Real Estate Investment Trusts (REITs) stock splits,
operating performance and liquidity.
Abstract:
This study examines 89 U.S. REIT stock splits in the period from 1980 to 2017. We find significant 5-day cumulative abnormal returns around the announcement date. This supports the attention-grabbing and signaling theories that predict an increase in trading around the announcement date. However, no and weakly significant abnormal returns are found around the record date and ex-date respectively. In addition, no evidence of long term post-split operating performance increases are found as well. This is consistent with the hypothesis that the signaling theory will not hold for REITs, because of the relative transparency in this industry. Furthermore, liquidity increases around the announcement date and in the period after the ex-date. However, these liquidity increases are not statistically significant, which does not support the trading-range/liquidity theories. Therefore, the findings of this study suggest that REITs only split their shares to attract investors attention. Additionally, we find that the changes in liquidity can significantly explain the 5-day cumulative abnormal returns around the announcement date. Finally, no significant evidence of less transparency of hybrid- and diversified REITs is found.
University of Amsterdam/Amsterdam Business School
MSc Finance: Dual track Real Estate Finance and Asset Pricing
Master Thesis
Author: Jelte Bemelman 10479619
Thesis supervisor: M. Droës
2 Statement of originality:
This document is written by Student Jelte Bemelman who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Table of contents
1. Introduction 4
2. Literature review 7
2.1 Stock split theories 7
2.2 REITs are special corporate entities 8
2.3 Prior research on REIT stock splits 9
2.4 Long term performance after stock splits 11
2.5 Different REIT industries 12
3. Methodology 13
3.1 Methodology for the initial reaction of a REIT stock split 13
3.2 Methodology for the long term operating performance 14
3.3 Methodology for the liquidity measures 15
3.4 The determinants of the announcement period abnormal returns 17
4. Data and descriptive statistics 18
4.1 Data collection 18 4.2 Descriptive statistics 19 5. Results 23 6. Robustness checks 32 7. Discussion/Conclusion 36 7.1 Conclusion 36
7.2 Limitations and recommendations for further research 37
References 39
4
1. Introduction
Stock splits are interesting corporate actions management can execute, because stock splits increase the number of shares without affecting any fundamental value of the firm. So, why do firms perform stock splits if no fundamental firm value is affected?
Extensive research on stock splits aimed to answer this question and has led to four stock split theories. The signaling theory states that management uses stock splits to signal favorable future performance to the stock market. Furthermore, the optimal trading range- and liquidity theories explain that firms use stock splits to avoid too high share prices and improve liquidity. Finally, the attention-grabbing theory states that management performs stock splits to grab attention from small uninformed investors.
Despite the extensive research and development of above mentioned theories on stock splits, recent studies (e.g. Boehme et al., 2003; Byun and Rozeff, 2003), exclude real estate investment trusts (REITs) from the sample. Researchers claim this is because REITs are different entities in terms of corporate structure and operating environment. REITs have the obligation to distribute 90 to 95% of net earnings to their investors in forms of dividends in order to qualify for tax-exemption (Li et al., 2006). Because more than 90% of net earnings has to be paid in dividends, REITs have to acquire capital from external markets in order to expand. These external markets such as investment banks, funds and other investors, monitor the performance of REITs extensively. Furthermore, the REIT industry is known for its substantial disclosure on the assets REITs have in their possession. Hence, extensively monitoring from the capital markets and the information disclosure of REIT performance create substantial transparency compared to their industrial counterparts, which suggests that REITs do not have the need to signal future performance to the market. This allows for a cleaner evaluation of the other theories on stock splits. We therefore examine the following research question:
Do REITs split their shares to attract investor attention rather than to signal or improve trading and liquidity in the long run?
Hardin et al. (2005) and Huang et al. (2011) are the first to investigate REIT stock splits to gain insights in the factors that determine stock splits at the REIT asset class level. This is important for managers and investors in the REIT industry, since prior research tends to exclude REITs. Huang et al. (2011) examine if the trading range/liquidity theories will hold for REITs. This study finds no significant increase in liquidity after the split. However, Huang et al. (2011)
5
mention that the liquidity results might be biased, since they collected a limited sample of REIT stock splits.
Furthermore, Hardin et al. (2005) and Huang et al. (2011) find significant abnormal performance around the split announcement date, which support the predictions of the signaling and attention-grabbing theories. Although, in the long run both studies do not find significant stock abnormal performance, suggesting that the signaling theory is not likely to hold for REITs, because of the relative transparency in this industry.
In addition, there remains substantial debate in the finance field on which event methodology to use for long term stock performance pointed out by Fama (1998) and Mitchell and Stafford (2000). Since there is no clear methodology to test the long term stock performance of split firms, the signaling theory could also be empirically investigated by the change in operating performance. Studies from Lakonishok and Lev (1987) and Asquith et al. (1989) investigate the change in operating performance for industrial firms. However, for the REIT industry specific, this has never been studied before. Hence, examining the REIT operating performance change, contributes to existing literature if the signaling theory will hold for this transparent industry, since there is still no consensus which event methodology to use for long run stock performance.
Moreover, this study divides the stock split sample into two groups. Hybrid and diversified REITs as one group and other REIT property types as the other group. This separation is to address the findings of Danielsen and Harrison (2007) that these REIT types have microstructure measures, suggesting that these REIT types are difficult to value, which decreases the transparency of these REITs. This is of importance for further research on dividend and stock split signaling theories, because excluding these less transparent REIT types might allow for an even cleaner evaluation.
The third contribution of this study addresses the concerns of Huang et al. (2011) who point out that the liquidity results might be biased, because of the limited sample collection. They advocate that further research examines the liquidity change again with a larger REIT stock split sample. In this research, 89 REIT stock splits are collected which is significantly larger than the 45 splits Huang et al. (2011) collected.
Using a sample of 89 REIT stock splits in the period 1980-2017, we find significant 5-day cumulative abnormal returns around the announcement date. This supports the
attention-6
grabbing and signaling theories that predict an increase in trading around the announcement date. However, no and weakly significant abnormal returns are found around the record date and ex-date respectively. In addition, no evidence of long term post-split operating performance increases are found. This is consistent with the hypothesis that the signaling theory will not hold for REITs, because of the relative transparency in this industry. Furthermore, the liquidity increases around the announcement date and in the period after the ex-date. However, no significant evidence is found that liquidity increases around and after the stock split. Hence, the findings of this study suggest that REITs split their shares to attract investors attention. Finally, no significant evidence of less transparent hybrid- and diversified REITs is found.
The remainder of this study is organized as follows. In the following section, the related literature on stock splits and REIT stock splits is described. Section 3 discusses the methodologies used to answer the research question. In section 4, the descriptive statistics and data collection are presented. Section 5 discusses the results of the research. In section 6, robustness checks on several methodologies are performed. The study concludes with a discussion in section 7.
7
2. Related literature
In this section, the literature related to this topic is elaborated further and hypotheses are developed to answer the research question. First, the several theories of stock splits are described. Then the peculiarities of the REIT industry is are explained, followed by prior research on REIT stock splits. Eventually, the long term performance of stock splits and differences in REIT industries are explained.
2.1 Stock split theories
Stock splits are interesting corporate actions to investigate. When managers perform a stock split, the number of shares outstanding increases without affecting the market value and the share price of the firm. Hence why do managers perform stock splits?
In the field of finance, Fama et al. (1969) were the first to indicate that management uses stock splits and dividend policy to signal positive future prospects of the firm to the stock market. Subsequent research on stock splits focused on whether abnormal returns around stock split announcements were consistent and the reasons why management performs these splits (Hardin et al., 2005).
This subsequent research have led to the development of four main theories why firms split their shares. These theories are the signaling theory, liquidity theory, trading-range theory and the attention-grabbing theory. The signaling theory is developed by Brennan and Copeland (1988). This theory states that managers use stock splits to convey favorable inside information about the firm’s future prospects to investors. These favorable prospects of the firm attract investors who buy the stock and this will increase the stock price of the firm. Hence according to the signaling theory, stock splits are associated with positive excess returns around the announcement date and in the long term after the stock split. These excess abnormal returns around the announcement date are empirically verified by the extensive study by Nazar and Rozeff (2001) for industrial firms. Evidence of long term outperformance of split firms is rather mixed and will be elaborated in the long term performance subsection in this chapter.
The trading range theory conducted by Copeland (1979) states that managers want their share price to be in an optimal price range. If securities trade in this optimal price range, the brokerage costs relative to the value of the stock are lowest which improves liquidity of the stock. Furthermore the liquidity hypothesis (Lakonishok and Lev, 1987; Maloney and Mulherin, 1992;
8
Anshuman and Kalay, 2002) argues that management strives to avoid too high stock prices, because when stock prices are too high, small or uninformed investors cannot trade in large quantities. This limitation of smaller investors trading in smaller quantities reduces the liquidity of the stock. Hence, splitting the stock reduces the share price and improves liquidity. Hence, the trading range/liquidity hypotheses predict that liquidity increases around the split and that this increase is permanent in the long run.
The trading range/improved liquidity hypotheses are supported by several empirical researches. Dhar et al. (2003) find that small individual investors trade more in stocks who conducted a split. Moreover, Easley et al. (2001) find an increase in uninformed trades after stock splits, which implies that firms attract groups of smaller investors to trade their shares. Finally, Baker and Gallagher (1980) conduct a survey among financial management and find that 94% of respondents state that they use stock splits to improve liquidity.
The signaling and liquidity theories were the prominent theories explaining stock splits during the last era. However since 2008, another theory established in the field of behavioral finance: the attention-grabbing theory. According to the attention-grabbing theory developed by Barber and Odean (2008) individual investors have limited attention. They do not have the resources to follow the performance of each individual stock. Therefore, small investors tend to buy stocks that recently grabbed their attention for example if the stock was in the news, regardless if the news was positive or negative. For this reason, managers conduct stock splits to grab attention from individual investors who then buy these stocks and improve trading activity (Barber and Odean, 2008). The earlier empirical research of Schultz (2000) and Brennan and Hughes (1991) find respectively an enlarged ownership base and more analysts following the stock after the stock split, which suggests that securities receive more attention and more trading from investors around a split announcement. So, the attention-grabbing theory predicts significant positive abnormal returns around the announcement date.
2.2 REITs are special corporate entities
Although stock splits are investigated extensively, most empirical research tends to exclude REITs from their sample. The reason that REITs are excluded is because of their corporate structure and operating environment. In order to qualify for tax-exemption, REITs have to pay out 90% to 95% of net earnings in dividends. This is in sharp contrast to industrial firms who pay much less or no dividends at all (Li et al., 2006). Because 90% of net income is paid to shareholders, REITs have to acquire capital from external sources to expand their business.
9
Hence, REITs are required to actively manage their assets and capital to offset the inability to generate internal capital accumulation. They have to lend from the capital market, such as commercial banks, bond underwriters and CMBS investors to acquire new properties and development projects (Hardin et al., 2005). Therefore REITs generally have higher debt ratios than industrial firms and are extensively monitored by the capital market.
Furthermore, the REIT market is known for its substantial disclosure on the assets these property companies own. Disclosure about which property types REITs own, geographical regions they invest in and information about the cash flows the properties generate (Li et al, 2006). Because of this transparency on REIT performance, institutions like the NAREIT, MSCI and EPRA are able to set up performance indices for different property types and regions. These indices are practical for institutional investors who strive to diversify within the real estate asset class. Moreover, the indices are also used as benchmarks were property managers can evaluate their performance with industry peers. Because of this information disclosure, investors in the listed property market are able to compare the underlying assets with the market value of the REIT to reveal if there are premiums or discounts to NAV.
All in all, extensively monitoring from the capital market and the information disclosure of REIT performance create substantial transparency compared to their industrial counterparts. Hence, this transparency restrains the urge of management to signal positive future prospects to investors, which suggests that the signaling theory is less likely to hold for REITs.
2.3 Prior research on REIT stock splits
There is abundance of research on stock splits and on REITs in general. However, the current literature on REIT stock splits is limited.
Hardin et al. (2005), Li et al. (2006) and Huang et al. (2011) are the first and only studies that investigate REIT stock splits. Hardin et al. (2005) find a positive reaction around the announcement date and a muted reaction around the ex-date, which is consistent with prior research on stock splits (e.g. Nazar and Rozeff, 2001; Anshuman and Kalay, 2002; Byun and Rozeff, 2003; Boehme et al., 2003). In addition, Li et al. (2006) and Huang et al. (2011) also find significant positive abnormal returns around the announcement date. Hence, these three studies on REIT stock splits find significant abnormal returns around the announcement date, which supports the attention-grabbing hypothesis. Therefore the following hypothesis is formed:
10 Hypothesis Ι: REITs perform stock splits to attract attention from small uninformed investors.
Additionally, significant positive returns around the announcement date also support the signaling theory. However, this theory predicts positive stock performance in the long run as well. Hardin et al. (2005) do not find significant positive stock market reactions in the long run after the stock split. In other words, REITs that performed a split did not significantly outperform specific REIT indices in their research. They argue that this is the result of a REIT split being quickly absorbed by the stock market and therefore produces normal returns in the long run. This quick absorption is the consequence of the relative transparency of REIT performance and the extensive monitoring of the capital markets. Therefore they conclude that the signaling theory is not likely to hold for REITs. Furthermore, Huang et al. (2011) also investigate the long run stock performance for ½, 1, 2 and 3 years after the split ex-date and also find insignificant results. Therefore Huang et al. (2011) also conclude that REITs do not outperform their industry peers in the long run and suggest that the signaling theory is not likely to hold for these listed real estate firms. In conclusion, because REITs are relatively transparent entities and therefore require less need to signal positive prospects, combined with no evidence of REIT split long term abnormal returns, the following hypothesis is constructed.
Hypothesis ΙΙ: The signaling theory is not likely to hold for REITs, because REIT performance is relatively transparent and therefore management is not intended to signal favorable prospects to the market.
Moreover, Hardin et al. (2005) do find some evidence that REITs perform stock splits after a significant stock price increase relatively to other REITs. They suggest that REITs use stock splits to improve the trading range of their stock, which supports the trading range/liquidity theories. For this reason, Huang et al. (2011) include liquidity measures in their study to test if the liquidity and trading range theories hold for REIT stock splits. They find that most of the liquidity measures increase after the stock split announcement. However, the increased liquidity is limited to several days after the announcement of the stock split. In fact, they find that the liquidity after the ex-date has reverted back to the same level as before the stock split. Thus, these results are not in line with the liquidity and trading range theories, which predict that the liquidity improves in the long run. However, because the majority of empirical research (e.g. Lakonishok and Lev, 1987; McNichols and Dravid, 1990; So and Tse, 2000), finds improved liquidity after stock splits and Huang et al. (2011) mention their limited REIT stock split sample, the following hypothesis is formulated:
11 Hypothesis ΙΙΙ: REITs use stock splits to improve the stock price trading range and liquidity.
2.4 Long term performance after stock splits
The signaling theory predicts that trading activity and returns increase around stock split announcements and this will hold in the long run after the split. The findings of the initial positive market reaction around the announcement date are consistent among the majority of empirical research. Evidence of long term stock outperformance is mixed. Ikenberry et al. (1996) and Desai and Jain (1997) find significant positive long term abnormal returns. On the contrary, Byun and Rozeff (2003) find no significant evidence of long term stock outperformance when this is measured from the ex-date instead of the announcement date. Both studies of Hardin et al. (2005) and Huang et al. (2011) on REIT stock splits also do not find significant abnormal performance in the long run. Hence, there is no clear evidence for long term abnormal performance after stock splits. Furthermore, there remains substantial debate in the finance field on which methodology to use for long term stock performance pointed out by Fama (1998) and Mitchell and Stafford (2000). Since there is no clear methodology to test the long term stock performance of split firms, the signaling theory could also be empirically investigated by the change in operating performance. Lakonishok and Lev (1987) investigate this for industrial firms and find some empirical evidence for an improvement in operating performance. They find that firms who conducted a stock split have higher growth in earnings one-year after the split than a group of control firms. Similar research from Asquith et al. (1989) find some evidence for higher earnings growth, however, not longer than one year after the split. The signaling theory for REIT splits has never been studied using changes in operating performance. Therefore, in this study, this gap will be filled by analyzing the operating performance change after the stock split of REITs.
2.5 The different REIT industries
Studies of Bradley et al. (1998), Kallberg et al. (2003) and Li et al. (2006) document that REITs unique dividend policy allows for cleaner evaluation of dividend and stock split signaling theories, since the transparency in this industry is relatively high. However, Danielsen and Harrison (2007) find that diversified and hybrid REITs have microstructure measures, suggesting that these REIT types are difficult to value, which decreases the transparency of these REITs. Therefore, this study will divide the sample in hybrid and diversified REITs and the other REITs types separately to test if there are different results. This might be of fundamental importance, because excluding these types allows for an even cleaner evaluation
12
of dividend and stock split theories. Nevertheless, despite the findings of less transparent REIT types of Danielsen and Harrison (2007), REITs are relatively more transparent than industrial firms. Hence, the following hypothesis is formulated:
Hypothesis ΙV: The initial reaction on stock splits and operating performance changes of the hybrid and diversified REIT group will not deviate from the whole sample and the other REIT type groups.
13
3. Methodology
In this section the methodology of this research is described. First, the methodology of the initial reaction of the REIT stock split is explained. Then the methodology for the comparison of the operating performance and liquidity with control firms are described. Finally, the regression of the determinants on the announcement 5-day CAR is reported.
3.1 Methodology of the initial reaction of a REIT stock split
Following the methodology of Nayar and Rozeff (2001) and Hardin et al. (2005) the abnormal returns are calculated 2 two days before and 2 days after the relevant stock split dates: the announcement date, the record date and the ex-date. To allow comparison with previous research on stock splits and REIT stock splits, the abnormal returns are first related to the CRSP value-weighted and equal-weighted indices. However, the value-weighted and equal-weighted CRSP indices may not be representative for the real estate sector and in specific for REIT stock performance. Therefore, the CRSP/Ziman value-weighted and equal-weighted REIT indices are also used for performance comparison.
To allow for further comparison with prior research on stock splits and REIT stock splits, three methodologies are performed to measure the initial effect of the corporate event. In the results section, the market model is used to measure the average abnormal return. The formula for the market model average abnormal return is the following:
(1)
where,
AARt = the market model average abnormal return on any day t,
Rit = the daily return for security i on day t,
Rmt = the daily return for the CRSP or Ziman-REIT value- or equal-weighted index on day t,
ἆi = the estimated intercept of the market model,
Ḃi = the estimated slope of the market model, and
N = the number of REITs with stock splits.
Moreover, the adjusted-market model and the mean-adjusted model are presented in the robustness check section. These two methodologies are added to address the concerns pointed out by Fama and French (1992) that the market model is not valid.
14
The adjusted-market model average abnormal return on day t is:
(2)
The mean-adjusted model average abnormal return on day t is:
(3)
3.2 Methodology for the long term operating performance
Researches from Lakonishok and Lev (1987) and Asquith et al. (1989) investigate operating performance for stock splits. However, both studies use only earnings as a proxy for operating performance. Therefore, this study will follow the methodology of a recent, more elaborate study from Guo et al. (2011) which investigates operating performance after leveraged buyouts. The operating performance measures for profitability are FFO/sales and net income/sales.1 For
the return on assets FFO/total assets and net income/total assets are used.2
The operating performance change is measured as follows. First, the operating performance change between the last quarter before the stock split and the same quarter from the previous year (t-2 to t-1) is calculated.3 Then the percentage change between these quarters is compared
with the changes between the same quarters for the next three years (t+1, t+2 and t+3). To test if the operating performance improves after the stock split, the percentage changes of the operating performance measures of the sample group are compared with the percentage changes of the control group. This is tested using the Wilkinson matched-pair signed rank test for median changes.
The study from Guo et al. (2011) matches each observation with 4 control firms based on the same last four-digit SIC code. In this study the control sample is even more thoroughly selected
1 Guo et al. (2011) use EBITDA for the operating measures. However, FFO is a better proxy for the operating
performance of REITs and the real estate sector in general.
2 Funds from operations (FFO) is calculated as net income, plus depreciation and amortization and
extraordinary items, minus gains and losses from sales of properties. Depreciation and amortization, extraordinary items, gains and losses from sales of properties are treaded as zero if they are missing in COMPUSTAT (following Huang et al., 2011).
3 For example: if observation i announced a stock split in June 2000 (quarter 2), then the operating
15
as follows: Control firms are selected from the CRSP/Ziman REIT database, since REITs are different corporate entities than standard industrial firms. For each company and observation, the control firms are first selected if they are in the same real estate subtype industry. Then the four firms with the most similar size are assigned to each observation. Total assets is used as a proxy to measure the size of the firms. Furthermore, some REIT subindustries consist of limited amount of firms, because they are relatively new. For observations in these subtype industries, too few control firms could be selected. Therefore, for these subtype industries, control firms are assigned from the larger property type industries and then matched with size. Finally, a thorough check is performed whether the control firms did not have a stock split at the same time, which would bias the results. This was sometimes the case and these control firms were replaced by firms with less similar size.
3.3 Methodology for the liquidity measures
To measure liquidity changes after the stock split, the methodology of Huang et al. (2011) is followed. In this research, three different measures are used to measure the liquidity changes around and after a REIT split. These liquidity ratios measure price impact, trade and spread. The formulas and the explanations of these liquidity measures are obtained from the review of Ametefe et al. (2016). The liquidity measures are the following: the turnover ratio, Amihud’s illiquidity measure and zero-return trading days. The first liquidity measure is the turnover ratio, which is a volume-based liquidity measure. The turnover ratio is the inverse of transaction volume and scales transaction volume with the market or firm size. Amihud et al. (1997) find a negative correlation between the turnover ratio and illiquidity costs. In addition, when the turnover ratio is high, market makers tend to charge lower transaction costs for a security (Ametefe et al., 2016). So, the higher the turnover ratio, the higher the liquidity of a security. The turnover ratio is measured by the transaction volume, divided by the shares outstanding times the share price as in the following formula:
(4)
The second liquidity measure is the Amihud illiquidity measure. This is a price impact measure, which means that it measures the impact of a dollar in trading volume on the stock price. The higher the trading volume, the lower the price change, the higher the liquidity of the market. Therefore, the lower the Amihud’s illiquidity ratio, the higher the liquidity of a security or market. The Amihud ratio is calculated as follows:
16
(5)
Where, n is the number of days, which are not zero return days. TRi, is the absolute return of
the stock on each day i and Voli is the trading volume on each day i.
The third measure is the zero return trading days variable. This is the number of zero return trading days divided by the total number of trading days in a certain period. This variable is also a price impact measure and is a proxy for transaction costs. The line of thought of this variable is that investors reduce trading if transaction costs exceed the benefit of signaling private information. Hence, the higher the transaction costs of a security, the more zero return trading days are expected. Therefore, the lower the zero return trading days variable is, the higher the liquidity of a security.
(6)
To measure if the liquidity increases around and after the split event, a control sample for each observation is created. The selection of the control group is the same as described in the methodology section for the operating performance.
Furthermore, five time intervals around the announcement date and ex-date are used to test the liquidity change following Huang et al. (2011). These time intervals are the pre-split period (a-122 to a-3), the announcement period (a-2 to a+2), the announcement-to ex-date (a+3 to e-3), the ex-date period (e-3 to e+3) and the long term post ex-date (e+3 to e+122). These time intervals combined cover approximately a one year period around the split event. To test if the liquidity improves during these time intervals, the coefficients of the three liquidity measures of the sample group are compared with the coefficients of the control group. This is tested using the paired sample t-test for means which is described in the following formula:
𝑡 = 𝑑
𝑆𝐸(𝑑) (7)
And, 𝑆𝐸 =𝑆𝑑
√𝑛 (8)
Where d is the mean difference of all paired observations and Sd is the standard deviation of the
17 3.4 The determinants of the announcement period abnormal returns
In this analysis the cumulative abnormal returns around the five day announcement period are regressed on several determinants. The main purpose of this regression analysis is to see whether the positive abnormal returns around the announcement date are explained by the changes in liquidity. Furthermore, the cumulative abnormal returns are regressed on several control variables. The regression model is the following:
CARi = β0 + β1*∆LIQi + β2*FACTOR + β3*RUNUP + β4*INTF + β5*HYBRID
β6*PROPERTY (9) Where, CARi is the five-day cumulative abnormal return around the announcement date for
each REIT split event. ∆LIQi is the change in liquidity between the pre-split period (122 to
a-3) and the announcement period (a-2 to a+2) and the change in liquidity between the pre-split (a-122 to a-3) and the post-split long term period (e+3 to e+122). To measure the liquidity, the turnover ratio is used as the proxy, because it measures the propensity to trade the most accurate of other liquidity measures (Huang et al., 2011). Furthermore, the Amihud illiquidity measure is used as a robustness check for the liquidity change. FACTOR is the REIT stock split factor, which is also considered to be an informative signal (Brennan and Hughes, 1991). RUNUP is the return calculated from 122 days before the stock split announcement to 5 days before the stock split announcement. This variable is included as a proxy for the deviation of the stock price from its target range following Grinblatt et al. (1984) and Huang et al. (2011). INTF is dummy variable which is 0 if the split factor is smaller than 1, which indicates 4 for 3 stock splits for example. INTF equals 1 if the split factor is 1 or larger, for example 2 for 1 splits. HYBRID is a dummy variable that equals 1 if it is a hybrid REIT. PROPERTY is set of dummy variables for each property type. Some property types are more transparent in their activities than others (retail versus diversified). Therefore dummies for each property type are included to test whether this has an impact on the cumulative abnormal returns around the announcement date.
18
4. Data and descriptive statistics
In this section the data collection of the REIT stock split sample is described. Furthermore, descriptive statistics of the splits are reported.
4.1. Data collection
4.1.1 Stock split data from CRSP
The data for the Real Estate Investment Trust (REIT) stock splits and returns are retrieved from the Center for Research in Security Prices (CRSP). The SIC code 6798 is used to gather the data which includes only but all listed REITs in the United States. The stock splits are collected using the distribution code 5523 in CRSP, which defines a stock split for a REIT. The time period is from January 1980 to December 2017. This time period is chosen, because in the 1980s relative transparency increased significantly for U.S. REITs (Li et al., 2005). The stock splits for REITs in this time frame resulted in 258 split events. However, following Hardin et al. (2005) and Huang et al. (2011) for REIT stock splits to enter the sample, the following criteria are defined: (1) the minimum split factor is 0.25, which is equivalent to a 5 to 4 stock split or greater. (2) the estimation period covers 255 trading days ending on day -251 relative to the declaration date and a minimum of 150 trading days with non-missing daily returns. (3) At least four trading days separate the announcement date and the record date and at least four trading days separate the record date and the payable date, to prevent a contagion effect. (4) Finally, it is required for a REIT to re-enter the sample after one year to prevent overlapping data. Consistent with Hardin et al. (2005), Li et al. (2006) and Huang et al. (2011) REIT stock split are rare corporate events. After applying the criteria the final sample contains 89 stock splits, which is on average something more than two REIT splits a year. There are 37 splits in the 1980s, 26 splits in the 1990s, 22 splits in the 2000s and 4 splits after 2010.
4.1.2 Operating performance data from Compustat
Operating and financial ratios are collected from the Compustat database that collects accounting values and fundamentals. REITs which do not have accounting data 2 years prior and 3 years after the split are excluded. For REITs that perform multiple stock splits within 4 years of each other, one observation is excluded. This is because including the second stock split within 4 years creates an overlap with the pre-split operating performance change (t-2 to t-1) which would bias the results. Excluding these observations yields a total of 68 splits.
19
4.1.3 Event dates associated with stock splits
Corporate events like dividend distributions and stock splits have a certain chronological order of event dates. The announcement date, the record date, the payable date and the ex-date. The announcement date is the first event date in the process. This is the date when the stock split is announced (CRSP code is dclrdt). The record date is the second event date, this is the date when the ownership rights to the newly split shares are allocated (CRSP code is rcrddt). Following a few days from the record date is the payable date, this is the date the firm distributes the new stock (CRSP code is paydt). Finally, one trading day after the payable date is the ex-date, this is the date the newly split stock is traded for the first time (CRSP code is exdt).
4.1.4 Liquidity measures
To measure the liquidity changes around and after the REIT stock splits, three liquidity measures of Huang et al. (2011) are used. These liquidity measures are based on price impact, trade and spread. Daily stock data is used to conduct the liquidity measures. Intraday and quote data are not used, because these databases became available after 1993, which do not cover the whole sample period. Since these daily data liquidity measures are calculated from stock prices, the liquidity data is retrieved from CRSP.
4.1.5 REIT types and subtypes
Since the recent establishment of the CRSP/Ziman REIT database, it is currently possible to divide the REIT sample in different subsamples. Not every observation was classified under each type. Therefore, the missing observations were collected manually from the REIT specific sites. The main subsample is the differentiation in REIT types. In the sample there are 74 equity REIT stock splits, equity REITs solely invests in developing and exploiting income producing properties. Furthermore, there are 10 mortgage REIT stock splits, which only invests in mortgages and mortgage-backed securities. Finally, there are 5 hybrid REIT stock splits in the sample. Hybrid REITs are a combination of the investment strategies of equity and mortgage REITs. In addition, REITs can be divided into property type and sub-property type categories. These property types and the number of observations for each sector are presented in Table 3.
4.2 Descriptive statistics
In Table 1 the statistics of the number of trading days between the declaration (announcement date), the record date and the payable date is presented. The average between the declaration
20
date and the record date is 21.20 trading days, while the record date to payable date is on average shorter with 16.45 trading days between these periods. In this sample, both periods are longer than in the sample of Hardin et al. (2005) which are 14.97 and 12.47 trading days for the declaration to record date and record to payable date respectively. This difference is explained by the relatively longer period for REIT stock splits after 2000.4
Table 1:
Number of trading days between REIT stock split event dates Statistic Declaration to record date Record to payable date Mean Median St. dev. Observations 21.20 14 18.09 89 16.45 15 7.84 89
Note: In this table the descriptive statistics of the periods between the declaration-, record- and payable dates of the REIT stock split sample are presented . The numbers in the table present the mean, median and standard deviations of the number of trading days between each period. To be included in the sample, the split factor is at least 0.25 (which equals a 5 for 4 split) for the period 1980 to 2017.
Table 2:
Split factors and REIT stock split periods Panel A: Split ratios
Two-for-one Three-for-one Three-for-two Five-for-one Five-for-four Other Number of splits 38 3 32 7 5 4 Cumulative sum 42.70% 46.07% 82.02% 89.89% 94.38% 100%
4 The period Hardin et al. (2005) investigate is 1964-2000. I divided the REIT stock split before and post 2000
and found that on average the declaration to record date period and record to payable date periods from splits after 2000 were relatively longer than before 2000. This explains the longer periods of the sample in this research than from the research of Hardin et al.
21 Panel B: REIT split periods
1980s 1990s 2000s 2010s Total Number of splits 33 26 24 6 89 Cumulative sum 37.08% 66.29% 93.26% 100%
Note: In this table the REIT split factors and the stock split periods are provided. The sample consists of 89 REIT stock splits between 1980 and 2017. For a stock split to be included, the minimum split factor is 0.25 which is equivalent to a 5-for-4 stock split. The most common split factors are two-for-one and three-for-two splits. The most REIT stock splits occurred in the 1980s, followed by the 1990s and the 2000s.
In Table 2, the split factors and the REIT stock splits per subperiod are described. The most common split factors are two-for-one and three-for-two splits with 38 and 32 splits respectively. This is consistent with the findings of Huang et al. (2011) who also find approximately 90% of the sample to be a two-for-one or a three-for-two split. The most REIT stock split took place in the 1980s with 33 splits, followed by the 1990s with 26 splits and 2000s with 24 splits.
In Table 3, the REIT stock splits for each property type sector are presented. In total there are 11 property types and 19 property subtypes.5 REITs in property type sectors as unknown,
mortgage-backed securities or self-storage have never performed a stock split. The most common property types to perform a stock split is retail with 30 splits and diversified REITs with 20 splits. If these ratios are consistent with prior research cannot be checked, since this is the first study that collects the REIT stock splits by property type.
22 Table 3:
Real estate investment trusts property types
Property type Description Number of observations
0 1 2 3 4 5 6 7 8 9 10 Unknown Unclassified Diversified Healthcare Industrial/Office Lodging/Resorts Mortgage Mortgage-backed securities Residential Retail Self Storage 0 9 20 6 7 2 8 0 9 30 0
Note: In this table the different property types of Real Estate Investment Trusts are described. In total there are 11 different property type sectors where REITs could be active in. REITs in property type sectors as unknown, mortgage-backed securities or self-storage have never performed a stock split. The most common property types to perform a stock split is retail with 30 splits and diversified REITs with 20 splits.
23
5. Results
In this section the results of this research will be discussed. First, the findings of the initial reaction around the stock split are presented. Followed by the results of the operating performance- and liquidity change around the split. Finally, the regression of the determinants of the 5-day CAR around the announcement date is discussed.
Table 4:
Abnormal returns around the event dates of REIT stock splits using aggregate market and REIT specific indices.
Index CRSP VW index CRSP EW index REIT VW index REIT EW index
Announcement date Abnormal return (%) (t-statistic) Record date Abnormal return (%) (t-statistic) Ex-date Abnormal return (%) (t-statistic) 2.98*** (5.67) 0.24 (0.37) 0.87 (1.39) 3.03*** (5.81) 0.31 (0.46) 0.94 (1.50) 3.05*** (6.09) 0.49 (0.72) 1.06* (1.73) 2.94*** (5.82) 0.46 (0.70) 0.90 (1.51)
Note: In this table the 5-day cumulative abnormal returns around the announcement date, the record date and the ex-date of REIT stock splits are reported. The sample consists of 89 REIT stock splits between 1980 and 2017. For a stock split to be included, the minimum split factor is 0.25 which is equivalent to a 5-for-4 stock split. To calculate the 5-day abnormal returns the market model is used.6 The CRSP
value-weighted, CRSP equal-weighted, REIT value-weighted and REIT equal-weighted are used as proxy for the market. These four proxies for the market are used to check for robustness. Inside the parentheses are t-statistics of each coefficient. The ***, ** and * denote the 1%, 5% and 10% significance levels respectively.
In Table 4, the 5-day cumulative abnormal returns around the announcement date, the record date and the ex-date of REIT stock splits are reported. The proxies for the market are the CRSP value-weighted and equal-weighted index in order to compare the results with prior research of
24
Nayar and Rozeff (2001) and Hardin et al. (2005). Moreover, the REIT value- and equal-weighted indices are used as a proxy for the market, because this better reflects the market performance for REITs, since REITs are different corporate entities than standard industrial firms. In the 5-day event window around the announcement date, the average cumulative abnormal return is around 3% for all market indices. These cumulative abnormal returns are significant on a 1% level for all market proxies. This is in line with the studies of Hardin et al. (2005) and Huang et al. (2011) who also find positive statistically significant 5-day CARs of approximately 3%. This indicates that there is a positive initial market reaction to REIT stock split announcements. This favorable market reaction can be attributed to the buying pressure of investors, which supports the predictions of the signaling and attention-grabbing theories. Furthermore, investors require a lower return of the stock, because long term liquidity should improve, pointed out in Huang et al. (2011).
The 5-day window around the record date also provides a positive market reaction of 0.3 to 0.4%. However, this is not significant for each of the market indices. This is in line with the research of Hardin et al. (2005) which also find positive, though insignificant reactions around the record date. On the other hand, these results are not consistent with the research of Nazar and Rozeff (2001), who find negative reaction around the record date. The explanation they provide, is that it requires holdings costs for short-term investors and traders to hold split stocks. A possible explanation that REIT splits do not have negative reaction around the record date, is that REIT investors are predominantly long term investors who do not the bear the short-term holding costs.
The 5-day CAR around the ex-date also provides favorable returns around 1%. This is not significant, except for the REIT value-weighted index which is significant on a 10% level. Hardin et al. (2005) find similar positive abnormal returns, which are not or weakly significant. Huang et al. (2011) however, find a negative reaction around the ex-date. Their explanation is that brokers do not promote REIT shares to investors, especially not to small investors.
25 Table 5:
Operating performance changes after REIT stock splits
-2 to -1 -1 to 1 -1 to 2 -1 to 3 A.1 Profitability
FFO/sales Net income/sales A.2 Return on assets FFO/total assets Net income/total assets
3.33% (-0.40) -2.32% (0.12) 0.25% (-1.42) -4.72% (-1.04) -3.37% (0.07) -4.68% (0.52) -3.36% (-0.85) -4.68% (-0.20) -0.92% (-1.13) -17.34% (-0.72) -5.40% (-1.42) -11.66% (-0.27) -15.44% (-0.97) -20.82% (-0.12) -17.02% (-1.13) -18.64% (-0.33) Note: In this table the operating performance median changes before and after 68 REIT stock splits in the period 1980-2017 are reported. All percentages are the median changes of the operating performance ratios between these periods. (-2 to -1) is the operating performance change of the last quarter before the stock split announcement and the quarter of the previous year. (-1 to 1), -(1 to 2) and (-1 to 3) are the operating performance changes of the last quarter before the stock split announcement and the same quarter of the first, second and the third year after the split announcement respectively. Z-statistics of the Wilkinson matched-pair signed rank test for medians are reported in parentheses next to the percentage changes. The ***, ** and * denote the 1%, 5% and 10% significance levels respectively.
Table 5 reports the median operating performance changes (in percentages) of the REIT stock split sample for the different time intervals. Panel A.1 reports the changes in profitability (measured as FFO/sales and net income/sales). The median profitability change the first year after the split (-1 to 1) is 3.37% and 4.68 % decrease for the FFO/sales and net income/sales respectively. Additionally, except for the period -2 to -1, the median profitability performance changes for the REIT split sample are all negative. Hence, for the majority of the REIT split sample, profitability decreased after the split. However, these negative changes are not significantly different than the split control group profitability changes. On the contrary, despite the negative profitability changes, the median profitability decrease of the split sample is less than the control group in the period -1 to 1, although insignificant.7
Panel A.2 reports the changes of return on assets (measures as FFO/total assets and net income/total assets). Similarly to profitability, median return on assets decreased for each period after the stock splits as well. Median return on assets decreases are almost similar to the median profitability decreases (e.g. for the period -1 to 1, -3.37% and -4.68% decrease against
26
-3.36% and -4.68%). Furthermore, this decrease was larger than the control group for each post-split period, however all not significant. Hence summarized, median operating performance decreases for each post-split period, though not significant. These results are not consistent with the studies of Lakonishok and Lev (1987) and Asquith et al. (1989) who find significant positive operating changes relative to their peers for several post-split periods. Nevertheless, these results are in line with the long run stock performance findings of Hardin et al. (2005) and Huang et al. (2011) who find no evidence of long run outperformance. Despite the evidence of significant CARs around the announcement date which support the signaling theory, these findings indicate that REITs do not signal favorable prospects to investors. This supports the hypothesis that the signaling theory is not likely to hold for REITs, because of the relative transparency in this industry.
Table 6:
Liquidity changes around and after REIT stock splits
Event window Split REITs Control REITs Difference
Panel A: Turnover ratio (×10-2)
Pre-announcement period (a-122 to a-3) Announcement period (a-2 to a+2) Announcement-to ex period (a+3 to e-3) Ex-date period (e-3 to e+3)
Long-term post-ex period (e+3 to e+122)
Panel B: Amihud ratio (×10-6)
Pre-announcement period (a-122 to a+2) Announcement period (a-2 to a+2) Announcement-to ex period (a+3 to e-3) Ex-date period (e-3 to e+3)
Long-term post-ex period (e+3 to e+122) 8.99 (4.02) 11.78 (2.99) 9.95 (4.70) 5.65 (3.38) 13.94 (6.76) 2.51 (0.09) 4.86 (0.19) 3.33 (0.20) 2.73 (0.12) 4.03 (0.10) 11.46 (5.29) 17.14 (4.92) 13.89 (4.77) 12.83 (6.65) 13.33 (5.40) 4.36 (0.39) 3.65 (0.20) 3.33 (0.20) 2.95 (0.27) 4.84 (0.41) -0.725 (-0.567) -1.98* (-1.51) -0.923 (-0.542) -3.124*** (-1.34) 0.612 (0.285) -1.85 (-0.003) 1.21 (0.000) 0.000 (0.000) -0.220 (-0.005) -0.81 (-0.005)
27 Note: This table compares the liquidity measures of the sample of REIT stock split firms with their matched REIT-control sample for several event windows: the pre-announcement period (a-122 to a+2), the announcement period (a-2 to a+2), the announcement-to ex period (a+3 to e-3), the ex-date period (e-3 to e+3) and the long-term post-ex period (e+3 to e+122). The three (il)liquidity measures are the turnover ratio, the Amihud illiquidity measure and the zero-return trading days ratio. For each split, four control firms are matched on the basis of property subtype and then the REITs are picked with the most comparable size.8 Then the turnover ratios, Amihud illiquidity measures and the zero return trading days
of the control group are averaged. Split REITs and control REITs report the coefficients of the means of the three liquidity measures, medians are reported in parentheses. Difference is the t-statistic of a paired sample t-test for means which is described in formulas 8 and 9. Difference in parentheses are the Wilkinson signed rank test for the medians. The ***, ** and * denote the 1%, 5% and 10% significance levels respectively.
In Table 6, the liquidity measures of the split sample, the control group and the difference of the five event windows around the stock split are reported. Panel A shows that the mean turnover ratio increases from 8.99 to 11.78 from the pre-split period to the announcement date window. This is consistent with the attention-grabbing and signaling theories, because increased interest in the stock increases trading activity and thereby increases liquidity. Then between the announcement- and ex-date period the turnover ratio decreases to 9.95 and further decreases to 5.65 around the ex-date window. An explanation for this result is that if firms perform a stock split to improve the trading-range and liquidity of the stock, liquidity should only increase after the ex-date. Therefore, trading unsplit shares is inconvenient for investors
8 Total assets is used as a proxy for size, since the market capitalization is not always a reliable measure in
CRSP.
Table 6 (continued)
Event window Split firms Control firms Difference
Panel C: Zeros Pre-announcement period (a-122 to a+2) Announcement period (a-2 to a+2) Announcement-to ex period (a+3 to e-3) Ex-date period (e-3 to e+3)
Long-term post-ex period (e+3 to e+122) 0.171 (0.134) 0.114 (0.094) 0.231 (0.168) 0.190 (0.156) 0.123 (0.106) 0.152 (0.140) 0.161 (0.000) 0.103 (0.095) 0.127 (0.000) 0.152 (0.124) 0.031 (0.006) -0.068 (0.000) 0.104* (0.009) 0.074) (0.005) -0,021 (0.000)
28
which delay purchases and reduce liquidity. Finally, in the six months period after the ex-date, the turnover ratio increases to 13.94. This post-split turnover ratio (13.94) that is larger than the pre-split ratio (8.99) indicates that liquidity has improved after the split, which is consistent with the trading-range/liquidity theories. However, controlling for industry peers results in negative insignificant coefficients, except for the ex-date period which is significantly lower than the control group on a 1% level. This implies that the matched-control group on average earn higher turnover ratios. This is consistent with Huang et al. (2011) who find negative t-statistics for the difference as well, except in this study the mean turnover ratios from the split sample are higher than the control group in de post-split period. Hence, on average the split firms have lower liquidity than their matched-control REITs.
In panel B, the mean and median Amihud illiquidity measures are reported. The Amihud illiquidity measure increases from 2.51 in the pre-split period to 4.86 in the announcement date window. This seems consistent with the findings in panel A of the turnover ratio, although the Amihud is an illiquidity measure which implies that liquidity has decreased between these periods. After the announcement date, the Amihud decreases back to the pre-split levels till the ex-date. In the long term period after the ex-date the Amihud increases again to 4.03. Hence, according to the mean Amihud illiquidity measure, liquidity decreased after the stock split. The medians of the Amihud follow the same pattern, although the pre-split and post-split are similar eventually. Furthermore, insignificant differences between the split REITs and matched-control group are found.
In panel C, the means and medians of the zero-return trading day ratio are presented. The means (medians) of the zeros-ratio decreases from 0.171 (0.134) to 0.114 (0.094) from the pre-split period to the announcement date period. Then the ratio increases again between the announcement date and the ex-date. This is in line with the findings in Panel A of the turnover ratio, because the zeros ratio is an illiquidity measure as well. In the long term period post ex-date, the mean (median) ratio has decreased to 0.123 (0.106), which indicates that the liquidity has increased. In addition, no strongly significant differences between the split sample and the control-group are found. In short, liquidity increases from the pre-split to the long term post-split levels for the turnover ratio and the zeros ratio. However, these increases are not significant. The liquidity for split REITs is on average even lower than their matched-control groups. These findings comply with the results of Huang et al. (2011) for REITs, who find no significant liquidity increases as well.
29 Table 7:
The regression coefficients which explain the 5-day CAR around the announcement date Period for ∆LIQi
(a-122 to a-3) to (a-2 to a+2) Turnover Amihud
Period for ∆LIQi
(a-122 to a-3) to (e+3 to e+122) Turnover Amihud Intercept ∆LIQi FACTOR RUNUP INTF Hybrid Diversified Healthcare Industrial/Office Lodging/Resorts Mortgage Residential 0.022* (1.98) 0.020*** (4.18) 0.006 (0.68) -0.009 (-0.23) 0.008 0.53) 0.017 (0.78) 0.005 (0.42) -0.012 (-0.89) -0.024* (-1.84) -0.017 (-0.74) 0.026 (1.20) -0.022 (-1.36) 0.023* (2.35) 0.014 (1.04) 0.010 (1.10) 0.016 (0.36) -0.005 (-0.32) 0.014 (0.61) 0.009 (0.51) -0.009 (0.59) -0.009 (-0.51) -0.017 (-0.59) 0.021 (0.81) -0.012 (-0.69) 0.025** (2.14) 0.005* (1.88) 0.009 (1.04) 0.012 (0.29) -0.008 (-0.47) 0.019 (0.78) -0.001 (-0.04) -0.014 (-0.83) -0.016 (-0.85) -0.025 (-1.14) 0.022 (0.86) -0.021 (-1.17) 0.026** (2.14) -0.002 (-0.11) 0.009 (0.95) 0.010 (0.22) -0.003 (-0.17) 0.019 (0.80) 0.003 (0.15) -0.013 (-0.92) 0.017 (-0.92) -(-0.021 (-0.79) 0.019 (0.76) -0.022 (-1.26) R2 N 0.312 78 0.110 78 0.097 78 0.083 78
Note: Table 7 presents the regression coefficients of the 5-day cumulative abnormal return on the change in liquidity and other split control variables. The regression model is: CARi = β0 + β1*∆LIQi +
β2*FACTOR + β3*RUNUP + β4*INTF + β5*PROPERTY. Where, CARi is the five-day cumulative
abnormal return around the announcement date for each REIT split event. In the first column, ∆LIQi is
the change in liquidity between the pre-split period (a-122 to a-3) and the announcement period (a-2 to a+2). In the second column ∆LIQi is the change in liquidity between the pre-split (a-122 to a-3) and the
post-split long term period (e+3 to e+122). To measure the liquidity change, the turnover ratio and Amihud measure are used. FACTOR is the REIT stock split factor. RUNUP is the return calculated from 122 days before the stock split announcement to 5 days before the stock split announcement. INTF is dummy variable which is 0 if the split factor is smaller than 1, which indicates 4 for 3 stock splits for example. INTF equals 1 if the split factor is 1 or larger, for example 2 for 1 splits. HYBRID is a dummy variable that equals 1 if it is a hybrid REIT. PROPERTY is set of dummy variables for each property type omitting the largest category which is Retail. Inside the parentheses are t-statistics of each coefficient. The ***, ** and * denote the 1%, 5% and 10% significance levels respectively.
30
In Table 7, the regression results of the determinants of the 5-day cumulative abnormal return around the announcement date are presented. In the first column, the coefficient of the change in liquidity/turnover is positive and significant on a 1% level. This result suggests that improvements in liquidity around the announcement date lead to a positive market reaction, which is consistent with the attention-grabbing and signaling theories that predict an increase in trading activity and returns around the announcement date. However, the increase in trading activity around the announcement date is more likely to be explained by the attention-grabbing theory than the signaling theory, since REITs are relatively more transparent than industrial firms. Huang et al. (2011) test if the signaling theory will hold, adding future operating performance. However, they find no significant relationship between the positive announcement returns and future operating performance, which indicates that the signaling theory for REIT stock split will not hold.9
The coefficient of the split factor is positive, indicating that larger stock splits (2-for-1 or 3-for-1) obtain higher CARs around the announcement date. However, this is not significant. The majority of the property types coefficients are negative, indicating that REIT stock splits in the Retail sector earn higher returns around the announcement date. However, neither of these coefficients are significant, except for the Industrial/Office sector on a 10% level. The mortgage, hybrid, and diversified REIT sectors have positive coefficients, although not significant. These positive coefficients of hybrid and diversified REITs could indicate that the abnormal returns are larger around the announcement date, because they are less transparent than other REIT sectors due to microstructure measures pointed out by Danielsen and Harrison (2007). These are rough indications and more thorough analysis on these sectors are performed in the robustness checks section.
Measuring the liquidity change using the Amihud illiquidity ratio, results in no significant findings whatsoever. This is might be the result of a bad proxy for the change in liquidity and/or the fact that the turnover ratio captures the propensity to trade better than other liquidity measures pointed out by Huang et al. (2011).
The second column of Table 7 tests if the positive market reaction around the announcement date could be explained by the change in long term liquidity after the stock split, which is predicted by the trading range/liquidity hypotheses. The coefficient for the liquidity change is
9 The findings of Huang et al. (2011) indicate that the signaling theory will not hold for REITs. However, because
31
positive and significant on a 10% level, which indicates that the long term liquidity change explains the positive market reaction. Although, there is no statistically significant evidence for these theories. Hence, there is no significant evidence that REIT split their shares to improve long term liquidity and support both the liquidity theories.
32
6. Robustness checks
In this section the robustness of the main results are tested. First, the validity of the initial reaction on stock splits is tested using two different models. Then the REIT stock split sample is divided into hybrid/diversified and other REITs to test if the signaling theory does not hold for all REITs.
Table 8:
Abnormal returns around the event dates of REIT stock splits using the market-adjusted and the mean-adjusted model
Note: In this table the 5-day cumulative abnormal returns around the announcement date, the record date and the ex-date of REIT stock splits are reported. The sample consists of 89 REIT stock splits between 1980 and 2017. For a stock split to be included, the minimum split factor is 0.25 which is equivalent to a 5-for-4 stock split. To calculate the 5-day abnormal returns the adjusted market model and the mean-adjusted model are used. The CRSP value-weighted, CRSP equal-weighted, REIT value-weighted and REIT equal-weighted are used as proxy for the market. These four proxies for the market are used to check for robustness. Inside the parentheses are t-statistics of each coefficient. The ***, ** and * denote the 1%, 5% and 10% significance levels respectively.
To check for the robustness of the results of the initial reaction on stock splits, the market-adjusted and the mean-market-adjusted models, described in formulas 2 and 3 in the methodology section, are used. This is to address the concerns of Fama and French (1992) who question the validity of the market-model, since they question the relationship between systematic risk and returns. Therefore, the initial reaction is measured using three different short term event methodologies.
Index CRSP VW index CRSP EW index REIT VW index REIT EW index Mean-adjusted model
Announcement date Abnormal return (%) (t-statistic) Record date Abnormal return (%) (t-statistic) Ex-date Abnormal return (%) (t-statistic) 3.26*** (6.03) 1.04 (1.58) 1.41* (1.89) 3.16*** (6.16) 0.88 (1.39) 1.09* (1.76) 3.34*** (7.01) 1.09* (1.67) 1.49** (2.39) 3.23*** (6.68) 1.49 (0.70) 1.33** (2.12) 1.59** (2.44) 0.51 (1.57) 0.96** (2.53)
33
As can be noticed in Table 8, the 5-day cumulative abnormal returns around the
announcement date of the market-adjusted model are around three percent for each market proxy. This is significant on a 1% level and is consistent with the findings of the market model in Table 4, which reports almost identical results. Furthermore, the mean-adjusted model finds a 5-day cumulative abnormal return of 1.59% around the announcement date, this is significant on a 5% level. This 5-day CAR is almost half of the outcome of the other two models. An explanation for this lower CAR is that the mean-adjusted model corrects for a return period of the same REIT and not another market, which is on average closer than market proxies. All in all, the results of both models are similar to the market model findings which provides support for the attention-grabbing theory and the signaling theory.
Additionally, abnormal returns around the record date are higher for both models. However, these returns are insignificant which is in line with the market model findings. Moreover, the 5-day CAR around the ex-date provides abnormal returns of above 1%. For two of the four market proxies in the market-adjusted model and for the mean-adjusted model, these results are 5% significant. These are stronger results than found for the market model, although the findings are less significant than found for industrial firms.10
In addition, the REIT stock split sample is divided into two groups, the hybrid/diversified REITs group and the other REITs group. This separation is performed to address the concerns of Danielsen and Harrison (2007) that hybrid and diversified REITs contain microstructure measures and that therefore the transparency of the performance of these REITs decreases. Hence, the signaling theory might hold for these REIT types, because of relatively less transparency and the need for management to signal favorable future prospects. To test if the signaling theory holds for these REIT types, two regression are performed. These are the regressions of the initial reaction on the stock split and operating performance changes, since the signaling theory predicts significant positive reactions around the announcement date and increases in operating performance after the stock split.
In Table A.2 in the appendix, the initial reaction on stock splits are represented for both groups. In panel A, the 5-day CARs around each event date of the hybrid/diversified REIT group is reported. Using the market-model, the 5-day CAR around the announcement date is more than five percent for each market proxy. These are all 1% significant. On the other hand,
10 Nazar and Rozeff (2001) find 5-day CARs of around 3% and statistically significant on a 1% level for industrial
