Back to the Beginning: Persistence and the Cross-Section of Corporate Capital Structure *

Transcription

1 Back to the Beginning: Persistence and the Cross-Section of Corporate Capital Structure * Michael L. Lemmon Eccles School of Business, University of Utah Michael R. Roberts The Wharton School, University of Pennsylvania Jaime F. Zender Leeds School of Business, University of Colorado at Boulder First Draft: February 14, 2005 Current Draft: November 16, 2005 Preliminary: Please do not quote without permission * We thank Franklin Allen, Mark Leary, Andrew Metrick, Roni Michaely, Vinay Nair, Ivo Welch, Bilge Yilmaz and seminar participants at Babson College, Cornell University, University of Colorado, University of Maryland, University of Pennsylvania, and the University of Western Ontario for helpful discussions. Roberts gratefully acknowledges financial support from a Rodney L White grant and an NYSE Research Fellowship. Lemmon: (801) , finmll@business.utah.edu; Roberts: (215) , mrrobert@wharton.upenn.edu; Zender: (303) , jaime.zender@colorado.edu.

2 Back to the Beginning: Persistence and the Cross-Section of Corporate Capital Structure Abstract: We examine the dynamic behavior of corporate capital structures in order to determine the extent to which leverage is persistent and the corresponding implications for various theories of capital structure. Our results identify two types of persistence in leverage: the persistent effect of shocks to leverage on future leverage and the persistence of differences in leverage across companies. We find that the latter effect is responsible for a majority of the variation in leverage and lies behind some of the conflicting evidence regarding the former effect. Our main finding is that, on average, firms that have high (low) leverage today tend to remain relatively high (low) levered for over twenty years. These persistent differences in leverage are driven by the presence of firm specific effects that are responsible for over 90% of the explained variation in capital structure, while previously identified determinants (e.g., size, market-to-book, industry effects, etc.) are responsible for approximately 6%. Further investigation reveals several insights beginning with the inability of recent theories predicated on equity market timing or stock price inertia to explain the observed cross-sectional variation. Additionally, differences in leverage persist back in time prior to firms' IPOs implying that the frictions driving much of the cross-sectional heterogeneity in capital structures are largely unaffected by a major change in the information environment, distribution of control and access to capital.

3 A fundamental challenge for corporate finance lies in understanding the determinants of capital structure. To this end, recent research has focused on the dynamic behavior of leverage ratios in order to distinguish among competing explanations for the observed heterogeneity in leverage. A key question arising from these investigations is: to what extent is leverage persistent? That is, how important is history in determining current capital structure? This question is important for two reasons. First, there is a lack of consensus in the literature over the answer. 1 Second, different answers to the persistence question have different implications for the determinants of capital structure. If leverage is not persistent, then one need only understand the current (and expected) costs and benefits associated with financial policy in order to understand the cross-sectional dispersion in capital structures. Such is the case in static tradeoff theories. However, if leverage is persistent then history plays a role in determining current capital structures that is increasing in the degree of persistence, and the relevance of theories predicated on optimizing behavior may come into question. More precisely, leverage persistence implies three possible explanations for capital structure. The first is that firms are indifferent toward an optimal capital structure because the costs of deviating from this optimum are negligible, or because there simply is no optimum. This explanation is behind the pecking order hypothesis of Myers and Majluf (1984), the market timing hypothesis of Baker and Wurgler (2002), and the equity price inertia hypothesis of Welch (2004). The second explanation suggests that market 1 Studies by Spies (1974), Jalilvand and Harris (1984), Flannery and Rangan (2005), Alti (2005), and Hovakimian (2005) suggest that firms respond relatively quickly to shocks perturbing their capital structures. On the other hand, studies by Taggart (1977), Shyam-Sunder and Myers (1999), Fama and French (2002), and Huang and Ritter (2005) suggest the opposite.

4 frictions (e.g., transaction costs) impede a continuous or immediate adjustment towards an optimal leverage. This explanation motivates dynamic tradeoff theories, such as those provided by Fischer, Heinkel, and Zechner (1989), Hennessy and Whited (2004), and Strebulaev (2004). 2 The final explanation lies in a possible deficiency in our empirical specifications, where leverage persistence is symptomatic of mismeasurement or an omitted determinant of leverage. In this case, the relevant question then becomes: what is the deficiency in our existing empirical specifications? In this paper, we investigate the importance of history for the cross-sectional distribution of leverage in order to distinguish among competing explanations. We do this by examining the evolution of capital structure and analyzing the corresponding implications of our results for various theories, as well as previous empirical findings. Our results, while shedding light on several issues, also present new challenges to our understanding of capital structure. We begin by uncovering a type of persistence in the data that is responsible the majority of variation in capital structure. Specifically, we find that differences in leverage across firms tend to be highly persistent. This result is illustrated in figure 1 which, despite showing significant convergence over time, illustrates that firms with relatively high (low) leverage at time t tend to maintain high (low) leverage for over 20 years. Moreover, these differences are statistically and economically large at all future dates and cannot be explained by differences in previously identified observable firm characteristics (e.g., size, profitability, market-to-book, industry, etc.) or survivorship issues. These results are supported by regression analysis identifying firms initial 2 In Fischer, Heinkel and Zechner and Strebulaev, adjustment costs preclude firms from adjusting their leverage continuously in response to shocks. In Hennessy and Whited s model persistence in leverage arises because firms get captured in different tax regimes. 2

5 leverage ratios as the single most important determinant of capital structure at almost any future date. Simply put, firms tend to maintain their relative rankings in terms of leverage ratios for a long time. While related to the traditional econometric notion of persistence, this finding is fundamentally different, both economically and statistically speaking. 3 What the above results suggest is that cross-sectional variation in capital structure is characterized by an important firm specific effect. How important? The adjusted R-square from a regression of leverage on firm fixed effects alone is almost 60%. This is in contrast to the adjusted R-squares from traditional leverage regressions consisting of previously identified determinants (e.g., size, market-to-book, profitability, industry fixed effects, etc.). Depending on the specification, these R-squares range from 16% to 30%. However, we then show that the explanatory power of existing specifications is, in fact, due largely to the correlation of the determinants with the omitted firm specific effect. When we incorporate firm fixed effects into the specification with existing determinants, our variance decomposition reveals that 92% of the explained sum of squares is attributable to the firm specific fixed effects. That is, most of the variation in capital structure that is captured by observable determinants is attributable to variation in firm specific means. There are several implications of these results. First, the disagreement between some earlier studies concerning the persistence of shocks is due in part to differences in the treatment of these firm specific effects. 4 Second, while the issue of intra-firm 3 The traditional econometric definition of persistence refers to the long lasting effect of shocks on future realizations of in this context - capital structure. 4 Studies by Shyam-Sunder and Myers (1999) and Fama and French (2002) estimate partial adjustment models ignoring the firm specific effect, while Flannery and Rangan (2004) incorporate firm specific effects. See Hsiao (2003) discusses the impact of these different specifications on estimates of the speed of adjustment. In a recent independent paper, Huang and Ritter (2005) consider some of the econometric 3

6 persistence is interesting in its own right, our results suggest that it is, perhaps, a secondorder concern because firm specific shocks appear to have a relatively small impact on the cross-sectional distribution of capital structure. Third, the persistent effect of firm characteristics (market-to-book) or firm activities (net issuing activity) found in previous studies is due largely to their correlation with the firm specific effect. Appropriately controlling for this effect drives out any lasting impact these variables might have had and is consistent with the variance decomposition discussed above. Finally, and perhaps most importantly, in order to understand cross sectional variation in capital structure, one must understand the economic mechanism behind this firm specific effect. To this end, we examine recently proposed explanations for cross-sectional variation in capital structures, as well as undertaking an exploratory investigation into this issue. We begin by considering the market timing hypothesis of Baker and Wurgler (2002), who suggest that capital structure reflects the cumulative outcome of historical efforts to time equity issuances with equity market conditions. As such, leverage at any point in time should be strongly linked to historical net equity issuing activity. 5 We find that independent of whether or not we account for the firm fixed effect, net equity issuing activity is at best only weakly related to future capital structure. 6 Thus, while market timing may be an element of the managerial decision process, it appears to be an unlikely explanation for the cross-sectional variation in capital structures. issues surrounding the estimation of dynamic leverage regressions and conclude that the persistence of within firm shocks is quite large. 5 Of course, this assumes that the motivation behind all issuing activity is to take advantage of security mispricing by investors, as opposed to other reasons (e.g., improved investment opportunities). 6 This result is consistent with recent empirical evidence in Alti (2005), Hovakimian (2005), and Leary and Roberts (2005a). 4

7 We then consider the claim by Welch (2004) that the primary known determinant of capital structure is equity returns since firms do little in the way of net issuing activity to counteract the effects of equity price shocks. Like market timing, equity returns should have a long lasting effect on future capital structures. We find this is not the case and, if anything, higher historical equity returns lead to higher future leverage ratios. 7 Additionally, while we find that most of the cross-sectional variation in leverage is largely attributable to a (unknown) firm specific effect, this result is different from Welch s claim that the motives for net issuing activity are largely a mystery. 8 However, our results are consistent with the spirit of Welch s message; namely, there is perhaps more to be learned about capital structure than what we currently understand. For further insight into what may (or may not) be behind the firm specific effect, we investigate how far back in time the firm effect persists by examining a sample of firms for which we have IPO information. The results in Figure 3 mimic those in Figure 1 and reveal an important insight. Leverage differences across firms persist back in time before the IPO. In other words, high (low) levered private firms remain so even after going public. This finding is particularly interesting for two reasons. First, it casts further suspicion on market timing and equity price inertia stories since these explanations are largely inapplicable for private firms. Second, the IPO represents a dramatic change in the information environment, the distribution of control, and the access to capital 7 This result is consistent with recent empirical evidence in Flannery and Rangan (2005) and Leary and Roberts (2005a). 8 What is unknown in our study is what determines the firm specific means. That is, why are high (low) mean leverage firms as such? At the same time, Leary and Roberts (2005b) suggest that we have a fairly good understanding of the motives behind security issue decisions. These authors construct a model of financial policy that accurately identifies over 70% of security issuance decisions using existing empirical determinants to do so. 5

8 markets. Thus, this finding suggests that the IPO, and its associated changes, do little to mitigate the frictions that lie behind the firm specific effect. Finally, although completely (and convincingly) identifying the economic motivations underlying the firm specific effects is beyond the scope of this paper, we close our analysis with a discussion of potential explanations aimed at motivating and partly directing future research towards achieving this goal. The remainder of the paper is organized as follows. The data and sample selection are discussed in Section 1, where we also present summary statistics and some preliminary analysis. In section 2, we examine and quantify the degree of persistence in capital structure. Section 3 presents a variance decomposition of leverage, and Section 4 examines the implications of persistence in capital structure for various theories. Section 4 investigates the implications of our findings for recent explanations of capital structure. Section 5 examines net issuing behavior and pre-ipo capital structure and discusses some possible explanations for our results. Section 6 concludes. 1. Data, Sample Selection and Preliminary Analysis Our primary sample consists of all nonfinancial firm-year observations in the intersection of the monthly CRSP and annual Compustat databases between 1971 and We further require that all firm-years have nonmissing data for book assets. All multivariate analysis implicitly assumes nonmissing data for the relevant variables. We require leverage both book and market to lie in the closed unit interval. We also set any market-to-book ratios in excess of 20 equal to missing. All other ratios are trimmed at the upper and lower 1-percentiles to mitigate the effect of outliers and eradicate errors 6

9 in the data. For some of our analysis, we impose the additional requirement of an identifiable IPO date. 9 The construction of all of our variables is detailed in Appendix A. Panel A of Table 1 presents summary statistics for all of our firms, as well as a subsample of survivors composed of firms that have at least 20 years worth of nonmissing book leverage data. The potential for survivorship bias in our analysis motivates our examination of this subsample in all subsequent analysis as a robustness check; however, because of space considerations and similar findings, we often suppress these results. A quick comparison between the samples reveals several unsurprising differences. Survivors tend to be larger, more profitable and have fewer growth opportunities (i.e., lower market-to-book) but more tangible assets relative to the general population. Interestingly, survivors tend to have higher leverage, both in terms of market and book measures. This may suggest that firm exits due to buyouts and acquisitions are potentially as important as those due to bankruptcy. Alternatively, it may be an artifact of confounding effects survivor firms are larger and larger firms tend to have higher leverage (Titman and Wessels (1988)). Ultimately, we merely note that these summary statistics are broadly consistent with intuition and enable a straightforward comparison with previous capital structure studies to ensure consistency. 2. The Evolution of Leverage 2.1 Event Time Evolution To quantify the hypothesized persistence in leverage, we begin by studying the evolution of leverage for our cross-section of firms. Figure 1 presents the average 9 The IPO information is obtained from SDC and Jay Ritter, whom we kindly thank. Additionally, we thank Malcolm Baker and Jeffrey Wurgler for providing the list of IPO firms used in their study (Baker and Wurgler (2002)). 7

10 leverage of four portfolios in event time. The figure is constructed in the following manner. Each year we rank firms according to their leverage ratios into quartiles (i.e., four portfolios) which we denote: Very High, High, Medium, and Low. This portfolio formation period is denoted event year 0. We then compute the average leverage for each portfolio in each of the subsequent twenty years holding the portfolios constant. 10 We repeat these two steps of sorting and averaging for every year in the sample period. This process generates 33 sets of event time averages, one for each calendar year in our sample. We then compute the average leverage to each portfolio across the 33 sets within each event year. We perform this exercise for both book leverage and market leverage, the results of which are presented as bold lines with different tick marks in Panels A and C. The light, dashed lines surrounding these portfolio averages correspond to a 95% confidence interval. 11 Several features of the graphs are worth noting. First, there is a great deal of cross-sectional dispersion in the initial portfolio formation period. The range of average book (market) leverage is 56% (62%). Second, there is noticeable convergence among the four portfolio averages over time. After twenty years, the Very High book leverage portfolio has declined from 60% to 36%, whereas the Low portfolio has increased from 3% to 21%. (We see a similar pattern for the market leverage portfolios.) However, despite this convergence, the average leverage across the portfolios twenty years later remains significantly different, both statistically and economically. The average book 10 Of course, because of firm exit the portfolio composition will inevitably change over time, which raises concerns over survivorship bias. We address this below. 11 The confidence interval is defined as a two-standard error interval around the estimated mean. The standard error is estimated as the average standard error across the 33 sets of averages. Estimating the standard error using the standard error of the average across the 33 sets would greatly underestimate the true standard error because of the overlapping observations. Thus, we choose to use a conservative estimate that ignores the effects of averaging across the 33 sets, effectively treating each set as redundant. 8

11 leverage ratios in the Very High, High, Medium, and Low portfolios after twenty years are 36%, 31%, 27%, and 21%, respectively. This implies an average differential of 5%, which, when compared to the average within firm unconditional standard deviation (14%), is economically large. Therefore, a preliminary examination of leverage ratios suggests leverage differences are highly persistent. One potential concern with this analysis is survivorship biases. First, as we progress further away from the portfolio formation period, firms will naturally drop out of the sample due to exit through bankruptcy, acquisitions, and buyouts, for example. Second, from 1984 onward, the length of time for which we can follow each portfolio is censored because we only have data through To address these issues, we repeat the analysis described above for a subsample of firms that have at least 20 years of nonmissing data for book (or market) leverage. We refer to this subsample as Survivors. The results for this subset of firms are presented in Panels B and D of Figure 1, which reveal negligible differences between the survivors and the general population in terms of the dynamic behavior of leverage. A second potential concern with the results in Figure 1 is that the sorting of firms by leverage may simply be capturing cross-sectional variation in some underlying factor(s) that is associated with cross-sectional variation in leverage (e.g., bankruptcy costs, agency costs, etc.). For example, previous research (e.g., Titman and Wessels (1988)) has shown that leverage is positively correlated with firm size, so that members of the Very High portfolio may simply correspond to large firms, while members of the Low portfolio correspond to small firms. To address this possibility, we modify the portfolio allocation procedure. Specifically, each calendar year we begin by estimating a 9

12 cross-sectional regression of leverage on one-year lagged factors that have been previously identified by the literature as being relevant determinants of capital structure (e.g., Titman and Wessels (1988), Rajan and Zingales (1995), Baker and Wurgler (2002), Fama and French (2002), Phillips and Mackay (2004) and others). 12 Specifically, we regress leverage on firm size, profitability, tangibility, market-to-book, and industry indicator variables (Fama and French 38). 13 We now sort firms into four portfolios based on the residuals from this regression, what we will term unexpected leverage, and then track the average actual leverage of each portfolio over the subsequent twenty years. An attractive feature of this approach is that it allows for a transparent analysis examining the four portfolios while simultaneously controlling for factors known to be correlated with leverage. Additionally, by running the regressions each year, we allow the marginal effect of each factor to vary over time. To the extent that the factors included in the regression capture the cross-sectional heterogeneity in capital structure, the expectation is for the average leverage levels across portfolios to rapidly converge. This is not the case. Figure 2 presents the graphs for the unexpected leverage portfolios and shows that the results are quite similar to those presented in Figure 1. In particular, leverage still varies over a large range in the portfolio formation period, suggesting that the most of the variation in capital structure is found in the residual of existing specifications. (We return to this issue below.) As time progresses, we see similar patterns of convergence across the portfolios. And, finally, 12 We also examined the effects of using contemporaneous determinants. The results are imperceptibly different. 13 We also examined an alternative specification suggested by Frank and Goyal (2004) consisting of firm size, market-to-book, collateral, intangible assets, an indicator for whether the firm paid a dividend, and year and industry indicators. In addition, we included a measure of expected cash flow volatility computed as the quarterly standard deviation of income before extraordinary items during the 10 years following the IPO. The results are largely unchanged by these modifications and, as such, are not presented. 10

13 while the spread in average leverage across the portfolios in each event year has decreased, there still remain significant differences for most periods. For example, even 20 years after the portfolio formation period, the average leverage of Low levered firms is significantly below that of all other portfolios, both in terms of book and market leverage. Additionally, the average leverage of Very High levered firms is significantly different from that of Medium levered firms. And, again, these differences are economically significant as well, with the range in leverage across the portfolios in event year 20 equal to 10% (13%) for book (market) leverage. Thus, even after removing all observable heterogeneity associated with traditional determinants of capital structure, leverage differences are still highly persistent. Before continuing, it is worth distinguishing the persistence that we are observing in the figures from that traditionally referred to in the econometrics literature, as well as previous capital structure studies (e.g., Shyam-Sunder and Myers (1999), Fama and French (2002), Kayhan and Titman (2004)). The figures suggest that firms tend to maintain their leverage ratios in relatively confined regions over long periods of time, on average. This is distinct from the traditional notion of persistence that shocks to firms leverage ratios have long lasting effects. The figures suggest that the effects of shocks to leverage are on average quite small when compared to the differences in firms average leverage ratios. This contention is supported by the fact that the within firm standard deviation of book (market) leverage is 14% (16%), while the between firm standard deviation is 20% (23%). Thus, variation across firms means appears to dominate variation within firms over time, a result we will return to below. 11

14 2.2 The Effect of Initial Leverage An alternative approach to identify the effect revealed by the figures is to examine this persistence in a regression setting. However, unlike traditional approaches that estimate partial adjustment models incorporating lagged leverage as an explanatory variable, we include the firm s initial leverage, which we proxy for with the first nonmissing value for leverage. Econometrically, the model that we estimate is: Leverage it = α + βx it 1 + γleveragei 0 + ε it, (1) compared to partial adjustment models estimating Leverage it = α + βx it 1 + γleverageit k + ε it, (2) where i indexes firms, t indexes years, X is a set of control variables often assumed to be strictly exogenous or predetermined, and k is typically taken to be equal to The random error, ε, is assumed to be correlated within firm observations but independent across firms. 15 The distinction between the two specifications is that equation (1) fixes the lagged leverage over time within firms. The results from estimating equation (1) using book and market leverage are presented in Table 2. Panel A presents the results using the entire sample of firms, while Panel B presents the results using the subsample of firms required to survive for at least 20 years. Because the results are so similar, we focus on the results for the full sample. In order to ease coefficient comparisons, we standardize the right hand side variables to 14 Partial adjustment models are expressed as: yt yt = α + λ ( µ yt 1 ) + ε t 1, which is simply a reparameterization of the AR(1) specification presented in equation (1). 15 We incorporate year indicator variables to capture any common component in leverage shared by firms at a given point in time. 12

15 have zero mean and unit variance. 16 We present results from three specifications. The first restricts β to equal 0, so that leverage at time t is regressed only on initial leverage. The estimate suggests that a one standard deviation change in a firm s initial book leverage ratios corresponds with an 8% change in future values of book leverage. An even larger effect is found for market leverage, though the volatility of market leverage is greater than that for book leverage (refer to Table 1 and the earlier discussion of within and between firm standard deviations above). These results are consistent with the findings of Figure 1. We then relax the restriction on the beta coefficients and incorporate two sets of determinants into the specification. The first set consists of those variables suggested by Rajan and Zingales (1995) and subsequently used in many capital structure studies (e.g., Baker and Wurgler (2002), Frank and Goyal (2003), and Lemmon and Zender (2004)), augmented with calendar year fixed effects. The coefficient estimates are largely consistent with previous evidence, in terms of sign and statistical significance. Yet, initial leverage remains highly significant and reveals a small (economic) change from 0.08 to 0.07, in the case of book leverage, and a slightly larger change from 0.14 to 0.11 for market leverage. Despite the change, however, the statistical and economic magnitude of these effects is dramatic, particularly when compared to the marginal effect of the other determinants. These results are consistent with Figure 2, though they present a more stringent test of persistence in leverage differences since the determinants in these 16 Clearly, we are taking a bit of liberty with the statistical assumptions of the model by performing this transformation. In particular, we are inducing a small amount of correlation between the centered initial leverage variable (and the determinants, unless we assume strict exogeneity) and the error term. However, we believe that the potential bias that this transformation brings is more than offset by the clarity of the results, a claim we confirm in unreported results using the raw, untransformed data. 13

16 regressions update continuously through time, whereas the figure conditions on the determinants in the initial portfolio formation period. The final specification incorporates additional variables suggested by Frank and Goyal (2004), who perform an exhaustive analysis of capital structure determinants. Despite statistically significant marginal effects, the inclusion of the additional variables does little to eliminate the effect of firms initial capital structures on future values. The estimated coefficient is still highly significant and larger in magnitude than all other determinants but for the indicator of whether or not a firm paid a dividend. The results show that historical leverage is an important determinant of future leverage: Initially high (low) levered firms tend to remain so for many years. 2.3 The Probability of Transitioning across Leverage States One concern common to the analysis in Figures 1 and 2, as well as that in Table 2 concerns the masking of firm-level behavior by averaging. Specifically, the averaging inherent in the figures and regressions may potentially conceal offsetting firm level behavior. To address this possibility, we examine the probability of firms switching from one unexpected leverage portfolio to another through time. We do so by first sorting firms each year into four unexpected leverage portfolios (using the regression approach outlined earlier in the context of Figure 2). We then compute the fraction of firms in state i at time t-k that transition to state j at time t, for i,j in {Low, Medium, High, Very High} and for k in {1,5}. We view this analysis as a non-parametric robustness test of our earlier findings. 14

17 The results are presented in Table 3. Panel A (B) presents the results for one (five) year transitions. The large diagonal elements in both panels reveal a strong propensity to remain in a particular unexpected leverage portfolio through time. Specifically, we find that firms have a 50% probability of staying in the same unexpected leverage portfolio from year-to-year. Even over five year time spans, firms are 40% likely to remain in the same unexpected leverage portfolio. When firms do change states, it is almost always to an adjacent state. In unreported analysis, we find that for the firms that switch to adjacent states, their leverage is generally already close to the boundary so that the magnitude of the leverage change is commonly quite small. Only 5% (11-13%) of the time do firms move to a non-adjacent state after one (five) year. Thus, even at a disaggregated level, leverage differences appear persistent. 3. Variance Decomposition of Leverage The fact that leverage differences are so highly persistent is indicative of an important firm specific effect. That is, corporate leverage tends to vary around firm specific means, which also appear to be responsible for a substantial fraction of the crosssectional variation in leverage. This claim is supported by the spread across portfolios in Figures 1 and 2, as well as the regression evidence in Table 2. In this subsection, our goal is to investigate just how important this firm specific effect is and its implications for existing empirical evidence. To accomplish this, we perform an ANOVA or variance decomposition. We begin by specifying the following general model of leverage using notation identical to that found in equations (1) and (2) above: 15

18 Leverage = α + β 1 + η + ν + ε, (3) it X it i t it where η is a time invariant component and ν is a firm invariant component of leverage. The disturbances, ε, likely violate the assumptions underlying ANOVA: normal distribution, independent, and homoskedastic. However, our goal with this analysis is not hypothesis tests of the means. Rather, we only use the ANOVA technique as a tool for decomposing the variance of leverage. Panel A of Table 4 presents the results of the variance decomposition for several different specifications (i.e., parameter restrictions on equation (3)). However, because of the large number of firms (19,777) in our panel, performing the variance decomposition on the entire sample is not feasible because of memory limitations. As such, we randomly sample 10% of the firms in the panel and perform the analysis for this subsample. To minimize sampling error, we repeat the process of sampling and variance decomposition 100 times. We then average over all the results. Each column in the table corresponds to a different specification. The numbers in the body of the table - excluding the last row - correspond to the fraction of the total Type III partial sum of squares for a particular model. 17 That is, we divide the partial sum of squares for each effect by the aggregate partial sum of squares across all effects included in the model. This simply provides a normalization and forces the columns to sum to one. Intuitively, what the figures in the table correspond to is the fraction of the model sum of squares attributable to a particular effect (e.g., firm, year, size, market-to-book, etc.). When only one effect is included in the model, all of the model (or explained) sum of 17 We use Type III sum of squares for two reasons. First, type I sum of squares are sensitive to the ordering of the covariates because the computation involves sequentially projecting the dependent variable onto each variable. Second, our data is unbalanced in the sense that the number of observations corresponding to each effect is not the same (e.g., some firms have more observations than others). For an introduction to ANOVA, see Scheffe (1959). 16

19 squares is attributable to that effect. For example, when we examine only the firm effect (η i ) in column (a), all of the explained variation in leverage is attributable to that effect. Hence, it takes a value of 1.00, or 100%, and similarly for column (b) which examines the year effect in isolation. Comparing the adjusted R-squares for these two specifications, we see that firm specific effects capture almost 60% of the variation in capital structure, compared to 1% captured by the time effects. When we examine both firm and year effects together (column (c)), we see a negligible increase in the adjusted R-square from 57% to 58% and that almost all (99%) of the explained variation in capital structure is attributable to the firm effect while only 1% is captured by the year effect. We emphasize that throughout this (and all) analysis, we examine only adjusted R-squares to account for variation in the model degrees of freedom. Columns (d) and (f) present the results from specifications motivated by previous studies such as Rajan and Zingales (1995) and Frank and Goyal (2004). Column (d) shows that asset tangibility and industry effects account for most of the explanatory power in this specification. Interestingly, despite the relatively small explanatory power of the year effects (the adjusted R-square in column (b)), it captures 8% of the overall explained variance for this specification, even more than that captured by firm size. Column (f) identifies median industry leverage as the statistically most important determinant of capital structure, followed by measures of collateral and intangible assets. More relevant to the point of this exercise, we note that the adjusted R-squares are 16% and 30%, respectively. These are significantly lower, both statistically and practically speaking, than the firm fixed effect regression. 17

20 Columns (e) and (g) augment the previous two specifications with firm fixed effects. We note two features of these findings. First, the adjusted R-square more than triples when moving from specification (d) to (e) and more than doubles moving from (f) to (g). Second, and perhaps most striking, almost all of the explanatory power in these specifications, 96% and 92% respectively, is captured by the firm fixed effects. These findings confirm that not only is the persistence of leverage differences across firms an important firm specific effect but it is so important that it captures most of the explanatory power previously attributed to firm characteristics. That is, the importance of factors such as market-to-book and profitability, for example, is due largely to their correlation with an omitted firm specific effect. We note, however, that this is not to say that the marginal effects of these firm characteristics are insignificant. Panel B of Table 4 presents the coefficient estimates and t-statistics for models (d) through (g). 18 However, because we can simply transform the left and right hand sides of the regression by subtracting the firm specific means from each variable, we are no longer constrained by an excessively large design matrix when estimating the firm specific specifications. As such, Panel B presents the estimates using the entire sample, although similar results are obtained if we instead estimate the regressions on the 100 random samples and then average the results across the samples. The results illustrate that even in the firm fixed effect specifications and after accounting for heteroskedasticity and serial correlation, each determinant is highly statistically significant. However, with the exception of firm size, every coefficient estimate experiences a dramatic decline in magnitude moving to the fixed effect 18 The t-statistics in Panel B are computed using standard errors corrected for within firm dependence and heteroskedasticity. 18

21 specification. For example, moving from specification (d) to (e) we see that the coefficient on market-to-book falls by 74%, that on profitability falls by 62%, and the coefficient on tangibility falls by 64%. These differences are both statistically and economically significant. 19 Similarly, moving from (f) to (g), we see that all coefficients but for firm size fall by at least 47%. To summarize, corporate capital structures are characterized by an important firm specific component that is responsible for a majority of the total variation in leverage. Additionally, many of the previously identified empirical determinants are correlated with leverage, in large part, because of their correlation with this firm specific component. Once firm level means are removed from the model, the explanatory power of these determinants falls dramatically, though they still maintain statistically significant marginal effects. The question now, however, becomes what is the economic mechanism(s) behind the firm specific component? While a complete answer to this question is beyond the scope of this paper, we take several steps towards this goal in what follows. 4. Recent Explanations of the Cross-Sectional Distribution of Capital Structure In this section we examine two recent explanations for the cross-sectional distribution of capital structure. The first is the market timing hypothesis of Baker and Wurgler (2002), which argues that leverage is the cumulative result of historical attempts to time security issuances primarily equity to market conditions. Simply put, when 19 We perform a Hausman test of the null hypothesis that all of the coefficient estimates are the same across both specifications. 19

22 stock prices are high, firms issue equity to take advantage of the mispricing. 20 The second is the equity inertia hypothesis of Welch (2004), which suggests that the primary known determinant of capital structure is stock returns: Firms with historically high returns have relatively low market leverage, while firms with historically low returns have relatively high market leverage. The word known is in quotations because, as Welch suggests, most of the variation in leverage is attributable to net issuing activity but the motives for this activity are largely a mystery. Thus, what little we know about capital structure is that it varies mechanistically with equity returns. This last point raised by Welch is worth further discussion before turning back to the analysis. In particular, our results suggest that there is an important firm specific component in capital structure whose economic underpinnings have yet to be convincingly identified. While consistent with the spirit of Welch s (2004) message, there is a subtle yet important difference between the two. In particular, the motivation behind why firms issue and retire securities may, in fact, be very well understood despite our ignorance of what lies behind the firm specific component. Leary and Roberts (2005b) show that previously identified empirical determinants can accurately identify the debt and equity issuance decisions of over 70% of the firms in their sample. This suggests that why firms vary around their firm specific means may be reasonably well known. The current examination asks why firms have the mean leverage that they do. Returning to the hypotheses, a key implication of market timing and equity price inertia is that leverage in any period should be closely related to historical net equity issuing activity or stock returns, respectively. Table 5 tests these hypotheses by 20 Baker and Wurgler (2002) do not distinguish between mispricing due to irrational behavior on the part of investors or mispricing due to information asymmetry that may occur in a dynamic version of Myers and Majluf (1984). 20

23 regressing leverage at time t on lagged control variables used in the previous analysis, along with lagged net equity issuing activity and stock returns. Though we focus on results obtained using Frank and Goyal s (2004) Tier I factors, we obtain similar results using the controls suggested by Rajan and Zingales (1995). For comparison, we include specifications with and without firm fixed effects to further highlight the main message of the paper. The estimated coefficients on the control variables are generally similar to those found in Panel B Table 4. What we focus on are the coefficients on the variables measuring net equity issuance and equity returns. At the 1-year lag, net equity issuance and equity returns have statistically significant negative coefficients on both book and market leverage. Moving to deeper lags, this effect attenuates and in some cases changes sign: higher net equity issuances and returns lead to higher leverage. Interestingly, this result holds independent of the specification (pooled OLS or firm fixed effects). In unreported results, we use five year cumulative net equity issuances and returns, as opposed to one year values. The results are generally quite similar and reveal that both issuances and returns have negligible long run effects on corporate capital structure, consistent with other recent papers by Hovakimian (2004) and Alti (2004). These findings are, however, quite different from those reported by Huang and Ritter (2005), who find that equity issuances appear to have a lasting effect on capital structure (see their Table 5). In particular, they show that net equity issuances have a significant effect on leverage five and ten years hence. There are several reasons for the apparent discrepancy. First, Huang and Ritter use different measures of debt, equity, and 21

24 leverage. 21 However, their results appear to be somewhat sensitive to their definitions of leverage. After closely approximating their findings, we substitute our measure of leverage into their specification. Doing so results in a sign change on the net equity issuance variable to significantly positive in both book and market leverage specifications. Second, the largest marginal effect they find is an estimated coefficient of in the book leverage specification with net equity issuances lagged five years. While statistically significant this estimate implies that moving from the 25 th percentile to the 75 th percentile of net equity issuances (~15%) corresponds to a 1% change in leverage, an economically negligible effect. Thus, even after ignoring measurement concerns, the marginal effect of equity issuances on future leverage is small at best. In sum, neither net equity issuances nor equity returns appear to be particularly important determinants behind the cross-sectional distribution of capital structure. Therefore, market timing, while a potential element of the managerial decisions process, is an unlikely explanation for the observed cross-sectional distribution of capital structure. Similarly, while equity returns exhibit a short term mechanistic relation with market leverage, they have little subsequent impact on future capital structures. 5. What Lies Behind the Firm Specific Effect? At this point, a key question remains: what is the economic mechanism(s) driving the heterogeneity in the firm specific effects? A complete answer to this question is simply beyond the scope of this paper; however, in this section we attempt to make 21 Specifically, Huang and Ritter (2005) define book debt as total liabilities plus preferred stock minus deferred taxes and convertible debt. This measure thus considers current liabilities as part of the firm s debt. Book equity is defined as total assets minus book debt. Net equity as the change in book equity minus the change in retained earnings. Book leverage is book debt divided by book assets and market leverage is book debt divided by market assets. 22

25 strides toward this goal through further investigation and discussion of potential explanations. 5.1 Differences in Issuance Behavior We begin by exploring whether the large differences in leverage that we relate to firm specific fixed effects are also related to persistent differences in net issuing activity. To address this issue, we examine the issuance activity of the unexpected leverage portfolios described earlier. Each event year we compute the average net debt and net equity issued by each of these four portfolios. The results are presented in Figure 3, Panels A and B, respectively. To ease the presentation, we suppress the confidence intervals and instead focus on the economic magnitude of the differences. Focusing first on Panel A and net debt issuing activity, we find that initially, the tendency to issue debt differs quite dramatically across the portfolios. However, what is interesting is that the propensity to issue debt is monotonically negatively related to firms leverage. So, firms issue progressively more debt as we move from the Very High portfolio to the Low portfolio. These differences remain for three or four years before becoming largely indistinguishable. This finding is consistent with Leary and Roberts (2005a) and Hovakimian (2005), who suggest that an important motivation behind debt policy is capital structure rebalancing. Panel B, which displays net equity issuing activity presents a very different story. In particular, we find that Low levered firms are significantly more likely to issue equity and for some time. This result appears counter-intuitive for a rebalancing story. Moreover, the fact that the firms issuing the most equity already have low leverage 23

26 reinforces our earlier findings that much of the equity issuing activity we observe does not lead to large changes in capital structure. We also note that Very High levered firms appear to issue a significant amount of equity, on average, relative to the Medium and High portfolios. This finding is consistent with highly levered firms using equity to reduce their leverage. As for the Medium and High portfolios, there is little systematic difference between the two though, relative to the other two portfolios, these firms appear to use less equity on average. In sum, the results are generally consistent with the persistence in leverage that we document in the sense that issuing activity across the portfolios is done in a way that would not alter the relative leverage rankings. Put another way, the financial policies of the low and high levered firms keep them close to their long run mean leverage ratios on average. 5.2 How Far Back Does the Firm Specific Effect Go? We next explore how far back in time the leverage differences observed in Figures 1 and 2 persist? Specifically, we ask whether these differences exist before the IPO. To investigate this issue we form unexpected leverage portfolios at the time of the IPO. Because of a significant reduction in the number of observations (approximately 5,000 IPO firms), we modify slightly the analysis in Figure 2. We begin by computing initial leverage as the average of the first three public observations on leverage in event years 0 (year of the IPO), 1, and 2. We compute initial values of the corresponding determinants (size, market-to-book, profitability, and tangibility) in a similar manner. Averaging helps minimize noise and mitigate the effect of extreme observations in this 24