Can Dividend Reinvestment Plans Affect Firm Payout Choices? Evidence from Real Estate Investment Trusts

Transcription

1 Can Dividend Reinvestment Plans Affect Firm Payout Choices? Evidence from Real Estate Investment Trusts Shaun Bond, Yu-Jou Pai, Peng Wang, Suyan Zheng Carl H. Lindner College of Business, University of Cincinnati, Cincinnati, OH

2 Can Dividend Reinvestment Plans Affect Firm Payout Choices? Evidence from Real Estate Investment Trusts Abstract This study investigates whether firms having Dividend Reinvestment Plans (DRIPs) influence their payout choices. Agency Theory and/or the Signaling Model indicate that dividend-paying firms with DRIPs use a conservative payout policy in order to incentivize shareholders to make long-term investments. We test this prediction using a sample of Real Estate Investment Trusts (REITs) that are required to pay dividends. We provide supportive evidence showing that in comparison to REITs without DRIPs, REITs with DRIPs are less likely to: (1) pay regular dividends with extra dividends and share repurchases and (2) omit all payouts, including dividends and repurchases. We also find that REITs with DRIPs have weaker stock price reactions to announcements of dividend changes due in part to managerial discretion in maintaining a conservative payout policy. 2

3 1. Introduction This study investigates whether firms having Dividend Reinvestment Plans (DRIPs) affect their payout choices of dividend changes or share repurchases. DRIP provides an opportunity for investors to reinvest cash dividends paid on common shares to purchase additional common shares and, in many cases, make optional cash payments to purchase shares directly from the sponsoring companies. DRIP may allow companies to raise new equity capital with much lower flotation costs and to use the proceeds received from sales of the shares for general corporate purposes. According to a file documented by the U.S. Securities and Exchange Commission (SEC), more than 1,000 U.S. firms had some forms of DRIPs by the year of However, only a handful of studies on DRIP have been documented so far for lack of data 2. Furthermore, typical industrial firms have a declining propensity to pay dividends in the most recent three decades (e.g., Fama and French, 2001; Grullon and Michaely, 2002; Skinner, 2008), a fact that discourages researchers to delve into the relation between DRIP and payout choice. However, the case of Real Estate Investment Trusts (REITs) provides a natural testing ground for this relation, because: (1) REITs are dividend-paying firms. Following the Internal Revenue Code, REITs are required to pay out at least 90 percent of their taxable income to shareholders as dividends, in order to qualify for tax-exempt status 3 ; ( 2) DRIPs can be an important source of capital for REITs. REITs distribute a large part of operation generated cash flow in the form of dividends. Implementing DRIPs can facilitate REITs to retain some of these cash flow. This is crucial for REITs given their frequent demand for external financing; (3) Data collection process for REITs is relatively feasible and precise. We collect an equity REIT sample 1 See 2 Companies are not required to report sources and amounts of shares used for DRIPs. 3 See the Internal Revenue Code Sections

4 starting from 1999, due to regulation and policy changes in the REIT industry 4. In addition, most REITs started to adopt DRIPs after the SEC simplified the procedures of implementing the plan in December of 1994; (4) DRIPs help REITs obtain wider ownership of shares. The plan requires participants to possess stock certificates issued in their own names, instead of in the name of their brokers. This is meaningful for REITs since these firms are required to be beneficially owned by 100 or more persons and must not be closely held. In contrast to prior studies on DRIP mainly using survey data, this study employs handcollected data from the SNL financial, the SEC Electronic Data Gathering, Analysis, and Retrieval (EDGAR) database, Google Search, Factiva, and other data sources. Based on the SNL Real Estate database, 262 out of 721 REITs have DRIPs. Many shareholders holding stocks of a particular REIT with DRIP (hereafter, DRIP REIT) elect to participate in the plan. DRIP is well known to many investors in practice. However, to the best of our knowledge, none of the existing literature has investigated any DRIP topics in REITs, and no extant payout studies have examined whether DRIP influences firm payout choices. This study aims to fill these gaps. Even though REIT dividend payouts are greatly constrained by their internally generated cash flow, REIT managers do not just make payout decisions in accordance with the legislation. These managers seek to smooth dividend payout over time. For example, Wang, Erikson, and Gau (1993) first show that REITs can pay dividends more than their net income. Hardin and Hill (2008) find that REIT dividend payments can be decomposed into mandatory payment and excess payment. The mandatory payment is essential for REIT to stay tax-preferred status and the excess payment (payment above the statutory minimum) is optional. Boudry (2011) proposes 4 See the REIT Modernization Act of 1999 on the website: 4

5 a new methodology that decomposes REIT dividends into discretionary and nondiscretionary components. All these studies indicate that REIT managers have concerns about firm payout choices. Firm managers tend to pay extra dividends or repurchase shares when having nonsustainable excess cash, and they increase regular dividends only if they expect sustainable higher future earnings (e.g., Baker and Wurgler, 2004; Li and Lie, 2006). This is because increases in regular dividends may be interpreted by shareholders as the long-term commitment of firm managers to distribute future cash flows (e.g., Brav et al., 2005). Failure to commit sustainable dividend payouts can result in unfavorable market reactions. Share repurchases can be interpreted as a signal of stock undervaluation and lead to temporary increases in stock prices. Scholes and Wolfson (1989) argue that implementing a DRIP may avoid a negative signal that a new equity offering provides; using DRIP to raise equity capital over a long period of time, rather than all at once, can mitigate the adverse selection problem and reduce information asymmetries. In attempt to attract shareholders to participate in DRIPs, managers aim to improve firm performance by investing in positive NPV projects and make conservative payout choices. Additionally, investors are likely to invest in DRIP REITs when these firms consistently pay stable dividends. Therefore, DRIP might align firm management and shareholder interests. We formulate hypotheses based on Agency Theory and /or the Signaling Model in the next section. Either theory indicates that dividend-paying firms with DRIPs employ a conservative payout policy in order to incentivize shareholders to make long-term investments. Specially, DRIP REITs rather than non-drip REITs tend to conduct a conservative payout policy due to the sensitivity of investor reactions to payout choices. 5

6 We examine the effect of DRIP on REIT payout choices using a multinomial logistic regression model. The sample includes 271 equity REITs over a period from 1999 to Among these REITs, 127 have DRIPs. After controlling for factors documented in prior studies that affect firm payout choices, we find that DRIP REITs, in comparison to non-drip REITs, are less likely either to pay regular dividends with extra dividends and repurchase shares or to omit all payouts, including dividends and repurchases. This indicates that DRIPs influence REIT managers to make conservative payout choices. We also apply event study methodology to examine differences of the dividend announcement date effect and the dividend payment date effect between DRIP REITs and non- DRIP REITs. We show that (1) the positive market reactions to announcements of regular dividend increases or extra dividends are weaker for DRIP REITs than for non-drip REITs and (2) announcements of dividend decreases at date T1 can only cause statistically significant and negative cumulative abnormal returns from (T1-30) to (T1 + 30) in non-drip REITs. We interpret these phenomena as indicating that DRIP REITs have weaker market reactions to dividend changes, due in part to managerial discretion in maintaining a conservative payout policy. We consider the dividend payment date as the date on which the administrators will invest reinvested dividends for the purchase of common shares 5. We find that price reactions around the dividend payment date are greater for DRIP REITs than for non-drip REITs, suggesting higher temporary price pressure around the event date in the former case. 5 Unless the dividend payment date is not a trading day, in which case the dividend reinvestment date will be the next trading day. If the administrators purchase common shares in the open market, then they will buy back stocks around the dividend payment date, in most cases, within 30 days after the dividend payment date. 6

7 Our study contributes to the existing literature by providing new evidence showing that DRIP can affect firm payout choices. Most REITs use shares newly issued to investors who participate in the DRIP, converting a portion of internally generated cash flows into new equity with low flotation costs 6. DRIPs become an important source of capital for REITs identified in this study. As a consequence, REIT managers tend to make conservative payout choices after adopting DRIP. Our findings may also help to explain (1) why less than 30 REITs choose to distribute elective stock dividends subsequent to the enactment of the new IRS rules during financial crisis and (2) why REIT investors do not seem to favor cash dividends over stock dividends, two facts documented by Devos, Spieler, and Tsang (2014). They attribute their findings to the fact that under efficient market investors can always convert stock dividends into cash by selling the stock; hence investors are generally not worse off receiving stock dividends. Alternatively, we interpret the behavior of few REITs distributing elective stock dividends as a reflection of REIT managers maintaining a conservative payout policy. The remainder of this paper proceeds as follows: Section 2 provides related literature review and develops the research hypotheses. Section 3 discusses the sample selection procedures and research design. Section 4 summarizes the empirical results and Section 5 presents robustness test results. Section 6 concludes. 2. Related Literature and Hypothesis Development Like industrial firms, REITs can distribute cash flow back to shareholders in the forms of regular dividends, extra dividends (including irregular dividends and excess dividends), and stock repurchases. In a sample of 8,290 REIT regular dividend announcements, 91 percent 6 According to security registration fillings in the SEC EDGAR database, companies use Form S-3D, Form S-3 or Form S-3ARS to self-register new shares under DRIPs without using underwriters. 7

8 (about 7,562) of regular announcements are made on a recurring quarterly basis and less than 10 percent of regular dividend announcements recur monthly, annually, or semiannually. Special dividends are usually one-time dividend distributions to shareholders and are considered not to be sustained in the future. Most REITs repurchase stocks using open market stock repurchase programs or tender offers. REITs, rather than industrial firms, have a lower overall level of repurchases because of the regulation on dividend payout policy (Boudry, Kallberg, and Liu, 2013). As such, repurchases cannot be the dominant form of REIT payout, in contrast to industrial firms that use repurchases to substitute for dividend payout in recent decades (e.g., Grullon and Michaely, 2002; Skinner, 2008). Existing literature has developed four possible hypotheses to explain firm payout policy: Agency Theory, the Signaling Model (or the Information Content of Dividends), the Clientele Effect, and the Tax Effect. Our study builds on Agency Theory and the Signaling Model to develop four main testable hypotheses. Extant studies provide mixed results on the tax and clientele effects on payout choices. While Brav et al. (2005) suggest that taxes play a secondorder role in firm payout choices, Hanlon and Hoopes (2014) provide evidence that managers adjust corporate payout choices in response to changed investor-level taxes. In addition, shareholders who participate in DRIPs have the same federal income tax obligation for dividends reinvested under these plans as for dividends received in cash. The Clientele Effect concerns the reactions of institutional investors. Interestingly, DRIPs limit the allowable periodic investment to attract individual clients, but to prevent institutional investors and corporate insiders from exploiting large economies of scale. These stylized facts suggest that estimating possible links between tax or clientele effects and REIT payout choices are complicated and thus deserve another independent study. We leave it to the future work. 8

9 2.1 Agency Theory REITs are capital-intensive in nature and need to raise funds from external capital markets frequently. If DRIP can be considered a source of capital, firm managers should maximize returns to shareholders in order to attract investors to participate in DRIP. The principal-agent problem arises when firm managers do not act in the interest of shareholders. Firms having DRIPs may help reduce asymmetric information, as managers need to keep investors informed about current investments and future prospects. Shareholders who participated in DRIP REITs can act as debt holders to monitor firms activities and make sure that firms can fulfill obligations. Agency Theory also argues that when faced with high asymmetric information, a firm will pay more of its cash flow in dividends. Consistent with this prediction, Ghosh and Sirmans (2006) provide evidence showing that dividend decisions reflect managerial motives and incentives. Good performing managers invest in positive NPV projects to maximize shareholder value and need not pay high dividends, but poorly performing managers pay generous dividends to appease disgruntled investors. Under Agency Theory, firm managers aim to invest positive NPV projects and improve stock performance, in order to encourage investors to participate in DRIPs. Thus, firm managers in DRIP REITs prefer a conservative payout policy. 2.2 The Signaling Model Theoretical models suggest that firm managers use cash flow distributions to signal changes in their expectations about future prospects of firms (e.g., Leland and Pyle, 1977; Bhattacharya, 1979; Miller and Rock, 1985). Thus, payouts as signaling devices (in the form of regular dividend, extra dividend, or share repurchase) require managers to pay due diligence on 9

10 conveying information about future prospects. Bradley, Capozza, and Seguin (1998) find that REIT managers make discretionary payout choices to mitigate the negative effect of dividend reduction on stock market. Brau and Holmes (2006) provide evidence that the managerial signaling hypothesis accounts for the abnormal returns observed around the share repurchase announcements. The strict regulation on payout policy disenables REITs to finance their growth via internally generated cash flow. Access to capital markets in order to finance investment projects is crucial for long-term REIT growth. With DRIPs, REIT managers are aware of the importance of consistently paying dividends. As a return, the equity market rewards REIT managers for considering investor demand for dividends. By adopting DRIPs, REITs not only promote longterm investment from existing shareholders but also encourage wider ownership of shares. Raising equity capital is also cheaper via DRIPs than through seasoned equity offerings, as DRIPs allow REITs to pay lower flotation costs on newly issued equity. Therefore, we make the hypothesis that DRIP REITs should have a conservative payout policy under the Signaling Model, due to high costs for decreasing or suspending dividends. In sum, the reduction in the agency problem and the signaling effect of payouts suggest that DRIP affects REIT payout choices. In our sample of 2,269 firm-year observations, REIT payout choices can be classified into seven categories: (1) paying stable regular dividends only; (2) paying regular dividends, and meanwhile distributing extra dividends, repurchasing shares, or both; (3) increasing regular dividends only; (4) increasing regular dividends, and meanwhile distributing extra dividends, repurchasing shares, or both; (5) decreasing and/or suspending regular dividends, with a possibility of distributing extra dividends, repurchasing shares, or both; 10

11 (6) distributing extra dividends, repurchasing shares, or both (without regular dividend payment); (7) omitting all payouts, including dividends and repurchases. First, we consider the payout policy that can cause extreme negative effects - the payout choice (7). Healy and Palepu (1998) find that firms with dividend omissions have decreasing earnings and investors could interpret announcements of dividend omissions as managers forecasts of future earnings decreases. Li and Lie (2006) also argue that payout omissions signal firm financial stress and uncertainty on future performance. Managers in DRIP REITs who conduct a conservative payout policy dislike uncertainty coupled with the payout choice (7). We then make the first testable hypothesis: H1 Relative to non-drip REITs, DRIP REITs have a lower tendency to omit all payouts during each of the sample years, as opposed to making other payout choices. Second, we discuss the payout choice (2) - REITs paying stable regular dividends with extra dividends and/or share repurchases. Compared with industrial firms, REITs tend to have less repurchases and more extra dividends. Extra dividends paid by industrial firms have gradually disappeared over the last half century (DeAngelo, et al., 2000). Since the dividend policy is sticky and REITs need financial flexibility to invest, DRIP REITs would prefer not to pay regular dividends with additional payouts such as extra dividends and/or stock repurchases. Thus, we state a second hypothesis as follows: 11

12 H2 Relative to non-drip REITs, DRIP REITs should be less likely to pay stable regular dividends with extra dividends and/or share repurchases, as opposed to making other payout choices 7. Payouts convey managerial confidence about future cash flows and thus help alleviate uncertainty faced by outside investors (Brav et al., 2005). After adopting DRIPs, firm managers make a conservative payout policy to cater to investors demand for a stable dividend policy, and as a return, such behavior is rewarded with less shock to the stock market valuations in dividend announcement events. Thus, we expect that H3 The stock price reactions to announcements of dividend changes are less significant in DRIP REITs than in non-drip REITs. In addition, since investors who participate in DRIPs will receive the new shares converted from cash dividends at each dividend payment date, increased demand for shares at that date leads to temporary price pressure. Berkman and Koch (2016) use a sample of regular firms from 2008 to 2012 and show that the payment date effect is concentrated in firms with DRIP, but not in firms without DRIP. Similarly, we state a fourth hypothesis as follows. H4 The dividend payment date effects are more pronounced in DRIP REITs than non-drip REITs. 3. Data and Research Design 3.1 Sample 7 Regarding the other five payout choices, we cannot easily infer different preferences between DRIP REITs and non-drip REITs. 12

13 We first select all the U.S. REITs, about 725 REITs, from SNL s Real Estate database 8. With a field search for dividend reinvestment plan in Data Wizard, we find REITs having DRIP in the most recent year in our sample (2014). We also identify additional DRIP REITs showed in the annual lists of DRIP firms since 1996 from the American Association of Individual Investors (AAII). These two sources only roughly tell us whether active or non-active REITs have DRIP. We then match these 725 REITs with REITs in CRSP/Ziman Real Estate database by ticker and/or company name and only select overlapped REITs showed in both databases. One channel that DRIP firms uses to communicate with investors is through filings with the SEC, for example, letters to shareholders, prospectuses, prospectus supplements (forms 424B3, 424B4, or 424B5), security registration forms (forms S-3, S-3D, or 3-11), and quarterly or annual forms (forms10-q or 10-K). To identify the initial offering date and suspension periods of DRIP for each REIT, we search the key terms reinvest and/or dividend in the above mentioned SEC filings and read through the designated firm report. When we are not sure whether a REIT has a DRIP, we further search the REIT name with the key terms reinvest and dividend in Google and Factiva. We find another eight REITs that release public news regarding the initial offering dates of DRIPs, which are earlier than the dates filed in the SEC reports. We also notice that many REITs have DRIPs but their names are not shown in the annual lists of DRIPs provided by the AAII; and most DRIP REITs gradually allow investors who make optional cash contributions to directly purchase stocks. Our sample period covers 1999 to The main reason is that accounting data for REITs might be limited before the year of 1999 (Hardin and Hill, 2008). We require the sample to have 8 The sample includes listed and delisted REITs. 13

14 positive market equity and non-missing institutional ownership data. The final sample includes 2,269 firm-year observations of 271 REITs. And 127 out of the 271 REITs have DRIPs. The data set is an unbalanced panel. The sample REITs have year-observations between 1 year and 16 years. Table 1 presents the distribution of DRIP REITs and Non-DRIP REITs by year. Columns (1) and (3) show that the number of sample DRIP REITs ranges from 71 to 81, with an average of 75; and the number of sample non-drip REITs ranges from 48 to 89, with an average of 66. Columns (2) and (4) indicate a slightly higher percentage of DRIP REITs than non-drip REITs on average. [Put Table 1 about Here] We collect dividend distribution information from SNL Real Estate s database. This database provides the dividend announcement date, the ex-dividend date, the dividend payment date, and the amount of distributed dividends. It also shows whether the dividends are distributed as regular, special, irregular, or suspended dividend payments and whether regular dividends are paid on a recurring quarterly, monthly, annual, or semiannual basis. In this study, we define special dividends and irregular dividends as extra dividends for the sake of simplicity. Following Boudry, Kallberg, and Liu (2013), we collect the number of share repurchases and the average price of repurchased shares from SNL Real Estate s database. The database has these two items available starting in the year Note that the SEC Exchange Act rule 10b-18 requires public firms to disclose the two items in their financial statements (i.e., forms 10-K/Q) beginning on March 15 th, 2004 (Dittmar and Field, 2015). The dollar value of share repurchases 14

15 is the product of these two variables. For share repurchase data before 2006, we obtain the line item common share repurchases from the cash flow statement. We define a REIT with increasing (or decreasing) regular dividends in a given fiscal year as the REIT with greater (or less) regular dividends in the current year than those in the prior fiscal year. A REIT is defined as having stable regular dividends if it distributes the same amount of regular dividends in both current and prior fiscal years. A REIT is defined as paying extra dividends if it pays any extra dividends during a fiscal year. A REIT is defined as having dividend omissions if it makes an announcement of dividend suspensions during a fiscal year. A REIT is defined as repurchasing shares if it buys back shares in a given fiscal year. In our sample of 2,269 firm-year observations, REIT payout choices can be summarized into seven categories as previous stated: (1) paying stable regular dividends only; (2) paying regular dividends, and meanwhile distributing extra dividends, repurchasing shares, or both; (3) increasing regular dividends only; (4) increasing regular dividends, and meanwhile distributing extra dividends, repurchasing shares, or both; (5) decreasing and/or suspending regular dividends, with a possibility of distributing extra dividends, repurchasing shares, or both; (6) distributing extra dividends, repurchasing shares, or both (without regular dividend payments); (7) omitting all payouts, including dividends and repurchases. Table 2 reports fractions of REIT payout choices during each of the years between 1999 and In Table 2, columns (1) and (2) show that an average of 12 percent of REITs only pay regular dividends each year, and an average of 6 percent of REITs pay regular dividends with extra dividends and/or repurchase shares; columns (3) and (4) show that combined, about half of REITs increase regular dividends; and columns (5) through (7) present that an average of 15 percent of REITs decrease and /or suspend regular dividends, with a possibility of distributing extra dividends, repurchasing shares, 15

16 or both; around 7 percent of REITs distribute extra dividends, repurchases shares, or both, but they do not pay regular dividends during the year; and about 11 percent of firms omit all payouts, including dividends and repurchases. [Put Table 2 about Here] 3.2 Research Methodology We examine the effect of DRIP on REIT payout choices using a multinomial logistic regression model. The payout choices have j = 7 alternatives. We try to explain the probability that REIT i chooses alternative j, Pij = β0 + β1 DRIPi + Σβm CONTROLsi + βn Fixed Effectsi + ε (1) The dependent variable Pij stands for seven possible REIT payout choices and is an indicator variable, which (1) takes a value of one if a REIT only pays stable regular dividends during a year; (2) takes a value of two if a REIT pays regular dividends, and meanwhile distributes extra dividends, repurchases shares, or both; (3) takes a value of three if a REIT only increases regular dividends; (4) takes a value of four if a REIT increases regular dividends, and in the same year, distributes special dividends, repurchases shares, or both; (5) takes a value of five if a REIT decreases and/or suspends regular dividends, with a possibility of distributing extra dividends, repurchasing shares, or both; (6) takes a value of six if a REIT pays special dividends and/or repurchases shares, but does not pay regular dividends; (7)takes a value of seven if a REIT omits all payouts. Note that the alternatives j = 1, 2, 3 7 are assigned arbitrarily and have no meaning, because the alternatives have no particular ordering. 16

17 The main variable of interest is the indicator variable DRIP, which takes the value of 1 for DRIP REITs and 0 otherwise. We follow Bradley, Capozza, and Seguin (1998), Lie (2005), Hardin and Hill (2008), and Case, Hardin, and Wu (2012) to select control variables that affect firm payout choices. The control variables include MTB, BLEV, SIZE, IO, DRAWN, ROA, FFO, BETA-1, BETA, PAYOUT RATIO, DIVIDEND YIELD, and CREDIT/AT. MTB is the market value of equity plus the book value of debt in year t scaled by book value of total assets in year (t - 1). BLEV is measured as the book value of total debt in year t over book value of total assets in year (t - 1). SIZE is the log of market capitalization of a firm in year (t - 1). The market capitalizations are deflated to 2014 dollars using the consumer price index. IO is the percentage of the total shares owned by institutional investors according to 13F filings in the event year t. DRAWN is the difference of the ratio of credit line drawn to total credit line between year (t - 1) and year t. ROA is measured as net income available to common shareholders in year t divided by book value of total assets in year (t - 1). FFO is funds from operations, defined as net income excluding gains or losses from sales of real estate, plus depreciation and amortization in year t, divided by book value of total assets in year (t - 1). BETA-1 is the equity estimated using daily returns from CRSP during the year (t - 1). BETA is the difference between the equity beta estimated using daily returns during the year (t + 1) and BETA-1. PAYOUT RATIO is the total dividends in year t scaled by net income available to common shareholders in year (t - 1). If the dividends exceed the net income, the payout ratio is set to 1. DIVIDEND YIELD is the splitadjusted regular dividends per share during the year (t - 1) scaled by the price at the end of the year (t - 1). CREDIT/AT is the ratio of credit line available in year t over book value of total assets in year (t - 1). All the variables are also defined in detail in Appendix A. We also include year fixed effect and property-type fixed effect to control for time-invariant unobserved 17

18 correlated variables. We follow Hardin and Hill (2008) and use seven property types: multifamily, retail, office, industrial, self-storage, hotel, and other. We employ the standard event study methodology to investigate differences of the dividend announcement date effect or the dividend payment date effect between DRIP REITs and non- DRIP REITs. A standard market model is used with the value-weighted Ziman REIT index being the market return index and the estimation period spans from 10 to 250 days prior to the dividend announcement date. Similar to Case, Hardin, and Wu (2012), we calculate the cumulative abnormal returns around the dividend announcement date T1 based on four different event windows: (T1-30, T1-1), (T1 + 0, T1 + 1), (T1 + 2, T1 + 30), and (T1-30, T1 + 30). Following Berkman and Koch (2016), we calculate the cumulative abnormal returns around the dividend payment date T2 based on four different event windows: (T2-10, T2-1), (T2 + 0, T2 + 1), (T2 + 2, T2 + 10), and (T2-10, T2 + 10). The estimation period is from (T2-250) to (T2-10). 4. Empirical Results 4.1 Descriptive Statistics In Table 3, Panel A reports descriptive statistics for the full sample. The average REIT has a market-to-book ratio of The sample means of BLEV, SIZE, IO, and ROA are respectively 0.63, 20.51, 0.56, and Although REITs have a small PAYOUT RATIO with a mean of 0.00, DIVIDEND YIELD has a mean of Since variable SIZE has a mean much greater than means of the other variables, we standardize all the selected variables to make our model behave better. Panel B presents descriptive statistics for two subsamples: DRIP REITs and non-drip REITs. When comparing firm fundamentals between the two subsamples, we make the null hypothesis that means (medians) of variables between the two subsamples are equal. The last 18

19 two columns of Panel B indicate that DRIP REITs and non-drip REITs have significantly different firm characteristics. Relative to non-drip REITs, DRIP REITs have higher MTB, larger SIZE, and a higher percentage of institutional ownership (IO). DRIP REITs also report higher ROA and FFO, accompanied with higher PAYOUT RATIO and DIVIDEND YIELD. DRIP REITs appear to have a significantly larger market beta in year (t - 1). Panel C reports the Pearson correlation coefficients. DRAWN is not significantly correlated with other selected firm characteristics. [Put Table 3 about Here] 4.2 Multinomial Logistic Regression Results We present the main empirical results of multinomial logistic regression analyses for specification (1) in Tables 4-5. All the independent variables except DRIP are standardized. Thus, we can interpret each estimated coefficient as the effect of a one standard deviation increase in the independent variable on the percentage change in the possibility of a specific payout choice, ceteris paribus. Odds ratios can be interpreted as the effect of a one standard deviation increase in the independent variable on the predicted odds ratio of making a specific payout choice, with the other variables in the model held constant. Year and property-type fixed effects are also included in all regressions. Table 4 reports Type Three Analysis of Effects, based on the Wald test. Each chi-square is to test the null hypothesis that an explanatory variable has no effect on payout choice. The significant level on DRIP is beyond the 0.01 level. Therefore, we cannot reject the null hypothesis that DRIP has no effect on payout choice. Other control variables also significantly impact REIT payout choice, except DRWAN. 19

20 [Put Table 4 about Here] In Table 5, Panel A presents results with REIT payout choice (7), in which REITs omit all the payout, as the reference category. The positive and statistically significant estimated coefficients on DRIP in columns (1) and (3) - (5) imply that having DRIP decreases the likelihood of REITs omitting all payouts during each of the sample years, as compared to other payout choices. The corresponding odds ratios, respectively, show that having DRIP results in a 2.2-fold increase in the odds of REITs paying stable regular dividends, as compared with REITs omitting payouts; a 2.1-fold increase in the odds of REITs increasing regular dividends; a 2.4- fold increase in the odds of REITs increasing regular dividends and meanwhile distributing special dividends, repurchasing shares, or both; and a 1.9-fold increase in the odds of REITs decreasing and/or suspending regular dividends. But columns (2) and (6) indicate that neither the odds of REITs paying regular dividends with additional payout, nor the odds of REITS only distributing extra dividends, repurchasing shares, or both, are significantly different from the odds of REIT payout omissions. Overall, rather than non-drip REITs, DRIP REITs have lower tendency to omit all payouts during each of the sample years, as opposed to making other payout choices (H1). Column 1 of Panel A, in Table 5, indicates that a one standard deviation increase in MTB produces, on average, a 1.4 decrease in the log odds of getting REIT payout choice (1) paying stable regular dividends only. In other words, higher MTB induces a greater possibility of REITs omitting all of the payments. The same conclusion can be drawn from the significant negative coefficients on MTB in columns (2) and (4) - (6). In addition, REITs are more likely to omit all of the payouts, when they have higher BLEV, smaller SIZE, lower FFO, greater BETA-1, lower DIVIDEND YIELD, or lower CREDIT/AT. 20

21 Panel B of Table 5 displays results with REIT payout choice (2), in which REITs pay regular dividends with extra dividends, share repurchases, or both, as the reference category. In column (1), the odds ratios for DRIP is 2.2, implying that the odds that DRIP REITs paying regular dividends, rather than regular dividends with other additional payouts, are about 2.2 times the odds for DRIP REITs. The odds ratio for DRIP is 2.1 in column (2), suggesting that the odds that DRIP REITs will increase regular dividends rather than regular dividends with other payouts are about 2.1 times the odds for DRIP REITs. Similarly, the odds that DRIP REITs will increase regular dividends, as shown in columns (3) - (4), rather than regular dividends with other payouts are about twice the odds for DRIP REITs. The coefficient of DRIP in column (5) has a positive sign but is not statistically significant, indicating that the possibility that only paying extra dividends and/or repurchasing share instead of paying regular dividends with other payouts is not significantly different between DRIP REITs and non-drip REITs. The positive and statically significant estimated coefficients and odds ratios in columns (1) through (4) indicate that DRIP REITs are less likely to make payout choice (2), relative to the other payout choices, such as paying stable regular dividends only, increasing regular dividends, and decreasing dividends (H2). Column (2) of Panel B, in Table 5, also indicates that a one standard deviation increase in MTB produces, on average, a increase in the log odds of getting REIT payout choice (3) - increasing regular dividends only. Lower MTB induces a greater possibility of REITs paying regular dividends with other additional payments, instead of only paying regular dividends. But the significant negative coefficient on MTB in column (4) of Panel B indicates that higher MTB is associated with a greater possibility of REITs paying regular dividends with other additional 21

22 payouts, instead of decreasing regular dividends. REITs with lower BLEV, larger firm size, or a lower percentage of institutional ownership (IO) are more likely to use payout choice (2). Panel C of Table 5 shows results based on comparisons of any other two possible payout choices. The coefficients on DRIP are all insignificant, suggesting no significant difference of any two of the payout choices (1) and (3) - (6) between DRIP REITs and non-drip REITs. We attempt not to over-interpret this result because DeAngelo et al. (1996) also suggest dividends not to be reliable signals in some cases. Firm managers might not have informative resources to estimate future cash flow. [Put Table 5 about Here] 4.3 Dividend Announcement Date Effect In this section, we study how the market responds to announcements of dividend changes. Stable regular dividend announcements are defined as quarterly regular dividends paid at date T1 are the same as the most recent quarterly regular dividends. Regular dividend increase (or decrease) announcements are defined as regular dividends paid at date T1 are greater (or less) than the most recent regular dividends. Extra dividend announcements include announcements of special dividends and irregular dividends. Increased (or decreased) extra dividend announcements are defined as extra dividends paid at date T1 are greater (or less) than the most recent regular dividends. DeAngelo et al. (2000) find that the stock market reacts favorably to announcements of extra dividends. Included in the analyses are 5,650 quarterly stable regular dividend announcements, 1,361 regular dividend increase announcements, 168 regular dividend decrease announcements, 79 dividend suspension announcements, and 217 extra dividend announcements. The extra dividend announcements consist of 111 increased extra dividend 22

23 announcements and 106 decreased extra dividend announcements. Announcement period returns are the abnormal stock returns measured from the date T1 a REIT makes dividend announcement through the day after the announcement (T1 + 1) using a market model, where the valueweighted Ziman REIT index is used to proxy overall market returns and the estimation period spans from 10 days to 250 days prior to the announcement date. Table 6 shows seven cases of dividend changes. The market reacts statistically significantly and positively to regular dividend increases or extra dividends. Panel B shows that in the announcements of regular dividend increases, CARs from date T1 to date (T1 + 1), or (0, +1) are 0.30 percent (t-stat = 4.80) for DRIP REITs and 0.73 percent (t-stat = 6.97) for non-drip REITs, respectively. The difference of CARs (0, +1) between non-drip REITs and DRIP REITs is 0.43 percent (t-stat = 2.75). Panels E and F show that in the announcements of extra dividends, the difference of CARs (0, +1) between non-drip REITs and DRIP REITs are all positive and statically significant. Thus, in a short term, the positive market reactions to dividend increases or extra dividends are stronger in non-drip REITs than in DRIP REITs. Panel C shows that in a prolonged period of time, CARs from (T1-30) to (T1 + 30) around the regular dividend decrease announcement date T1 are, respectively, 2.88 percent (t-stat = 0.11) for DRIP REITs and percent (t-stat = -3.28) for non-drip REITs. The difference between the two CARs is percent (t-stat = -2.48). Panel F shows that CARs (-30, 30) in the announcement of large extra dividends, are respectively, 3.67 percent (t-stat = 2.20) for DRIP REITs and 7.32 percent (t-stat = 3.85) for non-drip REITs, resulting in a difference of 3.65 percent (t-stat = 1.61). Therefore, CARs from (T1-30) to (T1 + 30) around the announcements of dividend decreases (or large extra dividends) are more negative (or positive) in non-drip REITs than in DRIP REITs. In dividend suspension announcements, there is no statistical difference of 23

24 CARs around the announcement dates. Overall, dividend announcement date effects are weaker in DRIP REITs than in non-drips REITs, consistent with H3. DRIP REITs have weaker market reactions to announcements of dividend changes, due in part to managerial discretion in maintaining a conservative payout policy. [Put Table 6 about Here] 4.4 Dividend Payment Date Effect In this section, we follow Berkman and Koch (2016) to investigate the dividend payment date effect in REITs. They find that the dividend payment date effect is concentrated among firms with DRIPs. In that the plan administrators need to purchase shares in the open market at each dividend payment date, leading to temporary price pressure. Therefore, we expect that DRIP REITs should have stronger dividend payment date effect than non-drip REITs. As in the previous section, we classify dividend changes into six different types, such as no change in regular dividends, regular dividend increase, regular dividend decrease, extra dividend increase, extra dividend decrease, and dividend omission. Each dividend payment, excluding the omitted one, has the corresponding dividend payment date 9. Shown in Table 7 are 5,469 stable regular dividend payments, 1,312 increased regular dividend payments, 165 decreased regular dividend payments, and 217 extra dividend payments. The extra dividend payments consist of 112 increased extra dividend payments and 105 decreased extra dividend payments. Table 7 reports cumulative abnormal returns (CAR) around the dividend payment date T2. CARs are separately estimated from four event windows: (T2-10, T2-1), (T2, T2 + 1), (T2 + 2, T2 + 10), and (T2-10, T2 + 10). A standard market model is used with the value-weighted Ziman 9 We have slight different observations of dividend payment date due to data availability on daily stock returns. 24

25 REIT index being the market return index, and the estimation period spans from 10 to 250 days prior the dividend payment date. In Table 7, Panel A shows that CARs (0, 1) around the stable regular dividend payment date are, respectively, 0.18 percent (t-stat = 4.21) for DRIP REITs and 0.09 percent (t-stat = 1.50) for non-drip REITs. But the difference of CARs between these two groups is not significant. Panel B shows that CARs (0, 1) around the increasing regular dividend announcements are, respectively, 0.23 percent (t-stat = 3.65) for DRIP REITs and 0.02 percent (t-stat = 0.26) for non- DRIP REITs, with a difference of percent (t-stat = -2.13) between the two groups of firms. We find no significant difference of CARs between DRIP REITs and non-drip REITs, in case of extra dividend payments or increasing extra dividend payments. Overall, we show a slightly stronger dividend payment date effect in DRIP REITs than in non-drip REITs, consistent with H4. [Put Table 7 about Here] 5. Robustness Check We also use a logistic regression model to model the probability of REITs that pay stable regular dividends (increase regular dividends, decrease regular dividend, pay extra dividends, or repurchase shares) taking on the value of 1, rather than the probability of REITs that do not pay stable regular dividends (do not decrease regular dividends, do not increase regular dividends, do not pay extra dividends, or do not repurchase shares) taking on the value of 0. When not considering alternative payout choices during the year, we show in Table 8 that DRIP REITs are more likely to increase regular dividends and to decrease regular dividends. This result might mislead us to interpret as more flexible payout choices in DRIP REITs. Thus, it is important to 25

26 take all the payout choices into account when we study the effect of DRIP on REIT payout choices. [Put Table 8 about Here] In Appendix B, we also use an alternative definition of payout choice to re-examine the dividend announcement date effect and the dividend payment date effect. Stable regular dividend announcements are defined as regular dividends paid in quarter Q are the same as regular dividends paid in the same quarter of last year, or quarter (Q - 4). Regular dividend increase (decrease) announcements are defined as regular dividends paid at quarter Q are greater (less) than the regular dividends paid at the same quarter of last year. According to the new definition of payout choice, we consistently show that dividend announcement date effect is weaker and dividend payment date effect is still slightly stronger in DRIP REITs than in non-drip REITs (see Tables I & II). 6. Conclusion This study investigates whether dividend-paying firms with DRIP have a conservative payout policy, predicted by Agency Theory and/or the Signaling Model. We use the case of REIT to test this predication. In that REITs are required to make distributions, other than capital gain dividends, to their shareholders each year in an amount at least equal to the sum of 90 percent of their income, in order to stay federal tax-preferred status. This greatly limits REIT managers to use internally generated cash flows and requires them to access capital market frequently to raise capital. Therefore, it is important for REIT managers to use DRIPs to retain some internally generated cash and thus to alleviate the demand for external financing. 26

27 We show that managers in DRIP REITs employ a conservative dividend payout policy. Relative to non-drip REITs, DRIP REITs are less likely to pay regular dividends with extra dividends and/or share repurchases, or to omit all payouts, including dividends and repurchases. We also interpret the fact of DRIP REITs having weaker market reactions to dividend changes, to some extent, as a reflection on the managerial effectiveness on maintaining a conservative payout policy. In addition, we find slightly stronger dividend payment date effect in DRIP REITs, but not in non-drip REITs, suggesting higher temporary price pressure in DRIP REITs around the dividend payment date. Future work will be of great interest if: (1) we can trace how and to what extent, firms raise new issues through DRIPs; (2) we can figure out why some firms adopt this plan while others not; and (3) we can figure out the risk and returns of those firms adopted DRIPs. We thank McKay Price, seminar participants at the 2016 European Real Estate Society annual conference, and seminar participants at the University of Cincinnati for helpful comments and suggestions. 27

28 References Baker, Malcolm, and Jeffrey Wurgler, 2004, Appearing and Disappearing Dividends: the Link to Catering Incentive, Journal of Financial Economics 73: Barclay, M.J., and C.W. Smith Corporate Payout Policy: Cash Dividends versus Open-market Repurchases. Journal of Financial Economics 22: Berkman, Henk, and Paul D. Koch, 2016, DRIPs and the Dividend Pay Date Effect, Journal of Financial and Quantitative Analysis, forthcoming. Blouin, Jennifer, and C. Bryan Cloyd, 2005, Price Pressure from Dividend Reinvestment Activity: Evidence from Closed End Funds, Wharton School Working Paper. Bhattacharya, S Imperfect Information, Dividend Policy, and the Bird in the Hand Fallacy. Bell Journal of Economics 10: Boudry I. Walter An Examination of REIT Dividend Payout Policy, Real Estate Economics 39(4): Boudry I. Walter, Jarl G. Kallberg, and Crocker H. Liu, 2013, Investment Opportunities and Share Repurchases, Journal of Corporate Finance 23: Bradley, Michael., Dennis R. Capozza, and Paul J. Seguin Dividend policy and cash-flow uncertainty. Real Estate Economics 26(4): 555-5d80. Brau, J.C. and A. Holmes Why Do REITs Repurchase Stock? Extricating the Effect of Managerial Signaling in Open Market Share Repurchase Announcements. Journal of Real Estate Research 28:1-23. Brav, A. J.R. Graham, C.R. Harvey, and R. Michaely Payout Policy in the 21 st Century 28

29 Case, B., W.Hardin and Z. Wu REIT Dividend Policies and Dividend Announcement Effects during the Liquidity Crisis. Real Estate Economics 40(3) Dammon M. Robert, and Chester S. Spatt, 1992, An option-theoretic approach to the valuation of dividend reinvestment and voluntary purchase plans, Journal of Finance 1, DeAngelo, H., L. DeAngelo, and D.J. Skinner Reversal of Fortune Dividend Signaling and the Disappearance of Sustained Earnings Growth. Journal of Financial Economics 40, DeAngelo, H., L. DeAngelo, and D.J. Skinner Special Dividends and the Evolution of Dividend Signaling. Journal of Financial Economics 57, Denis, D.J., D. K. Denis, and A. Sarin.1994.The Information Content of Dividend Changes: Cash Flow Signaling, Overinvestment, and Dividend Clienteles. Journal of Financial and Quantitative Analysis 29: Devos, E. A. Spieler, and D. Tsang, 2014, Elective Stock Dividends and REITs: Evidence from the Financial Crisis, Real Estate Economics 42, Dhillon, Upinder, Dennis Lasser, and Gabriel Ramirez, 1992, Dividend reinvestment plans: An empirical analysis, Review of Quantitative Finance and Accounting 2, Dittmar, A., and L.C. Field Can Managers Time the Market? Evidence Using Repurchase Price Data, Journal of Financial Economics 115(2), Eades, K Empirical Evidence on Dividends as a Signal of Firm Value. Journal of Financial and Quantitative Analysis 17: Fama E. F., and K.R. French Disappearing Dividends: Changes Firm Characteristics or Journal of Financial Economics 60, Lower Propensity to Pay? Ghosh, C. and C. F. Sirmans Do Managerial Motives Impact Dividend Decisions in REITs? The Journal of Real Estate Finance and Economics 32: Ghosh, C. and L. Sun Agency Cost, Dividend Policy and Growth: The Special Case of REITs. The Journal of Real Estate Finance and Economics 48: Grinstein, Y., and R. Michaely Institutional Holdings and Payout Policy. Journal of Finance 60: Grullon, G., and R. Michaely, Dividends, Share Repurchases, and the Substitution Hypothesis. Journal of Finance 57: Guay, W., and J. Harford The Cash-Flow Permanence and Information Content of Dividend Increases versus Repurchases. Journal of Financial Economics, 57 (2000), Hanlon, M., and J. L. Hoopes.2014.What do firms do when dividend tax rates change? An examination of alternative payout responses. Journal of Financial Economics 114, Hardin, W. and M. Hill REIT Dividend Determinants: Excess Dividends and Capital Markets. Real Estate Economics 36(2): Healy, P.M. and K. G. Palepu Earnings Information Conveyed By Dividend Initiations and Omissions. Journal of Financial Economics 21, Jagannathan, M., C. Stephens, and M. Weisbach Financial Flexibility and the Choice between Dividends and Stock Repurchases. Journal of Financial Economics 57,