Online Appendix. Spillovers of Non-Fundamental Risks in Securitized Real Estate Companies: Singapore s Evidence. National University of Singapore

Transcription

1 Online Appendix Spillovers of Non-Fundamental Risks in Securitized Real Estate Companies: Singapore s Evidence Chen, Zhen 1, Seah, Kiat Ying 2, Sing, Tien Foo 3 and Wang, Long 4 National University of Singapore Revised: 4 May Department of Real Estate, National University of Singapore. chenzhen77@gmail.com 2 Department of Real Estate, National University of Singapore. sky@nus.edu.sg 3 Corresponding author. Department of Real Estate/Institute of Real Estate Studies (IRES), National University of Singapore. rststf@nus.edu.sg 4 Department of Real Estate, National University of Singapore. wanglong@u.nus.edu

2 This online appendix supplements the main paper. We provide the empirical results on the nonfundamental risks spillovers using samples of Singapore s real estate investment trusts (REITs) and real estate operating companies (REOCs). Here, we report the results of our tests using the REITs index inclusion event in a structurally different market. The Singapore s REIT market is a relatively young market with the debut of its first public REIT only in Prior to its emergence, REOCs are the only publicly listed real estate vehicle for firms, which are involved in a wide range of real estate businesses ranging from land development, residential real estate sales to commercial real estate investments. Like the US REITs, Singapore s REITs (SREITs) are tax-exempt vehicles set up to invest (more than 75% of total assets) in income-generating real estate. High dividend pay outs (more than 90% of earnings) and leverage restrictions (borrowing limit at 45% of the total asset value) are two features that set REITs apart from REOCs in Singapore. There were only two other non-index REITs listed on the Singapore Exchange, when the first two REIT were added to the Singapore s general market index. However, REOCs existed long before REITs, and several REOCs were already included in the broad market index before REITs. We empirically test the cross-category asset return spillovers using the two structurally different public real estate assets in Singapore (REITs and REOCs). In contrary to US results, we find weak and near insignificant spillover effects of the REITs index inclusion event on REOC betas in Singapore s tests. We next use the REIT IPO events in Singapore to further test if spillover effects could be caused by other market frictions; and the results show that REIT betas are more positively correlated with market returns in the post REIT IPO periods compared to REOC betas. We also find significant evidence of captive influence of sponsor REOCs in explaining changes in REOC betas. REIT IPOs are used as a de-risk vehicle by sponsor REOCs, where we find significant decreases in the systematic risks of sponsor REOCs after the REIT spinoff events. The results imply that investors could harness captive relationships of REITs with their sponsor REOCs in Singapore to reduce systematic risks. 1

3 A1. Singapore REITs Singapore s REITs (SREIT) only emerged in 2002 following the listing of the first retail mall REIT, Capitaland Mall Trust1 on the Singapore Exchange (SGX). As on 2005, there were only 7 SREITs with a total market capitalization of S$11 billion representing 2.5% of the total market capitalization of the SGX. The strong recovery in real estate prices in 2006 has seen the rapid expansion of REITs in Singapore. More new REITs are listed during this period; and at the same time, the incumbent REITs also aggressively grew their portfolios through accretive acquisitions. Some REITs also started to expand regionally by acquiring foreign real estate. In a short span of two years as of 2007, the number of listed REITs increased to 20 and the total market capitalization grew to S$28 billion constituting 3.9% of the total SGX market capitalization. As of April 2013, 34 S-REITs (inclusive of 8 business trusts) are listed on the SGX with total market capitalisation of S-REITs estimated at approximately SGD 75 billion. We identify March 18, 2005 as the treatment date on which the two pioneered REITs CapitaLand Mall Trust (CMT) and Ascendas REIT (A-REIT), were concurrently added for the first time into the Singapore s general market index Straits Time Index (STI).2 We use the treatment event to test the spillover effects of the REIT index inclusion event on non-index REITs and REOC stock returns. Due to the small number of non-index REITs at the treatment date,3 the results on non-index REITs need to be interpreted with the limitations in mind. REOCs existed way before REITs were listed in Singapore, and some REOCs were included into STI constituents before the two index REITs inclusion. Therefore, in Singapore s experiment, we separately test the spillover effects of the REITs inclusion into STI event on both index-reocs and non-index REOCs. In Singapore s markets, REOCs are larger by market capitalization and also exist long before REITs. Institutional investors should thus also take into account the strong presence of REOCS, and also the captive relationships of REOCs with REIT sponsored by them. 1 CapitaMall Trust (CMT) was the first REIT listed on the Singapore Exchange in July It has been renamed in 2015 to CapitaLand Mall Trust (CMT). 2 In the two other emerging REIT markets in Asia, which are Japan and Hong Kong, no REITs were included in the general market indices. 3 CapitaCommercial Trust and Suntec REIT were the only two non-index REITs listed only shortly after inclusion event. 2

4 Singapore data are obtained from Bloomberg and Datastream. We collected daily and weekly total stock return and market value data for the universe of Singapore stock market, which contain a total of 720 stocks as at the end of Our sample period spans from July 1, 2003 to December 29, The sample period is winsorized at both ends to avoid two exogenous shocks: (a) the severe acute respiratory syndrome (SARS) outbreak, which was only fully eradicated in July 20034, and (b) the US subprime financial crisis following the filing of the bankruptcy protection by several US conglomerates including Lehman Brothers between January and February Table A.1 presents the summary statistics for Singapore s securitized real estate markets. There were 25 REOCs and 10 REITs listed on the Singapore Exchange during the study period In Singapore, REOCs with an aggregate market value of US$33, million were nearly 3 times larger than Singapore s REITs (US$10, million). Based on the performance statistics, Singapore REITs are defensive in nature with higher book to market value ratio, and low systematic risks. Another interesting distinguishing feature of the two market is that the high concentration of the REOCs in Singapore, which is reflected by the high Herfindahl Index of The presence of large developers with large controlling shares in the real estate market could have important influence on the dynamics in Singapore real estate markets. [Insert Table A.1 here] A2. Results A2.1 Tests of Comovements of Asset Returns The following specification tests for the treatment effect of the REITs index inclusion event for Singapore. The event date is March 18, R i,t = α + β 1 AFTER + β 2 R STI RE,t + β 3 (R STI RE,t AFTER) + ε t (1) where R i,t is asset i return (either daily or weekly) in the period t. R STI RE,t is the value-weighted return of all STI index constituent stocks in the period t excluding index REITs and index REOCs. 4 The World Health Organization (WHO) removed Singapore from the list of SARS infected areas on May 30, 2003, when no new cases were reported for 20 consecutive days. 3

5 AFTER is the treatment time dummy, which is defined by the date of the inclusion of the two REITs into the STI on March 18 th, The OLS estimator is used to estimate the models; and the results are summarized in Table 1. The difference in non-index REIT betas before and after the REITs index inclusion event are significant and positive. The weekly spillover effects of the REITs index inclusion event are stronger than the daily spillover effects in Singapore s non-index betas, which are estimated at (weekly) and (daily), respectively. The spillover effects are stronger in Singapore s non-index REIT market than in the US s non-index REIT market. [Insert Table A.2 here] We next use the REOC returns as proxies for assets that are not in the same category as REITs to test the sentiment views on the spillover effects. Unlike in the US, Singapore s REOCs are relatively larger by the market size, and some of them were included in the STI index even before the existence of REITs in Singapore. We sort Singapore s REOC samples into index REOCs and non-index REOCs, when the tests are repeated in the Singapore s setup. The results in Table A.2 show that increases in daily beta of non-index-reocs (0.196) and index REOCs (0.148) are significant at less than 1% and 10%, respectively, but increases in the weekly betas are insignificant for the two types of REOCs. The Singapore s results show that the spillover effects on both daily and weekly betas are weaker in the REOCs (both index and non-index REOCs) than in the non-index REIT markets. The differential beta changes could imply that magnitude of sentiment risk spillovers associated with REIT s index inclusion news into the two non-overlapping markets are different, though both REIT and REOC markets react to the inclusion event. A2.2 Robustness Tests A2.2.1 Removing non-real estate related shocks Like ALP, we purge possible market-wide shocks that are unrelated to real estate markets in a twostage process by first estimating the S&P index model (excluding real estate stocks), R STI RE,t, as follows: R STI RE,t = R non_re,non_sti,t + η t (2) 4

6 where R non_re,non_sti,t represents the value-weighted return of the non-index stock portfolios.5 The residual, η t, which represents the market-wide Straits Times Index (STI) shock that is orthogonal to non-real estate market shocks, is included in the following stage-two regressions: and R non_index_reocs,t = α + β 1 AFTER + β 2 η t + β 3 (η t AFTER) + ε t R non_index_reits,t = α + β 1 AFTER + β 2 η t + β 3 (η t AFTER) + ε t (3a) (3b) The results shown in Table A.3 are insignificant for both non-index REOCs and index REOCs; and the sign is negative on the beta of the index REOCs, which have longer history of existence on the Singapore Exchange and are larger firms by market capitalization. [Insert Table A.3 here] We use a variation of the market model by switching the value-weighted market return, R non_re,non_s&p,t, to the left-hand side of the model, and uses the three S&P indices to jointly purge the non-real estate shocks from the market wide shocks: R non_re,non_sti,t = R STI RE,t + ζ t (4) The residual, ζ t, extracted from Equation 4 is then used to replace the earlier residual, η t, in Equations (3a) in the second stage regression. The results are summarized in section d of Table A.3. The results are mixed, non-index REIT s beta change is positive, but insignificant; whereas, index-reocs return correlation with the purged market return (0.400) is significant and positive at less than 1% level. A2.2.2 Changing Composition of Index Constituent Stocks 5 The ALP study uses the return of the value-weighted portfolio of all non-index REIT stocks, R non_reits,non_s&p,t. 5

7 To minimize possible survival bias caused by contemporaneous changes to the constituent stocks in the general market index, we reconstruct the general market indices using only those stocks that stay (survive) in the index for the entire study period. The results are shown in Table A.4. The results remain largely positive and significant for only the daily returns but not for the weekly returns. [Insert Table A.4 here] A2.2.3 Gradual Increases in the Betas Here, we run the same regressions as ALP, but using daily data instead to increase the degree of freedom in our estimation: 5 5 R i,t = α + d=1 δ d D d + βr STI RE,t + d=1 φ d (D d R STI RE,t ) + ε t (6) We set a time dummy variable, D d (0,1), around the treatment event date of March , which include the pre-event time dummy, D 1 = [October 1, 2004 to March 18, 2005]; and other post-event time dummy: D 2 = [March 19, 2005 to September 30, 2005], D 3 = [Oct 1, 2005 to February 28, 2006], D 4 = [March 1, 2006 to July 31, 2006], and D 5 = [August 1, 2006 to December 31, 2006]. The results in Table A.5 show that the pre-event and the event period betas of the nonindex REOCs were negative and insignificant, and the spillover effects become highly significant and positive, but delayed by one period. We only observe the positive increases in beta after D2 periods. [Insert Table A.5 here] A2.2.4 Market Frictions Like Barberis et al (2005) and ALP (2007), we use the Dimson (1979) betas to measure contributions of market frictions to the observed beta increase in non-index REIT and REOCs. The Dimson betas associated with the 5-day prior and the 5-day after returns of the S&P general index (excluding all REITs and REOCs) are estimated using the following model: 6

8 5 5 R i,t = α + βafter + β d R S&P RE,t d + γ d (R S&P RE,t d AFTER) + ε t d= 5 d= 5 (7) The details of the estimation are outlined in the main paper and the results are shown in Table A.6. Market frictions appear to dominate the beta changes for the REOCs in Singapore. We observe that the inclusion of CMT and A-REIT STI to have no significant spillover effects on the nonindex REOCs. [Insert Table A.6 here] A2.2.5 Alternative Treatment Events Singapore s REIT industry is a relative new, and also a highly captive market, where large developers sponsor and retain controlling stake in some listed REITs. The two S-REITs (A-REIT and CMT) included in STI are affiliated to the Government and its entity - Temasek Holdings. A- REIT was sponsored and controlled by Ascendas Group, which is a private investment vehicle of JTC Corporation, a government agency under the Ministry of Trade and Industry; whereas CMT was sponsored by CapitaLand Limited, which is a public listed firm with substantial shareholdings controlled by Temasek Holdings. At the time of index inclusion, CapitaLand Limited, has already been a constituent stock of STI. City Developments Limited (CDL) and Keppel Land are other two REOCs in the STI constituent stocks that sponsor REITs listed on Singapore Exchange. The strong presence of the large-cap REOCs in the STI index could weaken the spillovers of the REITs index inclusion news onto the REOC betas in Singapore, which may potentially be the root for the differences observed in the earlier spillover results between the Singapore and the US markets. As the number of initial public offering (IPO) for new REITs is relatively small and easily tractable at the time of the study, we use the IPO events as alternative shocks to test if spillover effects are significant with the REIT, and also across to the REOC markets. Like Barberis et al (2005), we model the returns of REITs and/or REOCs as a simple market model: R i,t = α i + β i R Market Index,t + ε t (9) 7

9 where R Market Index,t denotes the market index, which is represented by the STI. We set the effective IPO date, T, as the first time the new REIT appears in the weekly time series, and remove the weekly return in the week prior to the IPO event to reduce the IPO pre-announcement noise. We then run two regressions, one covering the 52-week sample window before the effective IPO event, [t< T], and another covering the 52-week sample window after the effective IPO event, [t T]. Changes in beta are estimated as the difference in the post-ipo beta and the pre-ipo beta estimated from the models, i.e. [ βi =βi,t T βi,t<t]. IPO events that are too close to each other cause clustering problem that may bias the results. We remove IPO events that are within 4 weeks from the previous one. We use two set of aggregate data in evaluating the spillover effects, which include: (1) commercially (FTSA) published REIT and REOC indices; and (2) value-weighted of REIT and REOC portfolio return series. Based on 23 REIT IPO events in Singapore, Table A.7 computes the average differences in changes in the pre- and post-ipo betas between REITs and REOCs, [ β = Δβ REITs Δβ REOCs ]. The results show average positive differences of when using the returns of the FTSA indices and when using the value-weighted portfolio returns, and both the results are statistically significant with t-statistics of and 2.843, respectively. The REIT IPO events in Singapore induce stronger reaction in REIT market (larger increases in the pre- and the post-ipo betas of REITS) than in REOCs. The results affirm the earlier findings on the REITs index inclusion event that the cross-category spillovers of REIT market news onto REOC betas were weak and insignificant compared to the spillovers effects on REOC betas. [Insert Table A.7 here] Next, we conduct further tests to separate the sponsor-captive effects, by including REOC sponsors that are already constituent stocks in the STI index. Non-real estate sponsors in the STI constituents (e.g. Singapore Press Holdings and Parkway Holdings Limited, etc.) are excluded because of their business lines are not related to real estate, and their market betas could be influence by other market shocks that are orthogonal to the real estate market shocks. Base on the sample criteria, and also data availability, we are only able to use one REIT IPO (Ascott Residence Trust (ART)) 8

10 and two spin-off events (CapitaLand Commercial REIT (CCT) and Keppel REIT (K-REIT)) in our tests, where the sponsors of the two events are CapitaLand Limited and Keppel Land Limited. The results (in the lower panel of Table A.7) show that sponsor-reocs betas are negative for the CCT (-0.45) and K-REIT (-0.68) events, and the CapitaLand s beta is marginally positive estimated at Compared to the responses from other non-sponsor-reocs over the events, we take the difference of the beta changes between the sponsor REOCs and other non-sponsor REOCs. The results show that the relative changes in betas between the sponsor-reoc vis-à-vis the nonsponsor-reocs are negative ranging from to Sponsor-REOCs that use REIT IPOs as a vehicle to unlock illiquid assets and reduce leverage from their books show significant larger negative spill-over effects than other non-sponsor REOCs, which may not have direct benefits from the REIT IPO events. The results imply that sponsor-reocs experience negative systematic shocks from the IPO/spinoff events, which are deemed as a de-risk vehicle for REOC sponsors. The results also verify that sponsor-captive effects could influence spillover effects; and the strong presence of large sponsor-reocs in Singapore could explain the differences in sentiment risk spillovers between the US and Singapore. 3. Discussions We replicate the spillover tests using the REITs index inclusion event in Singapore, which is a different market characterised by the strong presence of large REOCs that have existed on the Singapore Exchange long before the emergence of REITs. We find weak and near insignificant spillover effects of the REITs index inclusion event on REOC betas in Singapore s tests. The contrasting outcomes between the US and Singapore experiments may imply that the categorization of assets is not a necessary condition for the spill-over effects; where the spillover effect could be caused by other correlated market frictions. We use REIT IPO events in Singapore as the alternative treatments in our tests, and find that REIT betas are more positively correlated with market returns in the post REIT IPO periods compared to REOC betas. The differences in REIT and REOC betas widen after the REIT IPO news. We also use three sponsored REIT IPO events and find positive evidence on the captive effects of REOCs in Singapore s market. REIT 9

11 IPOs are used as a de-risk vehicle by sponsor REOCs, which experience decreases in the systematic risks after the REIT spinoff events. Reference: Ambrose, B.W., Lee, D.W. and Peek, J Comovement after Joining an Index: Spillovers of Nonfundamental Effects. Real Estate Economics 35(1): Barberis, N., A. Shleifer and J. Wurgler Comovement. Journal of Financial Economics 75(2): Dimson, E Risk Measurement When Shares are Subject to Infrequent Trading. Journal of Financial Economics 7(2):

12 Table A.1: Characteristics of the REIT and REOC Markets in Singapore Singapore REIT REOC Established 2002 N.A. Sample period July 1, 2003 to December 29, 2006 Number of listed REITs Market Value (MV) (US$ million) 10, , Total Return (TR) (%) Book to Market Value Ratio (bmk) Price to Book Ratio (PB) Relative Price to Book Ratio (RPB) Systematic Risk (beta) Herfindahl Index Note: The table shows the summary statistics for the sample real estate investment trusts (REITs) and real estate operating companies (REOCs) in Singapore. We include various statics, such as total return, book to market value ratio, price to book ratio, relative price to book ratio, systematic risk and Herfindahl Index. The data are obtained from Datastream. 11

13 Table A.2: Tests of Sentiment Risk Spillovers Country Singapore Asset Category Non-index REOCs Index REOCs Market Returns: Beta Sig Beta Sig Daily Weekly Note; This table shows the results on differences in Beta before and after the REIT s index inclusion events, Beta ; and Sig. shows the level of significance on the change in beta. We run the tests for the effects of REIT inclusion into index using Singapore s non-index REITs and Index REOCs, and the market return proxy is based on the Straits Times Index Return (STI). We use both daily and weekly returns. 12

14 Table A.3: Spillover Effects and Real Estate Related Market Risks Country Singapore Asset Category Non-index REOCs Index REOCs Beta Diff (p) Beta Diff (p) Market Returns: Residual of Straits Times Index (STI) Return Daily Residual of Non-Real Estate and Non-index Stock Portfolio Returns Daily Note; This table shows the results on differences in Beta before and after the REIT s index inclusion events, Beta ; and Sig. shows the level of significance on the change in beta. The inclusion of the REITs into Singapore s Straits Times index occurs on March 18, We test the effects of REIT inclusion into index using Singapore s non-index REITs and Index REOCs; and the market return proxy is based on the residual of Straits Times Index Return (STI). Only daily returns are used in the tests. 13

15 Table A.4: Controlling for the Effects Composition Changes in Market Indices Country Singapore Asset Category Non-index REOCs Beta Diff (p) Market Returns: Fixed STI Return Daily Weekly Note; This table shows the results on differences in Beta before and after the REIT s index inclusion events, Beta ; and Sig. shows the level of significance on the change in beta. We test the effects of REIT inclusion into index using Singapore s non-index REOCs, and the market return proxy is based on Fixed Straits Times Index Return (STI). Both daily and weekly return series are used in the tests. 14

16 Table A.5: Changes in Betas Before and After the REITs Index Inclusion Events Country Singapore Asset Category Non-index REOCs Estimate Diff (p) Market Returns: STI return Initial = Note; This table shows the results on differences in Beta before and after the REIT s index inclusion events, Beta ; and Sig. shows the level of significance on the change in beta. The changing beta periods are defined as: D 1 = [October 1, 2004 to March 18, 2005]; and other post-event time dummy: D 2 = [March 19, 2005 to September 30, 2005], D 3 = [Oct 1, 2005 to February 28, 2006], D 4 = [March 1, 2006 to July 31, 2006], and D 5 = [August 1, 2006 to December 31, 2006]. 15

17 Table A.6: Dimson Betas on Market Frictions Country Singapore Asset Category Non-index REOCs Dimson Beta: Beta Diff (p) Market Returns: STI Return Daily Note; This table shows the results on differences in Beta before and after the REIT s index inclusion events, Beta ; and Sig. shows the level of significance on the change in beta. We test the effects of REIT inclusion into index using Singapore s non-index REOCs, and the market return proxy is based on the Straits Times Index Return (STI). Only daily returns are used in the tests. 16

18 Table A.7: Alternative Treatment Using REIT IPO Events in Singapore a) REIT IPO Events Country Index Singapore Value-weighted Number of IPO events Average Difference,( β = Δβ REITs Δβ REOCs ) Standard error t-statistic b) Sponsor REOCs and Non-Sponsor REOCs Sponsor CapitaLand CapitaLand Keppel Land REIT IPO/Spin-off Events CapitaLand Ascott Residence Keppel REIT Commercial Trust Trust First day trading date May 17, 2004 March 23, 2006 April 28, 2006 Sponsor REOC beta change, (Δβ Sponsor ) Non-sponsor REOC beta change, (Δβ other REOCs ) Difference in beta change, (Δβ Sponsor - Δβ other REOCs ) Note; The Panel (a) above shows the REIT IPO events in Singapore during the study periods. During the sample periods, there were 23 REITs listed on the exchange. The average difference in beta changes before and after IPO events is represented by β. We also conduct the t- test of differences and the standard error and t-statistics are summarized in the table. Panel (b) repeats the same tests on the IPO events, where we use only the sponsor-affiliated REIT IPO events. There are three REIT IPOs that are afficiated with the sponsors, which are also listed REOCs at the time of IPOs (CapitaLand and Keppel Land). The beta changes for sponsor REOCs and non-sponsor REICS are shown in the last row of the table. Data are obtained from Datastream. 17