have equity reits experienced periodically collapsing bubbles? · 2013. 9. 12. · literature on...
TRANSCRIPT
-
Have Equity REITs Experienced Periodically Collapsing Bubbles?
James E. Payne Professor and Chair Department of Economics Illinois State University Normal, IL 61790-4200 [email protected] 309-438-8588
and
George A. Waters Assistant Professor Department of Economics Illinois State University Normal, IL 61790-4200 [email protected] 309-438-7301 Final Version Accepted November 2005
Abstract: This paper uses the momentum threshold autoregressive (MTAR) model and the residuals-augmented Dickey-Fuller (RADF) test to examine the possibility of Evans’ (1991) periodically collapsing bubbles in the equity REIT market. The results are mixed. The MTAR model indicates that overall real equity REIT prices and dividends are cointegrated with asymmetric adjustment towards the long-run equilibrium. However, the estimated coefficients of the MTAR model do not indicate the presence of periodically collapsing bubbles. Adjustment in the standard cointegration tests of bubbles for skewness and excess kurtosis via the RADF test fails to reject the null hypothesis of no cointegration, leaving the possibility of periodically collapsing bubbles. The MTAR and RADF results with respect to equity REIT sub-sectors are mixed. Lodging is the only sub-sector in which both the MTAR and RADF results support periodically collapsing bubbles. Moreover, market fundamentals proxied by two alternative measures of capacity utilization do not explain either real equity REIT prices or dividends.
-
2
Have Equity REITs Experienced Periodically Collapsing Bubbles?
I. Introduction
Within the standard present value model, stock prices are determined by the discounted
values of expected future dividends. However, asset prices that are in excess of what is viewed
as the asset’s fundamental value have been interpreted as speculative bubbles. A class of
speculative bubbles known as rational bubbles, do not violate the rational expectations
hypothesis and are consistent with the efficient markets hypothesis. Investors recognize the
overvaluation; however, investors are compensated with excess positive returns for the risk of a
bubble collapsing. Such rational bubbles are due to self-fulfilling expectations that can break the
connection between prices and dividends over the short term.
With respect to the REIT market there are a number of reasons to explore the possibility
of bubble formation in REIT prices. First, the empirical evidence indicates that REITs are
integrated with the stock market and share common risk factors.1 In light of the growing
literature on speculative bubbles with respect to the stock market, it is a natural extension to test
for bubble-like behavior in the REIT markets.2 Second, there is some evidence for the presence
of speculative bubbles in the housing market, but this issue has not been studied from the
standpoint of commercial real estate.3 Third, when prices continue to increase beyond
fundamental values there is an increase in short selling, a signal of overvaluation in the market.
In the case of REIT markets, Li and Yung (2004) argue that REIT markets are not liquid enough
to support such short selling as a means to signal overvaluation in the market and the formation
of a bubble. Fourth, due to informational problems and market inefficiency, there is an
underpricing of REIT seasoned equity offerings [Howe and Shilling, 1998 and Ghosh et al,
-
3
2000] which serves as a deterrent to the issuance of seasoned equity offerings to capture market
overvaluation (Wang et al, 1995).4
Recently, Jirasakuldech et al (2005) test for the presence of rational speculative bubbles
in the equity REIT market over the period 1973:01 to 2003:12 along with the sub-periods
1973:01 to 1991:10 and 1991:11 to 2003:12 with the results indicating the absence of rational
bubbles. Specifically, following the approach of Diba and Grossman (1984, 1988a,b),
Jirasakuldech et al (2005) implement the standard unit root and cointegration tests of equity
REIT returns and macro fundamentals along with tests of duration dependence in equity REIT
returns.5 Evidence against the presence of bubbles is supported if REIT prices and macro
fundamental variables are respectively integrated of order one and the existence of a
cointegrating vector between REIT prices and the macro fundamental variables. Moreover,
evidence against the presence of bubbles is supported if REIT prices do not exhibit negative
duration dependence.
However, Evans (1991) argues that this standard approach will not be able to detect a
class of periodically collapsing rational bubbles. For example, the sudden collapse of a bubble
may be mistaken by standard cointegration tests for mean reversion, resulting in a bias towards
rejection of the null hypothesis of no cointegration. In the case of equity REITs, the approach
followed by Jirasakuldech et al (2005) implicitly assumes the bubble component follows a linear
process whereas the bubble specification of Evans can follow a non-linear process. The task of
this study is to extend the recent work of Jirasakuldech et al (2005) by investigating whether or
not periodically collapsing bubbles exist within the equity REIT market. The existence of
periodically collapsing bubbles is examined using the momentum threshold autoregressive
(MTAR) model advanced by Enders and Granger (1998) and Enders and Siklos (2001) and the
-
4
residuals-augmented Dickey-Fuller (RADF) test by Taylor and Peel (1998). The MTAR model
will capture the possible asymmetries in the adjustment towards the long-run equilibrium,
particularly the sharp falls in the asset price after the price has reached a certain threshold
relative to dividends. The RADF test corrects the standard cointegration tests for the skewness
and excess kurtosis that may arise in the presence of periodically collapsing bubbles. The results
are mixed as to whether periodically collapsing bubbles exist in the equity REIT market.
The paper is organized as follows. Section II provides the theoretical framework for
periodically collapsing bubbles. Section III presents the empirical methodologies, the data and
empirical results. Section IV investigates the robustness of the results and the role of market
fundamentals with respect to equity REIT sub-sectors. Concluding remarks are presented in
Section V.
II. Theoretical Framework of Periodically Collapsing Bubbles
Solutions to the present value model represent both fundamental and bubble solutions for
asset prices. Evans (1991) provides a specific form for generating periodically collapsing
bubbles that might not be detected by simple unit root analysis.6 The asset price Pt at time t
depends on the expectation at time t of next period’s price Pt and dividend Dt such that
11t ttt DPEP (1)
where the discount factor is 0
-
5
variables a la McCallum (1983, 1997)) condition on (2), then the bubble term must be zero, B t =
0. In this case, the asset price Pt is determined solely by expected future dividends,
corresponding to the fundamental solution in the asset pricing literature.
While some have argued for the non-existence of bubbles on the basis of the conditions
mentioned above, much work has been advanced recently on ways to detect whether asset prices
are determined by dividends alone. Evans (1991) discusses a class of bubbles that would not be
detected by simple cointegration techniques. Such periodically collapsing bubbles may be
represented as follows.
,11
1
ttt vBB if tB (3a)
1111 tttt vBB , if tB . (3b) The parameters in the above equations satisfy and 0 < . The stochastic process
vt is iid and has conditional expectation Etvt+1 = 1, which ensures that a bubble will not switch
sign. The term t is a Bernoulli process that takes the value 1 with probability and the value 0
with probability 1-. Equation (3a) represents the phase when the bubble grows at mean rate
, but equation (3b) shows that if the bubble exceeds the threshold , it explodes at mean rate
However, this phase does not last indefinitely as the bubble collapses with probability 1-
each period.
Evan’s model of a bubble incorporates three important characteristics, Bt satisfies the
martingale property EtBt+1 = Bt, bubbles cannot be detected by simply examining
cointegration between asset prices and dividends, and bubbles that are initially positive stay
positive. The last observation that bubbles stay positive fits an assumption that, while not
universally accepted, is commonly made in the literature. In this paper we test for positive
bubbles, without denying the possibility of negative bubbles.
-
6
III. Data, Methodology, and Results
Monthly data on the prices and dividends for equity REITs were obtained from the
National Association of Real Estate Investment Trusts (NAREIT) for the period 1972:01 to
2005:03. The original data has been converted to natural logarithms. Following Jirasakuldech
et al (2005), we recognize the structural shift in the REIT markets in 1991 in testing for unit
roots in equity REIT prices and dividends, respectively. The hypothesized structural shift can be
attributed to the following: (1) the REIT market became more dominated by large institutional
investors along with increased market liquidity in the post 1992 period [Damodaran and Liu,
1993; Wang et al, 1995; Below et al, 1996; Chan et al, 1998; and Chan et al, 2003]; (2) analyst
and media coverage of the REIT markets increased, thereby increasing the transparency of the
market throughout the 1990s [Gentry et al, 2003 and Chui et al, 2003]; and (3) the creation of the
umbrella partnership REIT organization structure in 1991 allowed for more flexibility in
purchasing property; however, the lack of transparency has made valuation more difficult
[Damodaran et al, 1997 and Ling and Ryngaert, 1997].
For prices and dividends to be cointegrated, representing a long run connection between
the two, they must be integrated of the same order. We begin by testing the null hypothesis of a
unit root in real equity REIT prices, tp and real equity REIT dividends, td , incorporating the
possibility of structural breaks using Perron’s (1989) unit root test.
titk
i itbtycyTDDTtDUy 11)( (4)
where tt py or td ; )(1 bTtDU is a post-break constant dummy variable; t is a linear time
trend; tTtDT b )(1 is a post-break slope dummy variable; )1(1)( bb TtTD is the break
-
7
dummy variable; and t are white noise error terms.7 The null hypothesis of a unit root is given
by 1 . Table 1 reports the results of the unit root tests allowing for a structural break in
1991:11 as suggested by Jirasakuldech et al (2005). Neither real equity REIT prices nor
dividends yield statistically significant coefficients on the respective dummy variables.
Moreover, the results are unable to reject the null hypothesis of a unit root in either real equity
REIT prices or dividends.
Given the respective real equity REIT prices and dividends contain a unit root, the
momentum threshold autoregressive (MTAR) model advanced by Enders and Granger (1998)
and Enders and Siklos (2001) and the residuals-augmented Dickey-Fuller (RADF) test set forth
by Taylor and Peel (1998) are estimated in order to capture the dynamics of periodically
collapsing bubbles.8
MTAR Approach:
The following cointegration equation representing the relationship between real equity
REIT prices, tp , and dividends, td , is estimated.
ttt dp (5)
Indeed, if real equity REIT prices and dividends are integrated of order one, the residuals from
the cointegration equation, t , should be stationary in levels. The cointegration test with the
possibility of asymmetric adjustment is undertaken by the following regression of the residuals.
titp
i itttttvII 11211 )1( (6)
Heaviside indicator function, tI , is represented by:
1
1
ˆ0ˆ1
t
tt if
ifI (7)
-
8
where the threshold is the value that minimizes the residual sum of squares.9 The MTAR
model allows the speed and direction of adjustment, represented by and, to depend on the
previous period’s change in 1t . This model is especially valuable when the adjustment is
believed to exhibit more momentum in one direction than the other, as in the case of collapsing
bubbles. The null hypothesis of no cointegration is tested by the restriction,
Indeed, if real equity REIT prices, tp , and dividends, td , are cointegrated, the null hypothesis of
symmetry is tested by the restriction,. If the estimated coefficient, , is statistically
significant and negative and larger in absolute terms relative to the estimated coefficient, ,
there is evidence of a sharp correction when prices have risen above a certain threshold relative
to dividends. Therefore, if the null hypothesis of symmetric adjustment is rejected and 2 < 1 ,
we conclude that periodically collapsing bubbles are present in REIT prices.
The cointegration and momentum threshold tests are reported in Panel A of Table 2.
Real equity REIT prices and dividends appear cointegrated as evident from the significant ADF
test statistic (CR in Panel A of Table 2). According to the Diba and Grossman approach the
presence of a cointegrated relationship between real equity REIT prices and dividends can be
interpreted as evidence against the presence of speculative bubbles in the REIT market. To
address the possible asymmetries in the adjustment towards the long-run equilibrium relationship
between real equity REIT prices and dividends that might occur in the presence of periodically
collapsing bubbles, the MTAR model is explored. Note from equations (6) and (7), the MTAR
specification provides point estimates of 1 and 2 . As in the case of the Engle-Granger
cointegration tests, the null hypothesis of no cointegration, , is rejected.
Furthermore, the null hypothesis of symmetry,, is rejected; however, the point estimates,
-
9
1 and 2 , do not satisfy the condition 2 < 1 . Therefore, the results from the MTAR
approach do not provide evidence in support of periodically collapsing bubbles in the equity
REIT market. However, this model requires the very restrictive assumption that the threshold
for the collapse is constant throughout the sample.
RADF Approach:
The argument of Diba and Grossman (1984, 1988a,b) and Jirasakuldech et al (2005) that
cointegration between price and dividends is evidence against bubbles assumes a linear process
for the growth of tB , implying normality in the residuals. However, a preliminary test on the
residuals from equation (5) shows both skewness and excess kurtosis in the residuals, which
could be caused by the presence of periodically collapsing bubbles (see Panel B of Table 2).
Taylor and Peel (1998) present Monte Carlo evidence to suggest that periodically collapsing
bubbles generate skewness and excess kurtosis in stock prices. Following the work of Im
(1996), Taylor and Peel (1998) argue that the presence of skewness and excess kurtosis can be
used to obtain a more efficient estimator in the cointegration tests of bubbles. Standard tests of
cointegration examine the stationarity of the residuals from equation (5) as follows:
ttt u 1 (8)
The null hypothesis of no cointegration is stated as 0 while the alternative hypothesis of a
stationary residual is 0 . Taylor and Peel (1998) argue that correcting the least squares
estimate in equation (8) for skewness and excess kurtosis provides a more efficient estimator of
and improves the ability to detect periodically collapsing bubbles.
Taylor and Peel (1998) suggest the following two-step estimator in the construction of
the residuals-augmented Dickey-Fuller (RADF) test of the null hypothesis of no cointegration.10
First, regress the first difference of the residuals of the cointegrating equation on their lagged
-
10
level (see equation 8 above) and use the new residuals, tû , and the estimated variance, 2̂ , to
construct the vector, )]ˆˆ(),ˆˆ3ˆ[(ˆ 2223 tttt uuuw . Next, re-estimate equation (8) with the
addition of the vector, tŵ , which corrects the estimate of for skewness and excess kurtosis of
the residuals as follows:
tttt w ˆ1 (9)
where t is white noise. The key test statistic is, )ˆ(/ˆ** VCR , where
*̂ is the estimator
in equation (9).11 Taylor and Peel (1998), Sarno and Taylor (2003), as well as Capelle-Blancard
and Raymond (2004) conduct Monte Carlo Studies to construct critical values and analyze the
power of this test against alternatives. In particular, Taylor and Peel (1998) show that the
adjustment for skewness and excess kurtosis has superior power over standard cointegration tests
to correctly reject a mean-reverting error model as a bubble. Panel B of Table 2 reports the
results of the RADF test on equity REITs. The value of CR is -0.79, far below the 10 percent
critical value, hence the null hypothesis of no cointegration (i.e. presence of a bubble
component) cannot be rejected. The results of the RADF test leaves room for the possibility of
periodically collapsing bubbles in the equity REIT market.
IV. Further Inspection of Equity REIT Sub-Sectors
Given the mixed results associated with equity REITs from the MTAR and RADF
specifications, a further inspection of equity REIT sub-sectors for robustness is warranted.
While the evidence on periodically collapsing bubbles is mixed for the overall equity REIT
market, it is possible that various sub-sectors may exhibit periodically collapsing bubbles.12 The
analysis begins with testing the null hypothesis of a unit root in the respective real equity REIT
-
11
prices and dividends along with two proxies for market fundamentals using the standard
augmented Dickey-Fuller (ADF, 1979) and Phillips-Perron (PP, 1988) unit root tests. Panels A
and B of Table 3 reports the results of the ADF and PP unit root tests for real equity prices and
dividends for overall equity REITs along with the equity REIT sub-sectors: apartments,
industrial, lodging, manufactured homes, office, and regional malls. Panel C of Table 3 reports
the corresponding unit root tests for two measures of capacity utilization, the National
Association of Purchasing Managers tnapm index and the actual capacity utilization rate tcu ,
which will be used as proxies for market fundamentals in explaining real equity REIT prices and
dividends. The unit root results indicate that the respective real equity prices and dividends
along with the capacity utilization measures are stationary after first-differencing, so there is the
possibility of cointegration between these variables. Next, tests for periodically collapsing
bubbles are performed for the overall equity REITs and the equity REIT sub-sectors using the
MTAR (equations 6 through 7) and RADF (equations 8 and 9) specifications and reported in
Table 4.
The Engle-Granger (1987) cointegration tests reveal that real equity REIT prices and
dividends are cointegrated at the 10 percent significance level for overall equity REITs and the
equity REIT sub-sector, regional malls (CR in Panel A of Table 4). The presence of
cointegration provides evidence of a connection between prices and dividends and the absence of
a bubble component, but, as we have noted, cointegration tests may not detect periodically
collapsing bubbles. The MTAR testing strategy, as shown in (6) and (7), has superior power
relative to the alternative assumption of symmetric adjustment associated with standard Engle-
Granger tests (Enders and Siklos, 2001). Thus, the null hypothesis of no cointegration,
, is tested allowing for the possibility of asymmetric adjustment. The column labeled
-
12
FC in Table 4 displays the test statistics associated with the MTAR adjustment in the
cointegration tests. It appears that the null hypothesis of no cointegration, associated
with the MTAR specification is rejected for the equity REIT sub-sectors, apartments, industrial,
lodging, and office.13 Thus, for those equity REIT sub-sectors in which the null hypothesis of no
cointegration is rejected using the MTAR specification, the null hypothesis of symmetry,
is tested.14 Indeed, the null hypothesis of symmetry is rejected for the equity REIT sub-
sectors, apartments, industrial, lodging, and office, indicating that prices above (or below) a
certain threshold, τ, exhibit differing speeds of adjustment toward the long-run equilibrium (i.e.
cointegrating) relationship between real equity REIT prices and dividends. However, upon
further inspection of the point estimates, 1 and 2 , the condition, 2 < 1 , for the presence of
periodically collapsing bubbles is not satisfied. Thus, it appears from the MTAR specifications
over the period, 1994:01 to 2005:03, that only the equity REIT sub-sector, lodging, exhibits
behavior consistent with periodically collapsing bubbles.
Panel B of Table 4 reports the results of the RADF test on the overall equity REIT market
and equity REIT sub-sectors. In every case the value of CR is far below the 10 percent critical
value, hence the null hypothesis of no cointegration (i.e. presence of a bubble component) cannot
be rejected. The results of the RADF test leave room for the possibility of periodically
collapsing bubbles in both the overall equity REIT market as well as the equity REIT sub-
sectors.
In light of the common criticism that bubble-like behavior may be attributed to changes
in market fundamentals, as opposed to self-fulfilling expectations as specified in section II, the
influence of such fundamentals are incorporated in the relationship between real equity prices
and dividends. Two alternative measures are used as proxies for market fundamentals: the
-
13
National Association of Purchasing Managers Index, a leading indicator of capacity utilization,
and the actual capacity utilization rate. The rationale for using either measure is that as firms
increase their capacity utilization, there is an increase in the demand for space. The increase
demand for space leads to a reduction in vacancy rates and upward pressure on rents which will
affect the respective REIT markets.15
The analysis proceeds by first establishing whether or not the respective proxies for
market fundamentals share a long-run cointegrating relationship with the respective real equity
REIT prices and dividends. Appendix A presents the results of the Johansen-Juselius (1990)
multivariate cointegration procedure to test the possible long-run relationships between the
respective real equity REIT prices, dividends, and the alternative measures of capacity
utilization, using the maximum eigenvalue and trace tests. As shown in Panels A through G of
Appendix A, the maximum eigenvalue and trace tests for cointegration proceed sequentially
from the first hypothesis of no cointegration to increasing numbers of cointegrating vectors.
More specifically, the maximum eigenvalue test max is based on the null hypothesis that the
number of cointegrating vectors is r against the alternative r + 1 cointegrating vectors. The trace
tests trace is based on the null hypothesis that the number of cointegrating vectors is less than
or equal to r against a general alternative. The results displayed in Panels A through G of
Appendix A fail to reject the null hypothesis of zero cointegrating vectors based on either the
maximum eigenvalue or trace tests. These results indicate that neither the NAPM index nor the
capacity utilization rate shares a common stochastic trend with real equity REIT prices and
dividends (i.e. not cointegrated).
Given the absence of a long-run relationship as indicated by the Johansen-Juselius
cointegration tests, the short-run dynamics are explored in Panels A through G of Appendix B
-
14
with the estimation of vector autoregressive models for the respective real equity REIT prices
and dividends using either the NAPM index or the capacity utilization rate. While there is
evidence of autoregressive behavior in many of the variables and the occasional feedback
between prices and dividends, neither the NAPM index nor the capacity utilization rate have
predictive power, as evident from their partial-F-statistics, in explaining either real equity REIT
prices or dividends. Thus, the results with respect to periodically collapsing bubbles cannot be
attributed to changing market fundamentals as proxied by the NAPM index or the capacity
utilization rate.
V. Concluding Remarks
This study extends the recent work of Jirasakuldech et al (2005) on rational speculative
bubbles in the equity REIT market by exploring the possibility of periodically collapsing
bubbles. Detection of periodically collapsing bubbles in asset markets, such as those in Evans
(1991), requires econometric tests beyond the standard cointegration tests proposed by Diba and
Grossman (1984, 1988a,b). Two approaches are undertaken to discern the whether periodically
collapsing bubbles are present in the equity REIT market: the momentum autoregressive
threshold (MTAR) model of Enders and Granger (1998) and Enders and Siklos (2001) and the
residuals-augmented Dickey-Fuller (RADF) test by Taylor and Peel (1998). The MTAR model
will capture the possible asymmetries in the adjustment towards the long-run equilibrium,
particularly the sharp falls in the asset price after the price has reached a certain threshold
relative to dividends. The RADF test corrects the standard cointegration tests for the skewness
and excess kurtosis that may arise in the presence of periodically collapsing bubbles. Over the
period 1972:01 to 2005:03, the MTAR results indicate that while overall real equity REIT prices
-
15
and dividends are cointegrated and exhibit asymmetries, the point estimates do not indicate the
presence of periodically collapsing bubbles in the equity REIT market. On the other hand, the
RADF results suggest the possibility of periodically collapsing bubbles over this period.
However, within the sub-period, 1994:01 to 2005:03, the MTAR approach provides
support for periodically collapsing bubbles for only the equity REIT sub-sector, lodging. The
RADF results fail to reject the null hypothesis of no cointegration in the overall equity REIT
market and equity REIT sub-sectors, hence the possibility of periodically collapsing bubbles.
Moreover, to account for the possibility that the relationship between real equity REIT prices and
dividends have been driven by market fundamentals, Johansen-Juseilus cointegration tests are
performed to infer the long-run relationship between real equity REIT prices, dividends, and two
alternative measures of market fundamentals, the National Association of Purchasing Managers
Index and the actual capacity utilization rate. Across equity REIT sub-sectors, the cointegration
results indicate the absence of a long-run relationship between real equity REIT prices,
dividends, and market fundamentals. In addition, vector autoregressive models for the
respective real equity REIT prices, dividends, and proxies for market fundamentals are estimated
to capture the short-run dynamics among these variables. It appears from the results that
changing market fundamentals proxied by two alternative measures of capacity utilization do not
explain either real equity REIT prices or dividends.
In summary, the results are mixed as to whether periodically collapsing bubbles exist in
the equity REIT market. An interesting extension would be to examine the connection
empirically between REITs and housing prices in the context of bubble formation.
-
16
Notes
1. See Myer and Webb (1992), Li and Wang (1995), Peterson and Hsieh (1997), Okunev and Wilson (1997), Ling and Naranjo (1999), Okunev et al (2000), and Glascock et al (2000) for studies on the relationship between REITs and the general stock market.. 2. See Diba and Grossman (1988a,b), Camerer (1989), Dezbaksh and Demirguc-Kunt (1990), Evans (1991), Topol (1991), Taylor and Peel (1998), Bohl (2003), Capelle-Blancard and Raymond (2004), and Harman and Zuehlke (2004) for studies on speculative bubbles in the general stock market. 3. See Case and Shiller (1990), Kim and Shuh (1993), Grenadier (1995), Abraham and Hendershott (1996), Clayton (1997), Bjorklund and Soderberg (1999), Hendershott (2000), and Roche and McQinn (2001) for studies on speculative bubbles in real estate markets. 4. Jirasakuldech et al (2005) also mention the latter three points with respect to bubble formation. 5. Harman and Zeuhlke (2004) discuss several shortcomings of using duration dependence tests to detect bubbles. In particular, duration dependence is sensitive to the method of correcting for discrete observation of continuous duration, the use of value-weighted versus equal-weighted portfolios, and the frequency of the observed data. 6. Charemza and Deadmen (1995) specify a more general model of periodically collapsing bubbles. 7. Equity REIT prices and dividends are converted into real terms by deflating by the consumer price index obtained from the St. Louis Federal Reserve Bank. 8. The use of the MTAR model parallels the approach undertaken by Bohl (2003) and the RADF test has been used in studies by Taylor and Peel (1998), Sarno and Taylor (1999, 2003), and Capelle-Blancard and Raymond (2004) for stock markets. 9. While the Chan (1993) method is typically used for choosing the threshold parameter, , the idea of eliminating extreme values when testing for bubble behavior seems counter intuitive. Specifically, Chan (1993) requires sorting the estimated residuals in ascending order, eliminating 15 percent of the largest and smallest values. The threshold parameter that yields the lowest sum of squared errors from the remaining 70 percent of the residuals is used in the MTAR model. 10. See Im (1996) and Taylor and Peel (1998) for the theoretical underpinnings of the two-step approach of the RADF test. Note that the basis of this test draws from the work of Im (1996) on residuals-augmented least squares estimators. 11. The covariance matrix of *̂ , )ˆ( *V , is estimated by 1ˆ
2* )~~()ˆ( XMXV wA where
-
17
2235
23
6246
44
44
244
235
443
23
6246
2322
)4()96)(()()3()4)(3(2)96(
A
and i denotes the ith central moment of tu . X
~ is the vector of the lagged series of centered
residuals and the idempotent matrix, wM ˆ , is given by WWWWIM Tw~)~~(~ 1ˆ
where TI is the
identity matrix and W~ is the matrix of the centered residuals of tŵ . 12. Upon the recommendation of an anonymous referee tests for periodically collapsing bubbles were examined for equity REIT sub-sectors. The equity REIT sub-sector data provided by the referee covered the period 1994:01 to 2005:03. The respective prices and dividends were converted into real terms by deflating by the consumer price index. 13. Note that for overall equity REITs along with the equity REIT sub-sectors, manufactured homes and regional malls, one fails to reject the null hypothesis of no cointegration (i.e. FC column of Table 4). Thus, as stated by Enders and Sikos (2001, p. 170) there is no need to pursue the tests of symmetry (i.e. FA column of Table 4). 14. Given the relatively short time horizon, the power of the cointegration tests are questionable (Hakkio and Rush, 1991 as well as Kremers et al, 1992). 15. We appreciate the suggestion of an anonymous referee to examine whether the NAPM index and/or the capacity utilization rate affect real equity REIT prices and dividends.
-
18
References
Abraham, J. and P.H. Hendershott (1996), “Bubbles in Metropolitan Housing Markets”, Journal of Housing Research, 7, 191-207. Below, S.D., J.K. Kiely, and W. McIntosh (1996), “REIT Pricing Efficiency: Should Investors Still be Concerned?”, Journal of Real Estate Research, 12(3), 397-412. Bjorklund, K. and B. Soderberg (1999), “Property Cycles, Speculative Bubbles and the Gross Income Multiplier”, Journal of Real Estate Research, 18(1), 151-174. Bohl, M.T. (2003), “Periodically Collapsing Bubbles in the US Stock Market?”, International Review of Economics and Finance, 12, 385-397. Camerer, C. (1989), “Bubbles and Fads in Asset Prices”, Journal of Economic Surveys, 3(1), 3-42. Capelle-Blancard, G. and H. Raymond (2004), “Empirical Evidence in Periodically Collapsing Stock Price Bubbles”, Applied Economics Letters, 11, 61-69. Case, K.G. and R.J. Shiller (1990), “Forecasting Prices and Excess Returns in the Housing Market”, AREUEA Journal, 18, 253-273. Chan, K.S. (1993), “Consistency and Limiting Distribution of the Least Squares Estimator of a Threshold Autoregressive Model”, Annals of Statistics, 21(1), 520-533. Chan, S.H., W.K. Lering, and K. Wang (1998), “Institutional Investment in REITs: Evidence and Implications”, Journal of Real Estate Research, 16, 357-374. Chan, S.H., J. Erickson, and K. Wang (2003), Real Estate Investment Trusts: Structure, Performance, and Investment Opportunities, Oxford University Press. Charemza, W.W. and D.F. Deadman (1995), “Speculative Bubbles with Stochastic Explosive Roots: The Failure of Unit Root Testing”, Journal of Empirical Finance, 2, 153-163. Chui, C.W.A., S. Titman, and K.C.J. Wei (2003), “Intra-Industry Momentum: The Case of REITs”, Journal of Financial Markets, 6, 363-387. Clayton, J. (1997), “Are Housing Price Cycles Driven by Irrational Expectations?”, Journal of Real Estate Finance and Economics, 14(3), 341-363. Damodaran, A. and C.H. Liu (1993), “Insider Trading as a Signal of Private Information”, Review of Financial Studies, 6, 79-120. Damodaran, A., K. John, and C.H. Liu (1997), “The Determinants of Organization Form Changes: Evidence and Implications from Real Estate”, Journal of Financial Economics, 45,
-
19
169-192. Dezbaksh, H. and A. Demirguc-Kunt (1990), “On the Presence of Speculative Bubbles in Stock Prices”, Journal of Financial and Quantitative Analysis, 25(1), 101-112. Diba, B.T. and H.I. Grossman (1984), Rational Bubbles in the Price of Gold, NBER Working Paper No. 1300. Diba, B.T. and H.I. Grossman (1988a), “Explosive Rational Bubbles in Stock Prices?”, American Economic Review, 78, 520-530. Diba, B.T. and H.I. Grossman (1988b), “The Theory of Rational Bubbles in Stock Prices”, Economic Journal, 98(392), 746-754. Dickey, D. and W.A. Fuller, (1979), “Distribution of the Estimators for Autoregressive Time Series with a Unit Root”, Journal of the American Statistical Association, 74, 427-431. Enders, W. and C.W.J. Granger (1998), “Unit Root Tests and Asymmetric Adjustment with an Example Using the Term Structure of Interest Rates”, Journal of Business and Economic Statistics, 16(3), 304-311. Enders, W. and P. Siklos (2001), “Cointegration and Threshold Adjustment”, Journal of Business and Economic Statistics, 19(2), 166-176. Engle, R. and C.W.J. Granger (1987), “Co-integration and Error-Correction: Representation, Estimation and Testing”, Econometrica, 55, 251-276. Evans, G.W. (1991), “Pitfalls in Testing for Explosive Bubbles in Asset Prices”, American Economic Review, 81, 922-30. Gentry, W.M., D. Kemsley, and C.J. Meyer (2003), “Dividend Taxes and Share Prices: Evidence from Real Estate Investment Trusts”, Journal of Finance, 58, 261-282. Ghosh, C., C.F. Sirmans, and R. Nag (2000), “The Pricing of Seasoned Equity Offerings: Evidence from REITs”, Real Estate Economics, 28(3), 363-384. Glascock, J.L., C. Lu, and R.W. So (2000), “Further Evidence on the Integration of REIT, Bond, and Stock Returns”, Journal of Real Estate Finance and Economics, 20(2), 177-194. Grenadier, S. (1995), “The Persistence of Real Estate Cycles”, Journal of Real Estate Finance and Economics, 10, 95-119. Hakkio, C.S. and M. Rush (1991), “Cointegration: How Short is the Long-Run?”, Journal of International Money and Finance, 10, 571-581.
-
20
Harman, Y.S. and T.W. Zuehlke (2004), “Duration Dependence Testing for Speculative Bubbles”, Journal of Economics and Finance, 28(2), 147-154. Hendershott, P.H. (2000), “Property Asset Bubbles: Evidence from the Sydney Office Market”, Journal of Real Estate Finance and Economics, 20(1), 67-81. Howe, J.S. and J.D. Shilling (1988), “Capital Structure Theory and REIT Security Offerings”, Journal of Finance, 43(4), 983-993. Im, K.S. (1996), “A Least Squares Approach to Non-Normal Disturbances”, Working Paper 9603, Department of Applied Economics, University of Cambridge. Jirasakuldech, B., R.D. Campbell, and J.R. Knight (2005), “Are There Rational Speculative Bubbles in REITs?”, Journal of Real Estate Finance and Economics, forthcoming. Johansen, S. and K .Juselius (1990), “Maximum Likelihood Estimation and Inference on Cointegration with Applications to the Demand for Money”, Oxford Bulletin of Economics and Statistics, 52, 160-210. Johansen, S. (1995), Likelihood-Based Inference in Cointegrated Vector Autoregressive Models, Oxford University Press, New York. Kim, K.H. and S.H. Shuh (1993), “Speculation and Price Bubbles in the Korean and Japanese Real Estate Markets”, Journal of Real Estate Finance and Economics, 6, 73-87. Kremers, J.J.M., N.L. Ericsson, and J. Dolado (1992), “The Power of Cointegration Tests”, Journal of Econometrics, 52, 389-402. Li, D.D. and K. Yung (2004), “Short Interest in Real Estate Investment Trusts”, University of Maryland Eastern Short and Old Dominion University Working Paper. Li, Y. and K. Wang (1995), “The Predictability of REIT Returns and Market Segmentation”, Journal of Real Estate Research, 10(4), 471-482. Ling, D.C. and A. Naranjo (1999), “The Integration of Commercial Real Estate Markets and Stock Markets”,Real Estate Economics, 27(3), 483-515. Ling, D.C. and M. Ryngaert (1997), “Valuation Uncertainty, Institutional Involvement, and the Underpricing of IPOs: The Case of REITs”, Journal of Financial Economics, 43(3), 433-456. McCallum, B.T. (1983), “On Non-Uniqueness in Rational Expectations Models: An Attempt at Perspective”, Journal of Monetary Economics, 11, 139-168. McCallum, B.T. (1997), “The Role of the Minimum State Variables Criterion in Rational Expectations Models”, International Journal of Tax and Finance, 6(4), 621-639.
-
21
Myer, F.C.N. and J.R. Webb (1993), “Return Properties of Equity REITs, Common Stocks, and Commercial Real Estate: A Comparison”, Journal of Real Estate Research, 8(1), 87-106. Okunev, J. and P.J. Wilson (1997), “Using Nonlinear Tests to Examine Integration between Real Estate and Stock Markets”, Real Estate Economics, 25(3), 487-503. Okunev, J., P.J. Wilson, and R. Zurbruegg (2000), “The Causal Relationship between Real Estate and Stock Markets”, Journal of Real Estate Finance and Economics, 21(3), 251-261. Osterwald-Lenum, M. (1992), “A Note with Quantiles of the Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statistics, Oxford Bulletin of Economics and Statistics, 54, 461-472. Perron, P. (1989), “The Great Crash, the Oil Price Shock, and the Unit Root Hypothesis”, Econometrica, 57(6), 1361-1401. Peterson, J.D. and C. Hsieh (1997), “Do Common Risk Factors in the Returns on Stocks and Bonds Explain Returns on REITs?”, Real Estate Economics, 25(2), 321-345. Phillips, P. and P. Perron (1988), “Testing for a Unit Root in Time Series Regression”, Biometrika, 75, 335-346. Roche, M.J. and K. McQinn (2001), “Testing for Speculation in Agricultural Land in Ireland”, European Review of Agricultural Economics, 28(2), 95-115. Sarno, L. and M.P. Taylor (1999), “Moral Hazard, Asset Price Bubbles, Capital Flows, and the East-Asian Crisis: The First Tests”, Journal of International Money and Finance,18, 637-657. Sarno, L. and M.P. Taylor (2003), “An Empirical Investigation of Asset Price Bubbles in Latin American Emerging Financial Markets”, Applied Financial Economics, 13, 635-643. Taylor, M.P. and D.A. Peel (1998), “Periodically Collapsing Stock Price Bubbles: A Robust Test”, Economics Letters, 61, 221-228. Topol, R. (1991), “Bubbles and Volatility of Stock Prices: Effect of Mimetic Contagion”, Economic Journal, 101(407), 786-800. Wang, K., J. Erickson, G. Gau, and S.H. Chan (1995), “Market Microstructure and Real Estate Returns”, Real Estate Economics, 23(1), 85-100.
-
22
Table 1 Perron Unit Root Tests Real Equity REIT Prices and Dividends 1972:01 to 2005:03 Panel A: Real Equity REIT Prices, tp DU t DT )( bTD 1tp k
2.RAdj )36(Q 0.175 -0.012 2.2E-5 4.5E-5 -0.018 0.964 1 0.946 32.60 (0.06)a (0.02) (3.8E-5) (8.1E-5) (0.04) (0.01) [0.63] Panel B: Real Equity REIT Dividends, td DU t DT )( bTD 1td k
2.RAdj )36(Q 0.121 -4.0E-4 6.2E-6 -3.5E-6 -0.003 0.974 4 0.961 43.40 (0.06)a (0.02) (2.7E-5) (5.5E-5) (0.03) (0.01) [0.19] Notes: Standard errors are denoted in parentheses and probability values in brackets. k is the number of augmented lags of the first-differences of tp and td , respectively. )36(Q denotes the Ljung-Box Q-statistic distributed as 236 . Critical values to test the null hypothesis of a unit root, 1 , is drawn from Table VI.B p. 1377 of Perron (1989) for 60. as follows: 1% -4.88, 5% -4.24, and 10% -3.95. For equity REIT prices the t-statistic for the null hypothesis, 1 , is -3.6 and for equity REIT dividends the t-statistic for the null hypothesis, 1 , is -2.6.
-
23
Table 2 MTAR and RADF Results
1972:01 to 2005:03 Panel A: Momentum Threshold Autoregressive Model Results
CR τ ρ1 ρ2 FC FA Q(5) k Equity -3.39b 0.31 0.05 -0.07 12.70a 10.10a 1.20 4 (1.74)c (-4.11)a {0.95} Panel B: Residuals –Augmented Dickey-Fuller Results
Skewness Kurtosis JB CRτ Q(5) Equity -0.001 3.56 5.23 -0.79 3.75 {0.07}c {0.59} Notes: CR is the Dickey-Fuller test statistic applied to the residuals from the cointegration equation (5) under the null hypothesis of no cointegration with critical values: a(1%) -3.73, b(5%) -3.17, and c(10%) -2.91 (Engle and Granger, 1987). CRτ is the residuals-augmented Dickey-Fuller test statistic applied to the residuals from the cointegrating equation (5) under the null hypothesis of no cointegration with critical values: a(1%) -3.98, b(5%) -3.44, and c(10%) -3.13 (Capelle-Blancard and Raymond, 2004). is the estimated threshold. 1 and 2 are the estimated parameters from the MTAR specification. Standard errors denoted by [ ], t-statistics denoted by ( ), and probability values in { } where a(1%), b(5%), and b(10%). FC represents the F-statistic corresponding to the null hypothesis of no cointegration (i.e. 021 ) with critical values provided by Enders and Siklos (2001, Table 5, p. 172, n = 250 and four lags): a(1%) 8.47, b(5%) 6.32, and c(10%) 5.32. FA represents the F-statistic corresponding to the null hypothesis of symmetry (i.e. 21 ) using the standard F distribution with critical values a(1%) 4.61 and b(5%) 3.00. k is the number of lags in equation (6). Q(5) denotes the Ljung-Box Q-statistic at 5 lags.
-
24
Table 3 Unit Root Tests 1994:01 to 2005:03 Panel A: Overall Equity REITs
tp tp td td ADF -1.58 -12.19a -2.55 -9.85a PP -1.67 -12.18a -2.99 -11.75a Panel B: Equity REIT Sub-Sectors Apartments tp tp td td ADF -3.04 -13.54a -0.79 -14.39a PP -3.12 -13.54a -1.48 -15.16a Industrial tp tp td td ADF -1.76 -13.95a -2.64 -6.87a PP -2.19 -13.88a -3.09 -29.32a Lodging tp tp td td ADF -1.79 -9.33a -2.47 -11.13a PP -1.79 -9.33a -2.45 -11.15a Manufactured Homes tp tp td td ADF -2.52 -12.05a -1.71 -4.15a PP -2.62 -12.05a -1.50 -15.41a Office tp tp td td ADF -1.90 -12.25a -1.53 -5.11a PP -1.98 -12.24a -2.78 -15.91a Regional Malls tp tp td td ADF -0.76 -13.87a -0.57 -13.27a PP -0.94 -13.75a -0.64 -13.35a Panel C: Measures of Capacity Utilization: National Association of Purchasing Managers Index, tnapm , and Capacity Utilization Rate, tcu tnapm tnapm tcu tcu ADF -2.43 -11.70a -1.66 -4.09a PP -2.57 -11.70a -1.55 -11.91a Notes: Lag length selection for the ADF unit root tests is based on Akaike’s information criterion while the PP unit root tests is based on Newey-West bandwidth using Bartlett kernel. ADF and PP unit root tests include constant and linear trend terms with the following critical values: a(1%) -4.03, b(5%) -3.44, and c(10%) -3.15.
-
25
Table 4 MTAR and RADF Results
1994:01 to 2005:03 Panel A: Momentum Threshold Autoregressive Model Results
CR τ ρ1 ρ2 FC FA Q(5) k Equity -2.94c 0.071 0.12 -0.06 2.21 0.86 4 (1.09) (-1.84)c {0.99} Apartments -2.03 0.041 -0.10 -0.12 5.67c 4.68b 0.49 4 (-0.12) (-2.67)a {0.99} Industrial -1.55 0.010 0.03 -0.13 6.59b 4.07b 0.63 4 (0.66) (-2.69)a {0.99} Lodging -2.41 0.149 -0.86 -0.05 15.67a 9.85a 1.78 4 (-4.20)a (-1.77)c {0.88} Manufactured Homes -2.69 0.017 -0.15 -0.05 1.54 0.61 4 (-2.59)a (-0.92) {0.99} Office -2.88 0.043 0.09 - 0.12 5.82c 6.08a 0.64 4 (1.11) (-3.25)a {0.99} Regional Malls -3.02c 0.002 -0.27 -0.21 4.07 4.84 4 (-0.40) (-3.42)a {0.44}
-
26
Table 4 (Continued) MTAR and RADF Results 1994:01 to 2005:03 Panel B: Residuals –Augmented Dickey-Fuller Results
Skewness Kurtosis JB CRτ Q(5) Equity 0.65 2.54 10.58 -0.41 5.15 {0.01}a {0.40} Apartments 0.48 3.16 5.27 -0.83 5.57 {0.07}a {0.35} Industrial 1.06 3.39 26.32 -0.49 8.82 {0.00}a {0.12} Lodging -0.18 2.45 2.48 -1.04 9.49 {0.29} {0.09}c Manufactured Homes -0.45 2.66 5.23 -1.28 2.35 {0.07}c {0.80} Office 0.36 3.06 2.87 -0.94 10.76 {0.24} {0.06}c Regional Malls 0.12 2.50 1.73 -1.85 4.74 {0.42} {0.45} Notes: CR is the Dickey-Fuller test statistic applied to the residuals from the cointegration equation (5) under the null hypothesis of no cointegration with critical values: a(1%) -3.73, b(5%) -3.17, and c(10%) -2.91 (Engle and Granger, 1987). CRτ is the residuals-augmented Dickey-Fuller test statistic applied to the residuals from the cointegrating equation (5) under the null hypothesis of no cointegration with critical values: a(1%) -4.36, b(5%) -3.61, and c(10%) -3.32 (Sarno and Taylor, 2003). is the estimated threshold. 1 and 2 are the estimated parameters from the MTAR specification. Standard errors denoted by [ ], t-statistics denoted by ( ), and probability values in { } where a(1%), b(5%), and b(10%). FC represents the F-statistic corresponding to the null hypothesis of no cointegration (i.e. 021 ) with critical values provided by Enders and Siklos (2001, Table 5, p. 172, n = 100 and four lags): a(1%) 8.91, b(5%) 6.56, and c(10%) 5.52. FA represents the F-statistic corresponding to the null hypothesis of symmetry (i.e. 21 ) using the standard F distribution with critical values a(1%) 4.79 and b(5%) 3.07. k is the number of lags in equation (6). Q(5) denotes the Ljung-Box Q-statistic at 5 lags.
-
27
Appendix A Johansen-Juselius Cointegration Tests Equity REIT Sub-Sectors 1994:01-2005:03
Panel A: Equity
Variables: tp , td , tnapm Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 10.74 20.97 25.52 r ≤ 1 r = 2 6.74 14.07 18.63 r ≤ 2 r = 3 2.69 3.76 6.65
trace r = 0 r ≥ 1 20.17 29.68 35.65 r ≤ 1 r ≥ 2 9.44 15.41 20.04 r ≤ 2 r ≥ 3 2.29 3.76 6.65
Variables: tp , td , tcu Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 13.51 20.97 25.52 r ≤ 1 r = 2 7.37 14.07 18.63 r ≤ 2 r = 3 1.76 3.76 6.65
trace r = 0 r ≥ 1 22.64 29.68 35.65 r ≤ 1 r ≥ 2 9.13 15.41 20.04 r ≤ 2 r ≥ 3 1.76 3.76 6.65 Panel B: Apartments
Variables: tp , td , tnapm Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 11.19 20.97 25.52 r ≤ 1 r = 2 3.79 14.07 18.63 r ≤ 2 r = 3 2.70 3.76 6.65
trace r = 0 r ≥ 1 17.68 29.68 35.65 r ≤ 1 r ≥ 2 6.49 15.41 20.04 r ≤ 2 r ≥ 3 2.70 3.76 6.65
-
28
Appendix A (continued) Johansen-Juselius Cointegration Tests Equity REIT Sub-Sectors
1994:01-2005:03 Variables: tp , td , tcu
Null Alt. Statistic 95%CV 99%CV max r = 0 r = 1 18.88 20.97 25.52
r ≤ 1 r = 2 3.78 14.07 18.63 r ≤ 2 r = 3 2.99 3.76 6.65
trace r = 0 r ≥ 1 18.88 29.68 35.65 r ≤ 1 r ≥ 2 3.78 15.41 20.04 r ≤ 2 r ≥ 3 2.99 3.76 6.65 Panel C: Industrial
Variables: tp , td , tnapm Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 11.60 20.97 25.52 r ≤ 1 r = 2 7.38 14.07 18.63 r ≤ 2 r = 3 0.46 3.76 6.65
trace r = 0 r ≥ 1 19.44 29.68 35.65 r ≤ 1 r ≥ 2 7.84 15.41 20.04 r ≤ 2 r ≥ 3 0.46 3.76 6.65
Variables: tp , td , tcu Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 8.12 20.97 25.52 r ≤ 1 r = 2 4.63 14.07 18.63 r ≤ 2 r = 3 1.58 3.76 6.65
trace r = 0 r ≥ 1 14.33 29.68 35.65 r ≤ 1 r ≥ 2 6.21 15.41 20.04 r ≤ 2 r ≥ 3 1.58 3.76 6.65
-
29
Appendix A (continued) Johansen-Juselius Cointegration Tests Equity REIT Sub-Sectors
1994:01-2005:03 Panel D: Lodging
Variables: tp , td , tnapm Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 13.95 20.97 25.52 r ≤ 1 r = 2 4.60 14.07 18.63 r ≤ 2 r = 3 0.67 3.76 6.65
trace r = 0 r ≥ 1 19.23 29.68 35.65 r ≤ 1 r ≥ 2 5.27 15.41 20.04 r ≤ 2 r ≥ 3 0.67 3.76 6.65
Variables: tp , td , tcu Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 17.70 20.97 25.52 r ≤ 1 r = 2 7.16 14.07 18.63 r ≤ 2 r = 3 1.73 3.76 6.65
trace r = 0 r ≥ 1 26.60 29.68 35.65 r ≤ 1 r ≥ 2 8.89 15.41 20.04 r ≤ 2 r ≥ 3 1.73 3.76 6.65 Panel E: Manufactured Homes
Variables: tp , td , tnapm Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 16.94 20.97 25.52 r ≤ 1 r = 2 10.56 14.07 18.63 r ≤ 2 r = 3 2.07 3.76 6.65
trace r = 0 r ≥ 1 29.57 29.68 35.65 r ≤ 1 r ≥ 2 12.63 15.41 20.04 r ≤ 2 r ≥ 3 2.07 3.76 6.65
-
30
Appendix A (continued) Johansen-Juselius Cointegration Tests Equity REIT Sub-Sectors
1994:01-2005:03
Variables: tp , td , tcu Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 7.60 20.97 25.52 r ≤ 1 r = 2 3.95 14.07 18.63 r ≤ 2 r = 3 3.50 3.76 6.65
trace r = 0 r ≥ 1 15.04 29.68 35.65 r ≤ 1 r ≥ 2 7.44 15.41 20.04 r ≤ 2 r ≥ 3 3.50 3.76 6.65 Panel F: Office
Variables: tp , td , tnapm Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 11.33 20.97 25.52 r ≤ 1 r = 2 6.22 14.07 18.63 r ≤ 2 r = 3 2.21 3.76 6.65
trace r = 0 r ≥ 1 19.76 29.68 35.65 r ≤ 1 r ≥ 2 8.43 15.41 20.04 r ≤ 2 r ≥ 3 2.21 3.76 6.65
Variables: tp , td , tcu Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 18.99 20.97 25.52 r ≤ 1 r = 2 2.62 14.07 18.63 r ≤ 2 r = 3 2.18 3.76 6.65
trace r = 0 r ≥ 1 23.79 29.68 35.65 r ≤ 1 r ≥ 2 4.80 15.41 20.04 r ≤ 2 r ≥ 3 2.18 3.76 6.65
-
31
Appendix A (continued) Johansen-Juselius Cointegration Tests Equity REIT Sub-Sectors
1994:01-2005:03 Panel G: Regional Malls
Variables: tp , td , tnapm Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 16.30 20.97 25.52 r ≤ 1 r = 2 8.31 14.07 18.63 r ≤ 2 r = 3 0.06 3.76 6.65
trace r = 0 r ≥ 1 24.67 29.68 35.65 r ≤ 1 r ≥ 2 8.37 15.41 20.04 r ≤ 2 r ≥ 3 0.06 3.76 6.65
Variables: tp , td , tcu Null Alt. Statistic 95%CV 99%CV
max r = 0 r = 1 18.82 20.97 25.52 r ≤ 1 r = 2 6.22 14.07 18.63 r ≤ 2 r = 3 2.03 3.76 6.65
trace r = 0 r ≥ 1 27.08 29.68 35.65 r ≤ 1 r ≥ 2 8.26 15.41 20.04 r ≤ 2 r ≥ 3 2.03 3.76 6.65 Notes: Critical values for the cointegration tests were obtained from Osterwald-Lenum (1992). Lag lengths were determined by Akaike’s information criterion. The Johansen-Juselius cointegration tests were performed with stochastic trends (Johansen, 1995 p.80-84).
-
32
Appendix B Granger Causality Tests Equity REIT Sub-Sectors and Capacity Utilization Measures 1994:01-2005:03 Panel A: Equity
Independent Variables Dependent Variable tp td tnapm F-statistic
2R k LM(4)
tp 0.36 0.33 0.48 0.43 0.010 1 3.84
td 0.18 0.12 0.18 0.15 0.004 (0.92)
tnapm 1.36 0.23 0.18 0.54 0.013
Independent Variables Dependent Variable tp td tcu F-statistic
2R k LM(4)
tp 0.77 2.50c 0.55 1.40 0.095 3 6.23
td 2.59c 2.24c 0.92 1.81c 0.119 (0.72)
tcu 2.05 2.08 6.98a 3.76a 0.219
Panel B: Apartments
Independent Variables Dependent Variable tp td tnapm F-statistic
2R k LM(4)
tp 2.96c 2.49 0.68 2.21c 0.049 1 7.41
td 0.01 5.60b 0.06 1.94 0.043 (0.59)
tnapm 0.55 0.02 0.13 0.22 0.005
Independent Variables Dependent Variable tp td tcu F-statistic
2R k LM(4)
tp 3.47c 2.41 0.57 2.17c 0.048 1 10.21
td 0.01 6.22b 2.49 2.79b 0.061 (0.33)
tcu 0.84 1.24 0.24 0.77 0.018 Panel C: Industrial
Independent Variables Dependent Variable tp td tnapm F-statistic
2R k LM(4)
tp 5.06b 0.04 0.00 1.75 0.039 1 12.76
td 0.44 0.01 0.01 0.15 0.004 (0.17)
tnapm 0.56 0.51 0.13 0.37 0.008
-
33
Appendix B (Continued) Granger Causality Tests Equity REIT Sub-Sectors and Capacity Utilization Measures 1994:01-2005:03
Independent Variables Dependent Variable tp td tcu F-statistic
2R k LM(4)
tp 1.37 2.34b 1.52 1.82b 0.195 5 10.37
td 0.45 4.52a 0.58 1.96b 0.206 (0.32)
tcu 1.35 1.99c 3.92a 2.39a 0.241
Panel D: Lodging
Independent Variables Dependent Variable tp td tnapm F-statistic
2R k LM(4)
tp 1.55 1.13 1.83 1.14 0.132 5 8.92
td 6.54a 0.20 1.59 3.64a 0.326 (0.45)
tnapm 2.31b 1.08 0.68 1.71c 0.185
Independent Variables
Dependent Variable tp td tcu F-statistic 2R k LM(4)
tp 1.03 0.52 0.38 0.74 0.070 4 10.24
td 9.68a 0.58 1.16 4.22a 0.302 (0.33)
tcu 1.94 0.56 5.52a 2.69a 0.216
Panel E: Manufactured Homes
Independent Variables Dependent Variable tp td tnapm F-statistic
2R k LM(4)
tp 0.63 3.27b 0.13 1.51 0.101 3 8.26
td 0.30 4.30a 0.77 1.98b 0.128 (0.51)
tnapm 0.85 0.34 0.65 0.60 0.042
Independent Variables Dependent Variable tp td tcu F-statistic
2R k LM(4)
tp 0.83 2.73b 1.33 1.95c 0.127 3 5.74
td 0.23 4.29a 1.77 2.36b 0.149 (0.77)
tcu 0.41 0.65 6.29a 2.71a 0.168
-
34
Appendix B (Continued) Granger Causality Tests Equity REIT Sub-Sectors and Capacity Utilization Measures 1994:01-2005:03 Panel F: Office
Independent Variables Dependent Variable tp td tnapm F-statistic
2R k LM(4)
tp 0.59 1.86 0.19 0.88 0.020 1 4.20
td 0.01 6.79a 0.16 2.36c 0.006 (0.90)
tnapm 0.09 0.87 0.13 0.35 0.008
Independent Variables Dependent Variable tp td tcu F-statistic
2R k LM(4)
tp 0.78 1.74 0.81 1.09 0.024 1 4.73
td 0.01 6.83a 0.72 2.56c 0.056 (0.86)
tcu 1.26 0.05 0.21 0.49 0.011 Panel G: Regional Malls
Independent Variables Dependent Variable tp td tnapm F-statistic
2R k LM(4)
tp 3.75c 0.65 1.77 2.07c 0.046 1 10.06
td 0.16 2.50 0.03 0.87 0.020 (0.35)
tnapm 2.01 1.53 0.16 1.32 0.030
Independent Variables Dependent Variable tp td tcu F-statistic
2R k LM(4)
tp 1.45 0.85 1.82 1.45 0.161 5 8.23
td 0.60 0.92 0.30 0.60 0.074 (0.51)
tcu 1.79 1.21 4.03a 2.35a 0.238
Notes: Partial F-statistics are denoted under the respective independent variables. F-statistic represents the overall F-statistic for the equation. Lag lengths, k , determined by Akaike’s information criterion. The multivariate Lagrange multiplier test for serial correlation at 4 lags is denoted by LM(4) with probability values in parentheses.
-
35