On the Capital Structure of Real Estate Investment Trusts (REITs)

Transcription

1 On the Capital Structure of Real Estate Investment Trusts (REITs) Zhilan Feng, Chinmoy Ghosh and C. F. Sirmans* Abstract Much of the literature on capital structure excludes Real Estate Investment Trusts (REITs) due mainly to the unique regulatory environment of these firms. As such, the issue of how REITs choose among different financing options when they raise external capital is largely unexplored. In this paper, we examine the capital structure of REITs to answer two questions: is there a relationship between market-to-book and leverage ratios? and, does market-to-book have a temporary or a long-term impact on leverage ratios? Our results suggest that REITs with high market-to-book ratios have high leverage ratios, and historical market-to-book has long-term persistent impact on current leverage ratio. We interpret these findings as supportive of pecking order theory. When financing costs of adverse selection exceed costs of financial distress, pecking order is more relevant in explaining the cross-sectional variation in capital structure. Zhilan Feng is at School of Management, the Graduate College of Union University, Chinmoy Ghosh and C. F. Sirmans are at the Center for Real Estate and Urban Economic Studies at the School of Business, University of Connecticut, Storrs, CT, USA. All correspondence may be addressed to Zhilan Feng at fengz@union.edu.

2 On the Capital Structure of Real Estate Investment Trusts (REITs) I. Introduction and Motivation This paper explores how the capital structure of Real Estate Investment Trusts (REITs) evolves over time. Much of the traditional literature in finance tends to exclude regulated firms because regulation is often designed to address the very same market imperfections that theory focuses on. The first motivation of our paper primarily draws from the unique regulatory environment REITs operate in. REITs were primarily created as an investment vehicle for institutions that tended to avoid investing in real estate assets because of lack of transparency and liquidity. In essence, the regulation on REITs are geared more to induce investment than to prevent neglect of fiduciary responsibility. There are mainly three good reasons to issue debt. One, it raises cash. Two, interest payments are tax deductible so that the tax shield adds value to the firm. Three, the mandatory interest payment on debt mitigates the agency cost of managerial proclivity to waste cash on poor investments. On the negative side, borrowing exposes the firm to bankruptcy costs; and, leverage may prompt managers to avoid profitable investments to minimize transfer of wealth to bondholders. Like debt, equity raises cash, but issue costs can be significant if investors discount the value of shares out of concern that managers issue shares only when they are overvalued. On balance, debt appears to be the less costly alternative. Over the years, the search for an optimal capital structure has been largely empirical, albeit elusive. A second, and potentially more interesting, motivation of our research is the recent controversy about which theory best describes the prevalent practice in firms financing choices. The trade-off theory states that an optimal capital structure exists and this is characterized by the trade off between benefits and costs of borrowing. Among the benefits, the most significant is the tax deductibility of interest payments, but costs of financial distress can be substantial. While firms deviate from the optimum capital structure in the short term, the long term capital structure is invariant. The theory implies that adjustments in capital structure in response to fluctuations in valuation of the firm, and capital needs are temporary; capital structure regresses to the optimal level in the long run. The evidence on actual capital structure choices lends only scant support to the trade off theory. 1

3 The pecking order theory developed by Myers and Majluf (1984) is more potent simply because it provides a better description of actual managerial behavior. The model, based on information asymmetry between shareholder and managers, says that if managers are more informed than shareholders about the firm s prospects, they would be tempted to sell new shares only when they are overvalued. Wary shareholders will anticipate this and revalue shares downwards. Under this scenario, stock prices will always react negatively to equity issues. Consequently, managers who act in shareholders interest will always avoid issuing new stock, and prefer issuing less risky debt instead. This implies that high growth firms, particularly those with insufficient free cash flow will have high debt ratios. A more dynamic version of the theory states that high growth firms may reduce leverage and use retained earnings for current investment to avoid issuing equity if and when need for additional funds arises in the future. An important implication of the theory is that no optimal capital structure exists, rather capital structure evolves in response to the firm s investment opportunities. Shyam-Sunder and Myers (1999) report evidence consistent with the pecking order theory. A third theory, known as the market timing theory [Baker and Wurgler (2003)], is more behavioral in nature and scope and simply states that long-term capital structure is merely a manifestation of manager s attempts to time equity issues to coincide with high market valuation. According to this theory, firms with high growth and investment opportunities have high market values and tend to issue equity more often, resulting in low leverage ratios. The idea is that if market associates high market values with low adverse selection costs, that presents high growth firms with the opportunity to issue equity at an advantage. It is noteworthy that the market timing theory and the simple pecking order theory have opposite implications for the relation between market values and debt ratios. Neither theory identifies an optimal capital structure, however. A final question that has attracted considerable attention is whether changes in capital structure are permanent. Trade off theory implies that any change in capital structure is temporary and firms regress to the long term optimum over time. There is no such implication under the pecking order or the market timing theories. An evidence of a permanent relation between the need (investment opportunities) and sources (financing choices) of capital is sufficient to unequivocally reject the trade off theory. While the theoretical underpinnings of the theories are well developed, the empirical evidence is mixed, at best. It is only recently that the empirical enquiries have focused on the dynamics of the evolution of capital structure over time. Shyam-Sunder and Myers (1999) claim support for the pecking order theory, Frank and Goyal (2002) refute it, and Fama and French (2002) report findings consistent with, and contrary to both trade off and pecking order stories. In 2

4 the most comprehensive analysis of market timing theory to date, Baker and Wurgler (2002) interpret the evidence to be in conformity with the market timing theory to the exclusion of the other two. It is worth noting all studies of capital structure decisions over time reject trade off theory unequivocally. Because of their unique regulatory environment, we contend, REITs are an ideal laboratory setting to provide additional evidence on these competing theories. First, REITs do not pay any taxes if 95% of taxable earnings are paid out as dividends. Second, high payout implies that REITs have low free cash flow, such that managers have little opportunity to waste cash on non value-maximizing acquisitions. REITs face the usual costs of financial distress, however. Absence of tax deductibility of interest payments, and reduced agency conflict, immediately suggest REITs should have no debt in their capital structure. The anecdotal evidence is clearly inconsistent with this notion. 1 The requirement that 95% of the taxable earnings be paid out as dividends forces REITs to raise funds from the capital market where debt is a less attractive alternative than taxable firms, and the agency cost benefit of debt is also muted. Turning to equity, however, entails the costs of adverse selection which must be borne by the existing shareholders. We argue that these costs are particularly severe for REITs. For example, monitoring REIT managers calls for special skills and knowledge about general and local economic trends, conditions of comparable properties, complex financing arrangements, other specialized skills, and even inside information [Han (2004)]. In addition, since REITs are involved in real property transactions that include a wide range of heterogeneous, illiquid assets, it is difficult for shareholders to determine the fair market values of these transactions. This results in lack of transparency which makes monitoring of managers critical. As Ghosh and Sirmans (2001, 2003, 2004), and Han (2004) observe, however, REITs must abide by special regulations that can weaken or render ineffective the standard governance mechanisms. To elaborate, to qualify as a REIT, the firm must maintain a diversified ownership with at least 100 shareholders, the five biggest of which may not own more than 50 percent of the total shares outstanding. Campbell et al. (2001) contribute the lack of hostile takeovers among REITs to this regulation. This unique ownership structure diminishes the effectiveness of monitoring by the market for corporate control, and exacerbates the lack of transparency. In essence, issuing equity is a particularly costly proposition for REITs. Under this scenario, pecking order theory predicts financing first with retained earnings, then debt, and 1 Brown and Riddiough (2003) reports that over the period September 1993 to March 1998, REITs made a 120 debt offerings of $133m each, on average. 3

5 equity last. Since retained earnings are very low for REITs, pecking order leans heavily towards debt financing. Finally, market timing theory suggests that managers look for opportunities to time equity issues when adverse selection costs are low. It may be argued that these opportunities are relatively scarce for REITs because of lack of transparency and incomplete monitoring. In summary, the trade off prediction of all equity contradicts anecdotal evidence, pecking order calls for predominantly debt financing, and market timing suggests selling equity if opportunities exist. We suggest that REITs will prefer to issue debt whenever the cost of discounted equity exceeds the cost of financial distress, and equity otherwise. The choice of financing is essentially an empirical issue. The clear advantage with REITs is that because of low retained earnings, the financing decision may come down to a simple choice between debt and equity. To determine how capital structure evolves over time, and the persistence of capital structure change, a time series analysis of the relation between needs and sources of financing must be conducted. Baker and Wurgler (2002) study the firms financing decision from the initial public offering (IPO). Conceivably, issuing debt is not appealing at the IPO stage because young firms are considered more risky. As the firm matures, financing decisions reflect both pure adjustments (if any) in capital structure and need for investment funds. It is generally assumed that need for funds is a function of the firm s investment and growth opportunities and the standard sources of capital include retained earnings and security issuances. The standard proxy for need for funds is the market value to book value ratio, the assumption being that high market value reflects market s assessment that the firm has access to profitable investment opportunities. We analyze the capital structure decisions of REITs over the period 1992 to 2003 using the same approach as Baker and Wurgler (2002). Over this period, we follow the REITs from the year they go IPO to the last surviving year. The sample size is therefore driven by the IPO activity of REITs. The most active years were 1994, 1995 and 1996 when over 50 REIT IPOs raised capital. The smallest sample size is 4 in 1992, steadily growing to the largest sample size of 108 in We use market to book ratio as a proxy for investment opportunities and firm s need for capital. The analyses based on contemporaneous data reveal a weakly significant positive relation between M/B ratio and leverage, strongly negative relation between M/B ratio and net equity issues, and weak relation between M/B and retained earnings. The long term weighted average M/B ratio is a strongly significant determinant for leverage ratio which suggests that the effect on M/B ratio on leverage is not transient and firms do not adjust their leverage ratios to a target level. The long-term persistence of leverage decisions is inconsistent with the trade off theory. The long-term as well as the contemporaneous evidence is not 4

6 consistent with the market timing theory. We find weak evidence in favor of the pecking order theory from the yearly analysis, but strong support from the long-term regressions results. The findings for REIT are intriguing in that they are contrary to the recent evidence in Baker and Wurglar (2002) for a broader data set, and subject to interpretation vis-a-vis the conclusions in Brown and Riddiough s (2003) analysis of REIT capital structure. Baker and Wurglar find evidence in line with the market timing theory. A potential explanation for the differential findings is that the window of opportunity when adverse selection costs are low is less frequent and narrower for REITs. The lack of transparency of real estate assets, and the consequent information asymmetry is a contributing factor. Reinforcing the problem is the restrictive ownership requirements in REITs which makes it difficult for blockholders to form ownership stakes, and diminishes their incentive to monitor management. A sheltered management widens the credibility gap between shareholders and managers. Brown and Riddiough (2003) analyze public issues by REITs over the years and document several stylized characteristics of these offers. An important finding is that offer spreads are positively related to maturity, which suggests that if performance improves and market value increases over time, market expects REITs to sell more debt so that, ex ante, offer spreads are higher for longer maturity issues. The authors further report that REITs use public issue proceeds primarily for investment purposes so that security sales often induce capital structure changes. These findings are consistent with the existence of a target leverage ratio as suggested by the trade off theory. However, testing the theories of capital structure requires that a time series analysis of financing decisions be undertaken. In the short run, trade off and pecking order (market timing) theories make similar predictions for firms that are underleveraged (overleveraged). Long run analysis facilitates separating the alternative theories. The paper proceeds as follows. Section II summarizes the prior research on capital structure. Section III develops the hypotheses under the regulatory environment of REITs. Section IV describes the data. We present and discuss our models and results in section V, and conclude in section VI. II. Literature Review A. Tradeoff Theory Tradeoff theory posits that the firm has a target debt ratio which is determined by the tradeoff between the costs and benefits of borrowing, with the firm s assets and investment plans held constant. The most significant benefit of debt financing is the tax shield of interest 5

7 payments. Mandatory interest payment reduces free cash flow which mitigates the agency conflict between securityholders and managers. This implies higher leverage and payout ratios for profitable firms, and the opposite for firms with more investments. The cost of financial distress is the major downside of debt financing. Further, from the shareholders perspective, leverage may induce managers of struggling firms to avoid profitable investments because most of the benefit accrues to the debtholders. Marsh (1982) studies the security issuances by UK companies between 1959 and He documents that companies which are below their long term or above their short term debt targets are more likely to issue debt. Firm size, cost of financial distress and asset composition were the significant determinants of firm s leverage ratio. He also finds some evidence for the market timing theory. Specifically, the results demonstrate that firms with large share price increases tend to issue equity, and prevailing market conditions influence firms finance decisions. Titman and Wessels (1998) find that debt ratio is negatively related to the uniqueness of a firm s line of business, and interpret this result as supportive of tradeoff theory. They also report that transaction costs are an important determinant of leverage ratios, and past profitability tends to reduce a firm s debt level. The latter evidence is more in line with the pecking order theory. Rajan and Zingales (1995) use a sample of corporations from G-7 countries to investigate the capital structure choices across countries. They find some evidence consistent with tradeoff theory. For example, tangibility is positively correlated with leverage in all countries. Consistent with market timing, the market-to-book ratio has a significant and consistently negative relationship with leverage in all countries. Size is positively correlated with leverage and profitability is negatively correlated with leverage in all countries except Germany. These can be interpreted as generally consistent with the tradeoff theory. As the authors point out, however, a deeper examination of the evidence suggests that the current capital structure models fail to fully explain the observed patterns. Under the tradeoff theory, deviations from target capital structure are only temporary. In a dynamic setting, firms make financing choices to adjust the debt ratio to the long-term optimum which implies that no systematic relation between debt ratio and the firm s investment opportunities is predicted. However, if costs of financial distress varies across firms, a cross sectional variation in optimum capital structure is expected. For example, high-growth firms that are more sensitive to fluctuations in business outlook and are therefore more vulnerable due to the costs of financial distress, choose to use less debt financing. Highly profitable firms, on the other hand, can risk higher debt ratios. The findings in Smith and Watts (1992) and Barclay, Morellec, and Smith (2001) are consistent with this notion. 6

8 B. Pecking Order Theory Developed by Mayers (1984) and Mayers and Majluf (1984), the pecking order assumes that managers have privileged information regarding the firm s value that investors do not have. This raises the potential that opportunistic managers will sell equity only when it is overvalued. New shareholders will therefore avoid or discount equity which implies that only poorly performing firms will have the incentive to issue equity. The avoidance of adverse selection cost is the main motivation for firms to prefer the safest security available, which means that firms always choose debt over equity if bankruptcy costs are not an immediate concern. Hence, high growth firms that need more external capital end up with high leverage ratio. The dynamic pecking order theory, however, predicts that, holding profitability constant, firms with more investment opportunities keep payout low to conserve funds, and maintain low leverage to preserve debt capacity so as not to be forced into high debt in the future. These firms are forced to have high leverage only if adjustments in dividend payout are difficult, and investment commitments are persistently large. On the other hand, holding investments fixed, more profitable firms have higher payout ratios and lower leverage ratios because they have larger cash reserves, and can withstand adversities better. Shyam-Sunder and Myers (1999) study a sample of mature corporations with continuous data on flow of funds between 1971 and They find that pecking order theory has much greater time-series explanatory power than a static tradeoff model. They conclude that pecking order is an excellent first-order descriptor of corporate financing behavior. Clearly, if companies have well-defined optimal debt ratios, managers are not much interested in getting there. One criticism of the Shyam-Sunder and Myers (1999) study is that the inferences are based on a rather small sample. Fama and French (2002) present a comprehensive analysis of the complementary and contrasting implications of the tradeoff and pecking order theories for both dividend payout and leverage ratios. They identify profitability and investments and the interaction thereof as the key determinants of financing and dividend decisions. Consistent with both trade off and dynamic pecking order theories, they find firms with more investments are less levered. Next, their finding that more profitable firms have less leverage supports the pecking order, and contradicts the trade off story. Further support for the pecking order theory draws from the evidence that for dividend paying firms, short-term variation in investment and earnings is mostly absorbed by debt. Finally, the authors report that the least-levered and non-dividend paying firms (typically 7

9 small, growth firms) make the largest net new issues of equity which is contrary to the pecking order theory. Among other authors to report evidence inconsistent with pecking order theory are Helwege and Liang (1996) and Frank and Goyal (2002). Using a panel of IPO firms, Helwege and Liang (1996) find no relationship between the decision to raise external funds and the shortfall of internally generated funds. Studying the financial activities of US firms from 1971 to 1988, Frank and Goyal (2002) conclude that new equity issues track the financing deficit more closely than debt issues, a clear contradiction of the pecking order model. C. Market Timing Theory Market Timing Theory suggests that firms tend to issue stock when the market condition is favorable, and issue debt when the stock market is under the cloud. Graham and Harvey (2001) report that most CFOs agree that prior stock price movement and perception of under- or over-valuation of firms stock play important roles in their decision to raise external funds. Assuming that the ratio of market value to book value reflects investment opportunities, the market timing theory [Baker and Wurgler (2002)] asserts a negative relation between market value to book value ratio and the firm s leverage ratio. This is contradictory to the simple form of the pecking order theory, but consistent with the more dynamic form. Baker and Wurgler (2002) demonstrate that leverage is negatively related to external finance weighted-average market-tobook ratio which implies that past market valuation has a significant and persistently negative impact on firm s leverage ratio. Their data further reveal that most of the financing is done by issuing equity, not through retained earnings. The authors reject the trade off and pecking order models and interpret the result as supportive of the notion that current leverage ratio is a cumulative outcome of firm s previous attempts to time the market. In summary, while the three theories have several overlapping implications, they also make some predictions that can be useful to infer which one best fits observed capital structure choices. We highlight the aspects of each theory that is unique. The trade off theory predicts a target capital structure that firms regress to in the long run, implying that any relation between capital structure and profitability or investments is transient. Neither the pecking order theory nor the market timing theory identifies an optimum capital structure. For dividend-paying (non dividend-paying) firms, the pecking order theory predicts a long run positive (negative) relation between market to book value ratio and leverage ratio. The market timing theory leads to a long run positive relation between market to book value ratio and leverage ratio for all classes of firms; 8

10 the difference between the two is that under pecking order, funds are drawn from retained earnings while for market timing, equity sales is the source for capital. III. REIT Regulatory Environment and Capital Structure In this section, we explore the implications of the various theories of capital structure from the perspective of REIT regulatory environment, and develop the hypotheses. In addition, we provide a review of the extant evidence on capital structure of REITs. A. Theoretical considerations REITs are not required to pay corporate taxes if they distribute 95% of taxable income as dividends. This nullifies two significant benefits of debt financing. One, the tax deductibility of interest payments and the tax shield is non-existent. Second, since most of the earnings is distributed, debt servicing has only limited impact on agency cost of free cash flow. Accordingly, REITs should have one hundred percent equity under the trade off theory. Costs of financial distress further reinforce the preference for equity. The only effect that induces less than all equity capital is that asymmetric information between shareholders and managers causes valuation discounts. In the aggregate, if REITs have an optimum capital structure, it includes relatively low level of debt. The main motivation to prefer debt over equity issues is that managers may use privileged information to sell overvalued equity and shareholders are aware of it. So, an equity issue is always discounted by the market. Greater the information asymmetry, higher is the discount. Information asymmetry is particularly severe in REITs because the transparency of the underlying assets is less than perfect. For example, analysis of REIT assets may require special skills and knowledge about general and local economic trends, conditions of comparable properties, and complex financing arrangements. In addition, shareholders may find it difficult to determine the fair market values of real estate transactions because they often include heterogeneous, and illiquid assets. Restrictions on REIT s income sources and investment options may further exacerbate the information asymmetry. The restrictions that REITs derive their income largely from real estate activities, and that acquisitions and combinations be restricted to the real estate sector, allow managers but limited opportunity to acquire inter industry skills, makes them less employable, and induces them to avoid hostile takeovers [Campbell, Ghosh, and Sirmans]. The 9

11 requirement that no single investor owns more than 10 percent of REIT shares deters blockholder formation. In conjunction, these regulations make managers less vulnerable to the discipline of the takeover market, and render the board weak. Weak monitoring allows opportunistic managers to reveal less information. Under this scenario, REITs would be expected to avoid equity issues and prefer funding investment from retained earnings first, then sell debt if more capital is needed. A more complex form of pecking order, Myers (1984) notes, states that firms with generous reserves of cash may avoid issuing debt to preserve debt capacity for future capital needs, implying a negative long-term relation between leverage and market to book value ratio. REITs, however, are not expected to have big accumulation of cash and retained capital because of the payout requirement, which implies that they have to resort to debt financing more often. Thus, the pecking order theory predicts a long term positive relation between leverage and market to book ratio. Why would REITs issue equity despite the potential for high adverse selection costs? One reason, Baker and Wurgler (2003) assert, is that adverse selection costs vary over time and across firms, and managers take advantage of these opportunities to favorably time equity issues. Opportunities for timing equity sales also arise as irrational investors periodically bid up share prices to abnormally high levels. Because of the reasons narrated above, the opportunities of low adverse selection costs may be relatively infrequent for REITs. Under this premise, the market timing theory prediction of a long-term negative (positive) relation between market to book value ratio and leverage (equity issues) is questionable for REITs. The theories and hypotheses are summarized in table 1. Trade off theory predicts a low long term target debt ratio. The simple pecking order theory implies a positive relation between debt ratio and market value to book value ratio. The complex pecking order theory which implies a negative relation between market to book ratio and leverage for cash rich firms may not hold for REITs. The market timing theory predicts a negative relation between market value to book value ratio and leverage and requires opportunities when adverse selection costs are low, a less likely event for REITs. B. Empirical Evidence The first attempt to analyze capital structure of REITs from the perspective of valuation was made by Howe and Shilling (1988) who state that, under the trade off theory, as a non-taxpaying enterprise, the tax gain to corporate borrowing is strictly negative for REITs, such that a negative reaction to debt issues is predicted. A negative reaction is consistent also with the 10

12 pecking order implication that issuing a security constitutes a negative signal that it is overvalued, and the implied-cash-flow change hypothesis [Smith (1986)] which states that unexpected security offerings suggest that operating cash flows are lower than expected. A positive reaction to debt sales follows only from Ross s (1977) assertion that debt issues convey the favorable information that future earnings will be sufficiently large to support the mandatory interest payments. Extant literature [Mikkelson and Partch (1986), and Eckbo (1986)] documents nonpositive to significantly negative reaction to debt offerings. Contrary to these studies, Howe and Shilling (1988) find a significant positive reaction, which they interpret as weak support for Ross s signaling hypothesis. In an important and comprehensive piece, Brown and Riddiough (2003) study the public offerings by equity REITs between September 1993 and March 1998, and identify numerous patterns in the issuance behavior. While the scope of the research seems limited to identifying some stylized facts about REIT capital structure, certain results have bearing on our analyses. A significant finding is that maturity of public REIT debt is positively related to offer spread. The authors point out that if credit market participants assess that REITs issue debt when they are aggressively leveraged, and if they anticipate that REIT balance sheets will strengthen in the future, then credit spreads should decline with maturity. On the other hand, if REITs issue public debt at long-term target leverage ratios, then credit spreads are predicted to increase with maturity. The evidence therefore suggests the existence of a long-term target debt ratio. Two, the authors report that majority of the firms are clustered just above the investmentgrade rating, and REITs that issue public debt are debt capital constrained. While this result also suggests a target long-run debt ratio, an alternative explanation consistent with pecking order theory -- is that as long as REITs can attain minimum investment-grade credit rating, they prefer to issue debt instead of equity to boost their credit ratings. Further, a significant number of REITs that issue equity are highly leveraged and remain so subsequently. Apparently, firms issue equity only when bankruptcy threat looms large, and even at this juncture, they raise just enough equity capital to mitigate the funding pressure. Finally, REITs with higher total assets and revenues are more likely to issue debt, another indication that when bankruptcy risk is low, managers choose debt financing, just as pecking order prescribes. Also in conformity with the pecking order model, REITs largely fund investment with bank lines of credit and other sources of private debt. When these sources are exhausted, REITs access the public capital market and use the issue proceeds to pay down credit lines in order to prepare for the next round of financing. Overall, Brown and Riddiough s (2003) data suggest that despite no obvious tax advantage, the standard deadweight costs of financial distress, and the pecking order and free 11

13 cash flow rationales being muted by the dividend payout requirements 2, REITs prefer issuing debt and choose equity only as a last recourse. Howe and Shilling s (1988) analysis demonstrates that the market approves of this choice. While this is powerful evidence, it is based on static analysis. To our knowledge, the evolution of capital structure over time has not been explored for REITs. Our paper fills the gap. IV. Data and Summary Statistics Our study includes REITs that went IPO during 1991 to 2003 and for which all accounting and firm specific information required for analyses from 1991 to 2003 are available in the SNL database. We collect each firm s financial information, including total debt, total equity, total assets, total revenue, net income, depreciation, dividend amount, total investment in real estate, stock price, and the total number of shares outstanding. We also identify the IPO date for each firm during 1991 to Table 2 shows the number of REITs in SNL database that went IPO between , and number of REITs in the final sample by IPO year and calendar year. Most of REITs have accounting and financing information one year prior to the year of interest, and hence are included in the analysis to investigate the temporary impact of market-to-book on leverage ratio. However, missing values reduce the sample size when we test the long-term relationship between market-to-book and leverage ratios. This limits the scope and interpretation of our results somewhat. As shown in panel A of table 2, the number of REITs in the sample is only 4 in The IPO activity picks up between 1994 and 1996 when the sample size jumps to 83 and then stabilizes at 108 in That most firms survived during the entire period under study is apparent in the low attrition reported in panel B. By the tenth year after IPO, as many as 30 of the original 33 REITs remained in the sample. However, as evident in column 2, availability of data is very limited for young IPO firms during the early years of our study. For example, only 68 of the eligible 107 firms have data in the first year after IPO. The proportion is much higher as the firms mature. Leverage is measured as the ratio of book value of debt to book value of assets. As Myers (1977) points out, market values incorporate the value of the call option on firm s future growth opportunities. Debt issued against these values can distort future real investment decisions. As a result, in practice, managers have a good reason to calculate debt ratios using 2 Free cash flow rationale of debt financing states that mandatory interest payment on debt mitigates the agency cost of free cash flow. This is muted for REITs by the regulatory requirement to payout 95% of taxable income as dividends. 12

14 book values. Additionally, we investigate the impact of investment opportunities on leverage ratios. Titman and Wessels (1988) and Fama and French (2002) suggest that a negative relationship between market leverage and investment opportunities may simply be a manifestation of better investment producing higher market values, rather than the workings of trade-off and pecking order models. The definition and measurement of key variables follows Baker and Wurgler (2002). Book debt is total assets minus book equity. Book equity is defined as total assets less total liabilities and preferred stock plus deferred taxes and convertible debt. Book leverage is calculated as the ratio of book debt to total assets (D/A). Market leverage is book debt divided by total assets minus book equity plus market equity. Market equity is the product of number of shares outstanding and the stock price. Net equity issues (e/a) is defined as the change in book equity minus the change in retained earnings divided by assets. Newly retained earnings ( RE/A) is the ratio of net income minus dividend to total assets. We calculate the net debt issued (d/a) as the residual change in assets divided by total assets. Market-to-book value ratio (M/B) is defined as total assets minus book equity plus market equity divided by total assets. In table 3, we report the summary statistics of REITs leverage ratios and their financing by IPO year in panel A and by calendar year in panel B. Three patterns are worth noting. First, REITs have relatively high book leverage compared to non-reit firms in the compustat data base studied by Baker and Wurgler (2002). During 1991 to 2003, REITs maintain a debt ratio of above 50 percent. More recently, the book debt ratio is well above 60 percent. In contrast, non- REIT firms from 1974 to 1999 have an average debt ratio below 50 percent. Analyses over calendar years reveal the same pattern. Higher leverage ratio for REITs is consistent with the pecking order theory, but contrary to tradeoff and market timing models. Tradeoff theory predicts lower book leverage for REITs due to the tax exempt status and lower free cash flow problem. The business nature of REITs makes it harder for their shareholders to discover the market values of investment transactions, which usually involves a wide range of heterogeneous, illiquid assets. According to the pecking order model, firms with high asymmetric information tend to resort to debt when they need external funds, and are more likely to have high leverage ratios. Market timing certainly provides no rationale for this phenomenon. If REITs behave like other firms in choosing the source of external capital, and raise equity under favorable market conditions and debt under unfavorable market conditions, it is hard to reconcile why REITs, on average, have higher leverage ratios relative to other types of firms. 13

15 Second, we observe an increasing trend in debt ratio as the maturity of REIT firm increases. This trend is consistent with table 2 in Baker and Wurgler (2002). The average debt ratio is 52 percent one year after IPO and steadily grows to 66 percent ten years later. If firms have a target capital structure in mind, it is difficult to reconcile why debt ratio is growing continuously over the years. The average debt ratio also increases over the calendar years. This trend can be explained only as an age effect, not a survival effect. Most REITs in 2003 have a trading history of at least 4 to 5 years. This pattern contradicts the reversion to target capital structure over time prescribed by the trade off theory. Finally, REITs issue more debt than equity in nine out of ten years after IPO (seven out of ten years based on calendar year). Although the percentage of net debt issued is decreasing over the years, it is the driving force in the annual change in total assets. Consistent with the prediction of pecking order theory, this result suggests that REIT managers turn to debt financing first, before they consider equity financing. On the other hand, both tradeoff theory and market timing theories predict firms depend on debt financing in some years and on equity financing in other years depending on leverage ratio and cost of adverse selection cost. In table 4, we report the correlation between various variables. Most interesting is the persistently positive relation between book leverage ratio and market-to-book ratio over the past ten years. The 9-year back market-to-book ratio is still positively associated with the current book leverage ratio. The univariate analysis thus demonstrates a persistent and positive long-term impact of market-to-book on leverage ratio, which constitutes strong evidence against trade off and market timing theories, and strong support for the pecking order story. V. Models and Empirical Results Implications of trade off, pecking order, and market timing theories are usually expressed in terms of how leverage ratio varies with profitability and investment opportunities. We use market-to-book value ratio as a proxy for investment opportunities. Brown and Riddiough s (2003) finding that the market is more sympathetic to financing for investment purposes than adjustments in capital structure provides some justification for the use of M/B ratio as a determinant of debt ratio. Under trade off theory, bankruptcy costs are lower, and leverage higher for more profitable firms. To minimize agency cost of free cash flow, profitable firms use higher leverage to disgorge more of firm s excess cash. Conversely, firms with more investment opportunities have less free cash flow and can have low leverage ratio. Pecking order theory asserts that to avoid sale of discounted equity, firms fund investment with retained earnings first, 14

16 then debt, and finally equity. It follows that if profitability and investments are persistent, leverage is lower for profitable firms, and higher for firms with more investment opportunities. Dividend payment reinforces the relationship. Market timing theory predicts that managers time equity issues when equity is overvalued. If investment opportunities are persistent, a long-term negative relation between market to book ratio and leverage ratio is predicted. A. The relationship between market-to-book and leverage ratios In this section, we investigate the determinants of annual changes in leverage. Following Baker and Wurgler (2002) and Rajan and Zingales (1995), we include three variables that are correlated with leverage. D D A A t M = a + b B + c REinvestment A EBITDA + d A t 1 t 1 t 1 t 1 t 1 D + e log( S ) t 1 + f + ut (1) A The sign of coefficient b is the main focus of this equation. Both tradeoff and market timing theory predict a negative sign, while pecking order suggests the opposite. A more complicated version of pecking order asserts that firms with larger expected dividends may keep current leverage low to preserve debt capacity so as to avoid funding future investments with new risky securities [Myers (1984)]. For REITs, however, such a strategy may not be feasible because of the 95% payout requirement. We use percentage of real estate investment as proxy for asset tangibility in equation (1). Tangible assets may be used as collateral and hence may be associated with higher leverage. However, REITs are expected to have most of the assets as tangible assets, such that much variability is not expected in the data. Hence, we do not expect a relationship between tangible assets and leverage ratios. Profitability is associated with the availability of internal cash flows, which implies lower leverage ratio under the pecking order theory. However, REITs are required to pay out 95 percent of the earnings as dividends. Hence, there is limited free cash flow and a significant relationship between profitability and leverage may not emerge. The natural logarithm of total revenue is used as a proxy for firm size. As large firms are less likely to suffer financial distress, they might be associated with high leverage if the financial distress costs are considered to be of first-order importance to the firms (as tradeoff theory suggests). Fama and 15

17 French (2002) argue that larger firms may have less volatile earnings which will also induce a higher leverage ratio. Therefore, in accordance with trade off theory, we expect the coefficient of firm size to have a positive sign. Finally, we add the lagged leverage in our model, as Baker and Wurgler (2002) suggest, to control for the fact that debt ratio is bounded from 0 to 1. It is expected to have a negative sign 3. Panel A of table 5 reports the regression results from equation (1). We run this model for each year after IPO (IPO regressions). This allows us to study the changes (if any) in capital structure as the firm matures. We also construct the full 1991 to 2003 SNL sample for comparison. The regression estimates for the sample with all firms contains multiple observations on the same firm from different calendar year. First, the negative and significant coefficient of real estate investment and non-significance of the coefficient of firm size do not support the implications of the tradeoff theory. The evidence on profitability is mixed. Second, the market-to-book ratio is significant only in three of the ten IPO years. When it is significant, the sign is positive. The market-to-book is positive and significant in the all firms regression which is contrary to market timing and the trade off theory, but consistent with the pecking order story. It is also interesting to note that the non-negative coefficient for M/B is not consistent with the Myers (1984) dynamic form of pecking order model which states that in anticipation of funding requirements for higher investments in the future, firms may preserve debt capacity by using retained earnings for current needs. While regular dividend paying firms with stable cash flow are most capable of following this strategy, REITs are at a disadvantage because of high payout requirements. In essence, the positive sign for M/B ratio allows us to unequivocally reject the trade off, market timing and the dynamic form of pecking order theory. The evidence, however, can be interpreted only as preliminary and weak support for the simple pecking order theory. Further confirmation for the model must await analysis of the different funding sources. B. Components of Leverage change Following Baker and Wurgler (2002), we further decompose the change in leverage into equity issues, retained earnings, and the residual change in leverage, which depends on the total growth in assets from the combination of equity issues, debt issues, and newly retained earnings. 3 This coefficient is negative in 7 of 10 IPO years, as well as in the All Firms regression. When it is positive, it is never significant in any level. 16

18 D D A A t e t REt = At At Et 1 1 At At 1 t 1 1 (2) This decomposition allows us to identify more specifically the driving force behind the leverage change. We use each component as our dependent variable and run the same regressions for each IPO year, as well as for all firms from 1991 to The results are presented in panels B, C and D of table 5. As Braker and Wurgler (2002) point out, the coefficients in panel A should equal the summation of the corresponding coefficients from panels B, C and D. The results in panel B indicate that market-to-book ratio has impact -- through net equity issues -- on changes in leverage as predicted by market timing. In five out of ten IPO regressions as well as in the regression for all firms, we find net equity issues help to reduce the leverage ratio when the market valuations of the firms are high. However, these effects are not significant in the other half of the IPO regressions. Also, consistent with our expectation, panel C shows only weakly significant impact of internally generated funds on REITs leverage ratio. In contrast, we find significant impact of market-to-book on changes in leverage through asset growth (panel D). In nine out of ten IPO regressions as well as the all firms regression, the coefficients of market-tobook are significant at the conventional levels. These coefficients are larger in terms of absolute value compared to those in panel B. In the all firms regression, the coefficient for market-to-book is in panel D compared to 8.66 in panel B. Hence, the dominant factor for change in leverage ratio is the growth of total assets. To put it in perspective simply, the growth in total assets is mainly funded by net debt issues. Comparing panels B and D, we conclude that debt funding is the major source for external financing for REIT firms. No other coefficient in the other columns of panels B, C and D is significant. As a result, we believe that market-to-book does have at least a temporary impact on leverage ratio. Firms with higher growth opportunities are more likely to fund their investment through external debt issuance. This statistically and economically significant impact is predicted by pecking order theory, but not by tradeoff or market timing models. C. Long-term impact of market-to-book Next, we investigate whether the impact of market-to-book on leverage ratio is persistent over time as predicted by pecking order and market timing or just temporary as suggested by tradeoff theory. We use the three variables that were reported to be correlated with leverage ratio by Rajan and Zingales (1995) as control variables. 17

19 D A t M = a + b B M + c B REinvestment + d A efwa, t 1 t 1 t 1 EBITDA e f log( S ) + u (3) + + t 1 A t 1 t where M B efwa, t 1 = t 1 s t 1 s= 0 e + ds. e + d r = 0 r r M B s (4) The focus of this analysis is the external finance weighted-average market-to-book ratio (eqn. 4) suggested by Baker and Wurgler (2002). We use this weighted average to capture the long term effects of market-to-book on the leverage ratio. This variable takes high values for firms that raise external finance when the market-to-book ratio is high and vice-versa. If firms tend to raise equity when the market-to-book is high reflecting greater investment opportunities, as predicted by the market timing story, then we expect a negative relationship between this variable and the leverage ratio. If firms tend to raise debt when the market-to-book is high, as predicted by pecking order theory, then we expect a positive coefficient for this variable. If the market-tobook has only temporary impact on leverage ratio as suggested by tradeoff theory, then we should not find the coefficient of this variable to be significant. Also, we follow Baker and Wurgler (2002) and allow the minimum weight to be zero in order to ensure the weights are increasing in the total amount of external finance raised in each period. The lagged value of market-to-book is included to control for the current cross-sectional variation in the level of market-to-book. What is left for the weighted market-to-book ratio is the residual influence of past, within-firm variation in market-to-book. In panel A of table 6, we use only the traditional control variables from Rajan and Zingales (1995). We find that current cross sectional market-to-book value has a positive impact on the leverage ratio in three of IPO regressions and the all firms regression. This result is consistent with those reported in previous tables. We also find that other control variables do not have a significant relationship with leverage ratios in most of our IPO regressions as well as the regression including all firms. As discussed previously, this finding is consistent with the unique regulatory characteristics of the REITs. In panel B, we add the weighted average market-to-book ratio in our models to test the long-term relationship between market-to-book and the leverage ratio. We do not include this 18