The Journal of Real Estate Finance and Economics

, 36:165

Real Estate Risk Exposure of Equity Real Estate Investment Trusts

Authors

    • Department of Finance, College of ManagementNational Yunlin University of Science and Technology
  • Ming-Te Lee
    • Department of FinanceYuan Ze University
  • Kevin C. H. Chiang
    • School of Business AdministrationUniversity of Vermont
Article

DOI: 10.1007/s11146-007-9058-2

Cite this article as:
Lee, M., Lee, M. & Chiang, K.C.H. J Real Estate Finan Econ (2008) 36: 165. doi:10.1007/s11146-007-9058-2

Abstract

This study examines the linkage between equity real estate investment trust (REIT) returns and the private real estate factor. The results reveal a tighter connection between REIT and the private real estate market starting from 1993. In addition, large-cap REITs seem to behave more like real estate than do small-cap REITs. Overall, the results are consistent with three notions: (1) that institutional investors provide information-gathering services (Bradrinath et al., Rev. Financ. Stud., 8:401–430, 1995), (2) that a more sophisticated investor base improves information flow, and (3) that a high degree of participation from institutional investors strengthens the linkage between REIT returns and the underlying real estate factor (Ziering et al., The evolution of public and private market investing in the new real estate capital markets, Prudential Real Estate Investors, Parsippany, NJ, 1997).

Keywords

REITReal estateAsset pricingReal estate factor

Introduction

One of the main reasons the US Congress created real estate investment trusts (REITs) was to provide economies of scale to individual investors participating in the risk–return tradeoff of large-scale real estate properties (Block 2002). Because of this congressional intent, it is not surprising that Giliberto (1990) finds a fundamental link between equity REIT returns and unsecuritized real estate returns. The author shows that Russell–NCREIF property returns and the National Association of Real Estate Investment Trusts (NAREIT) returns are significantly, positively correlated over the 1978–1989 period after stock and bond market influences are removed from the two return series.

The recent development of REIT markets has added challenges to the study of the link between REIT returns and underlying, unsecuritized real estate returns. According to the NAREIT, the total market capitalization of equity REITs increased from $0.33 billion in 1971 to $151.27 billion in 2002. In addition, increasing participation by institutional investors resulted in a structural change in the early 1990s (Glascock et al. 2000).1 Wang et al. (1995) document that institutional ownership in REITs between 1979 and 1990 ranged from 6.66 to 15.60%. By 1995, institutional participation had increased to 30%, according to Chan et al. (1998).

Two recent studies examine the linkage between REIT returns and unsecuritized real estate returns for the periods before and after the structure change. Glascock et al. (2000) hypothesize that the structure change may allow REITs to behave more like traditional stocks than like real estate. They find that equity REITs behave more like stocks after 1992, although equity REITs are cointegrated with private real estate during their entire sample period. Their results imply that equity REITs provide less real estate exposure to investors after 1992.2 In contrast, Clayton and Mackinnon’s (2003) regression analysis shows that equity REIT returns become increasingly sensitive to the performance of underlying real estate. Their results are consistent with the notion that REIT prices are better linked to real estate market fundamentals and behave more like real estate after 1992 (Ziering et al. 1997). One puzzling finding in Clayton and Mackinnon is that small-cap REITs behave more like real estate than do large-cap REITs after 1993. The result is difficult to reconcile with the idea that a more sophisticated investor base in large-cap REITs improves information flow and helps REIT prices better reflect the performance of underlying real estate (Clayton and Mackinnon 2003, p. 40; Ziering et al. 1997).

The purpose of this study is to use the Fama and French (1993) model to examine the real estate risk exposure of equity REITs for periods before and after the structural change. Specifically, this study asks two questions:
  1. (1)

    Is there a real estate factor in equity REIT returns when the Fama–French stock and bond factors are controlled for?

     
  2. (2)

    If there is such a real estate factor, how does the role of this factor vary over time?

     

The Fama and French model is used for three reasons: (1) it is widely used in real estate studies, including Hsieh and Peterson (2000) and He (2002); (2) its popularity makes it an ideal tool to check the robustness of existing results; and (3) its empirical usefulness is well-documented.

Consistent with Giliberto (1990), this study finds evidence suggesting that the real estate factor plays an important role in explaining quarterly equity REIT returns. In addition, this study documents that large-cap REITs appear to behave more like real estate than do small-cap REITs.

Data

This study uses the following return series during the 1978–2003 period:
  1. 1.

    NAN = the monthly returns on the NYSE/ASE/NASDAQ value-weighted index from the Center for Research in Security Prices (CRSP) stock files.

     
  2. 2.

    SMB = the monthly returns on the size mimicking portfolio. They are the differences between the simple averages of the returns on the three small stock portfolios and the three big stock portfolios.

     
  3. 3.

    HML = the monthly returns on the book-to-market mimicking portfolio. They are the differences between the simple averages of the returns on the two high book-to-market portfolios and the two low book-to-market portfolios.

     
  4. 4.

    LONG = the monthly returns on long-term US government bonds.

     
  5. 5.

    CORP = the monthly returns on long-term corporate bonds.

     
  6. 6.

    SHORT = the monthly returns on 1-month Treasury bills.

     
  7. 7.

    EREIT = the monthly returns on the NAREIT value-weighted equity REITs.

     
  8. 8.

    NC = the quarterly Russell–NCREIF property returns.

     

Kenneth French provides the SMB and HML factor returns. Return series on US government bonds, corporate bonds, and Treasury bills are collected from the 2005 Stocks, Bonds, Bills, and Inflation (SBBI) Yearbook. Equity REIT returns are provided by the NAREIT. The Russell–NCREIF property return series is collected from the National Council of Real Estate Investment Fiduciaries. Because NC is a quarterly series, all other returns are compounded into quarterly series so that they can be analyzed at the same frequency.

Methodology

The baseline analysis of this study consists of two sets of regressions whose specifications are based on the Fama and French (1993) model. Peterson and Hsieh (1997) show that equity REITs and stocks share common risk factors. Therefore, this study first utilizes the following four-factor models, which is similar to the one used in Hsieh and Peterson (2000):3
$$ {\text{EXREIT}} = \alpha + \beta _{{\text{M}}} {\text{MKT}} + \beta _{{\text{S}}} {\text{SMB}} + \beta _{{\text{H}}} {\text{HML}} + \beta _{{\text{N}}} {\text{NCR}} + \varepsilon _{p} $$
(1)
where EXREIT = EREIT − SHORT, MKT = NAN − SHORT, and NCR = NC − SHORT.

Tuluca et al. (2000) argue that private real estate markets seem to informationally lead public real estate markets. Specifically, insider trading may incorporate the prices established by the appraisal process in private real estate markets into the public prices of REITs.4 In contrast, Block (2002), Graff and Young (1997), Linneman (2000), and Murphy et al. (2003) suggest capital switching activities between public and private real estate markets.

According to Liziere and Satchell (1997), the capital switching concept originates from the urban social science literature (Harvey 1975, 1978, 1982, 1985, and 1989). Following Marx’s view of financial markets, Harvey speaks of “capital switching” when capital (investment) flows from a less profitable sector to a more profitable one (a sector taking in cash flows and acting as a substitute for the capital out-flowing sector) (Aalbers 2006). Thus, falling returns in one sector should be reflected in rising returns in the competing sector under a capital-switching scenario (Liziere and Satchell 1997). In a causality framework, capital switching expects a negative correlation between lead-lag returns of the sectors (Liziere and Satchell 1997).

King (1989a, b, c), Haila (1991), Aalbers (2006) further argues that capital not only switches between different sectors of the economy, but also within sectors of the economy, between forms of property and between places.5 Linneman (2000) and Clayton and MacKinnon (2002) suggest that liquidity plays a key role in capital switching between REITs and private properties. When disenchanted with the performance of their private property investments, institutional investors look into REITs for a more liquid way to gain exposure to the real estate asset class (Clayton and MacKinnon 2002).6 In contrast, when investors are content with the performance of their private real estate investments, they are less concerned with liquidity and may have a weaker demand for REITs.

This line of reasoning implies that since early 1990s institutional investors have been viewing REITs as potential substitutes for private real estate investments (in commingled or separate accounts) (Linneman 2000). Given the substitution relationship, if private real estate investments depreciate in value, investors may extrapolate that private real estate no longer represents an attractive investment opportunity and rebalance into the REIT market.7 This would cause a negative relation between the lead-lag returns of private real estate and REITs.8 In short, institutional investors may have been move in and move out the REIT market as a group conditioning upon the performance of private real estate markets (Graff and Young 1997; Linneman 2000).

To capture the possible informational lead-lag relation and capital switching between public and private real estate markets, this study includes lagged NCR into the specification.9 This leads to the following equation:10
$$ {\text{EXREIT}} = \alpha + \beta _{{\text{M}}} {\text{MKT}} + \beta _{{\text{S}}} {\text{SMB}} + \beta _{{\text{H}}} {\text{HML}} + \beta _{{\text{N}}} {\text{NCR}} + \beta _{{{\text{LN}}}} {\text{Lag}}{\left( {{\text{NCR}}} \right)} + \varepsilon _{p} $$
(2)
where Lag(NCR) is the one-quarter lagged value of NCR. Intuitively, the informational lead-lag relation suggests positive influences of lagged NCR on EXREIT, whereas capital switching activities imply negative influences.
Clayton and Mackinnon (2003) show that bond factors drive equity REIT returns over the 1993–1998 period. This study extends the six-factor model of He (2002) as follows:
$$ {\text{EXREIT}} = \alpha + \beta _{{\text{M}}} {\text{MKT}} + \beta _{{\text{S}}} {\text{SMB}} + \beta _{{\text{H}}} HML + \beta _{{\text{T}}} {\text{TERM}} + \beta _{{\text{D}}} {\text{DEF}} + \beta _{{\text{N}}} {\text{NCR}} + \varepsilon _{p} $$
(3)
where TERM = LONG − SHORT and DEF = CORP − LONG.
Extending Eq. 3 to capture the possible informational lead-lag relation and capital switching between public and private real estate markets leads to the following equation:
$$\begin{array}{*{20}c} {{\text{EXREIT}} = \alpha + \beta _{{\text{M}}} {\text{MKT}} + \beta _{{\text{S}}} {\text{SMB}} + \beta _{{\text{H}}} {\text{HML}} + \beta _{{\text{T}}} {\text{TERM}} + \beta _{{\text{D}}} {\text{DEF}}} \\ { + \beta _{{\text{N}}} {\text{NCR}} + \beta _{{{\text{LN}}}} {\text{Lag}}{\left( {{\text{NCR}}} \right)} + \varepsilon _{p} } \\ \end{array} $$
(4)

To examine whether there is a real estate factor in equity REIT returns and how the role of this factor evolves over time, this study estimates Eqs. 1, 2, 3, and 4 over both the full sample period (1978–2003) and three sub-periods—1978–1990, 1978–1992, and 1993–2003. To mitigate the concern about data mining, this study selects the 1993 cutoff point used in Clayton and Mackinnon (2003; Table 4). By doing so, our results can be directly related to existing results. The reason for excluding the 1991–1992 data for the first sub-period will become obvious when we present summary statistics for our data.

The price discovery of REITs with respect to their underlying fundamentals is a function of information gathering. A popular thought is that a more sophisticated investor base would improve information flow and that a higher degree of participation by institutional investors would facilitate information gathering (Ziering et al. 1997). Because institutional investors are interested primarily in large-cap REITs, one would expect that large-cap and small-cap REIT returns are linked to the performance of underlying real estate in a different way. For this reason, this study follows Graff and Young (1997) and groups REITs into two size categories: “large-cap REITs” with a market capitalization of $100 million or more, and “small-cap REITs” with a market capitalization of less than $100 million.

For this size-based analysis, we use the CRSP/Ziman Real Estate Database whose data starts in 1980. Two equal-weight portfolio excess return series, EXBREIT and EXSREIT, are computed for big- and small-cap equity REITs, respectively. The study repeats the regression analyses with the replacement of EXREIT by EXBREIT and EXSREIT for the period from 1993 to 2003. The size-based regression starts from 1993 because the REIT industry was dismissed as insignificant by institutional investors until 1993–1994 (Block 2002). In addition, by doing so, our results can be directly related to the results in Clayton and Mackinnon (2003; Table 4). Equal weighting is used for two reasons. First, this study is interested in addressing the question of whether big-cap (small-cap) REIT returns on average reflect the performance of underlying real estate (Loughran and Ritter 2000). Second, Clayton and Mackinnon (2003) use equal weighting to construct their large-cap and small-cap REIT portfolios. This study follows this convention so that our results can be directly related to existing results.

Description of Variables

Table 1 presents summary statistics for the nine variables in our analysis during the following three sub-periods: 1980–1990, 1991–1992, and 1993–2003.11 This table also reports the correlation structures relative to the excess return on NAREIT equity REITs (EXREIT). The reason for separating the 1991–1992 period from the full sample is that during this time period there was market expectation about the Revenue Reconciliation Act of 1993. The tax legislation included in the Act made large-scale investments in REITs more desirable to institutional investors. In addition, a unique market condition during the period 1991–1992 requires particular attention. Commercial real estate prices reached their lows during this period. At the same time, REITs were able to pick up properties at fire-sale prices from banks, insurance companies, real estate limited partnerships, and the Resolution Trust Corporation. Consequently, REITs were doing well from 1991 to 1993 (Block 2002). This market condition may temporarily distort REITs’ exposure to the real estate factor. This may also be particularly true for large-cap REITs because they are in better financial conditions to buy properties during the period 1991–1992.
Table 1

Summary statistics

 

1980–1990

1991–1992

1993–2003

Mean (standard deviation)

 EXREIT

1.24

4.71

2.25

(6.67)

(6.97)

(6.90)

 EXBREIT

2.21

6.24

2.98

(7.01)

(8.67)

(7.21)

 EXSREIT

0.42

2.51

3.07

(8.76)

(12.54)

(6.64)

 MKT

1.86

3.81

1.93

(9.02)

(5.64)

(8.99)

 SMB

−0.15

2.49

0.57

(4.40)

(6.55)

(5.56)

 HML

1.14

1.31

1.27

(5.77)

(7.28)

(7.37)

 TERM

1.46

2.16

1.25

(7.56)

(4.01)

(4.62)

 DEF

0.00

0.20

−0.14

(1.77)

(1.42)

(1.93)

 NCR

0.39

−2.36

1.24

(1.08)

(1.86)

(0.91)

Correlation coefficient with EXREIT

 EXREIT

100.00

100.00

100.00

 EXBREIT

87.54

96.67

97.25

 EXSREIT

88.27

55.14

66.89

 MKT

82.23

86.72

33.86

 SMB

64.56

51.72

38.43

 HML

−32.50

−45.82

30.31

 TERM

40.44

7.43

1.39

 DEF

−13.85

21.32

10.00

 NCR

27.43

7.73

−2.26

 Lag (NCR)

8.92

−8.06

−42.63

Correlation coefficient with EXBREIT (EXSREIT), 1993–2003

 EXREIT

  

97.25

   

(66.89)

 EXBREIT

  

100.00

   

(65.55)

 EXSREIT

  

65.55

   

(100.00)

 MKT

  

31.41

   

(28.68)

 SMB

  

43.90

   

(29.90)

 HML

  

31.67

   

(23.22)

 TERM

  

−2.15

   

(0.30)

 DEF

  

16.47

   

(10.14)

 NCR

  

−2.35

   

(−19.23)

 Lag (NCR)

  

−42.44

   

(−55.02)

The reported numbers are in percentage and their calculations are based on quarterly returns. EXREIT is the difference between the return on the NAREIT equity REIT index and the return on 1-month Treasury bills. MKT is the difference between the return on the CRSP value-weight portfolio and the return on 1-month Treasury bills. SMB is the difference between the return on small-cap stocks and the return on large-cap stocks. HML is the difference between the return on high book-to-market stocks and the return on low book-to-market stocks. TERM is the difference between the return on long-term US government bonds and the return on 1-month Treasury bills. DEF is the difference between the return on long-term corporate bonds and the return on long-term US government bonds. NCR is the difference between the return on the Russell–NCREIF property index and the return on 1-month Treasury bills. EXBREIT is the difference between the return on the large-cap equity REIT index and 1-month Treasury bills. EXSREIT is the difference between the return on the small-cap equity REIT index and 1-month Treasury bills.

Over the first sub-period, 1980–1990, the average quarterly excess return for big-cap REITs is 2.21%, which is higher than that for small-cap REITs, 0.42%. The broader-based NAREIT Index yields an average excess return of 1.24% per quarter. The average value of NCR is 0.39% per quarter.

The 1991–1992 period exhibits unusual return patterns. Big-cap REITs yields an average quarterly excess return of 6.24% while direct commercial real estate has an average quarterly excess return of −2.36%. Casual observations suggest that market expectation about the potential impacts of the Revenue Reconciliation Act of 1993 introduces a transitory surge in demand for REIT shares at the time when underlying real estate does not perform well. This is particularly true for big-cap REITs because the Act intends to promote institutional participation. It appears that there is a spillover effect on small-cap REITs. The average quarterly excess return for small-cap REITs is 2.51%. Because this important event and the unusual return patterns in the 1991–1992 real estate market may skew test results, this study performs two sets of regression analyses for the vintage REIT era: one with the 1991–1992 data, and the other without the 1991–1992 data.

Over the new REIT era, 1993–2003, big-cap and small-cap REITs have similar average excess returns. They are 2.98 and 3.07% per quarter, respectively. During this time period, the average return for NCR is 1.24% per quarter.

There are considerable time variations in REIT and factor returns. The stock market factor has the highest standard deviations of 9.02 and 8.99% per quarter for the 1980–1990 period and the 1993–2003 period, respectively. During the 1991–1992 period, small-cap REITs have the highest standard deviation of 12.54% per quarter. The real estate factor proxy has the smallest standard deviations. The standard deviations of NCR are 1.08, 1.86, and 0.91% per quarter for the 1980–1990 period, the 1991–1992 period, and the 1993–2003 period, respectively.

It is not surprising that EXREIT and the stock market factor move fairly together because they are traded on the same platform. Also, consistent with Chiang et al. (2005) and Clayton and Mackinnon (2003), this study finds that the correlations between EXREIT and the stock market factor decline after 1993. The correlation coefficients between the two series are 0.8223, 0.8672, and 0.3386 over the three sub-periods. This study also confirms Clayton and Mackinnon’s (2003) result that EXREIT does not appear to co-move with NCR. The correlation coefficients between the two series are 0.2743, 0.0773, and −0.0226 over the three sub-periods.

Another notable observation is that the comovement structure between big-cap REITs and stocks is slightly tighter than that between small-cap REITs and stocks. The correlation coefficient between EXBREIT and MKT over the 1993–2003 period is 0.3141, whereas the correlation coefficient between EXSREIT and MKT is 0.2868.

The Baseline Results

Table 2 reports the test results using the NAREIT excess return as the dependent variable and NCR as the real estate factor. The results in “The four-factor model” are based on the four-factor model in Eq. 1, and the results in “The six-factor model” are based on the six-factor model in Eq. 3. Over the full sample period, 1978–2003, the estimates for MKT, SML, and HML are all statistically significant at the 1% level under both the four-factor model and the six-factor model. The result suggests that stock market factors are useful in explaining aggregate REIT returns. The estimate for TERM is significant at the 5% level under the six-factor model. The estimate for DEF is far from being significant at any conventional level. The results are in line with Peterson and Hsieh (1997; Table 3) whose findings indicate that equity REIT returns are somewhat sensitive to TERM, but hardly sensitive to DEF. Furthermore, this study confirms Clayton and Mackinnon’s finding that NCR alone is not useful in explaining the NAREIT returns.12 The estimates for NCR are not statistically significant at any conventional level.
Table 2

Regression of quarterly EXREIT on concurrent risk factors

 

1978–2003

1978–1990

1978–1992

1993–2003

The four-factor model

 Constant

−0.00

−0.01

0.00

0.00

(−0.01)

(−1.38)

(0.02)

(0.26)

 MKT

0.57

0.66

0.66

0.47

(8.58)**

(8.52)**

(8.05)**

(4.71)**

 SMB

0.39

0.29

0.31

0.47

(4.12)**

(2.35)*

(2.74)**

(3.35)**

 HML

0.50

0.36

0.26

0.70

(6.19)**

(3.42)**

(2.53)*

(6.06)**

 NCR

0.16

0.94

0.28

−0.11

(0.53)

(2.35)*

(0.98)

(−0.14)

 R2

0.58

0.76

0.72

0.58

The six-factor model

 Constant

−0.00

−0.01

0.00

0.00

(−0.33)

(−1.46)

(0.08)

(0.18)

 MKT

0.52

0.56

0.55

0.50

(7.48)**

(6.52)**

(6.08)**

(4.75)**

 SMB

0.45

0.34

0.37

0.51

(4.76)**

(2.82)**

(3.14)**

(3.47)**

 HML

0.47

0.28

0.19

0.71

(5.71)**

(2.64)**

(1.81)

(5.86)**

 TERM

0.18

0.20

0.20

−0.06

(2.01)*

(2.58)**

(2.36)*

(−0.23)

 DEF

−0.13

0.33

0.30

−0.56

(−0.44)

(1.08)

(0.88)

(−0.88)

 NCR

0.27

1.13

0.42

−0.11

(0.92)

(2.90)**

(1.45)

(−0.14)

 R2

0.61

0.79

0.74

0.59

The reported numbers are based on quarterly returns. EXREIT is the difference between the return on the NAREIT equity REIT index and the return on 1-month Treasury bills. MKT is the difference between the return on the CRSP value-weight portfolio and the return on 1-month Treasury bills. SMB is the difference between the return on small-cap stocks and the return on large-cap stocks. HML is the difference between the return on high book-to-market stocks and the return on low book-to-market stocks. TERM is the difference between the return on long-term US government bonds and the return on 1-month Treasury bills. DEF is the difference between the return on long-term corporate bonds and the return on long-term US government bonds. NCR is the excess return on the Russell–NCREIF property index. The t statistics are in parentheses.

**Significant at the 1% level. *Significant at the 5% level.

Table 3

Regression of quarterly EXREIT with lagged NCR

 

1978Q2–2003

1978Q2–1990

1978Q2–1992

1993–2003

The four-factor model

 Constant

0.00

−0.01

−0.00

0.01

(0.39)

(−1.28)

(−0.01)

(0.71)

 MKT

0.56

0.67

0.66

0.39

(8.45)**

(8.10)**

(7.73)**

(4.33)**

 SMB

0.37

0.28

0.31

0.42

(3.83)**

(2.08)*

(2.44)*

(3.30)**

 HML

0.49

0.36

0.26

0.57

(6.02)**

(3.16)**

(2.49)*

(5.13)**

 NCR

0.52

0.93

0.27

2.18

(1.40)

(2.04)*

(0.78)

(2.18)*

 Lag(NCR)

−0.61

0.03

0.01

−2.69

(−1.61)

(0.05)

(0.03)

(−3.31)**

 R2

0.59

0.76

0.72

0.67

The six-factor model

 Constant

0.00

−0.01

0.00

0.01

(0.11)

(−1.54)

(0.05)

(0.55)

 MKT

0.52

0.55

0.55

0.42

(7.40)**

(6.07)**

(5.76)**

(4.36)**

 SMB

0.44

0.34

0.38

0.46

(4.52)**

(2.50)*

(2.93)**

(3.46)**

 HML

0.46

0.25

0.19

0.57

(5.59)**

(2.16)*

(1.76)

(4.89)**

 TERM

0.18

0.22

0.20

0.01

(1.98)*

(2.62)**

(2.33)*

(0.05)

 DEF

−0.19

0.39

0.32

−0.42

(−0.62)

(1.19)

(0.90)

(−0.73)

 NCR

0.67

1.02

0.37

2.18

(1.83)

(2.34)*

(1.07)

(2.16)*

 Lag(NCR)

−0.67

0.33

0.09

−2.69

(−1.81)

(0.65)

(0.26)

(−3.26)**

 R2

0.63

0.79

0.74

0.68

The reported numbers are based on quarterly returns. EXREIT is the difference between the return on the NAREIT equity REIT index and the return on 1-month Treasury bills. MKT is the difference between the return on the CRSP value-weight portfolio and the return on 1-month Treasury bills. SMB is the difference between the return on small-cap stocks and the return on large-cap stocks. HML is the difference between the return on high book-to-market stocks and the return on low book-to-market stocks. TERM is the difference between the return on long-term US government bonds and the return on 1-month Treasury bills. DEF is the difference between the return on long-term corporate bonds and the return on long-term US government bonds. NCR is the excess return on the Russell–NCREIF property index. The t statistics are in parentheses.

**Significant at the 1% level. *Significant at the 5% level.

Applying the same baseline analysis to the three sub-periods of 1978–1990, 1978–1992, and 1993–2003 reveals that the risk exposures of the NAREIT returns evolve over time. A notable observation is that, regardless of whether the four-factor model or the six-factor model is used, the fit of regression declines considerably. Under the four-factor model, the R-squared values are 76, 72, and 58%. Under the six-factor model, the R-squared values are 79, 74, and 59%. This decline is driven largely by a substantial decrease in NAREIT return’s sensitivity to MKT. On the other hand, the return sensitivities to SMB and HML increase substantially after 1992. These patterns are in line with Clayton and Mackinnon (2003; Table 2) whose findings indicate that the sensitivity of the NAREIT returns to large cap stocks has declined over time and, conversely, the sensitivity to small cap stocks has emerged recently.

The estimates for TERM are statistically significant under the six-factor model over the sub-periods of 1978–1990 and 1978–1992. Nevertheless, TERM loses its explanatory ability in describing NAREIT returns after 1992. The results are in line with Clayton and Mackinnon (2003; Table 2) whose findings indicate that equity REIT returns are firmly sensitive to bond factors before 1992 and are likely to become insensitive after that. The estimates for DEF are not significant at any conventional level in any sub-periods. Overall, our results suggest that the aggregate exposures of equity REITs to bond market factors are time-varying.

The main finding from the use of the three sub-periods is that NCR has a time-varying explanatory ability in describing NAREIT returns. The estimates for NCR are positive and statistically significant at the 5% level and the 1% level during the 1978–1990 period under the four-factor model and the six-factor model, respectively. The result is consistent with Giliberto (1990) who argues that REIT returns reflect the performance of underlying real estate.13 In contrast, the estimates for NCR become insignificant and negative after 1993. The result is at odd with the notion that a more sophisticated investor base in the new REIT era improves information flow, and that a high degree of participation from institutional investors strengthens the linkage between REIT returns and private real estate returns (Ziering et al. 1997).

Table 3 reports the test results based on empirical models that consider possible informational lead-lag relation and capital switching activities. The results in “The four-factor model” are based on the specification in Eq. 2, and the results in “The six-factor model” are based on the specification in Eq. 4. Overall, the test results for MKT, SMB, HML, TERM, and DEF are similar to those reported in Table 2.

A notable observation is that Lag(NCR) has a time-varying explanatory ability in describing NAREIT returns. The estimates for Lag(NCR) are insignificant over the 1978Q2–1990 and 1978Q2–1992 sub-periods under both specifications. However, its estimates are negative and statistically significant at the 1% level during the 1993–2003 period. The inclusion of Lag(NCR) noticeably increases the fitness of regressions. R-squared values rise from 58 to 67% in “The four-factor model” and from 59% to 68% in “The six-factor model”. The signs and statistical significance of Lag(NCR) are consistent with the capital switching hypothesis. The result, on the other hand, does not support the informational lead-lag relation argument.

When capital switching activities are taken into consideration by including Lag(NCR) into the specification, NCR exhibits statistical significance in explaining NAREIT excess returns over the 1978Q2–1990 period and the 1993–2003 period.14 The significance of NCR extends Clayton and Mackinnon’s (2003) finding in the sense that REIT returns are not only sensitive to unsecuritized real estate in 1990s, but also sensitive to the underlying real estate factor in the old REIT era. Comparing the sizes of the estimates for NCR across sub-periods reveals a tighter connection between REIT and the private real estate market starting from 1993. The estimates are 2.18 under both the four-factor model and the six-factor model. Their counterparts during the 1978Q2–1990 period are 0.93 and 1.02, respectively. The result is consistent with the notion that REIT prices are better linked to real estate market fundamentals and behave more like real estate after 1992 (Ziering et al. 1997).

Overall, the significantly positive coefficients of NCR support Giliberto’s (1990) finding that REITs provide investors exposure to real estate markets. The time-varying estimates for Lag(NCR) reflect capital-switching activities that are introduced by institutional investors starting from early 1990s, as suggested by Block (2002), Graff and Young (1997), Linneman (2000), and Murphy et al. (2003). When Lag(NCR) is included into test specifications, R-squared values increase considerably and the coefficients of NCR become significant for the 1993–2003 regressions. These results suggest that the influence of capital-switching activities on REIT returns is significant in the new REIT era.15 The magnitudes of NCR estimates are consistent with Clayton and Mackinnon (2003) and Ziering et al. (1997) that REITs are more like real estate since the early 1990s.

Size-Based Results

Following Graff and Young (1997), this study uses $100 million as the cutoff point for grouping an equity REIT into either the large-cap portfolio or the small-cap REIT portfolio. Graff and Young (1997) point out that a $100 million market capitalization is usually the critical hurdle for determining whether to include an REIT in an institutional portfolio. We perform the size-based analysis only for the 1993–2004 period because the REIT industry was dismissed as insignificant by institutional investors until 1993–1994 (Block 2002). This choice makes our results directly related to the results in Clayton and Mackinnon (2003; Table 4).
Table 4

Regression of quarterly EXBREIT and EXSREIT without lagged NCR, 1993–2003

 

EXBREIT

EXSREIT

The four-factor model

 Constant

0.01

0.03

(0.65)

(2.32)*

 MKT

0.45

0.41

(4.54)**

(3.59)**

 SMB

0.60

0.29

(4.24)**

(1.79)

 HML

0.74

0.54

(6.44)**

(4.08)**

 NCR

0.03

−1.49

(0.04)

(−1.62)

 R2

0.62

0.41

The six-factor model

 Constant

0.01

0.03

(0.72)

(2.30)*

 MKT

0.45

0.41

(4.43)**

(3.69)**

 SMB

0.59

0.31

(4.02)**

(1.95)

 HML

0.74

0.53

(6.29)**

(4.14)**

 TERM

−0.11

0.18

(−0.42)

(0.63)

 DEF

−0.19

1.18

(−0.32)

(1.79)

 NCR

0.00

−1.54

(0.00)

(−1.71)

 R2

0.62

0.47

The reported numbers are based on quarterly returns. EXBREIT is the difference between the return on the big-cap REIT portfolios and the return on 1-month Treasury bills. EXSREIT is the difference between the return on the small-cap REIT portfolios and the return on 1-month Treasury bills. MKT is the difference between the return on the CRSP value-weight portfolio and the return on 1-month Treasury bills. SMB is the difference between the return on small-cap stocks and the return on large-cap stocks. HML is the difference between the return on high book-to-market stocks and the return on low book-to-market stocks. TERM is the difference between the return on long-term US government bonds and the return on 1-month Treasury bills. DEF is the difference between the return on long-term corporate bonds and the return on long-term US government bonds. NCR is the excess return on the Russell–NCREIF property index. The t statistics are in parentheses.

**Significant at the 1% level. *Significant at the 5% level.

Our size-based analysis repeats the regression analyses of Eqs. 1, 2, 3, and 4 with the replacement of EXREIT by EXBREIT and EXSREIT for the 1993–2003 period. Table 4 and Table 5 report the test results. Overall, these tests show similar results for the sensitivities to stock and bond market factors. Big-cap REITs are significantly sensitive to MKT, SML, and HML at the 1% level under both the four-factor model and the six-factor model. Small-cap REITs are significantly sensitive to MKT and HML at the 1% level under both models. The result suggests that stock market factors are useful in explaining both big-cap and small-cap REIT returns.
Table 5

Regression of quarterly EXBREIT and EXSREIT with lagged NCR, 1993–2003

 

EXBREIT

EXSREIT

The four-factor model

 Constant

0.01

0.04

(1.15)

(3.05)**

 MKT

0.37

0.32

(4.14)**

(3.12)**

 SMB

0.54

0.23

(4.30)**

(1.57)

 HML

0.61

0.38

(5.54)**

(3.06)**

 NCR

2.34

1.25

(2.36)*

(1.11)

 Lag(NCR)

−2.72

−3.22

(−3.37)**

(−3.52)**

 R2

0.70

0.55

The six-factor model

 Constant

0.02

0.04

(1.44)

(3.20)**

 MKT

0.38

0.34

(4.21)**

(3.37)**

 SMB

0.51

0.23

(3.95)**

(1.61)

 HML

0.60

0.38

(5.38)**

(3.10)**

 TERM

−0.30

−0.02

(−1.26)

(−0.08)

 DEF

−0.59

0.76

(−1.08)

(1.29)

 NCR

2.46

1.09

(2.46)**

(0.99)

 Lag(NCR)

−2.96

−3.46

(−3.55)**

(−3.46)**

 R2

0.72

0.60

The reported numbers are based on quarterly returns. EXBREIT is the difference between the return on the big-cap REIT portfolios and the return on 1-month Treasury bills. EXSREIT is the difference between the return on the small-cap REIT portfolios and the return on 1-month Treasury bills. MKT is the difference between the return on the CRSP value-weight portfolio and the return on 1-month Treasury bills. SMB is the difference between the return on small-cap stocks and the return on large-cap stocks. HML is the difference between the return on high book-to-market stocks and the return on low book-to-market stocks. TERM is the difference between the return on long-term US government bonds and the return on 1-month Treasury bills. DEF is the difference between the return on long-term corporate bonds and the return on long-term US government bonds. NCR is the excess return on the Russell–NCREIF property index. The t statistics are in parentheses.

**Significant at the 1% level. *Significant at the 5% level.

Big-cap REITs have higher market betas than small-cap REITs. In addition, big-cap REITs behave more like small, value stocks than do small-cap REITs as evident by big-cap REITs’ higher exposures to the SMB factor and the HML factor. We conjecture that this style is related to the information role of big-cap REITs in the new REIT era. It is well known that the dominant style of equity REITs is in small, value stocks (Chiang and Lee 2002). If institutional investors trade REITs like small, value stocks in the new REIT era, it is likely that big-cap REITs would show a higher purity of the small-value style because institutional investors invest predominately in big-cap REITs.

Consistent with our baseline (NAREIT) results and Clayton and Mackinnon’s (2003) sized-based results (Table 4), big-cap and small-cap REIT show weak exposures to bond market factors, TERM and DEF. Also in line with the baseline results, both big-cap and small-cap REITs are negatively and significantly related to Lag(NCR). The sensitivity of big-cap REIT returns to Lag(NCR) supports the notion that institutional investors view REITs as potential substitutes for direct real estate investments (in commingled or separate accounts) and switch their capital between public and private real estate markets. The negative sensitivity of small-cap REIT returns to Lag(NCR) may reflect the investing strategy of small investors in small-cap REITs. That is, small investors learn from institutional investors, and move in and out small-cap REITs conditional upon the performance of private real estate markets. This is consistent with the noisy rational expectation model of De Long et al. (1990). The model suggests that small investors are noisy traders, and that they attempt to mimic informed traders.

The main finding of the size-based analysis is that when Lag(NCR) is included to capture capital switching activities, big-cap REITs exhibits positive and significant association with concurrent NCR, whereas small-cap REITs do not.16 In addition, big-cap REITs have higher exposures to the unsecuritized real estate market than do small-cap REITs as evident by big-cap REITs’ higher estimates for NCR. Our results help to resolve the puzzling finding in Clayton and Mackinnon (2003) that small-cap REITs seemingly behave more like real estate than do large-cap REITs after 1993. Our results appear to support Ziering et al. (1997) argument that a more sophisticated investor base improves information flow.

Conclusions

This study examines the linkage between equity REITs and underlying real estate. Consistent with Giliberto (1990), this study demonstrates that REIT returns reflect the performance of underlying real estate. Specifically, after accounting for Fama and French’s (1993) stock and bond factors, the study finds that the real estate factor is useful in explaining large-cap REIT returns. This study also provides evidence suggesting capital switching activities between private and public real estate markets. One possible explanation for the results is that a high degree of participation from institutional investors strengthens the linkage between REIT returns and the real estate factor (Ziering et al. 1997).

Our results have implications. First, equity REITs are capable of providing investors real estate exposures. The evidence suggests that the creation of REITs by the US Congress fulfills part of its legislative intent. Second, our evidence supports a long list of studies, including He et al. (1996 and 1997) and Hsieh and Peterson (2000), which advocate the use of REITs to capture the real estate factor.

Footnotes
1

This structural change is concentrated on equity REITs (Chan et al. 1998).

 
2

The results of Glascock et al. (2000) are consistent with the notion that REITs and traditional stocks share some common risk factors (Peterson and Hsieh 1997).

 
3

Following He (2002) and Tuluca et al. (2000), this study does not unsmooth NC. Tuluca et al. (2000) give two reasons not to unsmooth the series: (1) investors have access to returns of commingled real estate funds that comprise NCREIF; and (2) the ways to correct the problems inherent in the appraisal-based series are still under refinement.

 
4

Tuluca et al. (2000) show the first and fourth lag returns of NCREIF index significantly influence REIT returns in the 1978–1995 period. Intuitively the informational lead-lag relation suggests positive influences of lag returns of NCREIF index on EXREIT. Interestingly, while the fourth lag returns of NCREIF index have a significant and positive influence, its first lag returns exert a counter-intuitive, significant, and negative influence on returns of NAREIT index in their study.

 
5

Liziere and Satchell (1997) provide evidence that capital switches between real estate and equity markets in the United Kingdom. Charney (2001) provides evidence suggesting capital switching within the real estate sector in Canada.

 
6

Institutional investors have focused on large-cap REITs in search of liquidity (Graff and Young 1997).

 
7

This is a direct analogy of Subrahmanyam’s (2006) argument about the relationship between the stock and real estate markets. Subrahmanyam (2006) finds that order flows and returns in the stock market negatively forecast REIT order flows, but not REIT returns. Case and Shiller (2003) mention capital switching between the stock and real estate markets as a popular theory.

 
8

A negative return relationship between two investment substitutes is suggested by Subrahmanyam (2006) and Liziere and Satchell (1997).

 
9

The authors would like to thank an anonymous referee for suggesting the inclusion of lagged NCR.

 
10

We also performed regression analysis with the inclusion of the fourth-lag NCR. We did not find the fourth lag helpful in explaining REIT returns in either the whole sample period or any sub-period. The largest absolute value of associated t statistics is 0.90, which is not statistically significant at any conventional significance level.

 
11

This analysis does not include the 1978–1979 data because EXBREIT and EXSREIT start in 1980. The authors would like to thank an anonymous referee for suggesting the treatment on the 1991–1992 data.

 
12

Our full-sample results are similar to Clayton and Mackinnon’s (2003, p.47) full-sample finding. In Table 2, the insignificance of NCR over the period 1978–2003 may reflect the influence of its small, negative estimates over the 1993–2003 period.

 
13

We also estimate the four-factor and six-factor models over the 1978–1989 period. Consistent with Giliberto (1990), equity REITs are significantly related to the unsecuritized real estate market over this period.

 
14

Similar to Claytkn and Mackinnon’s (2003) finding, the coefficients of NCR over the full sample period are insignificant in Table 3. This may reflect the facts that Lag(NCR) is not a relevant variable before 1990s and that, as a result, it is not properly treated in the full-sample regressions.

 
15

Given its influence, failure to incorporate Lag(NCR) is likely to under-estimate the coefficients for NCR (Greene 2000; Gujarati 2003). The insignificant and negative estimates for NCR in Table 2 confirm this.

 
16

In line with the NAREIT results, the regressions excluding Lag(NCR) in Table 4 under-estimate the coefficients of NCR when compared to their counterparts in Table 5.

 

Acknowledgements

Ming-Long Lee would like to acknowledge research support from Taiwan National Science Council grant NSC 93-2416-H-224-015. This research was initiated when Kevin C.H. Chiang was an assistant professor at University of Alaska Fairbanks. Kevin C.H. Chiang would like to acknowledge the database support from that university.

Copyright information

© Springer Science+Business Media, LLC 2007