the profits-leverage puzzle revisited
TRANSCRIPT
Electronic copy available at: http://ssrn.com/abstract=1863629
The Profits-Leverage Puzzle Revisited∗
Murray Z. Frank† Vidhan K. Goyal‡
May 26, 2011
ABSTRACT
It is well known that in a leverage regression, profits are negatively related toleverage. The literature (e.g., Myers, 1993; Fama and French, 2002) considers thisto be a key rejection of the trade-off theory. We disagree. Contrary to Myers (1993),highly profitable firms typically issue debt and repurchase equity, while the lowestprofit firms tend to raise external funds – particularly equity. The typical issuanceis in the direction predicted by the trade-off. It is also true that more profitablefirms experience an increase in both the book value of equity and the market valueof equity. The effect of profits on equity drives the negative coefficient in the usualleverage regression, thus giving a misleading impression. Transaction costs may beimportant because we find that large firms make more active use of debt, whilesmall firms make more active use of equity. Furthermore, poor market conditionslead to reduced use of external finance. The impact is particularly strong on smalland low profit firms.
JEL classification: G32
Keywords: Capital structure, Trade-off theory, Profits, Agency theory, Leverageratios
∗We thank Raj Aggarwal, Mark Flannery, Fangjian Fu, Paul Povel, Jay Ritter, Philip Strahan, IlyaStrebulaev, Michael Roberts, as well as seminar audiences at the 2009 AFA, Universita Bocconi, BostonCollege, City University of Hong Kong, University of Florida, Imperial College, Korea University, Univer-sity of Minnesota, University of Pittsburgh, Oxford University, and Singapore Management Universityfor helpful comments. Murray Z. Frank thanks Piper Jaffray for financial support. Vidhan K. Goyalthanks the Research Grants Council of Hong Kong for financial support (Project #641608). We aloneare responsible for any errors. c© 2011 by Murray Z.Frank and Vidhan K. Goyal. All rights reserved.†Murray Z. Frank, Carlson School of Management, University of Minnesota, Minneapolis, MN 55455.
Tel.: (612) 625-5678. [email protected]‡Vidhan K. Goyal, Department of Finance, Hong Kong University of Science and Technology, Clear
Water Bay, Kowloon, Hong Kong. Tel.: +852 2358-7678. [email protected]
Electronic copy available at: http://ssrn.com/abstract=1863629
I. Introduction
The trade-off theory of capital structure predicts that more profitable firms ought
to borrow more and have higher leverage.1 The profits-leverage puzzle is the empirical
evidence that the predicted sign is backwards. “The most telling evidence against the
static trade-off theory is the strong inverse correlation between profitability and financial
leverage. Within an industry, the most profitable firms borrow less, the least profitable
borrow more” (Myers, 1993, page 6). This relationship between corporate profits and
leverage is widely regarded as a particularly serious defect of the trade-off theory (e.g.
Fama and French (2002)).
In this paper we revisit the profits-leverage relationship. We show that the standard
empirical methodology has interacted with the usual approach to simplifying the theory,
in a particularly misleading way. The empirical methodology has focused on leverage
ratios, but interpreted them as if they were the result of debt market actions. In fact,
the equity component is very important when considering the impact of profits on firm
capital structure empirically. The theory has focused on intuition derived from static
models with inadequate attention given to the dynamic impact of the form of transaction
costs.2
The empirical problem arises in the standard leverage regression. To see this, let D
denote corporate debt and E corporate equity, then leverage is L = DD+E
. Let xit denote a
factor such as profits, with the subscript i for each firm, and t for the date. Let εit denote
the error term, and α and β be the coefficients to be estimated. It is common to run a
panel regression such as Lit = α + βxit + εit. If xit is profits, then under the trade-off
theory a more profitable firm has a greater need to shield profits from taxation, and so
1The term ‘trade-off theory’ is used in different ways by different authors. For some authors it meansthat bankruptcy and taxes are being balanced (Kraus and Litzenberger, 1973). For other authors itincludes agency-based arguments (Fama and French, 2002). Some authors simplify by assuming thatinvestment is unaffected, even though the cost of capital is changed by the leverage choice. Other authorsactually analyze the impact of the leverage choice on investment. A recent review is provided by Frankand Goyal (2008).
2Our point is related to that in Leary and Roberts (2005). The difference is that we focus on profitsand we emphasize the importance of the simultaneous presence of fixed and variable transactions costs.
1
it is predicted that β > 0. Empirically, however, β < 0, and so the trade-off theory is
rejected. This rejection is the profits-leverage puzzle.
This rejection could arise if the trade-off theory is indeed wrong. This is the usual
interpretation. But the rejection could arise for a number of other reasons too. It could
come from debt or from equity. It could come from corporate inaction, action in the
wrong direction, or simply action that is in the right direction but not strong enough.
The standard interpretation is that the regression result is due to more profitable firms
borrowing less. We will show that this is actually incorrect.
We present a sequence of closely related analyses that decomposes the standard re-
gression. We start by replicating the standard leverage regression and showing that the
standard result continues to hold in more recent data. The standard result holds for both
book- and market-based definitions of leverage.
Since our goal is to trace the source of the ‘incorrect’ sign on profits, we begin with
simple sorts (firm size and profits) and descriptive statistics. In this way, we document
the frequency and magnitude of various capital structure rebalancing actions. The general
pattern is that more profitable firms tend to issue more debt and are much more likely to
repurchase equity. On the contrary, the lowest profit firms tend to retire debt and raise
more equity capital.
These basic patterns are very much in line with the traditional interpretation of the
static trade-off theory, but contradict the profits-leverage puzzle. So the next task is
to reconcile these findings with the standard leverage regression results. To that end
we examine the impacts of alternative conditioning factors, fixed firm effects, fixed year
effects, alternative standard error assumptions, the use of changes instead of levels, and
alternative normalizations of the issuing regressions.
We find that the defect is not with the theory, but with the use of scaled measures
of leverage in which profitability can affect both the numerator and the denominator
of the ratio. This makes the sign of the relationship between leverage and profitability
theoretically ambiguous.
The main findings are as given below.
2
1. When a firm makes extra profits, the book value of equity increases unless the firm
takes some sort of offsetting action. Similarly, when a firm makes extra profits,
unless there is some sort of offsetting action, the market value of the firm’s equity
increases. Thus more profitable firms will automatically have more book equity and
more market equity, unless the firm takes offsetting actions.
2. Among large firms: The highest profit firms increase their debt the most. Those
with high profits experience large increases in both the book and the market value
of equity. The highest profit firms tend to repurchase equity while the lowest profit
firms tend to issue more equity.
3. Among small firms: Profit seems to have only a very minor effect on debt. Those
with high profits experience some increases in both the book and the market value of
equity. Those with low profits experience negative effects on market equity. There
is a tendency to issue equity, with the lowest profit firms issuing the most equity.
4. Almost any optimizing model of an interior optimal capital structure will imply that
the use of debt and equity will vary as market conditions vary. Empirically there is
time variation in the corporate use of external financing. In good times, firms issue
more net debt and net equity than in bad times. In particular, issuing equity in bad
times is more of a problem than issuing equity in good times. The negative effect
of profits on equity issuing is much stronger in good times than in bad times.
5. When firms adjust leverage, the magnitude of the adjustment is not sufficient to
fully undo the impact of the underlying shocks. Firms do not return to a unique
static optimum. They seem to ‘underadjust’. In section VII we show that this is
exactly what ought to be expected in a trade-off model in which there are both fixed
and variable costs of adjustment. Full adjustment is costly so partial adjustment is
typical as the firm balances the costs and benefits at the margin.
There is a huge prior literature on our topic, and so we cannot review all related
studies. For a review of the literature see Frank and Goyal (2008). The fact that there is
3
an inverse relationship between profitability and leverage ratios has generated a variety
of responses from scholars.3
One response in the literature has been to argue that we should not consider static
models. The trade-off theory predictions can be more complex in a dynamic model such
as that of Fischer et al. (1989) or that of Strebulaev (2007).4 Empirically, the response
has been to argue that leverage and profitability are negatively related because firms
passively accumulate profits (see Kayhan and Titman, 2007).5 This implies that at the
time of rebalancing, leverage should be positively related to profitability. Mackie-Mason
(1990) shows that companies with tax loss carryforwards are more likely to issue equity.
Hovakimian et al. (2001) and Gomes and Phillips (2007) show that highly profitable firms
are indeed more likely to issue debt. This is consistent with what we find. By contrast,
Jung et al. (1996) report finding no relation between the likelihood of equity issuance and
profitability.
Welch (2004, 2007) makes the important point that changes in debt and equity values
and changes in debt ratios are conceptually different. This distinction also plays a role
in our analysis. Welch (2007) stresses the idea that non-financial liabilities should not
be implicitly mistreated as if they were equity by paying excessively narrow attention to
financial liabilities in a leverage ratio. We have adopted his approach in the empirical
work reported here.
This paper is organized as follows. Section II describes the construction of the data
and variables and provides summary statistics. Section III provides results from the fixed
effects estimates of leverage ratios. Section IV provides the main results on debt and
equity issuance regressions. Section V explores the debt and equity issuance responses of
3A partial list of papers documenting an inverse relation include Auerbach (1985), Graham and Tucker(2006), Long and Malitz (1985), Titman and Wessels (1988), Fischer et al. (1989), Rajan and Zingales(1995), and Booth et al. (2001). Frank and Goyal (2009) show that the inverse relation between leverageand profitability has become weaker in the recent decades.
4Of course, every model leaves some things out. For instance one might consider only general equilib-rium models, or only models that account for the nature of industrial competition, or those that accountfor imperfect decision making, or any of a host of other plausible considerations. This is simply a matterof taste.
5However, Chen and Zhao (2005) conclude that neither transaction costs nor taxes can properlyexplain the negative relation between leverage and profitability.
4
firms during good and bad times. Section VI examines the debt and equity issuances when
scaled by total issuances and total capital. Section VII shows how the smooth pasting
and value matching conditions from dynamic optimization have important implications
for leverage. This helps explain why the reactions to shock are only partial. Section VIII
concludes the paper.
II. Data
We use conventional data sources, starting with the merged Compustat-CRSP data.
The data are annual and are converted into constant 2000 dollars using the GDP defla-
tor. We exclude financial companies (SIC 6000-6999), firms involved in major mergers
(Compustat footnote code AB), firms reporting financial data in currencies other than the
U.S. dollars, and firms with missing data on our key variables.6 The ratio variables are
trimmed at the 1% level in both tails of the distribution. This serves to remove outliers
and the most extremely misrecorded data. The final sample consists of 179,021 firm-year
observations from 1971-2009.
Table I provides definitions of financial variables and reports summary statistics. The
average debt (in constant US$) is about $653 million while the median is $24 million.
A significant fraction of firms have zero debt (the 10th percentile is 0). Book equity is
slightly larger than book debt. Market equity is almost three times larger than book debt.
Book assets average $2,191 million although the medians are considerably smaller.
If issuing or retiring securities incurs no fixed costs, then we would expect to see
many small actions and very few large actions (Leary and Roberts, 2005). If there were
significant fixed costs involved in issuing or retiring outstanding securities, then small
issues might not be worthwhile. Table I shows that although most firms issue little debt
or equity in a given year, the averages are large. In other words, when firms actually enter
debt and equity markets, they intervene massively.
6These include debt, book value of equity, market value of equity, assets, book and market leverage,profitability, market-to-book assets ratio, and tangibility.
5
The mean constant dollar debt issue is $165 million (the median is $1.6 million). In
unreported tables, we find that the average debt issue is about 8.1% of assets (the median
is 1.8%). About 38% of the firms issue no debt; 8% issue between 0 and 1% of the value
of their assets as debt; another 16% issue between 1 and 5% of the value of their assets
as debt; and the remaining 38% issue debt in excess of 5% of the value of their assets.
The mean constant dollar equity issue is about $26 million (the median is about $0.4
million). As a fraction of assets, the mean and median equity issues are about 7% and
0.2%, respectively. About 33% of the firms issue no equity; 34% of the firms issue between
0 and 1% of the value of their assets as equity; another 14% issue between 1 and 5% of
the value of their assets as equity; and the remaining 19% of the firms issue equity that
is in excess of 5% of the value of their assets. Average debt repayments are larger than
equity repurchases. This perhaps reflects the finite maturity of debt and its contractual
repayment. The median firm does not repurchase equity.
We construct both book and market leverage ratios. Book leverage is defined as debt
over debt plus book equity. Market leverage is defined as debt over debt plus market
equity. The median book leverage is 0.36 (the average is 0.31). The median market
leverage is 0.27 (the average is 0.20).
Profitability is defined as the ratio of operating income before depreciation to assets.
While the average firm is profitable (the ratio of ebitda to assets is 0.05), the median
firm is even more profitable (with a profitability ratio of 0.11). The sample includes a
large number of unprofitable firms as the 10th percentile is -0.18. The table also reports
descriptive statistics on the market-to-book ratio and the tangibility ratio. The market-to-
book ratio (M/B), defined as the ratio of the market value of assets to book assets, averages
at about 1.65. Tangibility, defined as the ratio of net property, plant and equipment to
assets, averages at about 31%.
6
III. Estimating a Leverage Ratio
The previous literature focuses on estimates obtained using leverage ratios. Hence,
we start with a similar estimation to check whether our results match those of previous
studies. Table II presents the results. Some scholars advocate book leverage ratios while
others advocate market leverage ratios. We report results for both.
The regressions include leverage factors following the previous capital structure liter-
ature (see, for example, Rajan and Zingales (1995) and Frank and Goyal (2009)). The
factors include (i) profitability, (ii) industry median leverage, (iii) market-to-book assets
ratio, (iv) tangibility of assets, and (v) firm size. Industry median leverage is estimated
as the median leverage of all other firms in the industry (excluding the firm under con-
sideration). Firm size is defined as the natural log of assets. Other factors are defined in
Section II. We cluster standard errors by firm and estimate these regressions both with
and without fixed effects.
In Table II we run conventional cross section leverage regressions. Since we employ
the usual data, it is not surprising that our results match those reported in the existing
capital structure literature, i.e., profitability has a negative sign in both the book leverage
regression and in the market leverage regression. The coefficients on other factors largely
match those reported in earlier studies. Firms operating in industries with high lever-
age tend to have high leverage. A higher market-to-book ratio is associated with lower
leverage. Larger firms are typically more highly levered. The coefficients on tangibility
are positive. The conclusion from Table II is clear. The leverage regression results match
those reported in previous studies.
At times there is a concern that in a regression what is being estimated is a conditional
mean, but some parts of the distribution might behave differently from other parts. If
so, then ’accidents’ like sample selection issues, or asymmetries in the underlying distri-
butions might play more of a role than is recognized. The source of identification could
be misinterpreted. To guard against this potential problem we estimated the same model
7
using quantile regressions.7 The basic model results are robust in terms of both the sign
and statistical significance.
The impact of profits does appear to be somewhat stronger among the high leverage
firms. For instance in a book leverage regression the coefficient on profits for the first
quartile is -0.105 with a t-ratio of -37.8. For the third quartile the coefficient on profits is
-0.406 with a t-ratio of -101.5. This difference is curious and might deserve further study.
However, the fact that in each case we get the negative sign and statistical significance is
sufficient for the purposes of the current paper. So the results on profits are rather robust
across the distribution.
A. Frequency of Financing Activity
Table III tabulates the percentage of firms issuing or repurchasing debt or equity for
annual sorts based on lagged profitability. We employ the conventional 5% cut-off rule to
exclude minor fluctuations. Firms ‘issuing debt’ are therefore those that issue debt (both
short-term and long-term) in excess of 5% of the value of their assets. Other decisions
are similarly defined using a 5% cutoff.
How does security issuance behavior vary with firm profitability? To answer this
question, Table III sorts firms by profitability and reports the percentage of firms issuing
or repurchasing debt or equity. The results show that the likelihood of issuing debt is
mostly independent of firm profitability. Similarly, there is only a weak relation between
the profitability and the likelihood that the firm retires debt. Unlike debt issues, the
probability of issuing equity is strongly related to profitability.
Low-profit firms are much more likely to issue equity than are high-profit firms. Fur-
thermore, high-profit firms are much more likely to repurchase equity. Accordingly, low
profitability firms are much more likely to be net equity issuers than are high profitabil-
ity firms. Among the firms in the lowest quintiles of profits, roughly 30% of firms issue
7Cameron and Trivedi (2010) provide an extensive discussion of quantile regressions in Stata.
8
net equity that exceeds five percent of their assets. By contrast, for the most profitable
quintile of firms, only about 8% issue net equity.
Firm size is an important variable in the recent literature – it is often used as a proxy
for access to capital markets (as in Faulkender and Petersen (2006), and Leary (2009)).
Small firms are bank-dependent, risky, and informationally opaque. They have restricted
access to public debt markets and consequently face more severe supply constraints in
their ability to issue debt. Thus, we expect small firms to be more sluggish in adjusting
their debt and equity in response to shocks to profitability. Large firms, by contrast,
have much easier access to public debt markets and they face fewer obstacles in accessing
securities markets.
How do size and profitability interact? To examine this question, we first sort firms
annually by firm size and then, within each size quintile, we sort firms based on prof-
itability. The bottom part of Table III reports results for the smallest and the largest
size quintiles. Among the small firms, there is little relation between profitability and the
likelihood of issuing debt. However, larger firms exhibit a small increase in the likelihood
of net debt issuance with increasing profitability.
The effects of profitability on equity issuance and repurchases are much more consistent
across size quintiles. As profitability increases, firms are generally less likely to issue equity
and more likely to repurchase it. Across the two extreme size groups, we note a striking
difference between the proportion of firms issuing equity and that issuing debt. Among
low-profit small firms, almost 41% are equity issuers, whereas only about 9% of low-profit
large firms are equity issuers. However, regardless of size, we note a monotonic reduction
in the likelihood of issuing equity as profitability increases.
Importantly, Column (8) shows that the likelihood of issuing debt and simultaneously
repurchasing equity increases with profitability. Column (9) shows that, conversely, the
likelihood of issuing equity and retiring debt declines with profitability. The effects of
sorting on firm size and firm profitability mirror those reported for all firms. Again, we
find that low-profit firms are less likely to issue debt and repurchase equity; they are
9
instead more likely to issue equity and retire debt. Firms with high profitability exhibit
the reverse pattern.
B. Magnitude of Financing Activity
In the previous section, we considered the probability of having a nontrivial level of
debt or equity activity. The next question is how large are the dollar values involved. In
Table IV, we sort the firms according to profits and then tabulate the levels and changes
in both debt and equity. We do this first for all firms and then for small versus large
firms.
For the analysis on all firms, we observe in Column (1) that debt peaks at the middle
of the distribution. This is because firms with medium profitability are also the largest, as
seen in Column (8) which reports average asset values for different profitability quintiles.
It is further confirmed when we sort first by firm size and then examine firms with different
profitabilities within size quintiles. For small firms, debt is roughly independent of profits,
while for large firms low profit firms tend to have a higher level of debt.
As expected, Columns (3) and (5) show that more profitable firms have higher equity
values. Columns (1), (3), and (5) show that firms in our sample are quite typical of those
used in previous studies, which stands to reason because we are studying firms from the
most commonly used data set for such studies.
In Column (2) of Table IV, we consider the relationship between the change in debt
and firm profitability. Consistent with the predictions of the trade-off theory, we find that
debt issuances are significantly larger for more profitable firms. Also, consistently, less
profitable firms issue more equity while the most profitable firms repurchase equity.
The fact that more profitable firms issue debt and repurchase equity while the least
profitable firms retire debt and issue equity is consistent with the predicted relation be-
tween profitability and financing decisions under the trade-off theory. Columns (4) and
(6) provide an explanation of why the leverage ratio regression results contradict those
from the basic profitability sorts presented here. As we can see, profitability indirectly
10
affects leverage ratios by increasing equity values. Changes in both the book value of
equity and the market value of equity are positive and large for highly profitable firms.
By contrast, these changes are negative for less profitable firms.
We also examine two-way sorts by size and profitability and report results for the
smallest and largest quintile of firms. For the smallest quintile of firms, the change in
debt is largely unrelated to profits. But for large firms, there is a positive relation between
profits and debt issuances. High profit firms have a big positive change in debt. Low profit
large firms have a negative change in debt.
The changes in both the book value of equity and the market value of equity across
profit quintiles are illustrated in columns (4) and (6). Large and highly profitable firms
are associated with a big positive change in equity. For small firms the effect is much
weaker. However the impact from the large firms is sufficiently strong that the patterns
persist for the case of all firms.
Column (7) of Table IV is of particular importance for the trade-off theory. It shows
that small and low profit firms tend to issue more equity than do small and high profit
firms. Among large firms, the low profit ones tend to issue equity, while the high profit
ones tend to repurchase equity.
The finding that more profitable firms tend to repurchase equity comes as expected
from the basic trade-off theory. The fact that in general high profit firms tend to issue
more debt is also as predicted.
This evidence also illustrates an important issue concerning the use of leverage ratios.
Such ratios are often interpreted as essentially reflecting the use of debt by the firm. This
interpretation, while common, is empirically misleading.
For the typical firm, the change in the value of equity is larger than the change in
debt. For example, in the third profit quintile for large firms, the mean equity issue is
just $9 million, but the change in the market value of equity is $242 million. At the
same time the mean change in debt is $27 million. This suggests that a fair bit of the
observed variation in the leverage ratios is primarily driven by the changes in the market
value of equity in the denominator, rather than by the changes in debt in the numerator.
11
Since equity issues are often small, this implies that the variation in the leverage ratio is
primarily driven by internal operations rather than by external financing actions. This
again points to the fact that leverage ratios can provide a misleading account of actual
patterns in the data.
There is always a potential concern that the averages may be misleading due to the
impact of outliers. To address this concern, in Table V we present median values of the
profitability sorts. This table generally reinforces the findings in Table IV. The first
observation concerning the medians is that an average profitability firm typically reduces
debt. Low profit firms have a negative change in both book equity and market equity.
Equity issues are mostly found among low profit firms.
When we decompose small and large firms, important differences emerge at the me-
dians. Small firms generally have no change in debt and some reduction in both the
book and the market value of equity. Large unprofitable firms reduce debt and experience
declines in both book equity and market equity. Large profitable firms at the median
experience increases in both book and market equity. However experiencing an equity
increase is quite different from issuing equity. For both large and small firms equity is-
suance is found primarily among the low profit firms. The scale of equity issues relative
to firm size is much larger for small firms than for large firms.
IV. Estimating Debt and Equity Regressions
So far we have demonstrated that in our data the conventional leverage regressions
have the usual signs. We then demonstrated that in simple sorts of the data, strikingly
different financing patterns emerge. The next task is to reconcile these differences.
Table VI presents simple regressions predicting changes in debt, book equity, and
market equity, as well as equity issuances. Consistent with the sorts, we control for
changes in firm size, and we also include year dummies. There may be a concern about
the speed of a firm’s responses. Accordingly we consider the lagged change in profits as
the main case, but we also include the contemporaneous change in profits. Whether the
12
contemporaneous change in profits is included or not does not in fact matter. We report
panel-robust standard errors adjusted for clustering at the firm level.
The results in Table VI support the findings in the sorts. More profitable firms in-
crease debt. More profitable firms experience increases in book equity and market equity.
However, profitable firms show negative equity issues, meaning that when profitability
increases, firms tend to repurchase equity. All of these effects are statistically significant,
and seem intuitively reasonable.
The fact that more profitable firms do actually issue debt is crucial from the perspective
of the trade-off theory, as is the fact that these firms repurchase equity. However, these
firms apparently experience a large increase in the value of equity.
In Table VI all firms are pooled, and there are year fixed effects. Thus the time effects
have been removed, and identification is possible by observing the differences between
firms. Some scholars have argued that firms have financial targets that are key to under-
standing their financing behavior. Empirically defining the target is controversial since
there is no accepted model of the target. However, Lemmon et al. (2008) have argued
that there is a great deal of persistence in capital structure, so that a long term average
does a good job of predicting what firms will do. Thus, we run all of our regressions using
firm fixed effects. The long term average ought to reflect this targeting behavior.
We find that empirically very little changes whether we include firm fixed effects or
leave them out. Thus the targeting behavior, to the extent that it happens, does not
account for what we are reporting. Table VII provides an illustration. As is readily
apparent nothing much changes when we include the firm fixed effects. Accordingly
we do not think that the lack of concern for capital structure target seeking drives our
findings.
The next step is to bring the sorts and the simple regressions together into a common
setting. We therefore include factors in addition to profitability: (i) median industry
leverage, (ii) the market-to-book assets ratio, (iii) tangibility of assets, and (iv) firm size
(measured by the log of assets). Rajan and Zingales (1995) show that these factors are
related to leverage in G7 countries. A number of studies have used these factors to
13
estimate leverage targets. Frank and Goyal (2009) show that these factors are robustly
related to leverage in the U.S. In these regressions, we use indicators for the quintile that
the firm is in for each factor.
Table VIII reports these results. In Columns (1) and (2), we examine debt issues
and again show that profits positively affect debt issuances. The effects are large and
statistically significant at the 1% level. Columns (3) and (4), which examine book equity
show a strong effect of profits. This effect is only slightly reduced by the inclusion of
other factors. Columns (5) and (6) illustrate that the impact of profits on the change
in market equity is also robust to the inclusion of the conventional factors. Columns
(7) and (8) examine equity issuances. Here we again find that profits have a significant
negative impact on equity issuances. As before, the results are robust to the inclusion of
the conventional factors.
In unreported tables, we estimate debt and equity changes using quantile regressions.
In terms of the signs and significance of the profitability variable, the results are quali-
tatively similar to the OLS results reported in Table VIII. Profitability positively affects
debt issuances and negatively affects equity issuances across various quartiles.
While it is not our main focus, we note that the market-to-book ratio positively affects
debt issues and changes in the book value of equity. However, its impact on equity issues is
not statistically significant.8 In the existing literature, finding a negative relation between
the market-to-book ratio and leverage is common.9
8We are not testing the market timing hypothesis. However, these results are rather surprising fromthe perspective of the market timing theory of capital structure.
9The market-to-book ratio is commonly considered as a proxy for growth opportunities. Growth firmsare expected to finance with relatively more equity to avoid debt-related agency conflicts (Myers, 1977).Previous studies have typically found a negative relation between leverage and growth opportunities (seeSmith and Watts, 1992; Rajan and Zingales, 1995; Hovakimian et al., 2001; Goyal et al., 2002; Barclayet al., 2006; Frank and Goyal, 2009). The market timing literature also predicts a negative relationbetween the market-to-book ratio and leverage (Baker and Wurgler, 2002). Flannery and Rangan (2006),Kayhan and Titman (2007), and Liu (2009) suggest that it is difficult to disentangle the effect of growthopportunities and market timing in cross-sectional regressions of the market-to-book ratio on leverage.Graham and Harvey (2001) provide a useful survey of executives’ opinions about the importance ofvarious factors.
14
Firm size has a substantially larger effect on debt issuances than on equity issuances.
This may help to explain the commonly observed positive relation between firm size and
leverage ratios.
Table VIII also shows that there is nothing all that special about using sorts or regres-
sions to explain debt and equity issues. In either case we find that more profitable firms
tend to increase their debt, experience an increase in the value of equity, and repurchase
shares. Thus the control factors are not responsible for the usual rejection of the trade-off
theory. It would appear that what matters is the size of the issuing activity relative to
the organic increases in equity value due to profits. In other words what matters in the
leverage regression seems to be coming from the presence of E in DD+E
, and not from D.
Overall, the results from the debt and equity issuance regressions are consistent with
the trade-off theory. The coefficient estimates on the leverage factors generate several
interesting results. Profits positively affect debt issuances. Equity issues are negatively
related to profits. The market-to-book ratio positively affects the change in debt. Firm
size positively affects the change in debt and book equity. Firm size negatively affects
equity issues.
V. Are Financial Market Conditions Important?
According to the trade-off theory, capital structure is determined by considering the
impact of costs and benefits of debt. The time variation in the costs and benefits of debt
imply that issuance decisions would also vary over time. Furthermore, market conditions
affect the transaction costs of adjusting. Adverse selection is a greater problem in a
cold market than in a hot one. Accordingly it may be easier to adjust in good market
conditions. Related ideas have been developed by Huang and Ritter (2009), Faulkender
et al. (2010), and Halling et al. (2011). Faulkender et al.’s paper examines cross-sectional
variation in adjustment speeds including market valuations and how they may affect
adjustment speeds. It is natural to think that in good market conditions, it will be less
15
costly to issue both debt and equity, and adjusting leverage to respond to profitability
shocks would be much easier.
To test the importance of market conditions, we require a definition of good times
and bad times. Our empirical strategy is to estimate good times versus bad times at
the 4-digit industry level. We define an industry as having “good times” if the median
firm in that industry has a market-to-book ratio that is higher than the 67th percentile of
the time-series distribution of the industry median market-to-book ratios. Conversely, an
industry is defined as having bad times if the median firm in the industry has a market-
to-book ratio that is lower than the 33rd percentile of the time-series distribution of the
industry median market-to-book ratios.
Panel A of Table IX tabulates issuance activity for profitability sorts in both good
times and bad times. As might be expected, active debt and equity issuances are larger
during good times. There is also somewhat more active swaps between debt and equity
during good times.
Panel B reports the issuance activity for the smallest and the largest firm during the
good times and bad times. Within each size quintile, firms are sorted on profitability.
This is a three way sort. As before we find that there is much more active use of external
markets during good times. Small, low-profit firms are more likely to issue equity in good
times than in bad times. Debt issuances are significantly higher in good times. Large,
high-profit firms are significantly more likely to issue debt and repurchase equity in good
times than in bad times.
Panel C reports the magnitudes of the financial variables rather than the frequencies.
During bad times, less profitable large firms retire substantial amounts of debt and they
show a tendency to issue equity. Small firms do not seem to engage in similar debt
reduction activities. However, like the larger firms, there is a tendency for small firms to
issue more equity – at least to the extent possible.
During good times, large profitable firms raise significant amounts of debt as they
experience an increase in the value of their equity. Such firms also engage in share re-
purchases. Debt issues by smaller firms are much less affected by their own profitability
16
during good times and their change in debt is more or less independent of their prof-
its. However, equity issuance is primarily found among the low profit small firms during
good times. Table IX shows that, empirically, profitability has a time-varying impact on
leverage choices.
In Table X, the basic regressions for changes in debt and for equity issuance are
presented for both good times and bad times. Comparing good times with bad times,
we see that the effects of profits are much stronger in good times. In bad times the
impacts are rather weak statistically. This difference is interesting and deserves further
study. However for our purposes recall that the static trade-off theory implies time-
varying capital structure choices even when the target is time invariant. Empirically we
do observe time-varying choices.
VI. Is Scaling Important?
In the preceding section, we provided results for unscaled debt and equity issues.10
This seems appropriate to us. However it involves making two changes to the standard
regression. We would like to determine if both changes are crucial. Accordingly, in this
section, we take a step back to consider prior literature in which variables are scaled. The
questions are, how does scaling affect the results and why.
Table XI examines alternative scalings. Columns (1) and (2) scale the debt and equity
issuances by the total firm net issuance. Columns (3) and (4) scale the debt and equity
issuances by the sum of debt and book equity. Column (5) considers the change in debt
scaled by the sum of debt and book equity. Column (6) considers the change in leverage
due to external financing. Since we are examining changes, we use the changes in the
standard regression factors as controls. The primary interest is in the coefficient on the
change in profits.
10More than a century ago the famous statistician Pearson pointed out that scaling two independentvariables by the same third variable induces spurious correlation. Barraclough (2007) points out that thisidea extends to capital structure regressions. If you scale both the right-hand and left-hand side variablesin a regression by the same variable, that can induce a spurious correlation.
17
Column (1) shows that a change in profits is associated with an increase in debt issues.
But the sign reverses in Column (3) when we scale the debt issues by the sum of debt
plus book equity. This is telling. What is really driving the results is the change in
equity in the denominator, not the change in debt in the numerator. This result is further
substantiated in Column (5) where the change in debt to capital ratio is used instead of
debt issuance.
To further explore the impact of active issuance relative to passive experiencing of
changes we decompose leverage in Column (6). We separate out the effect of debt and
equity issues from those due to changes in retained earnings by defining leverage change
due to external financing as:
∆LeverageEF =Dt−1 + d
Dt−1 + Et−1 + d+ e− Dt−1
Dt−1 + Et−1, (1)
where D is debt, E is book value of equity , d is debt issues net of retirements, and e is
equity issues net of repurchases. Garvey and Hanka (1999) use a similar measure in their
study. This measure directly captures the net leverage effect of debt and equity issuances.
Column (6) reports results from regressions of the net leverage effect of debt and equity
issues on changes in profitability after controlling for other leverage factors in differences
and year indicator variables. The key effect is the coefficient on the change in profits,
which is positive. The effect of profitability on leverage change due to external financing
is positive and statistically significant. This comparison suggests that changes in total
debt and equity are only partly a result of debt and equity issuance decisions. Other
balance sheet adjustments complicate the inferences from leverage ratio regressions.
In Table XII, we examine change in leverage and change in leverage due to external
financing during good and bad times separately. The traditional negative effect of a change
in profits on a change in leverage is found both in good times and in bad times. However
when we decompose the impact in order to isolate the external component, as before, the
result disappears. In good times the external financing component of a change in profits
is fairly weak, but still positive. During bad times it is not statistically significant. This
implies that distinguishing between good and bad times does not fundamentally alter
18
the basic conclusions. The traditional leverage result on profits is driven by increases in
equity that are experienced and then partially, but not completely, offset by firms.
To the extent that firms are engaging in active financing they are following the pre-
scriptions of the trade-off theory. Thus we have overall a coherent picture of how firms
are responding to profits. Basically, more profitable firms experience an increase in the
value of their equity. They respond to this by issuing debt and repurchasing equity. But
the magnitude of the increase in equity is larger than the magnitudes of the active steps
taken in the debt and equity markets. This means that the firms’ actual actions are as
predicted by the trade-off theory. But it leaves open the question of why they do not act
more strongly.
In the next section we argue that partial adjustment to shocks is what ought to be
expected generally. The idea is that both fixed and variable adjustment costs are real and
need to be minimized in the firm’s optimization problem.
VII. Trade-Off with Adjustment Costs
The capital structure literature has long been interested in dynamics and capital struc-
ture adjustment paths. In recent years this interest has increasingly resulted in explicit
dynamic optimizing models such as those of Goldstein et al. (2001), Hennessy and Whited
(2005), Strebulaev (2007). These models commonly make specialized assumptions leading
some, such as Welch (2011), to question their generality.
In contrast, we argue that dynamic optimization (Dixit (1993), and Stokey (2009))
does offer some fairly general observations about the required conditions for optimality
when there are both fixed and variable adjustment costs. These conditions have important
implications for tests of the trade-off theory that have largely been ignored in the literature
on capital structure adjustment. As a result some rather misleading impressions have
become widespread.
The basic conditions for optimality are known as: “smooth pasting” and “value match-
ing”. Standard analysis of cash management (Constantinides and Richard (1978)), invest-
19
ment (Bertola and Caballero (1990)) and inventory (Stokey (2009)) all have an essentially
similar structure. This structure can be reinterpreted as a trade-off theory of corporate
capital structure. This works because the original models use fairly general assumptions
about the shapes of the underlying functions, rather than making problem-specific as-
sumptions. The presentation here largely follows that of Bertola and Caballero (1990).
Consider a model in which excessive, or inadequate leverage is costly. Think of these
costs as reflecting the taxation benefits of debt and deadweight bankruptcy costs. Lever-
age is the choice variable. Assume also that the firm has a nice concave production
function with a unique optimal scale of operations.
The risk-neutral firm chooses leverage in order to maximize the discounted present
value. Let L = leverage (the choice variable), r = the discount rate, ε = an exogenous
Brownian shock process, π = profits (twice differentiable, and strictly concave), V = the
firm’s value function. The unconstrained maximum is given by
x∗(ε) = arg maxL
π(L , ε).
If there were no adjustment costs then the firm would always pick leverage to ensure that
xτ = x∗(ε) for all τ . This is the static ‘leverage target’.
Now suppose that there are adjustment costs. The optimization problem can be
expressed as
V (Lt, εt) = max{Lt}
Et
∫ ∞t
e−r(τ−t)(π(L , ε)− ‘adjustment costs’).
To actively adjust leverage requires decision making. Adjusting takes time and effort
on the part of executives, along with board members and investment bankers in some
cases. Issuing securities also involves underwriting costs as documented by Altinkilic and
Hansen (2000) and Chen and Ritter (2000). There will normally be both a fixed cost
component and a variable cost component. Similarly, repurchasing securities will involve
both fixed and variable costs (fees).
20
In general, there is no reason for the fixed costs and the variable costs of increasing
and reducing leverage to be equal. In fact, there are good reasons to expect there to be
four distinct parameter values.11 Let cfi = the fixed cost of increasing leverage, cfr = the
fixed cost of reducing leverage, cvi = the variable cost of increasing leverage, cvr = the
variable cost of reducing leverage. All of these values are positive numbers.
If the firm wants to increase leverage by some amount ∆L, then it costs the firm
cfi + cvi∆L. Similarly if the firm wants to reduce leverage by some amount ∆L, then it
costs the firm cfr + cvr∆L. There is a discontinuity in the total cost of adjustment right
at the point of zero adjustment. Zero adjustment of leverage by the firm involves zero
adjustment cost. Any actual active adjustment entails a strictly positive fixed cost.
Consider what happens as the size of an adjustment shrinks towards zero. The average
unit cost becomes arbitrarily large since the fixed cost component is divided by something
incredibly small. But the profit function itself is smooth and has no such discontinuity.
Clearly it must be the case that a small enough (but strictly nonzero) leverage adjustment
cannot pay. Furthermore, the point at which it becomes too small to pay will generally
differ on the two sides of zero since the cost parameters differ.
With a setup that is essentially the same as this, Constantinides and Richard (1978)
and Bertola and Caballero (1990) show that the model can be depicted as in Figure 1.
The intuition is quite clear. If the firm is increasing leverage, then the marginal value of
a unit increase must equal the marginal cost of a unit increase. Otherwise the firm could
increase profits by making some other amount of adjustment. This pins down how large
an increase will be if it happens. When will the firm choose to increase leverage? It will
do so if the value of the firm is greater with a leverage increase than without. In other
words the firm undertakes the leverage increase at the point when the value of the firm
with the increase just matches the value of the firm without (‘value matching’). The same
reasoning applies to leverage reductions albeit with different parameter values.
11Going further, it ought to matter whether the action is in the debt market or in the equity market.Adding this would turn the four parameters discussed in the text into eight parameters. While thatwould be realistic, it would be overkill for our purposes. In this section we simply want to point out theimportance of smooth pasting and value matching for studies of capital structure.
21
Returning to Figure 1, the value function v(z) is rewritten in terms of zt = Lt − L∗t .The steps to make the conversion follow those of Bertola and Caballero (1990). There
are four critical boundary values Bla, Blr, Bua, Bur. Here B is a boundary value with the
subscript l denoting lower, u denoting upper, a denoting action point, and r denoting the
point to which the firm rebalances. Thus if Lt < Bla then the firm increases the leverage
to Blr. If Lt > Bua then the firm reduces leverage to Bur. If Bla < Lt < Bua, the firm
neither actively increases nor decreases leverage. The only changes to leverage reflect the
passive effect of the Brownian shocks.
As can be seen in Figure 1 the existence of the unit costs implies that increases and
decreases of leverage do not return the firm to the unconstrained maximum x∗. In each
case the movement is in that general direction, but in each case the movement is partial.
It does not pay to go all the way due to the unit adjustment costs.
This has crucial implications for tests of capital structure adjustment. Firms will not
adjust all the way to the static optimum. Instead there is movement in the direction
towards the optimum. Both increases and decreases are partial. There is a large zone of
inaction with a pair of outer boundaries at which it pays to readjust. There is also a pair
of inner readjustment levels with a gap in between. This inner gap includes the static
optimum. The larger the various adjustment costs the farther apart these critical values
will be.
It should be stressed that the analysis in this section is essentially a simple rein-
terpretation of previous models. Since those models have a very general structure the
interpretation in terms of leverage is not difficult. But by the same token, the ideas are
not novel. What is novel is that the models also apply to capital structure.
A limitation is that the model assumes that the choice variable is one dimensional
leverage. But interpreting leverage as one dimensional is a bit extreme. In reality, at least
debt and equity can be distinguished. Formally analyzing a model with fixed and variable
costs and with debt and equity would entail doubling the number of cost parameters.
The algebra would be messier. But the basic ideas would be the same for both debt and
equity. We would need to characterize the smooth pasting and value matching conditions
22
for both debt and equity. Assuming that the parameters differ, we would then have eight
cost parameters.
The optimal actions would have more complex rebalancing conditions and more com-
plex zones of inactivity. We would also need to decide in each case whether adjusting
debt or equity made more sense. Thus, some inequalities would need to be presented.
For our purposes the details of such analytics is not worth it. The existence of zones of
inactivity will remain. The fact that readjusting is only part way towards the static target
will remain. The fact that the theory implies the directions of the changes will remain.
As the number of choice variables increases, so will the number of rebalancing points.
Instead of two rebalancing points, for well behaved functions we ought to find four. As
still greater realism is introduced the complexity and the detail will grow. But the ideas
of value matching and smooth pasting will retain their basic importance.
The magnitudes of adjustments will depend on a number of parameters, and some
details of the functional forms. Characterizing these goes well beyond the scope of the
current paper. What we want to stress is that the trade-off theory has clear implications
for the directions of typical leverage adjustments. The implication for the magnitudes are
much more tenuous and detail dependent.
VIII. Conclusion
The connection between corporate profits and capital structure has been very influ-
ential in the assessment of the trade-off theory. The standard evidence has pushed the
literature away from the trade-off, and towards much more complex models and ideas.
As a result it is important to make sure that the evidence has been correctly interpreted.
Unfortunately, it has not.
In fact more profitable firms really do borrow more (not less). They tend to repurchase
equity. They experience an increase in both the book value of equity and the market value
of equity. Less profitable firms really do tend to reduce their debt and to issue equity.
23
Firm size and market conditions also matter. Larger firms tend to be more active
in the debt markets while smaller firms tend to be relatively more active in the equity
markets. During good times there is more use of external financing.
The usual profits-leverage puzzle result is primarily driven by the increase in equity
that is experienced by the more profitable firms. Accordingly the puzzle should be restated
as asking: why do firms not take sufficiently large offsetting actions to fully undo the
change in equity? What limits the magnitudes of the typical leverage response to profit
shocks?
In a frictionless model the partial response appears puzzling. But there is good empir-
ical reason to believe that rebalancing entails both fixed and variable costs. These costs
can be fully avoided by doing nothing. Accordingly, the firm must decide whether any
given shock is big enough to be worth responding to. If it is, then the firm must decide
how big a response is called for. These technical conditions are known as ‘value matching’
and ‘smooth pasting’.
Thus optimization implies that some shocks will be ignored. Even if the shock is not
ignored, the optimal response will only partially undo the shock. The magnitude of the
leverage response must balance the marginal cost and the marginal benefit of an extra
unit of leverage. Since the marginal cost of adjusting leverage is strictly positive, the
adjustment towards that static leverage optimum will only go part way. This is true both
for leverage increases and for leverage reductions.
Surely nobody ought to be surprised that more profitable firms are more inclined
to repurchase equity, while unprofitable firms tend not to do so. The fact that more
profitable firms typically experience an increase in the value of equity is equally natural.
Were it not for the contrary claims in the literature (i.e. the profits-leverage puzzle), we
would argue that nobody ought to find it surprising that more profitable firms are more
inclined to issue debt. The fact that leverage adjustments typically only partially offset
profit shocks, should also sound quite natural. These facts all fit together easily within
the trade-off theory.
24
References
Altinkilic, O., and R.S. Hansen, 2000, Are there economies of scale in underwriting fees?Evidence of rising external financing costs, Review of Financial Studies 13, 191–218.
Auerbach, A.J., 1985, Real determinants of corporate leverage, in B.M. Friedman, (ed.)Corporate capital structures in the United States (University of Chicago Press, Chicago,IL).
Baker, M., and J. Wurgler, 2002, Market timing and capital structure, Journal of Finance57, 1–32.
Barclay, M.J., E. Morellec, and Jr. C.W. Smith, 2006, On the debt capacity of growthoptions, Journal of Business 79, 37–59.
Barraclough, K., 2007, Scaling regression equations: Solution or problem? Working Paper,Vanderbilt University.
Bertola, G., and R.J. Caballero, 1990, Kinked Adjustment Costs and Aggregate Dynam-ics, NBER Macroeconomics Annual 5, 237–288.
Booth, Laurence, Varouj Aivazian, Asli Demirguc-Kunt, and Vojislav Maksimovic, 2001,Capital structures in developing countries, Journal of Finance 56, 87–130.
Cameron, A.C., and P.K. Trivedi, 2010, Microeconometrics Using Stata (Stata Press).
Chen, H.C., and J.R. Ritter, 2000, The seven percent solution, The Journal of Finance55, 1105–1131.
Chen, L., and X. Zhao, 2005, Profitability, mean reversion of leverage ratios and capitalstructure choices, working paper, Michigan State University and Kent State University.URL: http://ssrn.com/abstract=614082
Constantinides, G.M., and S.F. Richard, 1978, Existence of optimal simple policies fordiscounted-cost inventory and cash management in continuous time, Operations Re-search 26, 620–636.
Dixit, A.K., 1993, The art of smooth pasting (Routledge).
Fama, E., and K.R. French, 2002, Testing trade-off and pecking order predictions aboutdividends and debt, Review of Financial Studies 15, 1–33.
Faulkender, M., M. Flannery, K.W. Hankins, and J.M. Smith, 2010, Cash flows andleverage adjustment, Working paper, University of Maryland, University of Florida,and University of Kentucky.
Faulkender, M., and M.A. Petersen, 2006, Does the source of capital affect capital struc-ture?, Review of Financial Studies 19, 45–79.
25
Fischer, E., R. Heinkel, and J. Zechner, 1989, Dynamic capital structure choice: theoryand tests, Journal of Finance 44, 19–40.
Flannery, M., and K. Rangan, 2006, Partial adjustment towards target capital structures,Journal of Financial Economics 79, 469–506.
Frank, M.Z., and V.K. Goyal, 2008, Trade-off and pecking order theories of debt, in B.E.Eckbo, (ed.) Handbook of Corporate Finance: Empirical Corporate Finance, Vol. 2. In:Handbook of Finance Series, Chapter 12, – (Elsevier/North-Holland, Amsterdam).
Frank, M.Z., and V.K. Goyal, 2009, Capital structure decisions: Which factors are reliablyimportant?, Financial Management 38, 1–37.
Garvey, G.T., and G. Hanka, 1999, Capital structure and corporate control: The effectof antitakeover statutes on firm leverage, Journal of Finance 54, 519–546.
Goldstein, R., N. Ju, and H. Leland, 2001, An ebit-based model of dynamic capitalstructure, Journal of Business 74, 483–512.
Gomes, A.R., and G.M. Phillips, 2007, Private and public security issuance by publicfirms: The role of asymmetric information, Working Paper, Washington University inSt. Louis and University of Maryland.
Goyal, V.K., K.M. Lehn, and S. Racic, 2002, Growth opportunities and corporate debtpolicy: the case of the U.S. defense industry, Journal of Financial Economics 64, 35–59.
Graham, J.R., and C. Harvey, 2001, The theory and practice of corporate finance: evi-dence from the field, Journal of Financial Economics 60, 187–243.
Graham, J.R., and A.L. Tucker, 2006, Tax shelters and corporate debt policy, Journal ofFinancial Economics 81, 563–594.
Halling, M., J. Yu, and J. Zechner, 2011, Leverage dynamics over the business cycle, Work-ing paper, University of Utah, University of New South Wales, and Vienna Universityof Economics and Business Administration.
Hennessy, C.A., and T.A. Whited, 2005, Debt dynamics, Journal of Finance 60, 1129–1165.
Hovakimian, A., T. Opler, and S. Titman, 2001, The debt-equity choice, Journal ofFinancial and Quantitative Analysis 36, 1–24.
Huang, R., and J.R Ritter, 2009, Testing theories of capital structure and estimating thespeed of adjustment, Journal of Financial and Quantitative Analysis 44, 237–271.
Jung, K., Y. Kim, and R. Stulz, 1996, Timing, investment opportunities, managerialdiscretion, and the security issue decision, Journal of Financial Economics 42, 159–185.
26
Kayhan, A., and S. Titman, 2007, Firms’ histories and their capital structures, Journalof Financial Economics 83, 1–32.
Kraus, A., and R.H. Litzenberger, 1973, A state-preference model of optimal financialleverage, Journal of Finance 33, 911–922.
Leary, M.T., 2009, Bank loan supply, lender choice, and corporate capital structure,Journal of Finance 64, 1143–1185.
Leary, M.T., and M.R. Roberts, 2005, Do firms rebalance their capital structures?, Journalof Finance 60, 2575–2619.
Lemmon, M.L., M.R. Roberts, and J.F. Zender, 2008, Back to the beginning: Persistenceand the cross-section of corporate capital structure, Journal of Finance 63, 1575–1608.
Liu, X.L., 2009, Do firms have target leverage ratios? Evidence from historical market-to-book and past returns 15, 602–612.
Long, M.S., and I.B. Malitz, 1985, The investment-financing nexus: Some empirical evi-dence, Midland Finance Journal 53–59.
Mackie-Mason, J.K., 1990, Do taxes affect corporate financing decisions?, Journal ofFinance 45, 1471–1493.
Myers, S.C., 1977, Determinants of corporate borrowing, Journal of Financial Economics5, 147–175.
Myers, S.C., 1993, Still searching for optimal capital structure, Journal of Applied Cor-porate Finance 6.1, 4–14.
Rajan, R., and L. Zingales, 1995, What do we know about capital structure: Someevidence from international data, Journal of Finance 50, 1421–1460.
Smith, C.W., and R.L. Watts, 1992, The investment opportunity set and corporate financ-ing, dividend and compensation policies, Journal of Financial Economics 32, 263–292.
Stokey, N.L., 2009, The economics of inaction: Stochastic control models with fixed costs(Princeton University Press).
Strebulaev, I.A., 2007, Do tests of capital structure theory mean what they say?, Journalof Finance 62, 1747–1787.
Titman, S., and R. Wessels, 1988, The determinants of capital structure choice, Journalof Finance 43, 1–21.
Welch, I., 2004, Capital structure and stock returns, Journal of Political Economy 112,106–131.
27
Welch, I., 2007, Common flaws in empirical capital structure research Working Paper,http://ssrn.com/abstract=931675.
Welch, I., 2011, A critique of recent quantitative and deep-structure modeling in capitalstructure research and beyond Working Paper.
28
Fig
ure
1:Sm
oot
hP
asti
ng.
The
stat
icop
tim
alle
vera
gera
tio
isat
the
pea
kof
the
valu
efu
nct
ion.
The
pos
itiv
ely
slop
edline
segm
ents
hav
eth
esa
me
slop
eas
each
other
.Sim
ilar
ly,
the
neg
ativ
ely
slop
edlines
segm
ents
hav
eth
esa
me
slop
eas
each
other
.If
leve
rage
isst
rict
lyb
etw
eenBla
andBua
then
the
firm
willnot
acti
vely
adju
stle
vera
ge.
The
real
ized
leve
rage
isth
enth
epas
sive
resu
ltof
the
Bro
wnia
nsh
ock
s.If
leve
rage
isle
ssth
anBla
,th
enth
efirm
will
opti
mal
lyin
crea
sele
vera
geto
Blr
.If
leve
rage
isgr
eate
rth
anBua,
then
the
firm
will
opti
mal
lyre
duce
leve
rage
toBur.
Thes
ere
bal
anci
ng
choi
ces
equat
eth
em
argi
nal
cost
wit
hth
em
argi
nal
ben
efit
ofan
extr
aunit
ofle
vera
gech
ange
.D
ue
toth
eunit
cost
ofad
just
ing,
when
the
firm
adju
sts
itw
ill
adju
stin
the
dir
ecti
onof
the
stat
icop
tim
um
,but
itw
ill
only
gopar
tof
the
way
.G
oing
furt
her
wou
ldnot
be
wor
thth
eex
tra
cost
.
29
Table I
Data description
Distribution
Variable N Mean SD 10th 50th 90th
Debt ($ millions) 179,021 653.01 4,591.84 0.00 24.10 1,047.90Book equity ($ millions) 179,021 814.69 4,122.31 3.31 69.97 1,372.58
Market equity ($ millions) 179,021 1,783.77 9,589.09 8.56 118.33 2,697.70
Assets ($ millions) 179,021 2,191.30 11,530.12 10.12 153.00 3,574.98
Debt issuance ($ millions) 179,021 165.29 1,335.28 0.00 1.58 244.54
Equity issuance ($ millions) 179,021 25.67 186.21 0.00 0.38 39.65
Debt repayment ($ millions) 179,021 131.96 1,123.38 0.00 2.87 181.19
Equity repurchase ($ millions) 179,021 24.32 257.48 0.00 0.00 9.19
Book leverage 179,021 0.36 0.34 0.00 0.31 0.72
Market leverage 179,021 0.27 0.25 0.00 0.20 0.66
Profitability 179,021 0.05 0.27 -0.18 0.11 0.23MB 179,021 1.65 2.06 0.52 1.02 3.24
Tangibility 179,021 0.31 0.24 0.05 0.26 0.70
Data sources: The sample comes from the annual Compustat files. The sample period is 1971–2009. Weexclude financial companies (SIC 6000-6999), firms involved in major mergers (Compustat footnote codeAB), firms reporting financial data in currencies other than the U.S. dollar, and firms with missing dataon the key variables. All financial variables are deflated to year 2000 using the GDP deflator. The ratiovariables are trimmed at the 1% level in both tails of the distribution. This serves to remove outliers andthe most extremely misrecorded data.
Variable definitions:
Debt = Long-term debt (dltt) + Short-term debt (dlc)Book equity = Common shareholder equity (ceq)
Market equity = Number of outstanding shares (csho) × Closing share price (prcc f)Assets = Book assets (at)
Debt issuance = Issuance of long-term debt (Max(dltis, 0)) + Increase in current debt (Max(dlcch, 0))Equity issuance = Sale of common stock (Max(sstk, 0))Debt repayment = Reduction of long-term debt (Max(dltr, 0)) + Decrease in current debt (−Min(dlcch, 0))
Equity repurchase = Purchase of common stock (Max(prstkc, 0))Cash balance = Cash and marketable securities (che)
Book leverage = Debt/(Debt + Book equity)Market leverage = Debt/(Debt + Market equity)
Profitability = EBITDA (oibdp)/Total assets (at)Market-to-book ratio = Market value of assets (MVA)/Assets,
where MVA = Debt + Market equity + Preferred-liq. value (pstkl)- Deferred taxes (txditc)
Tangibility = Net property plant and equipment (ppent)/Total assets (at)
30
Tab
leII
:R
egre
ssio
ns
of
Book
and
Mark
et
Levera
ge
Rati
os
Th
eta
ble
pre
sents
esti
mat
esof
the
leve
rage
rati
ore
gre
ssio
ns
on
firm
chara
cter
isti
cs.
Th
esa
mp
leco
mes
from
the
an
nu
al
Com
pu
stat
file
sd
uri
ng
the
per
iod
1971
-200
9.F
inan
cial
firm
sare
excl
ud
ed.
Pan
elA
pre
sents
esti
mate
sof
the
book
leve
rage
esti
mate
das
the
rati
oof
deb
tov
erd
ebt
plu
sb
ook
equ
ity.
Pan
elB
pre
sents
esti
mat
esof
mark
etle
vera
ge,
esti
mate
das
the
rati
oof
deb
tov
erd
ebt
plu
sm
ark
eteq
uit
y.T
he
exp
lan
ato
ry
vari
able
sProfitability
t−1,( M B
) t−1,Tangibility t−1,
an
dLn
(Assets)
t−1
are
des
crib
edin
Tab
leI.IndMedianLev
t−1
ises
tim
ate
das
the
med
ian
book
leve
rage
ofal
loth
erfi
rms
inth
esa
me
ind
ust
ryin
Pan
elA
,an
das
the
med
ian
mark
etle
vera
ge
of
all
oth
erfi
rms
inth
esa
me
ind
ust
ry
inP
anel
B.
Th
ein
du
stry
isd
efin
edat
the
level
of
the
4-d
igit
SIC
cod
e.A
llsp
ecifi
cati
on
sin
clu
de
the
year
fixed
effec
ts.
Th
esp
ecifi
cati
on
s
inco
lum
n(2
)ad
dit
ion
ally
incl
ud
eth
efi
rmfi
xed
effec
ts.
We
rep
ort
t-st
ati
stic
sw
her
eth
est
an
dard
erro
rsare
clu
ster
edat
the
firm
level
in
par
enth
eses
.aS
ign
ifica
nt
atth
e1
per
cent
leve
l.
PanelA:BookLevera
ge
Qu
anti
leR
egre
ssio
ns
OL
Sw
ith
Fix
edE
ffec
tsw
ith
25th
%il
eM
edia
n75th
%il
eC
lust
ered
SE
Clu
ster
edS
ER
egre
ssio
nR
egre
ssio
nR
egre
ssio
n(1
)(2
)(3
)(4
)(5
)
Profitability
t−1
-0.2
55a
-0.1
97a
-0.1
05a
-0.1
70a
-0.4
06a
(-29
.3)
(-19.2
)(-
37.8
)(-
51.8
)(-
101.5
)
IndMedianLev
t−1
0.42
2a
0.1
23a
0.2
96a
0.5
24a
0.5
24a
(36.
9)(1
0.7
)(9
2.6
)(1
25.2
)(9
5.0
)( M B
) t−1
-0.0
16a
-0.0
09a
-0.0
10a
-0.0
16a
-0.0
23a
(-20
.3)
(-10.2
)(-
39.2
)(-
42.6
)(-
41.8
)
Tangibility t−1
0.18
4a
0.2
46a
0.2
39a
0.1
93a
0.1
23a
(20.
6)(1
5.4
)(9
6.4
)(5
8.0
)(2
7.6
)
Ln
(Assets)
t−1
0.01
1a
0.0
22a
0.0
24a
0.0
18a
0.0
05a
(11.
4)(7
.6)
(89.9
)(4
7.8
)(9
.6)
Constant
0.11
9a
0.0
98a
-0.1
19a
0.0
16a
0.2
94a
(15.
6)(6
.0)
(-28.4
)(2
.7)
(38.7
)
Yea
rF
ixed
Eff
ects
Yes
Yes
Yes
Yes
Yes
Fir
mF
ixed
Eff
ects
No
Yes
No
No
No
R2-A
dju
sted
0.13
00.5
41
Pse
ud
oR
20.1
60
0.1
50
0.0
92
Observations
158,
824
158,8
24
158,8
24
158,8
24
158,8
24
31
Table
IIC
onti
nued
PanelB:M
ark
etLevera
ge
Qu
anti
leR
egre
ssio
ns
OL
Sw
ith
Fix
edE
ffec
tsw
ith
25th
%il
eM
edia
n75th
%il
eC
lust
ered
SE
Clu
ster
edS
ER
egre
ssio
nR
egre
ssio
nR
egre
ssio
n(1
)(2
)(3
)(4
)(5
)
Profitability
t−1
-0.1
29a
-0.1
29a
-0.0
71a
-0.1
05a
-0.1
97a
(-27
.5)
(-26.2
)(-
34.9
)(-
42.5
)(-
53.8
)
IndMedianLev
t−1
0.45
6a
0.1
99a
0.3
22a
0.5
77a
0.6
33a
(50.
6)(2
3.9
)(1
29.9
)(1
63.0
)(1
04.2
)( M B
) t−1
-0.0
30a
-0.0
14a
-0.0
13a
-0.0
21a
-0.0
35a
(-50
.9)
(-30.3
)(-
74.3
)(-
70.9
)(-
53.4
)
Tangibility t−1
0.12
9a
0.1
82a
0.1
32a
0.1
43a
0.1
02a
(19.
1)(1
8.1
)(7
6.2
)(5
7.1
)(2
3.2
)
Ln
(Assets)
t−1
0.00
9a
0.0
48a
0.0
13a
0.0
10a
0.0
01a
(11.
2)(2
5.8
)(6
8.2
)(3
4.2
)(2
.9)
Constant
0.10
0a
-0.1
12a
-0.0
49a
0.0
50a
0.2
90a
(17.
1)(-
9.8
)(-
17.3
)(1
1.8
)(3
9.8
)
Yea
rF
ixed
Eff
ects
Yes
Yes
Yes
Yes
Yes
Fir
mF
ixed
Eff
ects
No
Yes
No
No
No
R2-A
dju
sted
0.30
30.6
74
Pse
ud
oR
20.1
57
0.2
25
0.2
05
Observations
158,
578
158,5
78
158,5
78
158,5
78
158,5
78
32
Tab
leII
I:D
ebt
and
equit
yis
suers
:P
rofita
bil
ity
sort
s
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
alC
om
pu
stat
file
sfo
rth
ep
erio
dfr
om1971
to2009.
Th
eta
ble
pre
sents
the
per
centa
ge
offi
rms
issu
ing
and
reti
rin
g(o
rre
pu
rch
asin
g)d
ebt
an
deq
uit
y.A
firm
iscl
ass
ified
as
‘iss
uin
gd
ebt’
ifit
issu
esd
ebt
inex
cess
of
5%
of
the
valu
e
ofit
sas
sets
;as
‘iss
uin
geq
uit
y’
ifit
issu
eseq
uit
yin
exce
ssof
5%
of
the
valu
eof
its
ass
ets;
as
‘ret
irin
gd
ebt’
ifit
reti
res
deb
tin
exce
ssof
5%
of
the
valu
eof
its
asse
ts;
and
as‘r
epu
rch
asin
geq
uit
y’
ifit
rep
urc
hase
seq
uit
yin
exce
ssof
5%
of
the
valu
eof
its
ass
ets.
Inad
dit
ion
,w
eals
ore
port
net
-deb
tis
suer
s,w
hic
har
efi
rms
that
issu
en
etd
ebt
over
5%
of
the
valu
eof
thei
rass
ets,
an
dn
eteq
uit
yis
suer
s,w
hic
hare
firm
sth
at
issu
en
et
equ
ity
inex
cess
of5%
ofth
eva
lue
ofth
eir
asse
ts.
We
an
nu
all
yso
rtfi
rms
on
lagged
pro
fita
bil
ity
an
dre
port
the
per
centa
ge
of
firm
sin
each
of
thes
eca
tego
ries
.T
he
bot
tom
par
tof
the
tab
lere
port
sth
ep
erce
nta
ge
of
firm
sis
suin
gor
reti
rin
gse
curi
ties
by
pro
fita
bil
ity
wit
hin
the
small
est
and
larg
est
asse
tqu
inti
les.
33
Table
III
Conti
nued
Per
cen
tage
of
firm
s
Issu
ing
Ret
irin
gIs
s.N
etIs
suin
gR
epu
rch
.Is
s.N
etIs
s.B
oth
Iss.
Deb
tIs
s.E
qu
ity
Iss.
Non
eD
ebt
Deb
tD
ebt
Equ
ity
Equ
ity
Equ
ity
D&
ER
ep.
Equ
ity
Ret
.D
ebt
Rep
.N
on
e
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Sort
son
firm
pro
fita
bili
ty
Allfirms
Low
π27
.624
.518
.630.1
1.3
29.7
7.1
0.1
3.8
28.7
231
.929
.918
.111.7
2.1
11.2
2.7
0.2
2.0
38.6
338
.534
.419
.48.6
2.9
8.0
2.1
0.5
1.5
36.9
438
.633
.021
.38.1
4.3
7.5
2.0
0.7
1.4
37.2
Hig
hπ
32.9
27.3
19.8
9.2
8.2
8.1
1.7
1.3
1.4
40.3
Smallfirms
Low
π27
.520
.420
.141.1
1.1
40.5
10.9
0.1
4.7
16.0
229
.123
.321
.337.9
1.0
37.8
9.9
0.1
4.5
24.0
328
.124
.620
.426.5
1.4
26.1
6.6
0.1
4.1
32.1
427
.527
.218
.616.2
1.5
15.9
3.4
0.2
3.5
38.6
Hig
hπ
27.5
28.1
17.8
13.6
2.7
13.1
2.5
0.2
2.6
40.1
Largefirms
Low
π41
.436
.317
.78.7
2.2
7.9
2.8
0.3
1.1
35.4
244
.937
.918
.06.2
3.0
5.7
1.7
0.7
0.9
35.8
343
.936
.018
.65.6
4.1
4.8
1.6
0.9
0.7
37.6
440
.332
.019
.44.6
6.7
3.9
1.3
1.2
0.6
39.3
Hig
hπ
37.4
26.4
20.7
5.0
12.9
3.5
1.0
2.8
0.5
41.1
34
Tab
leIV
:M
agnit
ude
of
Fin
anci
ng
Act
ivit
y
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
al
Com
pu
stat
file
sfo
rth
ep
erio
dfr
om
1971
to2009.
Th
eta
ble
pre
sents
aver
age
deb
tan
deq
uit
y,ch
ange
sin
deb
tan
deq
uit
yan
deq
uit
yis
suan
ces
(in
mil
lions
of
doll
ars
)fo
rfi
rms
sort
edon
pro
fita
bil
ity
wit
hin
size
class
es.
Th
eta
ble
rep
orts
info
rmat
ion
for
all
firm
sso
rted
on
pro
fita
bil
ity
an
dfo
rp
rofi
tab
ilit
yso
rts
wit
hin
the
small
est
an
dla
rges
tfi
rms.
Th
eso
rts
are
don
ean
nu
ally
.
D∆D
BV
E∆BVE
MV
E∆MVE
Equ
ity
Iss.
Ass
ets
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
AllFirms
Low
π82
.8-4
.777.8
-9.7
119.
7-4
.56.6
233.9
260
4.9
-13.
4508.6
-23.8
720.
6-7
.516.8
1,6
90.8
31,
226.
78.
71,1
22.7
-21.9
1,8
33.6
3.2
8.5
3,5
50.9
483
3.4
23.3
1,1
66.0
6.9
2,4
07.
673.7
-7.9
2,9
94.1
Hig
hπ
570.
253
.41,1
70.5
71.1
3,8
17.
167.7
-36.1
2,5
81.2
SmallFirms
Low
π2.
50.
44.2
0.1
33.
22.8
3.3
9.7
22.
90.
65.5
-0.2
27.8
1.6
2.1
11.6
33.
10.
76.3
0.1
21.2
0.5
1.2
12.7
43.
00.
67.0
0.5
17.9
1.0
0.8
13.6
Hig
hπ
2.4
0.6
8.4
1.4
23.4
1.7
0.7
14.2
Largefirms
Low
π3,
928.
6-5
5.4
2,9
89.9
-152.1
4,6
71.2
10.8
62.5
10,8
46.3
23,
825.
619
.13,1
77.7
-94.5
5,1
22.2
-66.1
11.0
10,7
21.4
33,
074.
326
.83,4
97.9
-41.1
6,1
73.0
241.6
9.4
9,6
97.3
42,
622.
762
.03,9
14.8
45.2
8,4
72.
794.4
-57.0
9,8
54.7
Hig
hπ
2,18
0.4
192.
94,6
41.6
261.3
14,9
24.
1202.1
-147.8
10,1
70.5
35
Tab
leV
:M
edia
nP
rofita
bil
ity
Sort
s
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
al
Com
pu
stat
file
sfo
rth
ep
erio
dfr
om
1971
to2009.
Th
eta
ble
pre
sents
med
ian
deb
t
and
equ
ity,
med
ian
chan
ges
ind
ebt
and
equ
ity,
and
med
ian
equ
ity
issu
an
ces
(in
mil
lion
sof
doll
ars
)fo
rfi
rms
sort
edon
pro
fita
bil
ity
wit
hin
size
clas
ses.
Th
eta
ble
rep
orts
info
rmat
ion
for
all
firm
sso
rted
on
pro
fita
bil
ity
an
dfo
rp
rofi
tab
ilit
yso
rts
wit
hin
the
small
est
an
dla
rges
tfi
rms.
Th
e
sort
sar
ed
one
annu
ally
.
D∆D
BV
E∆BVE
MV
E∆MVE
Equ
ity
Iss.
Ass
ets
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
AllFirms
Low
π2.
10.
07.5
-1.7
25.
6-2
.00.1
19.2
218
.6-0
.244.9
-1.8
59.8
-1.9
0.0
110.3
381
.6-0
.6127.4
-0.3
163.7
-0.5
0.0
322.4
478
.7-0
.2154.1
1.8
248.8
0.3
0.0
355.6
Hig
hπ
27.3
0.0
136.1
6.6
330.
01.4
0.0
257.8
SmallFirms
Low
π0.
50.
01.3
-0.7
14.
0-1
.40.6
4.0
20.
60.
02.5
-0.8
11.4
-1.2
0.1
5.6
30.
90.
03.2
-0.5
8.5
-0.7
0.0
6.7
41.
20.
03.9
-0.1
6.8
-0.3
0.0
7.9
Hig
hπ
1.0
0.0
5.1
0.4
9.8
-0.1
0.0
9.0
Largefirms
Low
π1,
148.
5-3
8.1
1,0
26.9
-37.1
1,3
22.7
-20.2
0.3
3,3
39.0
21,
305.
1-3
9.3
1,2
78.6
-14.0
1,7
57.
116.9
0.5
3,9
24.6
31,
155.
4-2
8.7
1,3
42.9
-2.4
2,1
62.
330.5
0.0
3,7
03.1
499
6.3
-20.
91,5
05.2
19.4
2,9
42.
430.8
0.0
3,6
70.5
Hig
hπ
736.
10.
01,6
38.0
61.3
4,9
14.
243.9
0.0
3,4
97.6
36
Tab
leV
I:D
ebt
and
Equit
yC
hanges
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
al
Com
pu
stat
file
sfo
rth
ep
erio
dfr
om
1971
to2009.
Th
eta
ble
pre
sents
esti
mate
s
from
regr
essi
ons
of
chan
gein
deb
t(∆D
),ch
ange
inth
eb
ook
valu
eof
equ
ity
(∆BVE
),ch
an
gein
the
mark
etva
lue
of
equ
ity
(∆MVE
),an
dn
et
equ
ity
issu
ance
s(EquityIss.)
.T
he
exp
lan
ator
yva
riab
les
incl
ud
ecu
rren
tan
dla
gged
chan
ge
inop
erati
ng
inco
me
bef
ore
dep
reci
ati
on
(∆Profits
and
∆Profits t−1
)an
dla
gged
chan
gein
book
valu
eof
ass
ets
(∆Assets t−1).
Th
ere
gre
ssio
ns
incl
ud
eyea
rin
dic
ato
rva
riab
les.
Th
ere
port
ed
t-st
atis
tics
are
corr
ecte
dfo
rcl
ust
erin
gat
the
firm
leve
l.a,b,
an
dc
mea
nsi
gn
ifica
nt
at
the
1%
,5%
,an
d10%
leve
l,re
spec
tive
ly.
∆D
∆BVE
∆MVE
EquityIss.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
∆Profits t−1
0.25
4a
0.23
6a
0.2
33a
0.3
80a
0.0
70
0.6
61a
-0.0
70a
-0.0
93a
(4.0
)(3
.2)
(3.6
)(3
.9)
(0.6
)(2
.9)
(-4.2
)(-
5.1
)
∆Profits
0.50
9b
0.8
20a
1.4
87a
-0.0
43a
(2.5
)(3
.3)
(4.9
)(-
3.3
)
∆Assets t−1
0.01
7-0
.053c
-0.2
42a
0.0
09b
(0.8
)(-
1.7
)(-
3.4
)(2
.3)
Constant
-9.2
83-1
7.55
5a
-4.5
12
-7.1
32
155.1
52a
170.3
95a
12.9
52a
13.9
60a
(-1.
2)(-
2.8)
(-0.7
)(-
0.9
)(4
.9)
(4.9
)(4
.5)
(4.5
)
Yea
reff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
R2−Adjusted
0.00
90.
047
0.0
09
0.1
07
0.0
13
0.0
54
0.0
15
0.0
21
Observations
162,
056
162
,056
162,1
30
162,1
30
157,5
50
157,5
50
162,1
54
162,1
54
37
Tab
leV
II:D
ebt
and
Equit
yC
hanges
–F
ixed
Eff
ect
s
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
al
Com
pu
stat
file
sfo
rth
ep
erio
dfr
om
1971
to2009.
Th
eta
ble
pre
sents
fixed
-eff
ect
esti
mat
esfr
omre
gres
sion
sof
chan
ges
ind
ebt
(∆D
),ch
an
ges
inth
eb
ook
valu
eof
equ
ity
(∆BVE
),ch
an
ges
inth
em
ark
etva
lue
of
equ
ity
(∆MVE
),an
dn
eteq
uit
yis
suan
ces
(EquityIss.)
.T
he
exp
lan
ato
ryva
riab
les
incl
ud
ecu
rren
tan
dla
gged
chan
ge
inop
erati
ng
inco
me
bef
ore
dep
reci
atio
n(∆Profits
and
∆Profits t−1
),an
dla
gged
chan
ge
inth
eb
ook
valu
eof
ass
ets
(∆Assets t−1).
Inad
dit
ion
tofi
xed
firm
effec
t,th
e
regr
essi
ons
incl
ud
eyea
rin
dic
ator
vari
able
s.T
he
rep
ort
edt-
stati
stic
sare
corr
ecte
dfo
rcl
ust
erin
gat
the
firm
leve
l.a,b
,an
dc
mea
nsi
gn
ifica
nt
atth
e1%
,5%
,an
d10
%le
vel,
resp
ecti
vel
y.
∆D
∆BVE
∆MVE
EquityIss.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
∆Profits t−1
0.21
6a
0.24
8a
0.1
51b
0.3
45a
-0.0
38
0.5
95b
-0.0
46b
-0.0
74a
(3.3
)(3
.1)
(2.5
)(3
.5)
(-0.3
)(2
.5)
(-2.5
)(-
5.3
)
∆Profits
0.50
0b
0.7
83a
1.4
04a
-0.0
33a
(2.4
)(3
.2)
(4.8
)(-
3.4
)
∆Assets t−1
0.00
0-0
.067b
-0.2
53a
0.0
12b
(0.0
)(-
2.1
)(-
3.5
)(2
.5)
Constant
-16.
699
-28.
300a
3.4
34
-5.1
35
114.4
70a
118.2
30a
60.4
91a
59.4
87a
(-1.
4)(-
3.0)
(0.4
)(-
0.5
)(3
.2)
(2.9
)(6
.9)
(7.4
)
Yea
reff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Fir
meff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
R2−Adjusted
0.00
70.
041
0.0
05
0.1
00
0.0
13
0.0
53
0.0
18
0.0
27
Observations
162,
056
162
,056
162,1
30
162,1
30
157,5
50
157,5
50
162,1
54
162,1
54
38
Tab
leV
III:
Deb
tan
dE
qu
ity
Issu
ance
s,P
rofita
bil
ity
and
Levera
ge
Fact
or
Qu
inti
les
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
al
Com
pu
stat
file
sfo
rth
ep
erio
dfr
om
1971
to2009.
Th
eta
ble
pre
sents
esti
mate
s
from
regr
essi
ons
ofch
ange
ind
ebt
(∆D
),ch
ange
inth
eb
ook
valu
eof
equ
ity
(∆BVE
),ch
an
ge
inth
em
ark
etva
lue
of
equ
ity
(∆MVE
),an
d
net
equ
ity
issu
ance
s(EquityIss.)
.T
he
exp
lan
atory
vari
ab
les
incl
ud
ela
gged
pro
fita
bil
ity
qu
inti
les
an
dla
gged
leve
rage
fact
or
qu
inti
les.
Th
e
regr
essi
ons
incl
ud
eyea
rin
dic
ator
vari
able
s.T
he
rep
ort
edt-
stati
stic
sare
corr
ecte
dfo
rcl
ust
erin
gat
the
firm
leve
l.a,b
,an
dc
mea
nsi
gn
ifica
nt
atth
e1%
,5%
,an
d10
%le
vel,
resp
ecti
vel
y.
∆D
∆BVE
∆MVE
EquityIss.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pro
fita
bil
ity
Qu
inti
le15
.019
a9.5
01a
20.6
11a
15.1
50a
19.1
18a
22.6
39a
-12.0
81a
-11.4
49a
(10.
2)(6
.2)
(9.3
)(1
0.1
)(3
.2)
(5.0
)(-
8.9
)(-
12.6
)
Ind
ust
ryM
edia
nL
ever
age
Qu
inti
le0.9
95
-0.7
04
-17.9
64a
3.8
93a
(0.6
)(-
0.4
)(-
3.2
)(3
.3)
Ass
etQ
uin
tile
13.7
95a
4.8
38c
-3.7
30
-6.0
09a
(3.6
)(1
.9)
(-0.6
)(-
3.3
)
Mar
ket/
Book
Qu
inti
le11.0
28a
25.9
50a
-39.3
66a
0.0
62
(4.8
)(7
.9)
(-6.0
)(0
.1)
Tan
gib
ilit
yQ
uin
tile
-1.3
58
3.3
21
22.8
78a
3.5
85a
(-0.8
)(1
.3)
(3.9
)(2
.9)
Constant
-68.
802a
-124.6
39a
-89.7
28a
-173.8
60a
53.6
46
156.7
15b
48.8
84a
42.6
42a
(-9.
0)(-
9.4
)(-
9.6
)(-
8.1
)(1
.4)
(2.5
)(9
.6)
(4.6
)
Yea
reff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
R2−Adjusted
0.00
20.0
02
0.0
03
0.0
04
0.0
12
0.0
13
0.0
10
0.0
11
Observations
155,
587
155,0
37
155,5
87
155,0
37
155,5
87
155,0
37
155587
155,0
37
39
Table IX
Debt and Equity Issues in Good and Bad Times
The table reports the frequency and magnitude of financing activity for sorts on profits for sub-samples
of firms in good and bad times. An industry is defined as having “good times” if the median firm in that
industry has a market-to-book ratio that is higher than the 67th percentile of the time-series distribution
of industry median market-to-book ratios. Conversely, an industry is defined as having bad times if the
median firm in that industry has a market-to-book ratio that is lower than the 33rd percentile of the
time-series distribution of industry median market-to-book ratios. Panels A and B report the percentage
of firms (a) issuing net debt in excess of 5% of the value of their assets, (b) issuing net equity in excess of
5% of the value of their assets, (c) issuing debt and repurchasing equity both in excess of 5% of the value
of their assets, and (d) issuing equity and retiring debt both in excess of 5% of the value of their assets.
Panel C tabulates the mean debt and equity levels, and the mean changes in debt and equity issuances.
The sample contains non-financial firms listed on the annual Compustat files for the period from 1971 to
2009.
40
Table IX Continued
Issuing Issuing Issuing Debt Iss. EquityNet Debt Net Equity Rep. Equity Ret. Debt
(1) (2) (3) (4)
Panel A: Sorts on Profitability
Bad
Tim
es Low Profits 14.5 13.1 0.1 2.0
2 17.0 4.9 0.2 0.8
3 17.3 4.6 0.4 0.7
4 18.2 4.7 0.3 1.0
High Profits 17.7 5.1 0.6 1.0
Good
Tim
es Low Profits 20.6 38.1 0.1 4.6
2 19.7 14.8 0.2 2.3
3 21.2 10.3 0.4 1.8
4 23.4 9.2 1.0 1.8
High Profits 21.6 9.2 1.6 1.4
Panel B: Sorts on firm size and profitability
Bad
Tim
es
Sm
all
Fir
ms Low Profits 15.1 31.8 0.0 3.1
2 18.2 27.6 0.2 2.5
3 15.1 14.0 0.2 2.9
4 18.3 7.3 0.0 1.9
High Profits 17.7 8.0 0.0 1.1
Lar
gefi
rms Low Profits 17.2 5.4 0.1 0.4
2 14.8 3.9 0.4 0.5
3 15.0 4.3 0.4 0.4
4 15.6 3.3 0.3 0.6
High Profits 14.7 2.8 0.5 0.3
Good
Tim
es
Sm
all
Fir
ms Low Profits 21.3 44.8 0.1 5.4
2 23.0 44.3 0.1 4.7
3 22.2 32.5 0.0 4.7
4 20.3 19.7 0.2 3.8
High Profits 19.0 15.3 0.2 3.1
Lar
gefi
rms Low Profits 20.3 9.6 0.3 1.2
2 20.8 6.6 0.9 1.3
3 23.2 5.4 1.6 0.6
4 21.5 3.9 1.9 0.5
High Profits 23.4 3.6 3.8 0.4
41
Table
IXC
onti
nued
D∆D
BV
E∆BVE
MV
E∆MVE
Equ
ity
Iss.
Ass
ets
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elC
:M
agn
itu
de
of
Fin
an
cin
gA
ctiv
ity
(Mea
ns)
BadTimes
SmallFirms
Low
π4.
00.3
5.8
-1.4
30.5
-0.6
2.2
14.0
24.4
0.5
8.2
-0.7
24.0
6.7
1.7
18.0
34.5
0.4
9.4
-0.2
17.6
1.8
0.5
18.7
44.3
0.9
9.8
0.2
13.7
-0.4
0.3
18.8
Hig
hπ
3.6
1.0
11.4
1.3
20.9
-1.6
0.4
19.9
Largefirms
Low
π3,
364.
7-7
2.9
2,5
22.8
-169.2
2,3
04.4
163.3
94.4
8,9
71.5
23,
030.6
-95.2
2,9
83.4
-35.4
3,2
84.0
277.5
56.0
9,0
67.5
32,
930.1
4.9
3,3
59.8
-13.9
3,9
47.1
307.6
56.8
9,2
24.7
42,
280.6
-6.4
3,8
16.7
19.4
5,2
13.1
441.5
-6.6
8,9
02.2
Hig
hπ
1789
.962.0
4,7
66.9
13.0
8,7
99.1
58.3
-31.4
9,5
73.4
GoodTimes
SmallFirms
Low
π2.
40.6
3.8
0.5
33.9
2.3
3.9
9.0
22.
60.8
4.8
-0.1
30.1
-0.4
2.5
10.4
33.
00.9
6.0
0.4
24.6
-0.3
1.7
12.5
42.
80.7
6.6
0.8
20.6
1.8
1.2
13.0
Hig
hπ
2.3
0.6
8.5
1.6
26.1
1.3
0.8
14.1
Largefirms
Low
π4,
532.
236.0
3,4
76.7
-103.1
6,5
57.2
-259.5
52.9
12,4
65.1
24,
580.
357.5
3,5
27.3
-138.8
6,9
10.8
-421.4
-23.4
12,5
22.3
32,
951.
5111.4
3,4
85.1
-2.4
7,5
85.5
-278.7
-31.7
9,5
34.7
42,
752.
4124.2
3,9
38.6
71.6
10,6
85.8
-32.5
-104.5
10,2
26.0
Hig
hπ
2,42
8.5
228.1
4,8
18.9
376.5
18,5
65.1
268.2
-221.6
10,8
29.9
42
Tab
leX
:D
eb
tan
dE
quit
yC
hanges
inG
ood
and
Bad
Tim
es
–R
egre
ssio
ns
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
al
Com
pu
stat
file
sfo
rth
ep
erio
dfr
om
1971
to2009.
Th
eta
ble
pre
sents
esti
mate
s
from
regr
essi
ons
ofch
ange
ind
ebt
(∆D
)an
dn
eteq
uit
yis
suan
ces
(EquityIss.)
ingood
an
db
ad
tim
es.
Th
eex
pla
nato
ryva
riab
les
incl
ud
e
chan
gein
pro
fits
,la
gged
chan
gein
pro
fits
,an
dla
gged
chan
ge
inass
ets.
Th
ere
gre
ssio
ns
incl
ud
eye
ar
ind
icato
rva
riab
les.
An
ind
ust
ryis
defi
ned
ash
avin
g“g
ood
tim
es”
ifth
em
edia
nfi
rmin
that
ind
ust
ryhas
am
ark
et-t
o-b
ook
rati
oth
at
ish
igh
erth
an
the
67th
per
centi
leof
the
tim
e-se
ries
dis
trib
uti
onof
ind
ust
rym
edia
nm
arke
t-to
-book
rati
os.
Conve
rsel
y,an
ind
ust
ryis
defi
ned
as
hav
ing
bad
tim
esif
the
med
ian
firm
inth
at
ind
ust
ry
has
am
arket
-to-
book
rati
oth
atis
low
erth
anth
e33rd
per
centi
leof
the
tim
e-se
ries
dis
trib
uti
on
of
ind
ust
rym
edia
nm
ark
et-t
o-b
ook
rati
os.
Th
e
rep
orte
dt-
stat
isti
csar
eco
rrec
ted
for
clu
ster
ing
at
the
firm
leve
l.a,b,
an
dc
mea
nsi
gn
ifica
nt
at
the
1%
,5%
,an
d10%
leve
l,re
spec
tivel
y.
Good
Tim
esB
ad
Tim
es
∆D
EquityIss.
∆D
EquityIss.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
∆Profits
0.94
2a
0.92
2a
-0.1
13a
-0.1
39a
0.7
25
0.7
37c
-0.0
11
-0.0
12
(3.3
)(3
.0)
(-2.8
)(-
3.6
)(1
.6)
(1.7
)(-
0.6
)(-
0.7
)
∆Profits t−1
0.22
8c
-0.2
08a
0.2
44
-0.0
21
(1.7
)(-
3.5
)(1
.4)
(-1.4
)
∆Assets t−1
0.03
1b
0.0
19b
0.0
66
0.0
05
(2.3
)(2
.4)
(1.0
)(0
.5)
Constant
92.3
46a
-33.
480c
13.1
73c
7.8
77c
-7.3
12
14.0
27
3.4
71b
7.5
45
(2.7
)(-
1.9)
(1.8
)(1
.7)
(-0.8
)(1
.5)
(2.1
)(1
.5)
Yea
reff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
R2−Adjusted
0.08
60.
098
0.0
20
0.0
58
0.2
23
0.2
57
0.0
12
0.0
13
Observations
74,8
2768
,729
74,8
91
68,7
87
26,9
65
26,0
13
26,9
71
26,0
18
43
Tab
leX
I:Sca
led
issu
ance
s
Th
esa
mp
leco
nta
ins
non
-fin
anci
alfi
rms
list
edon
the
an
nu
al
Com
pu
stat
file
sfo
rth
ep
erio
dfr
om
1971
to2009.
Th
eta
ble
pre
sents
esti
mate
sfr
omre
gres
sion
sof
deb
tis
suan
ceto
tota
lis
suan
ce(i
nco
lum
n(1
)),
equ
ity
issu
an
ceto
tota
lis
suan
ce(i
nco
lum
n(2
)),
deb
tis
suan
ceto
tota
lca
pit
al(i
nco
lum
n(3
)),
equ
ity
issu
ance
toto
tal
cap
ital
(in
colu
mn
(4))
,ch
an
ge
ind
ebt
toca
pit
ali
zati
on
rati
o(i
nco
lum
n(5
)),
an
dch
an
ge
inle
vera
gera
tio
con
sid
erin
gon
lyex
tern
alfi
nan
cin
g(i
nco
lum
n(6
)).
Th
ech
an
ge
inle
ver
age
rati
oco
nsi
der
ing
on
lyex
tern
al
fin
an
cin
gis
defi
ned
as:
∆LeverageE
F=
Dt−
1+d
Dt−
1+E
t−1
+d
+e−
Dt−
1
Dt−
1+E
t−1,
(2)
wh
ere
Dis
deb
t,E
isb
ook
valu
eof
equ
ity,
dis
deb
tis
sues
net
of
reti
rem
ents
,an
de
iseq
uit
yis
sues
net
of
rep
urc
hase
s.T
he
regre
ssio
ns
incl
ud
e
lagg
edle
vera
gefa
ctor
san
dye
arin
dic
ator
vari
able
s.T
he
rep
ort
edt-
stati
stic
sare
corr
ecte
dfo
rcl
ust
erin
gat
the
firm
leve
l.a,b,
an
dc
mea
n
sign
ifica
nt
atth
e1%
,5%
,an
d10
%le
vel,
resp
ecti
vely
.
DebtI
ss
TotI
ss
EquityIss
TotI
ss
DebtI
ss
(D+E)
EquityIss
(D+E)
∆D
D+E
∆LeverageE
F
(1)
(2)
(3)
(4)
(5)
(6)
∆Profitability
t−1
0.04
4a
-0.0
44a
-0.0
15a
-0.0
67a
-0.0
33a
0.0
14b
(3.5
)(-
3.5
)(-
3.4
)(-
8.3
)(-
3.5
)(2
.3)
∆Ind.medianbookleverage t−1
0.04
2c
-0.0
42c
-0.0
07
-0.0
15a
0.0
01
0.0
01
(1.9
)(-
1.9
)(-
1.5
)(-
3.2
)(0
.2)
(0.3
)
∆Market/bookt−
1-0
.007
a0.0
07a
0.0
02a
0.0
08a
-0.0
04a
0.0
00
(-4.
7)(4
.7)
(6.8
)(1
0.6
)(-
5.5
)(0
.5)
∆Tangibility t−1
-0.0
370.0
37
0.1
25a
0.1
02a
0.1
05a
0.0
33a
(-1.
2)(1
.2)
(14.4
)(9
.8)
(7.9
)(4
.4)
∆Assets t−1
-0.0
78a
0.0
78a
0.0
57a
0.0
17a
0.0
15a
0.0
25a
(-9.
7)(9
.7)
(28.3
)(7
.0)
(5.3
)(1
4.7
)
Constant
0.78
2a
0.2
11a
0.0
12a
0.0
15a
0.0
00
0.0
03b
(46.
3)(1
3.3
)(6
.7)
(11.8
)(0
.0)
(2.2
)
Yea
reff
ects
Yes
Yes
Yes
Yes
Yes
Yes
R2−Adjusted
0.01
50.0
15
0.0
22
0.0
25
0.0
09
0.0
12
Observations
133,
905
133,9
06
140,5
52
140,7
33
138,2
29
140,5
72
44
Table XII
Externally Financed Leverage Changes in Good and Bad Times
The sample contains non-financial firms listed on the annual Compustat files for the period from 1971 to2009. The table presents estimates from regressions of change in debt to capitalization ratio (in columns(1) and (3)) and change in leverage ratio considering only external financing (in columns (2) and (4)).The change in leverage ratio considering only external financing is defined as:
∆LeverageEF =Dt−1 + d
Dt−1 + Et−1 + d+ e− Dt−1
Dt−1 + Et−1, (3)
where D is debt, E is book value of equity, d is debt issues net of retirements, and e is equity issues net of
repurchases. The regressions include lagged leverage factors and year indicator variables. The reported
t-statistics are corrected for clustering at the firm level. An industry is defined as having “good times”
if the median firm in that industry has a market-to-book ratio that is higher than the 67th percentile of
the time-series distribution of industry median market-to-book ratios. Conversely, an industry is defined
as having “bad times ” if the median firm in that industry has a market-to-book ratio that is lower than
the 33rd percentile of the time-series distribution of industry median market-to-book ratios. a, b, and c
mean significant at the 1%, 5%, and 10% level, respectively.
Good Times Bad Times
∆ DD+E ∆LeverageEF ∆ D
D+E ∆LeverageEF
(1) (2) (3) (4)
∆Profitabilityt−1 -0.022c 0.016c -0.073b -0.007(-1.7) (1.8) (-2.5) (-0.4)
∆Ind.medianbookleveraget−1 0.000 -0.001 -0.010 0.001(0.0) (-0.1) (-0.7) (0.2)
∆Market/bookt−1 -0.004a 0.000 -0.008a -0.000(-4.0) (0.5) (-2.8) (-0.3)
∆Tangibilityt−1 0.097a 0.033a 0.127a 0.059a
(4.5) (2.7) (4.9) (4.3)
∆Assetst−1 0.014a 0.024a 0.037a 0.036a
(3.0) (8.4) (4.5) (8.8)
Constant -0.009 -0.000 0.028c -0.021(-1.2) (-0.0) (1.7) (-1.1)
Year effects Yes Yes Yes Yes
R2 −Adjusted 0.006 0.009 0.026 0.023
Observations 56,820 58,200 23,462 23,511
45