resco, Real Estate Market Review: Turkey, 2005, Issue 1).19 Switches in inflation occurred in other countries, too, as observed in Khadaroo’s (2005) interesting paper.
The inflation-hedging properties of Turkish REITs 2687
Fama–Schwert (1977) definition of expected inflation
rate in these new estimations.20
Table 7 presents empirical results for the hedging
performance of the sample assets against actual
inflation.21 None of the sample assets, expect for
Ihlas REIT, offer hedging against actual inflation
under moderate inflation. Meanwhile, a much stron-
ger than one-to-one hedging ability for all sample
assets, except for Ihlas REIT, is evident during the
high inflation period. So, the high-inflation period is
fundamentally the driving force behind the statisti-
cally significant empirical results for the entire sample
period in Table 4. Thus, studying separately sub-
periods allows us to unearth new evidence on the
hedging ability, which is masked in Table 4, of all ISE
common stock indices and some of the individual
REITs. Maybe the most interesting is Is-REIT’s
estimate. It goes from being insignificant in Table 4 to
being significant at the 1% level in Table 7.
Moreover, REITs offer a stronger hedge than do
ISE common stock indices during the sub-period of
high inflation. But, we remind the potential con-
founding influence of idiosyncratic risk in individual
REIT and the ISE REIT Price Index returns.Table 8 reports empirical results for the hedging
performance of the sample assets against the expected
and unexpected components of inflation under high
and moderate inflation in Turkey.22 All statistically
significant estimates in Table 5 become even more
significant and visibly greater in magnitude.23
Moreover, the insignificant estimates for the ISE
Financial Index, Garanti REIT, Is-REIT and Nurol
Table 7. Empirical results for hedging against actual inflation: Rit ¼ �þ ��t þ "it for sample sub-periods of high inflation and
moderate inflation.
High inflation period:January 2000 to January 2002
Moderate inflation period:December 2002 to December 2004
Dependent variable � � t-Statistic(�) R2DWstatistic � � t-Statistic(�) R2
DWstatistic
ISE-Stock
Rt (ISE_All) �0.147 3.706 1.537* 0.093 2.831 0.011 0.560 0.298 0.003 2.321Rt (ISE_100) �0.147 3.663 1.525* 0.092 2.843 0.012 0.491 0.253 0.002 2.294Rt (ISE_30) �0.153 3.819 1.619* 0.102 2.771 0.014 0.319 0.161 0.001 2.251Rt (ISE_Industrial) �0.115 3.085 1.596* 0.148 2.852 0.006 0.883 0.506 0.008 2.378Rt (ISE_Financial) �0.136 3.341 1.614* 0.113 2.591 0.015 0.335 0.145 0.001 2.177Rt (REIT_Index) �0.159 4.242 2.372** 0.197 2.867 0.008 1.243 0.503 0.008 2.299
Individual REITs
Rt (Alrko) 0.097 3.919 1.658* 0.107 2.625 0.036 �1.136 �0.517 0.008 2.231Rt (Atakule) – – – – – �0.003 1.726 0.577 0.010 2.370Rt (EGS) �0.272 7.386 2.208** 0.175 2.560 0.056 �2.310 �0.618 0.011 2.140Rt (Garanti) �0.144 4.730 1.788** 0.122 2.411 0.023 1.699 0.549 0.009 2.121Rt (Ihlas) �0.062 1.836 0.765 0.025 1.751 �0.060 7.152 1.577* 0.070 2.587Rt (Is) �0.157 4.002 2.388*** 0.199 2.841 0.017 1.056 0.371 0.004 2.335Rt (Nurol) �0.109 3.728 1.799** 0.123 2.212 0.019 0.086 0.030 0.000 2.400Rt (Vakif) �0.134 4.412 1.950** 0.142 2.826 0.007 1.138 0.451 0.006 2.145Rt (YapiKredi) �0.218 6.286 2.902*** 0.268 2.528 0.005 2.136 0.685 0.014 2.502
Notes: This table replicates Table 4 estimations for the high and moderate inflation sub-periods.*, ** and *** indicate statistical significance at the 10, 5 and 1% levels, respectively.DW denotes Durbin–Watson statistics. For n¼ 25 data (high inflation period) and �¼ 1%, DL¼ 1.13 and DU¼ 1.26.For n¼ 35 data (moderate inflation period) and �¼ 1%, DL¼ 1.25 and DU¼ 1.34.
20Our empirical results in Tables 5 and 6 are practically the same under the Fama–Schwert (1977) and Fama–Gibbons (1983)definitions.21DW statistics indicate no serial correlation.22DW statistics indicate no serial correlation.23Also, the statistical significance of the estimates for the regression intercept in Table 8 exhibits visible differences from thatin Table 5. All estimates, including those for the ISE indices are negative and significant, with the exception of those forAlarko REIT and Ihlas REIT, usually at the 5% level under the sub-period of high inflation. Under the sub-period ofmoderate inflation, these estimates are all positive, with the exception of that for Ihlas REIT, and attain significance for theISE REIT price index and all individual REITs, with the exception of those for Ihlas REIT and Yapi Kredi REIT, usually atthe 10% level. The only statistically significant estimate for the ISE common stock indices is for ISE-Financial Index. This isfurther evidence of a new and different economic era under the moderate inflation period.
2688 I. Erol and D. Tirtiroglu
Table
8.EmpiricalresultsfortheFama–Schwertmodel:R
it¼
aþ�E
ptj/t�
1
�� þ
cpt�E
ptj/t�
1
��
�� þ
e it,expectedinflationmodel1:pte¼
T�Bill
ðÞ t�1forsamplesub-periods
ofhighinflationandmoderate
inflation
Highinflationperiod:January
2000to
January
2002
Moderate
inflationperiod:February
2002to
Decem
ber
2004
Dependentvariable
Coefficient
t-Statistic
F-statistic
DW
Statistic
R2
Coefficient
t-Statistic
F-statistic
DW
Statistic
R2
ISE_Stock
Rt(ISE_All)
��0.257
�2.090**
2.343
2.674
0.176
0.070
1.246
0.856
2.568
0.051
�5.443
2.072**
�0.717
�0.339
g�0.027
�0.008
1.908
0.890
Rt(ISE_100)
��0.257
�2.101**
2.353
2.679
0.176
0.072
1.252
0.840
2.540
0.050
�5.413
2.071**
�0.820
�0.377
g�0.097
�0.028
1.874
0.849
Rt(ISE_30)
��0.265
�2.220**
2.624
2.595
0.193
0.077
1.293
0.817
2.490
0.049
�5.605
2.194**
�1.019
�0.458
g�0.012
�0.006
1.731
0.767
Rt(ISE_Industrial)
��0.180
�2.276**
2.918
2.710
0.215
0.049
0.928
0.606
2.533
0.037
�4.125
2.335**
�0.034
�0.017
g0.849
0.383
1.851
0.923
Rt(ISE_Financial)
��0.217
�2.172**
2.434
2.420
0.181
0.091
1.369*
0.957
2.446
0.056
�4.628
2.164**
�1.298
�0.518
g0.574
0.205
2.059
0.811
Rt(R
EIT
_Index)
��0.221
�2.378**
3.456
2.680
0.239
0.101
1.385*
1.304
2.638
0.075
�5.225
2.628***
�0.753
�0.274
g2.130
0.817
3.349
1.201
IndividualREIT
sRt(A
larko)
��0.152
�1.216
1.610
2.549
0.128
0.106
1.618*
0.978
2.489
0.058
�4.781
1.793**
�2.655
�1.074
g2.065
0.591
0.466
0.186
Rt(A
takule)
�–
––
––
0.104
1.178
1.230
2.635
0.071
��0.574
�0.172
g4.152
1.228
Rt(EGS)
��0.467
�2.835***
4.687
2.1995
0.299
0.174
1.563*
1.019
2.258
0.060
�10.480
2.976***
�4.866
�1.158
g0.735
0.159
0.387
0.091
The inflation-hedging properties of Turkish REITs 2689
Table
8.Continued
Highinflationperiod:January
2000to
January
2002
Moderate
inflationperiod:February
2002to
Decem
ber
2004
Dependentvariable
Coefficient
t-Statistic
F-statistic
DW
Statistic
R2
Coefficient
t-Statistic
F-statistic
DW
Statistic
R2
Rt(G
aranti)
��0.203
�1.451*
1.801
2.3264
0.141
0.141
1.540*
1.368
2.332
0.079
�5.660
1.893**
�0.837
�0.243
g2.730
0.697
4.376
1.255
Rt(Ihlas)
��0.104
�0.816
0.434
1.7308
0.038
�0.045
�0.322
1.215
2.598
0.071
�2.499
0.919
6.815
1.303*
g0.409
0.115
7.507
1.418*
Rt(Is)
��0.179
�2.010**
2.832
2.7518
0.205
0.117
1.387*
1.090
2.649
0.064
�4.349
2.381***
�1.094
�0.344
g3.255
1.302*
3.325
1.032
Rt(N
urol)
��0.159
�1.459*
1.872
2.1760
0.145
0.132
1.545*
1.267
2.719
0.073
�4.512
1.931**
�2.339
�0.727
g2.042
0.667
2.646
0.812
Rt(V
akif)
��0.260
�2.319**
3.8561
2.6865
0.260
0.143
1.995**
2.724
2.555
0.146
�6.413
2.671***
�1.778
�0.660
g0.111
0.035
4.215
1.544*
Rt(Y
apiKredi)
��0.301
�2.692***
5.076
2.3501
0.316
0.097
1.040
0.931
2.734
0.055
�7.605
3.180***
0.161
0.046
g3.449
1.100
4.221
1.184
Notes:Thistable
replicatesTable
5estimations,usingonly
theFama–Schwert(1977)definitionofexpectedinflationrate,forthehighandmoderate
inflationsub-periods.
*,**and***Indicate
statisticalsignificance
atthe10,5and1%
levels,respectively.
DW
denotesDurbin–Watsonstatistics.
Forn¼25data
(highinflationperiod)and�¼1%
,D
L¼1.05andD
U¼1.21.Forn¼35data
(moderate
inflationperiod)and
�¼1%
,D
L¼1.20andD
U¼1.31.
2690 I. Erol and D. Tirtiroglu
Table
9.EmpiricalresultsfortheFisheriandirectcausality
Model:qit¼
aþ�1pteþ�2½p
e tþ1�
pte�þ
e it,expectedinflationmodel1:pte¼ðT�BillÞt�
1forsamplesub-periodsof
highinflationandmoderate
inflation
Highinflationperiod:January
2000to
January
2002
Moderate
inflationperiod:February
2002to
Decem
ber
2004
Dependentvariable
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
ISE_Stock
�t(ISE_All)
��0.216
�1.595*
2.002
2.798
0.155
0.091
1.865**
6.512
2.557
0.289
�1
3.728
1.321*
�3.494
�2.185**
�2
�2.14
�0.709
�18.061
�3.320***
�t(ISE_100)
��0.214
�1.593*
2.034
2.808
0.156
0.095
1.885**
6.700
2.539
0.295
�1
3.676
1.310*
�3.631
�2.216**
�2
�2.225
�0.742
�18.775
�3.368***
�t(ISE_30)
��0.221
�1.684**
2.298
2.724
0.173
0.099
1.917**
6.498
2.509
0.289
�1
3.818
1.394*
�3.761
�2.230**
�2
�2.302
�0.786
�18.859
�3.291***
�t(ISE_Industrial)
��0.162
�1.851**
2.021
2.757
0.155
0.064
1.304*
3.255
2.430
0.169
�1
2.756
1.507*
�2.408
�1.505*
�2
�0.918
�0.47
�12.876
�2.366**
�t(ISE_Financial)
��0.191
�1.729**
1.819
2.488
0.142
0.119
2.142**
8.678
2.529
0.352
�1
3.128
1.357*
�4.561
�2.511***
�2
�1.353
�0.549
�23.72
�3.838***
�t(R
EIT
_Index)
��0.183
�1.787**
2.547
2.717
0.188
0.128
1.956**
5.963
2.517
0.272
�1
3.177
1.488*
�4.618
�2.173**
�2
�1.829
�0.802
�22.658
�3.134***
IndividualREIT
s
�t(A
larko)
��0.143
�1.025
0.968
2.527
0.081
0.121
1.902**
2.821
2.439
0.150
�1
3.387
1.169
�4.226
�2.041**
�2
�0.370
�0.119
�11.83
�1.679**
�t(A
takule)
�–
––
–0.133
1.624
4.476
2.471
0.219
�1
�5.002
�1.879**
�2
�24.62
�3.718***
�t(EGS)
��0.305
�1.843**
7.342
2.502
0.400
0.190
1.719**
1.803
2.204
0.101
�1
5.950
1.723**
�6.469
�1.792**
�2
�8.183
�2.217**
�12.858
�1.047
The inflation-hedging properties of Turkish REITs 2691
Table
9.Continued
Highinflationperiod:January
2000to
January
2002
Moderate
inflationperiod:February
2002to
Decem
ber
2004
Dependentvariable
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
Coefficient
t-Statistic
F-statistic
DW
statistic
R2
�t(G
aranti)
��0.193
�1.235
1.114
2.276
0.092
0.168
1.961**
4.019
2.376
0.201
�1
4.123
1.266
�5.319
�1.907**
�2
�0.367
�0.106
�23.724
�2.501***
�t(Ihlas)
��0.104
�0.739
1.095
2.477
0.087
�0.004
�0.03
3.282
2.598
0.170
�1
1.607
0.544
�0.218
�0.051
�2
�0.049
�0.016
�36.295
�2.496***
�t(Is)
��0.137
�1.376*
1.620
2.672
0.128
0.145
1.888
4.911
2.511
0.235
�1
1.994
0.964
�5.046
�2.014**
�2
�1.973
�0.892
�24.038
�2.821**
�t(N
urol)
��0.159
�1.302*
1.070
2.158
0.089
0.150
1.802**
2.655
2.568
0.142
�1
3.306
1.281
�5.301
�1.959**
�2
0.049
0.018
�15.251
�1.656*
�t(V
akif)
��0.213
�1.731**
3.480
2.863
0.240
0.155
2.162**
2.946
2.482
0.156
�1
4.519
1.764**
�5.273
�2.261**
�2
�2.481
�0.906
�11.189
�1.410*
�t(Y
apiKredi)
��0.271
�2.162**
3.555
2.310
0.244
0.126
1.457*
3.882
2.668
0.195
�1
5.503
2.103**
�4.349
�1.540*
�2
�1.298
�0.464
�25.243
�2.628***
Notes:ThistablereplicatesTable6estimations,usingonly
theFama–Schwert(1977)definitionofexpectedinflationrate,forthehigh-andmoderate-inflationsub-periods.
*,**,and***Indicate
statisticalsignificance
atthe10,5and1%
levels,respectively.
DW
denotesDurbin–Watsonstatistics.
Forn¼25data
(highinflationperiod)and�¼1%
,D
L¼1.05andD
U¼1.21.Forn¼35data
(moderate
inflationperiod)and
�¼1%
,D
L¼1.20andD
U¼1.31.
2692 I. Erol and D. Tirtiroglu
REIT in Table 5 attain statistical significance either
at the 5% or 1% levels in Table 8.24 Once again, thehigh inflation period drives fundamentally the statis-tically significant empirical results in Table 5. Thereare two interesting statistically significant results,
pertaining separately to two individual sampleREITs, from the moderate inflation period. Thefirst (second) is for Ihlas REIT (Vakif REIT),respectively. Ihlas REIT exhibits statistically signifi-
cant (at the 10% level) hedging abilities against boththe expected and unexpected components of inflation.This is the first piece of evidence of hedging againstthe unexpected component of inflation from our
sample entities. Vakif REIT also exhibits a statisti-cally significant (at the 10% level) hedging ability,interestingly, only against the unexpected componentof inflation.25
Table 9 results further demonstrate the distinctly
different economic dynamics of the two inflationarysub-periods during our sample period. Estimates for�1 for all ISE indices, including the ISE REIT PriceIndex, are all positive (negative) and statistically
significant at the 10% (mostly 5%) level during thehigh (moderate) inflation sub-period, respectively.Negative estimates of �1 under the moderate inflationera offset positive ones under the high inflation era,
leaving behind statistically insignificant results for �1for the entire sample period in Table 6. Furthermore,all �2 estimates for all ISE common stock indices are
negative, very large and statistically significant atthe 1% (5% for ISE Industrial Index) level under themoderate inflation sub-period,26 while they areinsignificant under the high inflation sub-period. All
of these results are inconsistent with the predictionsof the FDC model. Thus, we uncover new evidencethat real returns on all ISE common stock indices arenot independent of the level of and the changes in
inflationary expectations and that the relative influ-ence of these inflationary factors changes from highto moderate inflation sub-periods.
Empirical results for the sample individual REITsduring the moderate inflation sub-period are pretty
much the same as those for the ISE common stockindices during the same sub-period. Meanwhile andinterestingly, empirical results for Alarko REIT,
Garanti REIT, Ihlas REIT, and Is-REIT are
consistent with both predictions of the FDC model
during the high inflation period. The results for the
first three of these REITs (i.e., except for Is-REIT)
are consistent with those in Table 6. These four
REITs have jointly 77% share of the total net assets
and 58% share of the total REIT market capitaliza-
tion. With the exception of Alarko REIT, these
particular REITs invest mainly in residential prop-
erty, office buildings and projects. Furthermore, none
of the �2 estimates, except for that for EGS REIT, for
the sample individual REITs during the high inflation
sub-period are statistically significant. Thus, only the
level of inflation expectations exhibit influence on
sample individual REIT’s real returns.Overall, these are fascinating empirical results.
They indicate that Turkish REITs’ and ISE common
stock indices’ hedging abilities, mainly against actual
and expected inflation, were better under high
inflation than under moderate inflation and that
REITs performed better than ISE common stock
indices did. A number of our empirical results in
Table 9 exhibit inconsistency with the predictions of
the FDC model. The reasons for their inconsistency
with the FDC model are important, but beyond the
scope of our article. They require further research.
VI. Summary and Concluding Remarks
Extant literature reports mixed evidence on hedging
ability against inflation of REITs. On the one hand,
several studies, such as Murphy and Kleiman (1989),
Chan et al. (1990), Park et al. (1990), and Yobaccio
et al. (1995), show that REIT returns are either
unrelated or negatively related to inflation, indicating
that REITs are alike to other equities in their ability
to hedge against inflation. On the other hand, some
other studies, such as Gyourko and Linneman (1988),
Chen and Tzang (1988), Chatrath and Liang (1998),
and Ganesan and Chiang (1998), show that REITs
provide tend to hedge positively (perversely) against
actual and expected (unexpected) inflation.
24Once again, Is-REIT’s insignificant estimate in Table 5 attains significance at the 1% level in Table 8. It is evident thateither Is-REIT’s management style or its asset portfolio or both had a strong focus on high inflation period and protectionagainst its effects.25We observe Ihlas REIT’s different behaviour in Table 7, too. Table 1 provides a decomposition of proportions of bothsample REITs’ asset- and sector-based real estate portfolios. An eye-balling examination of the portfolios of both REITs doesnot reveal any distinct investment patterns vis-a-via the portfolios of other REITs in Table 1. We also note Is-REIT’sstatistically significant estimate at the 10% level against unexpected inflation under high inflation sub-period.26DW statistics indicate no serial correlation in either period. The statistical significance of the estimates for the regressionintercept in Table 9 exhibits visible differences from that in Table 6. Moreover, those under high (moderate) inflation sub-period are negative (positive), respectively.
The inflation-hedging properties of Turkish REITs 2693
Our article provides the first evidence on inflationhedging performance of Turkish REITs by compar-ing their performance against that of indices ofcommon stocks listed on the ISE between December1999 and December 2004. We borrow from Famaand Schwert (1977) to identify empirically expectedand unexpected inflation and from Solnik’s (1983)work to implement the Fisherian direct causalitymodel. Two points motivate our investigation. First,our sample period captures both high and moderateinflation sub-periods. We offer empirical evidenceboth from the entire sample period and from eachsub-period. Second, Turkish REITs are exempt fromboth corporation and income taxes and have thefreedom to choose their dividend policies. Thus, theyhave significantly more flexibility in managing theirportfolios than their counterparts in developedcountries.
A good number of Turkish REITs and the ISEREIT Price Index hedge against actual and expectedinflation. This finding comes out for the whole sampleperiod as well as for the high inflation period ofFebruary 2002 and December 2004. In addition,hedging performance of REITs and the ISE PriceREIT Index is visibly stronger than that of the ISEcommon stock indices, that hedge against expectedinflation, both during the entire sample period and thesub-period of high inflation, and actual inflationduring the sub-period of high inflation. REIT returnsoffer a more than one-to-one hedge against actual andexpected inflation. Interestingly, the sub-period ofhigh inflation accounts mainly for the empirical resultsfor the entire sample period as empirical resultsbetween the sub-periods of high and moderateinflation differ considerably. Two (one) sampleREITs exhibit some hedging ability against unex-pected inflation during the sub-period of moderate(high) inflation, respectively. Results from theFisherian Direct Causality model show that anumber of sample REITs exhibit a hedge againstexpected and unexpected inflation both during theentire sample period and the sub-period of highinflation. These sample REITs invest mainly inresidential real estate, office buildings and/or projects.Given these results and keeping in mind the idiosyn-cratic risk in individual REIT returns, we suggest thatincluding REITs, in particular those with a portfolioof office buildings, residential real estate and/orprojects in periods of high inflation, in a well-diversified portfolio is an appealing portfolio strategy.
These results from some of the individual sampleREITs put forward the following interrelated
questions27: Why residential property should offer asafe refuge against expected and unexpected infla-tion? Can we exclude as an alternative explanationthe specific abilities of the people who manage thedifferent REITs, given that our study examinesmainly REITs on an individual basis rather thanusing an aggregate measure?
Given our data and methodology, we can not ruleout as an explanation the specific abilities of REITmanagers for the empirical results. In fact, thisexplanation gains credibility under Onder’s (2000)results that housing in various Ankara submarketsdid not hedge against inflation. Her and our findingshighlight, to some extent, the contrast in home-owners’ and REIT managers’ inflation hedgingperformance. REIT managers are professionals withsector-specific expertise and have immediate access tomore reliable, more complete and better processedinformation and data than individual homeownersdo. Given that the high inflation sub-period drivesmainly our findings, it is not unexpected that theseprofessionals would feel a lot of pressure to paymeticulous and prompt attention to inflation and itseffects. Since they were fit enough to survive underhigh inflation conditions, we suggest that theirhuman capital in forming their portfolios andmanagement of them were/are likely to be importantfactors. Market efficiency and asset liquidity argu-ments also provide another, and somewhat comple-mentary, explanation. Securitized real estate in theform of REIT shares offers a huge liquidityadvantage over nonsecuritized real estate. Givenliquidity enhancement, REIT valuations and pricesshould reflect relatively fast all available informationand market participants’ expectations. As Gatzlaffand Tirtiroglu (1995) note, there is consensus thathousing markets are not as efficient as financialmarkets.28 Thus, inflation protection provided byREITs’ residential real estate investments may be anindication of the differential in informational marketefficiency between nonsecuritized and securitizedresidential real estate markets in Turkey.
Turkish REITs have corporate and income taxesadvantages, and have freedom to choose theirdividend policy. These factors are likely to affectour empirical results, too. Sample REITs’ completefreedom over their dividend policies may enhancetheir growth through 100% plowback of profits intonew investments. According to Myers and Majluf(1984), retained earnings (i.e., internal equity) con-stitute the cheapest source of capital. In fact, REITs,as traded assets with liquidity, are likely to have some
27We thank an anonymous referee for bringing up these questions to our attention.28 In fact, Case and Shiller (1989) point out inefficiencies in several housing markets in USA.
2694 I. Erol and D. Tirtiroglu
advantages over the ISE listed common stocks. Thefirms with floating common stocks do not havecurrently the same tax breaks that REITs enjoy. Notpaying corporate taxes along with some other taxbenefits and having access to 100% of profits,especially under high inflationary conditions, giveREIT managers a lot of inexpensive capital forinvestment and asset management purposes. Thus, wethink that these tax break and dividend policyadvantages contribute positively to REITs’ domi-nance in inflation-hedging over the ISE commonstock indices during our sample period.
Finally, we report a few REITs’ hedging abilityagainst the unexpected component of inflation,especially, under moderate inflation. This findingcontrasts starkly with the perverse hedging abilitydocumented for REITs in other countries in theliterature. Given that the moderate inflation is arelatively new process in Turkey and that REITmanagers and all other market participants arerelatively inexperienced in this new process, howREITs’ and common stocks’ hedging performanceagainst unexpected inflation in Turkey will endure inthe near and long term poses itself as an importantand interesting future research question.
Acknowledgements
We thank Professor Mark Taylor, the editor, and ananonymous referee for their constructive commentson our article and the participants of the 2006European Real Estate Society meetings for theirsuggestions. All remaining errors are ours.
References
Aydinoglu, C. (2004) Turkish REITs: an overview of theindustry and its performance, Master Thesis,Department of Architecture, Massachusetts Instituteof Technology.
Brooks, C. and Tsolacos, S. (2001) Linkages betweenproperty asset returns and interest rates: evidence forthe UK, Applied Economics, 33, 711–19.
Brueggeman, W., Chen, A. and Thibodeau, T. (1984)Real estate investment funds: performance and port-folio considerations, The Journal of the American RealEstate and Urban Economics Association, 13, 333–54.
Case, K. E. and Shiller, R. E. (1989) The efficiency ofmarket for single-family homes, American EconomicReview, 79, 125–137.
Chan, K. C., Hendershott, P. H. and Sanders, A. B. (1990)Risk and return on real estate: evidence for equityREITs, The Journal of the American Real Estate andUrban Economics Association, 18, 431–52.
Chatrath, A. and Liang, Y. (1998) REITs and inflation: along-run perspective, Journal of Real Estate Research,16, 311–25.
Chen, K. and Tzang, D. (1988) Interest-rate sensitivity ofreal estate investment trusts, Journal of Real EstateResearch, 3, 13–22.
Coleman, M., Hudson-Wilson, S. and Webb, J. R. (1994)Real estate in the multi-asset portfolio, in ManagingReal Estate Portfolios (Eds) S. Hudson-Wilson,C. H. Wurtzebach, and Richard D. Irwin, BurrRidge, IL and New York, Vol. 3, pp. 98–123.
Erol, I. and Patel, K. (2005) Default risk of wage-indexedpayment mortgage (WIPM) contract in Turkey,Journal of Housing Economics, 14, 271–93.
Fama, E. F. and Schwert, G. W. (1977) Asset returnsand inflation, Journal of Financial Economics, 5,115–46.
Fama, E. and Gibbons, M. (1982) Inflation, real returnsand capital investment, Journal of MonetaryEconomics, 9, 292–323.
Fisher, I. (1930) The Theory of Interest, MacMillan,New York.
Ganesan, S. and Chiang, Y. H. (1998) The inflationhedging characteristics of real and financial assets inHong Kong, Journal of Real Estate PortfolioManagement, 4, 55–67.
Gatzlaff, D. and Tirtiroglu, D. (1995) Real estate marketefficiency: issues and evidence, Journal of Real EstateLiterature, 3, 157–89.
Gultekin, N. B. (1983) Stock market returns and inflation:evidence form other countries, Journal of Finance, 38,49–65.
Gunay, A., Metin-Ozcan, K. and Yeldan, E. (2005) Realwages, profit margins and inflation in turkish manu-facturing under post-liberalization, Applied Economics,37, 1899–905.
Gyourko, J. and Linneman, P. (1988) Owner-occupiedhomes, income-producing real property, and REITs asinflation hedges, Journal of Real Estate Finance andEconomics, 1, 347–72.
Hartzell, D., Hekman, J. S. and Miles, M. E. (1987) Realestate returns and inflation, The Journal of theAmerican Real Estate and Urban EconomicsAssociation, 15, 617–37.
Hasbrouck, J. (1983) Stock returns, inflation, and economicactivity: the survey evidence, Journal of Finance, 39,1293–310.
Hoesli, M. (1994) Real estate as hedge against inflation:learning from the Swiss case, Journal of PropertyValuation and Investment, 12, 51–9.
Hoesli, M., MacGregor, B. D., Matsiyak, G. andNanthakumaran, N. (1997) The short-term inflationhedging characteristics of U.K. real estate, Journal ofReal Estate Finance and Economics, 15, 27–57.
Ibbotson, R. and Siegel, L. (1984) Real estate returns: acomparison with other investments, The Journal of theAmerican Real Estate and Urban EconomicsAssociation, 12, 219–42.
Khadaroo, A. J. (2005) A threshold in inflation dynamics:evidence from emerging countries, Applied Economics,37, 719–23.
Limmack, R. J. and Ward, C. W. R. (1988) Propertyreturns and inflation, Land Development Studies, 5,47–55.
Liu, C. H., Hartzell, D. J. and Hoesli, M. E. (1997)International evidence on real estate securities asan inflation hedge, Real Estate Economics, 25,193–221.
The inflation-hedging properties of Turkish REITs 2695
Miles, D. (1996) Property and inflation, Journal of PropertyFinance, 7, 6–20.
Miles, M. and Mahoney, J. (1997) Is commercial realestate an inflation hedge?, Real Estate Finance, 13,31–45.
Murphy, J. and Kleiman, R. (1989) The inflation-hedgingcharacteristics of equity REITs: an empirical study,Quarterly Review of Economics and Business, 29,95–101.
Myers, S. and Majluf, N. S. (1984) Corporate financing andinvestment decisions when firms have information thatinvestors do not have, Journal of Financial Economics,13, 187–221.
Nelson, C. (1976) Inflation and rates of return on commonstock, Journal of Finance, 31, 471–83.
Newell, G. (1995) Is Canadian real estate a hedgeagainst inflation?, The Canadian Appraiser, 39,25–27.
Newell, G. and Boyd, T. (1995) Inflation-hedging attributesof New Zealand commercial property, New ZealandValuers’ Journal, December, 50–54.
Onder, Z. (2000) High inflation and returns on residentialreal estate: evidence form Turkey, Applied Economics,32, 917–31.
Park, Y. J., Mullineaux, D. J. and Chew, I. (1990) AreREITs inflation hedges?, Journal of Real EstateFinance and Economics, 3, 91–103.
Rubens, J, Bond, M. and Webb, J. (1989) The inflation-hedging effectiveness of real estate, Journal of RealEstate Research, 4, 45–56.
Sing, T. and Low, S. Y. (2000) The inflation-hedgingcharacteristics of real estate and financial assetsin Singapore, Journal of Real Estate PortfolioManagement, 6:4, 373–85.
Solnik, B. (1983) The relation between stock prices andinflationary expectations: the international evidence,Journal of Finance, 38, 35–48.
Spyrou, S. I. (2005) Are stocks a good hedge againstinflation? Evidence from emerging markets, AppliedEconomics, 36, 41–8.
Stulz, R. (1986) Asset pricing and expected inflation,Journal of Finance, 41, 209–24.
Us, V. (2004) Monetary transmission mechanism in Turkeyunder the monetary conditions index: an alternativepolicy rule, Applied Economics, 37, 967–76.
Wurtzebach, C. H., Mueller, G. R. and Machi, D. (1991)The impact of inflation and vacancy on real estatereturns, Journal of Real Estate Research, 6:2, 153–8.
Yobaccio, E., Rubens, J. and Ketcham, D. C. (1995) Theinflation-hedging properties of risk assets: the case ofREITs, Journal of Real Estate Research, 10, 279–96.
Capital Markets Board (1998) Principles CommuniquePertaining to Investment Trusts, Available at http://www.spk.gov.tr (accessed November 1998).
Capital Markets Board (2004) Principles CommuniquePertaining to Investment Trusts, Available at http://www.spk.gov.tr (accessed May 2004).
Central Bank of Turkey http://www.tcmb.gov.trState Institute of Statistics http://www.die.gov.trIstanbul Stock Exchange http://www.imkb.gov.tr
2696 I. Erol and D. Tirtiroglu