Corporate Leverage:How Much Do Managers Really Matter?

Murray Z. Frank and Vidhan K. Goyal

August 21, 2007

ABSTRACT

This paper studies the effect of top managers on corporate financing decisions.

Differences among CEOs account for a great deal of the variation in leverage

among firms. After a CEO is forced out, leverage typically declines. Firms that

offer higher pay-for-performance to the top executives adjust leverage to target

more rapidly. CEO personal characteristics are not closely connected to corporate

leverage choices. To some extent the CEO may be serving as a proxy for an entire

management team. The CFO seems to play at least as important a role as the CEO

in determining corporate leverage.

JEL classification: G32

Keywords: Capital structure, behavioral finance, executive compensation con-

tracts, corporate governance.

Murray Z. Frank is the Piper Jaffray Professor of Finance, Carlson School of Management, University ofMinnesota, Minneapolis, MN 55455. E-mail: Murray.Frank@csom.umn.edu.

Vidhan K. Goyal is with the Department of Finance, Hong Kong University of Science and Technol-ogy, Clear Water Bay, Kowloon, Hong Kong. Phone: +852 2358-7678, Fax: +852 2358-1749, E-mail:goyal@ust.hk.

Acknowledgments: We thank Raj Aggrawal, Paul Povel, Michael Roberts, Masako Ueda, and partici-pants at the European Central Bank, Helsinki School of Economics, South Carolina, and the Minnesota-Wisconsin Corporate Finance Conference for helpful comments. Murray Z. Frank thanks Piper Jaffrayfor financial support. Vidhan K. Goyal thanks the Research Grants Council of Hong Kong for financialsupport. We alone are responsible for any errors. c© 2007 by Murray Z. Frank and Vidhan K. Goyal. Allrights reserved.

I. Introduction

When one CEO is replaced by another, the costs and benefits of corporate debt such as

tax savings or deadweight costs of bankruptcy do not change. Accordingly top man-

agement change should have little effect on corporate leverage. Yet there is a growing

literature claiming that CEOs are a key element for an understanding of firms (eg.

Bertrand and Schoar (2003), Cadenillas et al. (2004), Malmendier and Tate (2005), and

the survey by Baker et al. (2007)). How important is the CEO or the CFO for choosing

leverage? Are differences among executives a first order concern for corporate lever-

age?

If a manager runs a firm for a long time, then a managerial fixed effect will be hard

to tell apart from a firm fixed effect. If managers are optimally chosen by the board of

directors, it is possible that the manager characteristics are intended to be well suited to

firm characteristics. Worse still, we can only observe a subset of the characteristics that

the board of directors observes. This makes it unlikely that managerial characteristics

would be independent of the error term in a leverage regression.

To partly get around these problems we focus on corporate leverage and manager

turnover events.1 While there may be some change, the firm’s basic characteristics and

opportunities are not likely to be fundamentally changed around a turnover event. (A

steel company does not suddenly become a biotechnology firm when a new CEO takes

over the firm.) If the manager characteristics are unimportant then no change in lever-

age will be observed when the CEO changes. However, if managerial characteristics

drive leverage choices, leverage might change after the turnover event.

We distinguish cases of routine turnover from cases of forced turnover. Rou-

tine turnovers typically happen when the executive reaches normal retirement age or

leaves the firm because of health problems. Forced retirements are cases in which the

manager seems to have been fired. Incoming CEOs should be optimally selected in

1There are a number of previous studies of top manager turnover. For example, Denis and Denis(1995) and Huson et al. (2004) show that forced CEO replacements are often a result of poor operatingperformance. Much of the literature focuses on labor economics issues. Fee and Hadlock (2004) showthat other executives are often replaced when an outside CEO is brought in. They also study the type ofjobs that the displaced executives find. Parrino (1997) studies inside versus outside succession.

1

either case. Outgoing CEOs may have been fired due to suboptimal performance.2 If

the CEO was fired, then we expect the new CEO to try to fix things. If leverage is one

of the things that needs fixing, we should see changes in leverage. If the CEO turnover

was routine, then it is not clear that anything needs to be fixed and less change might

be observed.

In addition to leverage, we also examine leverage adjustment speeds. Organiza-

tional politics may also play a role here. A new CEO may have an initial ‘honeymoon’

period during which it is easier to make changes.3 Afterwards organizational inertia

may tend to set in. This suggests that leverage should adjust more rapidly after a new

CEO comes in. Offsetting this is the need for a new CEO to ‘learn the ropes’. Thus an

outside CEO might have a lagged adjustment relative to an inside CEO.

The evidence does support the idea that managerial behavior matters for corporate

leverage. This shows up in several ways.

1. A CEO fixed effect plays a strong role in a leverage regression. The CEO effect

and the firm fixed effect are empirically closely connected. Adding either to a leverage

regression increases the explanatory power substantially. Adding a firm fixed effect to

a leverage regression that includes a CEO effect adds less than a 1% improvement in

the R2, while adding a CEO effect to a leverage regression that includes a firm fixed

effect improves the R2 by about 4.5%. Parallel effects are found in leverage adjustment

models.

This finding potentially opens the door to an economic interpretation of the results

in Lemmon et al. (2007). They found that firm fixed effects dominate conventional

leverage factors in their ability to account for cross-sectional differences in leverage.

Our results support their claims. They do not provide an economic interpretation of

2Consistent with this idea we find that the characteristics and contracts given to the incoming CEOsare fairly similar in a forced turnover and in a routine turnover. In a forced turnover the departing CEOis generally more poorly paid than the incoming CEO. In a forced turnover the incoming CEO tends tohave a slightly higher pay for performance contract than does the incoming CEO in a routine turnover.

3For much the same reason it is common for newly elected politicians to attempt to affect changequickly. For instance think of the “first hundred hours” of the newly empowered Democrats in the UScongress after the election of 2006.

2

this fixed effect. Our evidence suggests that most of what they are finding probably

reflects the effect of managerial behavior.4

Further evidence of the importance of top management is found by looking at man-

agerial turnovers. Typically when a manager is forced out, leverage has been above

normal, particularly for the previous year or two. This is consistent with the idea that

debt might have been accumulating due to poor corporate performance as shown by

a number of papers including Denis and Denis (1995) and Huson et al. (2004). The

manager is replaced by the board in hopes of improved performance.

If the firm’s performance has been satisfactory, and a turnover event is needed,

it is likely to be a routine turnover rather than a forced turnover. Accordingly, it is

interesting to find that firms with routine turnovers have faster target adjustment of

leverage both before and after the turnover.

Additional evidence of the importance of managers can be found by looking at

the impact of executive compensation on leverage adjustments. Higher pay for perfor-

mance (Core and Guay, 2002) is associated with more rapid target adjustment behavior.

In other words, when the CEO’s pay check is at stake the firm seems to adjust leverage

more rapidly. When the CEO has less at stake, slower leverage adjustment is observed.

2. In an attempt to explore why managers make differing leverage choices we put

a fair bit of effort into an attempt to document the impact of managerial characteris-

tics on corporate leverage. Despite gathering considerable personal information about

the executives, this proves problematic. The panel data set that we gathered is large

enough that some statistically significant results are obtained. However these personal

attributes have limited ability to explain differences in financing decisions across firms.

Leverage choices are not all that closely connected to readily observable managerial

traits.

This could result from a fundamental disjunction between observable CEO charac-

teristics and CEO preferences over corporate leverage. If so, then there is little more

to be done beyond directly measuring the CEO’s beliefs. An alternative possibility is

4Baker and Wurgler (2002) show that differences in timing of security issuances have a persistenteffect on a company’s capital structure. But Lemmon et al. (2007) argue that capital structure differencesare permanent rather than time varying. Our evidence suggests there may be a very natural reconcilia-tion between the Baker and Wurgler (2002) and the Lemmon et al. (2007) results in terms of changes intop executives.

3

that the CEO fixed effect is actually serving as a proxy for something deeper and more

complex than the personality of a single person. It is likely that the managerial team

more broadly interpreted matters. (Fee and Hadlock, 2004) show that other top execu-

tives commonly change jobs at about the same time as the CEO. Hence the CEO fixed

effect might be serving as a proxy for a deeper and more complex set of relationships

among a group of top managers.

Such a hypothesis is inherently hard to measure. In order to get at it to some degree

we tried introducing the CFO into the analysis. If the CEO fixed effect is really mea-

suring the actions of the CEO then adding a CFO into the picture ought not to matter.

If however, leverage is chosen by more than just the CEO, then the CFO effect might

improve the fit.

3. The CEO and the CFO fixed effects are closely connected. To the extent that we

can tell them apart, it seems to be the CFO rather than the CEO that plays the more

important role in determining corporate leverage. This shows up in several ways.

Leverage adjustment equations fit better when both fixed effects are included. The pay-

for-performance sensitivity of leverage adjustment is quite a bit greater for the CFO

than it is for the CEO, and the CFO equations actually explain more of the variation in

the data. It is not just the CEO that matters.

Our bottom line is that the behavior of the top executives really matters for cor-

porate leverage. This is a first order effect. Some studies stress the impact of the CEO.

Given the high profile, and ready availability of data this emphasis is quite understand-

able. However, it may also be misleading. The CEO does not operate in a vacuum. It is

not just the CEO who is driving capital structure decisions. The CFO also matters. In

fact the CFO appears to matter more than does the CEO for how the firm is financed.

Given the job titles, perhaps that should not be too surprising.

II. Data Description

The sample consists of chief executive officers (CEOs) of S&P500, S&P MidCap400,

and S&P SmallCap 600 firms listed in Standard and Poor’s Execucomp database. The

sample period is from 1993 to 2004. The financial data are from the Compustat files

4

and are deflated to constant 2000 dollars using the GDP deflator. Financial firms, firms

involved in major mergers (Compustat footnote code AB), and firms with missing book

value of assets are excluded. The final sample has 16,929 observations on 2,248 firms

and 3,898 CEOs. Appendix A describes the key variables with relevant Compustat

data item numbers. The variables are winsorized at the 0.50% level in both tails to

replace outliers and the most extremely misrecorded data.

Appendix A also reports descriptive statistics for the leverage ratio and leverage

factors for the matched Execucomp-Compustat CEO sample. A comparison of these

statistics with a broader Compustat sample reported in Frank and Goyal (2007) shows

that Execucomp firms are relatively large, have higher profitability and collateral, and

more frequently pay dividends. However, leverage ratios are not all that different.

III. Corporate Leverage and CEO Fixed Effects

A. Cross-sectional estimates

Let TDMi,t denote the market leverage ratio for firm i at time t. Let Firm/ManagerFEi

denote the firm or manager fixed effect, as appropriate. Let FGi,t denote the seven

conventional leverage factors as in Frank and Goyal (2007). Then the regression is,

TDMi,t = α + µt + Firm/ManagerFEi + βF FGi,t + εi,t, (1)

Column (1) of Table I reports estimates from leverage regressions for the Execucomp

sample.5 The signs and statistical significance of the leverage factors is the same as in

Frank and Goyal (2007).

Lemmon et al. (2007) have found that a firm’s initial leverage is very important for

determining its subsequent leverage. They report that this effect matters even decades

later. In column (2), we therefore examine how important a firm’s initial leverage is

5Recently Petersen (2007) has argued for the use of clustered standard errors in this kind of setting.We therefore reports standard errors corrected for clustering at the firm level.

5

in determining leverage at any future point in time.6. As reported by Lemmon et al.

(2007), the initial leverage is statistically significant. Leverage is persistent. However

this does not have a large impact on the other factors, and it does not result in a huge

improvement in the fraction of the variation accounted for.

In column (3), we introduce firm fixed effects into the analysis. Consistent with

Lemmon et al. (2007), the results show that the firm fixed effects account for a great deal

of the variation in the data. The R2 increases from 0.461 to 0.774. All of the coefficients

on the conventional factors retain the same sign, and all remain statistically significant.

As might be expected, the firm fixed effect is much more powerful than the initial

leverage. The financial interpretation of the firm fixed effect is not entirely clear.

Columns (1) through (3) show that our data provides results very similar to those

reported in the previous literature. We therefore test for CEO effects by removing the

firm fixed effect and instead including a CEO fixed effect.

If managers are unimportant, then column (4) should fit much worse than columns

(2) and (3). In an extreme case one might even be back with a fit akin to column (1).

If managers are serving as noisy proxies for firm fixed effects, then the R2 should be

somewhere between the fit in column (1) and the fit in column (3). If, however, the

firm fixed effect is serving as a noisy proxy for a CEO effect, then the R2 in column (4)

should exceed that in column (3).

The introduction of CEO fixed effects does prove to be important. The R2 increases

to 0.811. In most cases the coefficients on the individual factors are largely unaffected

(except for dividends). The CEO effect is not just a noisy proxy for a firm effect. If

anything the reverse seems to be true. When we add both types of fixed effects at the

same time (column 5) the R2 = 0.819, which is almost the same as when we have only

the managerial fixed effects. From Table I we find that the firm fixed effect does not

contain very much information that is not already contained in the CEO fixed effect.

6Initial leverage is measured as the firm’s leverage in the first year for which it has reported Compu-stat data

6

B. Partial Adjustment Estimates with Manager Effects

A variety of studies have focused on leverage adjustments rather on leverage levels.

There has been a debate about how rapidly leverage converges towards the long run

target. Flannery and Rangan (2006) show that including firm fixed effects significantly

improves leverage adjustment speeds. The idea is that in a pooled regression it is

inevitable that the target is measured with error. The firm fixed effect will improve on

the estimated firm target and thus improve the estimate of the speed of adjustment.

Table II considers the impact of the CEO on the speed of leverage adjustment. As

with the analysis in Table I we start by verifying that our data provides estimates that

are basically similar to what is reported in the previous literature.

Column (1) reports estimates from pooled OLS and provide a benchmark for ad-

justment speeds in the absence of either firm-or manager-fixed effects. As found in

previous studies such as Fama and French (2002) the coefficient on lagged leverage is

fairly low (−0.150). This has been interpreted to mean that firms adjust leverage ex-

tremely slowly (‘a snail’s pace’). Column (2) is similar except that we correct standard

errors for clustering at the firm and year level.

Flannery and Rangan (2006) found that introduction of firm fixed effects lead to a

significant increase in the estimated speed of leverage adjustment. Adding firm fixed

effects substantially increases the adjustment speeds. In untabulated regressions we

excluded that leverage factors and find that the coefficient on lagged leverage is about

0.41. In column (4) we introduce a firm fixed effect, and find the same result that they

reported. There is a dramatic increase in the coefficient on lagged leverage (−0.458).

This is reassuring in that it means that our data is basically similar.

For completeness in column (3) we follow Lemmon et al. (2007) by including the

initial leverage as a factor. It proves to have very little effect on the explanatory power

and on the speed of adjustment. It seems that the adjustment is more a matter of last

period’s leverage rather than the distant past.

This positions us to examine the importance of the CEO effect relative to the firm

effect. This is carried out in columns (5) and (6). In column (5), we include CEO fixed

effects and find that it increases the adjustment speeds to about 55%. This is about 20%

(55% versus 46%) increase in relative terms. More importantly the amount of explained

7

variation increases quite a bit. With neither a firm fixed effect, nor a CEO fixed effect

(columns 1 and 2) a bit less than 11% of the variation is explained. Adding a CEO fixed

effect increases this to 28.6%. In other words it more than doubles. Once again the firm

fixed effect can explain some of the CEO fixed effect. Adding both fixed effects at the

same time (column 6) increases the adjustment speeds to 57% and the R2 to 0.302.

As in the cross section regressions, inclusion of a firm effect without the CEO effect

improves the fit of the equation quite a bit. Once again the bulk of this improvement

is accounted for by the CEO effect. Managers matter. Omitting managers from the

analysis is potentially misleading. Omitting firm fixed effects also matters somewhat.

IV. Leverage around CEO turnover

The importance of the CEO effects must be driven by changes in CEO since for any

given CEO the effect is a constant. Accordingly we now focus on the turnover process

itself. We consider the role of forced turnovers when compared to normal turnovers,

and also insider replacements versus outsider replacements. To do this we use dummy

variables and their interaction effects.

We follow CEOs of firms in Execucomp during the 1993-2004 period and define CEO

turnover as any change in the identity of the CEO. Of the 2,248 firms in the sample,

1,237 firms experienced at least one turnover. These 1,237 firms experienced a total of

1,885 CEO changes – 764 firms had one CEO change, 331 firms had two changes, 115

firms had three changes, 21 firms had four changes and 6 firms had five changes. The

annual CEO departure rate is a little over 11%, which is roughly consistent with that

found in previous studies such as Denis and Denis (1995) and Weisbach (1988).

For each of the 1,885 CEO changes, we search the Lexis/Nexis database for news

articles describing the turnover. We could find news articles describing 1,755 of these

changes. A full reading of these articles allows us to record the reason for the depar-

ture, the ages of both the departing and the incoming CEO, and the previous employ-

ment of the new CEO.

The departures are broadly classified as either normal or forced. In classifying these

changes, we follow the criteria described in (Fee and Hadlock, 2004; Parrino, 1997;

8

Huson et al., 2001; Lehn and Zhao, 2006; Kaplan and Minton, 2006; Bhagat and Bolton,

2006). Turnovers are classified as routine when the CEO (i) leaves the firm because of

poor health or dies suddenly, (ii) has reached age 63 and is retiring following normal

retirement policies, (iii) is less than 63 but a succession plan was in place for at least

6 months before the CEO departure date, and (iv) resigns but continues as chairman

for at least a year following the resignation. Turnovers are classified as forced if (i) the

departure is described as ‘forced’, ‘ousted’ or as a part of ‘management shakeup’, or if

the CEO leaves because of a fraud or a scandal, policy differences with the board, or for

poor performance, (ii) CEO is leaving the firm to pursue other interests or for personal

reasons, (iii) the company provides no reason for the departure of a CEO less than 63

years old, and (iv) CEO leaves following a spinoff, asset sale, merger or other corporate

restructuring transactions. We summarize the reasons for turnover in Appendix B.

In column (1) of Table III we see that leverage is not significantly different before

and after a normal turnover. This is not too surprising. The results for forced turnovers

are another matter entirely. Before a forced turnover, leverage is significantly elevated.

This is consistent with studies such as Huson et al. (2004) who show that poor firm

performance can lead to a forced turnover. After a forced turnover leverage is not

unusual.

In columns (2) and (3), we look at narrower definitions of forced turnovers. Column

(2) defines forced turnovers excluding CEO changes because of mergers and corpo-

rate restructuring transactions. Column (3) uses a more restrictive definition of forced

turnovers concentrating on the terminations only. These narrower definitions do not

fundamentally affect our conclusions.

In column (4) we consider the distinction between insider and outsider replace-

ments. When the turnover is routine, this distinction proves to be unimportant for

leverage. When the turnover is forced, we find that leverage is again significantly el-

evated prior to turnover. However, when outsiders are brought in following a forced

replacement, the new CEOs take the leverage back down to what would otherwise

have been expected. When insiders replace outgoing CEOs in forced turnovers, lever-

age goes down, but not quite as strongly. They seem to leave the leverage a little bit

elevated when compared to what would otherwise have been expected. So the distinc-

9

tion between insider and outside replacements does some to matter, at least to some

degree.

In Table IV, we consider the effect of managerial turnover on the speed of lever-

age adjustments. As might be expected from the previous table, when the turnover is

routine, the speed of leverage adjustment do not differ significantly from firms experi-

encing no CEO changes. By contrast, we observe large increases in leverage adjustment

speeds after forced turnover. Firms which experience forced turnover exhibit signifi-

cantly slower leverage adjustment speeds. The estimates suggest that all else equal,

firms with forced turnovers have adjustment speeds that are about 16% slower (49%

versus 58% for no turnovers). Following the turnover, the adjustment speeds become

normal.

The last column of Table IV examines if changes in adjustment speeds around

turnovers depend on whether the new CEO is an insider or an outsider. The results

show that adjustment speeds are slower than normal prior to forced turnovers regard-

less of whether the replacement is from inside the firm or from outside. The adjustment

speeds increase to normal in both cases following the CEO change.

From Tables III and IV we learn that forced replacements of CEOs have materially

different effects on leverage when compared to routine replacements. We also learn

that insider replacements of CEOs are materially different from outsider replacements.

This serves to enhance our confidence in the importance of managers for corporate

leverage choices.

V. Leverage and Incentives

Results in Section III show that manager fixed effects are important. But they cannot

explain why. This is a hard thing to tie down. There is not much theory to guide us,

and we do not know all that much about what motivates CEOs. The main exception

comes from the literature on executive compensation. From that literature we have the

basic idea that executives respond to incentive contracts. Accordingly in this section,

we analyze the effects of incentives and ownership on leverage levels and on leverage

adjustment speeds.

10

We use information on managerial compensation contracts to construct proxies for

managerial incentives. Following existing literature, incentives are measured as the

change in CEO’s firm-specific wealth for every one thousand dollar increase in share-

holder wealth. The higher the pay-performance sensitivity, the greater are the man-

agerial incentives to increase firm value. Both stockholdings and option portfolios are

used to estimate pay-performance sensitivities since Core and Guay (2002) and Jensen

and Murphy (1990) argue that incentives from stock and options grants and holdings

are much greater than those provided by direct compensation in the form of salary and

bonus.

The option pay-performance sensitivity is the change in the value of option portfo-

lio for every $1000 increase in shareholder value. The stock pay-performance sensitiv-

ity is the change in the value of managerial stockholdings for every $1000 increase in

shareholder value. An alternative measure of incentives from stockholdings is the frac-

tion of shares owned by executives see (Demsetz and Lehn, 1985; Jensen and Murphy,

1990; Yermack, 1995).

We value new option grants and option portfolios of unexercised exercisable and

unexercised unexercisable options following Core and Guay (1998). Their method is

relatively simple and they show that the method yields estimates of equity incentives

that are unbiased and 99% correlated with values that would be obtained if the param-

eters of CEO’s option portfolio were actually known. The procedures and data sources

are described in Appendix C. The option portfolio sensitivities and stockholding sen-

sitivities are combined to estimate total pay to performance sensitivities.

Appendix C provides summary statistics on total direct compensation, wealth gains

on option and stock portfolios, and pay-performance sensitivities for the CEOs. The

median annual flow compensation for CEOs is over $2 million consisting of $525 thou-

sand in salary, $300 thousand in bonus, and about $640 thousand in option grants

(valued using the Black and Scholes method). The rest comes from other compen-

sation and long-term payouts. While the median wealth gains from both stock and

option holdings are zero, the averages are quite large suggesting that some executives

are paid very large amounts.

The descriptive statistics on pay-performance sensitivity suggest that stockholding

provide much larger incentives compared to option holdings. The average stock pay-

11

performance sensitivity is about $26 for every $1,000 increase in shareholder wealth

while that from option holdings is about $7.7. The medians are considerably smaller.

The median stock and option pay-performance sensitivities are roughly $2.5 and $1.6,

respectively. Together, stocks and option holdings increase CEO wealth by a median

of $9.6 for every $1,000 increase in shareholder wealth. CEOs own a median of 0.36%

stock (the average is about 3%).

A. Effect of incentives and ownership on leverage

Table V examines the effect of incentives on the market leverage ratio. The model

is essentially the same as in Table I. However the pay-for-performance factors are

included. All of the regression estimates include firm and year fixed effects and report

t-statistics corrected for heteroscedasticity. The coefficients on the standard factors are

very similar to what is reported in Table I, and so to save space those coefficients are

not reported in the table.

In Table I we see that most of the effect of incentives on leverage ratios comes

from option holdings. In column (3) we see that the coefficient on the total pay-to-

performance sensitivity is negative and statistically significant at the 1% level. Man-

agers who have high powered incentives seem to choose more conservative leverage.

Columns (1) and (2) show that much of the managerial effect is driven by the incentives

provided by option holdings rather than by stocks.

In column (4), we directly examine if managerial stock ownership affects leverage.

Ownership is highly non-linear with a majority of executives owning small amounts of

equity but a few who own substantial amounts. There is a large literature that argues

that ownership has non-linear effects on firm performance with ownership providing

incentives at low levels and entrenching managers at higher levels. The results re-

ported in column (4) are consistent with this non-linear effect of ownership. At low

leverage levels (managerial stockholdings < 1%), leverage declines as ownership in-

creases. At intermediate ownership levels (between 1% and 5%), leverage is unrelated

to ownership and at high ownership levels (more than 5%), leverage increases with

ownership.

12

B. Effect of incentives and ownership on adjustment speeds

Table VI examines the effect of incentives and ownership on adjustment speeds. We

include interactions between various measures of pay-performance sensitivity and

lagged leverage to the baseline partial adjustment regression.

The results in columns (1) to (3) show that CEO incentives have statistically signif-

icant and economically large effects on leverage adjustment speeds. Higher stock and

option pay-performance sensitivities both significantly increase adjustment speeds.

The effect of option pay-performance sensitivity on leverage adjustment speeds is

larger than the stock pay-performance sensitivity. So as in the results in Table I, we

again find that the CEO behavioral response seems to be being driven by options.

In Column (4), we interact the ownership structure with lagged leverage to examine

the effect of ownership on adjustment speeds. As before, we construct three different

interaction terms to examine the non-linear effect of ownership on leverage adjustment

speeds. The first interaction term multiplies ownership with lagged leverage for man-

agers who own less that 1% of the equity in the firm. This interaction is zero for other

managers. The second interaction term captures managerial ownership in the range of

1% to 5% while the last interaction term captures managerial ownership greater than

5%. CEO ownership does not have much of an effect on leverage adjustment speeds.

In summary, while incentive contracts for CEOs, particularly option-based incentives,

do affect speed of leverage adjustment, CEO ownership does not.

VI. Manager Characteristics and Leverage

A number of behavioral finance studies have focused attention on the characteristics of

managers (Bertrand and Schoar, 2003; Malmendier and Tate, 2005; Baker et al., 2007).

To get a deeper understanding of why different managers make different choices, one

potential avenue open to us is to look at managerial characteristics. It has been argued

that certain managerial characteristics are associated with conservatism. For instance

it has been suggested that female managers, older managers, and managers with fi-

nance background might behave differently. Could one associate certain managerial

13

characteristics with different leverage levels or faster or slower leverage adjustment

speeds?

This section presents results on how observable executive characteristics affect

leverage. We analyze the effects of age, educational background and employment his-

tories. We hand-collect personal information on CEOs’ and CFO’s of Execucomp firms

from various editions of “Who’s Who in Finance and Industry” executives bios on

company web-sites, biographical information contained in news releases, and profiles

posted on Internet. The key data items are the year in which the executive was born,

tenure with the firm and in the current position, employment histories and educational

backgrounds. The variables are described Appendix D.

The CEO characteristics sample is an unbalanced panel (1993-2004) with 11,797

firm-years. There is information on 2,702 executives at 1,716 firms. Executive age is

estimated from information about the year in which the person was born. Similarly,

tenure with the firm and tenure as CEO/CFO is estimated from the year in which the

executive joined the firm and the year in which the executive assumed his/her current

position. Executives are classified based on their educational backgrounds and their

employment histories.

An executive is coded as having a finance education if they have an undergraduate

or graduate degree in finance and/or accounting. They have a technical education if

they hold undergraduate or graduate degree in natural sciences, math or engineering.

We separately identify CEOs with an MBA. Based on their employment background,

CEOs are classified as having a finance career in the finance function with their cur-

rent employer or in their previous employment or worked for a finance, auditing or

accounting firm. They have a technical career if they worked as an engineer or as a

technically oriented professional.

Table VII presents summary statistics for the CEO characteristics. The median CEO

in this sample is about 56 years old, has been at the firm for about 14 years, five of

which are in the CEO position. About 1.3% of the CEOs are female. Almost 38% have

an MBA, 6% have an undergraduate or graduate degree in finance or accounting, 40%

have a technical education and about 9% have a law degree. More than 20% have

previous finance exposure in their career. We tabulate routine versus forced turnover

CEOs, incoming versus outgoing CEOs, insiders versus outsiders. Not surprisingly

14

incoming CEOs are younger than outgoing CEOs. While the average CEO has a tenure

of 7.8 years, those who are forced out have shorter tenure than those who are replaced

in a routine manner.

Perhaps the most striking observation in VII is the comparisons of the incentive

packages. As a reflection of their time at the firm, outgoing CEOs commonly have

more stockholdings and more options than do the incoming CEOs. Accordingly the

pay-for-performance sensitivity typically drops dramatically when a CEO is replaced.

Presumably in an effort to mitigate this the Back-Scholes value of options given to

the incoming CEO is typically larger than the value of the outgoing CEO. This effect

is particularly strong when the incoming CEO is an outside replacement in a forced

turnover.

How much of an effect do the CEO characteristics have on leverage? Table VIII

presents the results. The strongest effect is found in the length of CEO tenure. The

longer the CEO has been in charge, the lower the leverage. CEOs who have worked at

more companies previously, have an MBA degree, or a law degree tend to have greater

leverage. These effects are statistically significant and account for some variation in

the data, but not all that much.

Table IX presents partial adjustment regressions with interactions between CEO

characteristics and lagged leverage. The results show that CEOs with an MBA have

slightly higher speeds of adjustment. An MBA degree adds about 9% to adjustment

speed. Older CEOs adjust less. Employment background has no apparent effects.

VII. CFOs and leverage

The fact that CEO characteristics do not account for all that much of the variation in

the data is worrying. Perhaps the effects that we have been attributing to the CEO are

not truly determined by the CEO. Perhaps the CEO fixed effect is actually proxying for

something else. But what could that be? Recall that Fee and Hadlock (2004) report that

other executives are often replaced when an outside CEO is brought in. It is possible

that leverage is actually being driven the Chief Financial Officer, and we then attribute

this role to the CEO because they are correlated.

15

To test this idea, we identify finance officers from among the list of highest-paid

executives reported in Execucomp (taken from a firm’s proxy statements). Not every

firm has a finance officer among the highest paid executives. We are able to identify

12,848 Finance Officer/Years. This consists of 2,140 firms and 3,743 CFOs over the pe-

riod 1993-2004. Finance officers are most often designated as ‘CFOs’ (or ‘Chief Finance

Officers’ or ‘Chief Financial Officers’). In our sample, about 90% of all finance officers

had a CFO designation. If a CFO is not listed, we search for ‘VP-Finance’ and then

‘Treasurer’ and then ‘Controller’. The VP-Finance (but no CFO) appears in the list of

highly paid executives about 6.5% of the times. Treasurer (but no CFO or VP-Finance)

accounted for another 2.4% of the cases. Controllers (without a CFO, or VP-Finance,

or Treasurer) in the top executive list are not common. Given the dominance of CFO

designation, we will refer to finance officers as CFOs in the rest of the paper.

How well are the CFOs compensated? They receive less than half as much as CEOs

– the median annual flow compensation for CFOs is $863 thousand consisting of $255

thousand in salary, $115 thousand in bonus, and $268 thousand in new option grants.

The total pay-performance sensitivity of CFOs is a median of $1.4 for every $1000 in-

crease in shareholder wealth. This is much lower than $9.6 for the CEOs. These dif-

ferences are consistent with Aggarwal and Samwick (2003) who argue that the level of

responsibility determines the pay-performance sensitivity.

Compared to CEOs, the CFOs are younger (median CFO age is 49 years). The

median CFO tenure with the firm is 5 years, which is considerably shorter than the

median CEO tenure. The median CFO has been in the CFO position for 4 years. Almost

all of them have a finance career (about 97 percent). In terms of education, about 54%

have an MBA and 27% have a degree in finance and accounting. CFOs own very little

stock in their firms (a median of 0.04%). Over 5% of CFOs are females.

Columns (1) to (4) of Table X examine CFO fixed effects in leverage regressions.

In this sample, a cross-sectional leverage regressions with firm fixed effects has an

R2 = 0.768. Including both firm and CEO effects increases the R2 = 0.811. If instead

of including a CEO fixed effect, we include a CFO fixed effect, the R2 = 0.821 while

including CEO and CFO effects along with firm fixed effects results in the R2 = 0.836.

The CFO is at least as important as the CEO for determining leverage.

16

Columns (5) to (8) of Table X examine the role of the CFO effect in a partial adjust-

ment regression with firm fixed effect. For the sample of firms with CFOs, the speed

of adjustment in the presence of firm fixed effect is 48% (R2 = 0.24). With both a firm

and CEO effect the speed is 59.6% (R2 = 0.305). With a firm and a CFO fixed effect the

speed of adjustment is 61.0% (R2 = 0.334). With all three fixed effects the adjustment

speed is 66.3% (R2 = 0.364). Again it seems that the CFO is the more important person

in the leverage decisions.

Table XI examines the leverage changes in situations when CEO and CFO turnover

occurs together. As the table shows, leverage declines significantly after turnover and

leverage adjustment speeds increase by 0.16 (more than a 20% increase in adjustment

speeds) around the departure of both CEO and CFO at about the same time. Thus it

seems that there are ‘managerial team’ effects that may be at work.

The CFO compensation contracts do not affect the level of leverage but they do have

significant effects on adjustment speeds. The effects come mainly from CFO holdings

of stock options. CFO characteristics do have some effects on leverage: CFOs with

longer tenure have lower leverage, female CFOs have lower leverage, and CFOs with

technical education and legal education have lower leverage. While these effects are

found in the data, it is a little unclear at this stage just how they fit together. What

is clear is that the CFO is at least as important as the CEO for leverage choice. In

particular, there is some evidence of managerial team-effects.

VIII. Conclusion

Differences among CEOs are a first order concern for corporate leverage choices. It

may help to provide an account for the persistence in leverage that has been stressed

by Baker and Wurgler (2002), Welch (2004) and by Lemmon et al. (2007).

The CEOs compensation structure seems to affect the firm’s leverage choices. When

the CEO has higher pay-for-performance, their firm tends to have lower leverage.

Since our data is from 1993-2004 it may not be too surprising that the options granted

to the CEO are of particular importance.

17

Beyond the direct compensation-based effects, understanding how and why dif-

ferent CEOs have different beliefs is, however, not easy. For interpretative purposes

it would be nice to provide clear empirical links between various managerial char-

acteristics and the manager fixed effect. We are able to document that a variety of

characteristics, such as having an MBA, length of job tenure, and educational back-

ground, do have statistically significant impacts. However, the observed managerial

characteristics do not account for a great deal of the variation in leverage or in leverage

adjustment.

The difficulty in fully explaining the CEO fixed effects with characteristics may

itself point to something deeper. Perhaps the CEO fixed effect is actually proxying for

more than the personality of the CEO. A CEO does not run a firm in isolation. Many

other people are involved. For example the CFO might naturally be responsible for

corporate financing. Empirically we find that the CFO matters more than does the

CEO, for the firm’s leverage choices. Greater attention to the managerial team rather

than just to the CEO might prove interesting.

18

References

Aggarwal, R.K., and A.A. Samwick, 2003, Performance incentives within firms: Theeffect of managerial responsibility, Journal of Finance .

Baker, M., R.S. Ruback, and J. Wurgler, 2007, Behavioral corporate finance, in B.E.Eckbo, (ed.) Handbook of Corporate Finance: Empirical Corporate Finance, Vol. 1. In:Handbook of Finance Series, Chapter 4, – (Elsevier/North-Holland, Amsterdam).

Baker, M., and J. Wurgler, 2002, Market timing and capital structure, Journal of Finance57, 1–32.

Bertrand, M., and A. Schoar, 2003, Managing with style: The effect of managers on firmpolicies, Quarterly Journal of Economics 118, 1169–1208.

Bhagat, S., and B.J. Bolton, 2006, Corporate governance and firm performance, workingPaper, University of Colorado at Boulder.

Black, F., and M. Scholes, 1973, The pricing of options and corporate liabilities, Journalof Political Economy 81, 637–654.

Cadenillas, A., J. Cvitanic, and F. Zapatero, 2004, Leverage decision and manager com-pensation with choice of effort and volatility, Journal of Financial Economics 73, 71–92.

Core, J.E., and W.R. Guay, 1998, The stock and flow of CEO equity incentives, workingpaper, University of Pennsylvania.URL: http://ssrn.com/abstract=138272

Core, J.E., and W.R. Guay, 2002, Estimating the value of employee stock option port-folios and their sensitivities to price and volatility, Journal of Accounting Research 40,613–630.

Demsetz, H., and K.M. Lehn, 1985, The structure of corporate ownership: causes andconsequences, Journal of Political Economy 93, 1155–1177.

Denis, D.J., and D.K. Denis, 1995, Performance changes following top managementdismissals, Journal of Finance 50, 1029–1157.

Fama, E., and K.R. French, 2002, Testing trade-off and pecking order predictions aboutdividends and debt, Review of Financial Studies 15, 1–33.

Fee, C.E., and C.J. Hadlock, 2004, Management turnover across the corporate hierarchy,Journal of Accounting and Economics 37, 3–38.

Flannery, M., and K. Rangan, 2006, Partial adjustment towards target capital structures,JFE 79, 469–506.

19

Frank, M.Z., and V.K. Goyal, 2007, Capital structure decisions: Which factors are reli-ably important?, working Paper, University of Minnesota and HKUST.

Huson, M., R. Parrino, and L. Starks, 2001, Internal monitoring mechanisms and CEOturnover: A long term perspective, Journal of Finance 56, 2265–2297.

Huson, M.R., P.H. Malatesta, and R. Parrino, 2004, Managerial succession and firmperformance, Journal of Financial Economics 74, 237–275.

Jensen, M., and K.J. Murphy, 1990, Performance pay and top management incentives,Journal of Political Economy 98, 225–264.

Kaplan, S.N., and B.A. Minton, 2006, How has CEO turnover changed? increasinglyperformance sensitive boards and increasingly uneasy CEOs, NBER working paper12465 (http://www.nber.org/papers/w12465).

Lehn, K.M., and M. Zhao, 2006, CEO turnover after acquisitions: Are bad biddersfired?, Journal of Finance 61, 1759–1811.

Lemmon, M.L., M.R. Roberts, and J.F. Zender, 2007, Back to the beginning: Persistenceand the cross-section of corporate capital structure, Journal of Finance, Forthcoming .

Malmendier, U., and G.A. Tate, 2005, CEO overconfidence and corporate investment,Journal of Finance 60, 2661–2700.

Parrino, R., 1997, CEO turnover and outside succession: A cross-sectional analysis,Journal of Financial Economics 46, 165–197.

Petersen, M., 2007, Estimating standard errors in finance panel data sets: Comparingapproaches, Review of Financial Studies Fortcoming, working Paper, Northwestern.

Weisbach, M.S., 1988, Outside directors and ceo turnover, Journal of Financial Economics20, 431–460.

Welch, I., 2004, Capital structure and stock returns, Journal of Political Economy 112,106–131.

Yermack, D., 1995, Do corporations award CEO stock options effectively?, Journal ofFinancial Economics 39, 237–269.

20

Appendix A. Variable definitions

Total debt/market value of assets (TDM): the ratio of total debt (item 34, debt in current lia-bilities + item 9, long-term debt) to MVA, market value of assets. MVA is obtained as thesum of market value of equity (item 199, price-close item 54, shares outstanding) + item34, debt in current liabilities + item 9, long-term debt + item 10, preferred- liquidationvalue, - item 35, deferred taxes and investment tax credit.

Total debt/book assets (TDA): the ratio of total debt (item 34, debt in current liabilities + item9, long-term debt) to item 6, book assets.

Median industry leverage (IndustLev): the median of total debt to market value of assets bySIC code and by year. Industry is defined at the four-digit SIC code level.

Market to Book ratio (Mktbk): the ratio of market value of assets (MVA) to Compustat item6, assets. MVA is obtained as the sum of the market value of equity (item 199, price-closeitem 54, shares outstanding) + item 34, debt in current liabilities + item 9, long-term debt+ item 10, preferred-liquidation value, - item 35, deferred taxes and investment tax credit.

Collateral (Colltrl): the ratio of (Compustat item 3, inventory + item 8, net property, plantand equipment) to item 6, assets.

Profitability (Profit): the ratio of Compustat item 13, operating income before depreciation,to item 6, assets.

Dividend Paying Dummy (Dividend): dummy variable that takes a value of one if item 21,common dividends, is positive and it is otherwise zero.

Log of Assets (Assets): the natural log of Compustat item 6, assets, where assets are deflatedby the GDP deflator.

Descriptive statistics on leverage and leverage factors for matched Execucomp-Compustat firms..

Variable Variable Definition N Mean Median 25th %ile 75th %ile

TDM Total debt to market assets 16,593 0.231 0.172 0.041 0.365

IndustLev Industry Median Leverage 16,714 0.212 0.222 0.087 0.313

(MB

)Market-to-Book ratio 16,593 1.825 1.253 0.856 2.053

Colltrl Collateral to total assets ratio 16,532 0.458 0.456 0.273 0.641

Profit Profitability ratio 16,641 0.134 0.137 0.093 0.190

Dividend Dividend paying dummy 16,714 0.536 1.000 0.000 1.000

Assets Assets in Millions of Constant $ 16,714 4,892.2 1,050.0 396.2 3,257.9

21

Appendix B. CEO Changes

CEO changes are broadly classified as normal or forced based on a reading of news articlesdescribing the change.

Normal turnovers are:

• CEO changes due to poor-health/Death: CEO resigns because of poor health or CEO diessuddenly. N = 67 (3.6%)

• Normal retirement: CEO reaches retirement age (63 years or higher as stated in the com-pany’s normal retirement policy). N = 552 (29.3%)

• Resignation with succession in place: CEO resigns but the company has a succession policyin place announced three months prior to CEO departure. N = 170 (9%).

• CEO stays as chairman/executive following resignation: CEO resigns (not forced) but contin-ues to remains as chairman of the board or as executive for at least one year followingthe resignation.

Forced turnovers are:

• Terminated: CEO is forced out or he/she leaves because of policy differences, or followingdisclosure of fraud, accounting irregularities or poor performance. N=313 (16.6%)

• Leaving to pursue other interests: CEO (age¡63 years) resigns for personal reasons or topursue other interests. N=161 (8.5%).

• No reason provided: The news article describing the CEO change; Resigning abruptly with-out any reason and no succession plan is in place. (CEO is less than 63 years old). N=156(8.3%). Even when a successor is announced, if the CEO is resigning and leaving in lessthan 3 months, it is classified as no reason provided. If the CEO reasons, but stays withthe company as chairman for less than a year or if it cannot be determined how longhe/she stayed, then we classify the turnover in this category.

• Corporate restructuring: CEO leaves following a merger, asset sale, spinoff or other re-structuring transaction. N=70 (3.7%).

Appendix C. Compensation contracts

Valuing newly granted optionsModified Black-Scholes option valuation method is used to value newly granted options. Theinputs requires are obtained as follows:

• Option exercise price per share: Exercise price per share for newly granted options fromExecuComp (EXPRIC).

22

• Option maturity for newly granted options: Options are assumed to be granted on July 1stof the particular year. Option maturity is the time-span in years between the optionexpiration date (EXDATE) and option grant date, rounded to the nearest whole year.

• Risk-free rate: The risk-free rates corresponding to various option maturities are read-offthe Treasury yield curve, constructed monthly from the historical interest rate providedby the Federal Reserve Statistical Release.7 Yield on one-year bond is used for optionsmaturing in one year; yield on two-year bonds is used for options maturing in two years;three year bond yield for options maturing in three years; five year bond yield for op-tions maturing between four and five years; seven year bond yield for options maturingbetween six and eight years; 10 year bond yield for options maturing beyond nine years.

• Stock price per share at fiscal-year end: Fiscal-end closing stock price is from CRSP.

• Stock price volatility: Stock return volatility over a 60-month period obtained from Execu-Comp (BS VOLAT).

• Dividend Yield: Average dividend yield over a three-year period obtained from Execu-Comp (BS YIELD).

Valuing unexercised optionsWe follow Core and Guay (2002) approximation method to value unexercised options held byexecutives. The inputs are obtained as follows:

• Exercise price for unexercised options: To estimate the average exercise price for unexer-cised exercisable options, follow the two-step process in Core and Guay. First, computethe ratio of the realizable value of in-the-money exercisable options (INMONEX) and thenumber of unexercised exercisable options (UEXNUMEX). In the second step subtractthis ratio from the fiscal year-end stock price. The resulting number is an estimate of theaverage exercise price for unexercised exercisable options held by executives. Similarly,an estimate of average exercise price of unexercised unexercisable options can be ob-tained by subtracting the ratio of (in-the-money unexercisable options (INMONUN) tothe number of unexercised unexercisable options (UEXNUMUN)) from the fiscal year-end stock price.

• Option maturity for unexercised exercisable options: The maturity of unexercised exercisableoptions is assumed to be four years less than the average maturity of the new grants.In case no grants are made this year, it is set at 6 years. The maturity of unexercisableoptions is set at one less than the average maturity of the new grants. In case no grantsare made this year, it is set at 9 years.

• Stock price, risk-free rate, dividend yield and volatility: These are obtained as above fromCRSP and ExecuComp database.

Total flow compensation comprises of salary, bonus, other annual compensation, totalvalue of restricted stock granted, total value of stock options granted (using Black Scholes

7The data can be downloaded from http://132.200.33.130/releases/h15/data.htm.

23

method described above), long-term incentive payout and all other total. It is estimated byadding Black-Scholes value of newly granted options to the difference in two ExecuComp vari-ables (TDC1 - BLK VALU).

Wealth change from stock holdings: The change in stock related wealth is estimated bymultiplying the value of CEO/CFO stock holdings at time t-1 with stock returns (in %) duringthe year.

Wealth change from option holdings: The change in option related wealth is estimated asthe value of newly granted options plus the change in the value of unexercised options (bothexercisable and unexercisable).

Pay-performance sensitivity: The pay-for-performance sensitivity is the change in exec-utive’s firm specific wealth (from both options and stocks) for every one-thousand dollar in-crease in shareholder wealth (measured by multiplying the beginning of year market valuewith the annual stock returns including distributions). The compounded annual returns areobtained by cumulating CRSP monthly stock returns with dividends for each firm during itsfiscal year.

Descriptive statistics on compensation and pay-performance sensitivities for the CEOs..

Variable Variable Definition N Mean Median 25th %ile 75th %ile

Salary Salary ($’000) 16,929 588.1 525.0 365.4 750.0

Bonus Bonus ($’000) 16,929 571.8 300.0 50.0 682.3

BSOptionV al Value of option grants ($’000) 16,929 3,837.3 640.1 0.0 2,918.3

ExecF lowComp Total flow compensation 16,827 5,750.0 2,141.9 955.7 5,370.1

WlthChgOwn Wealth gain from stock ownership 16,929 10,665.7 0.0 -62.7 961.6

WlthChgOpt Wealth gain from option holdings 16,929 2,103.0 0.0 0.0 901.2

StockPPS Stock Pay-Performance Sensitivity 14,440 0.0259 0.0025 0.0004 0.0150(per $ change in shareholder wealth)

OptionPPS Option Pay-Performance Sensitivity 14,440 0.0077 0.0016 0.0000 0.0115(per $ change in shareholder wealth)

TotalPPS Total Pay-Performance Sensitivity 14,440 0.0337 0.0096 0.0011 0.0374(per $ change in shareholder wealth)

ExecOwn CEO Ownership 16,633 0.0297 0.0036 0.0009 0.0186

24

Appendix D. Defining Managerial Attributes

Gender: dummy variables that takes a value of one if the CEO/CFO is a female and equal tozero if the CEO/CFO is a male (ExecuComp variable PGENDER and biographical datafrom Who’s Who in Finance and Industry (various editions) and Internet).

Age: age in years obtained by subtracting the year in which the executive was born from theyear in the data. The year of birth is from Who’s Who in Finance and Industry (variouseditions), Executive Biographies and profiles on the Internet.

Executive founder: dummy variable that takes a value of 1 if the executive is a founder/co-founder of the firm (Who’s Who in Finance and Industry (various editions), ExecutiveBiographies and profiles on the Internet).

Tenure as CEO/CFO: number of years the executive has been in the CEO or the CFO posi-tion (ExecuComp database for CEOs and Who’s Who in Finance and Industry (variouseditions), Executive Biographies and profiles on the Internet for CFOs).

Tenure at the firm: number of years the executive has been working at the firm (ExecuCompdatabase for CEOs and Who’s Who in Finance and Industry (various editions), ExecutiveBiographies and profiles on the Internet for CFOs)

Number of companies previously worked for: number of companies the executive hasworked for prior to joining this company (Who’s Who in Finance and Industry (variouseditions), Executive Biographies and profiles on the Internet).

Master Bus.Adm. (MBA): dummy variable equal to one if the CEO/CFO has an MBA degreeand equal to zero otherwise (Who’s Who in Finance and Industry (various editions),Executive Biographies and profiles on the Internet).

Finance education: dummy variable that takes a value of one if the CEO/CFO has a Bache-lor degree in Business Administration; or Bachelor in Accounting or in Economics; or aMaster degree in Finance and Accounting or an MBA and equal to zero otherwise (Who’sWho in Finance and Industry (various editions), Executive Biographies and profiles onthe Internet).

Technical education: dummy variable that takes a value of one if the CEO/CFO has a Bache-lor degree in Science; or a Bachelor degree in Engineering; or a Master degree in Scienceor in Engineering and no MBA degree and zero otherwise (Who’s Who in Finance andIndustry (various editions), Executive Biographies and profiles on the Internet).

Finance career: dummy variable equal to one if the CEO/CFO has previously worked in afinancial institution, or if previously worked as a CFO, Treasurer, Accountant, or financerelated professions (Who’s Who in Finance and Industry (various editions), ExecutiveBiographies and profiles on the Internet).

Technical career: dummy variable equal to one if the CEO/CFO previously worked as anengineer or other technically oriented professions (Who’s Who in Finance and Industry(various editions), Executive Biographies and profiles on the Internet).

25

Table IFirm and manager effects on corporate leverage policies

Cross-sectional leverage regression estimates with firm and CEO fixed effects. The dependent variableis market leverage. The variables are described in Appendix A. The sample is matched Execucomp-Compustat firms for the period from 1993 to 2004. Reported in the table are coefficient estimates andt-stats where the standard errors are clustered at the firm and year level. Also reported are the F-testsfor joint significance of firm fixed effects (column 3), CEO fixed effects (column 4), and CEO and firmfixed effects (column 5).

OLS with OLS with Firm CEO Firm and CEOclustered SE clustered SE Fixed Effects Fixed Effects Fixed Effects

(1) (2) (3) (4) (5)

AdjustedR2 0.444 0.461 0.774 0.811 0.819

Observations 16,130 16,130 16,130 16,130 16,130

F-test on fixed effects 11.6 9.2 9.5p-value (0.00) (0.00) (0.00)

Regression estimates:

InitialLev 0.161a

(0.019)

IndustLevt−1 0.660a 0.588a 0.284a 0.244a 0.226a

(0.025) (0.026) (0.031) (0.033) (0.034)(MB

)t−1

-0.023a -0.022a -0.008a -0.008a -0.007a

(0.002) (0.001) (0.001) (0.001) (0.001)

Colltrlt−1 0.085a 0.074a 0.066a 0.051b 0.048c

(0.015) (0.015) (0.025) (0.025) (0.026)

Profitt−1 -0.322a -0.319a -0.188a -0.180a -0.171a

(0.026) (0.026) (0.025) (0.025) (0.025)

Dividendt−1 -0.031a -0.031a -0.013 0.006 0.013(0.006) (0.006) (0.009) (0.009) (0.009)

Assetst−1 0.029a 0.028a 0.047a 0.050a 0.052a

(0.002) (0.002) (0.005) (0.005) (0.005)

Constant -0.072a -0.071a -0.160a -0.176a -0.188a

(0.016) (0.016) (0.040) (0.039) (0.042)

Year fixed effects Yes Yes Yes Yes Yes

26

Table IILeverage adjustment and manager fixed effects

This table reports results from estimations of the following partial adjustment model:

∆TDM t = α + λTDMt−1 + β1IndustLevt−1 + β2

(M

B

)

t−1

+

β3Colltrlt−1 + β4Profitt−1 + β5Dividendt−1 + β6 log (Assets)t−1 + Firm/ManagerFEi + µt + εi,t,

where TDM is the ratio of total debt to market value of assets. Other variables are as defined in AppendixA. The sample consists of nonfinancial Compustat firms with data on CEO compensation in Execucomp.t-statistics reported are robust to heteroscedasticity and within firm correlation. All regressions includeyear fixed effects.

Pooled Clustered Clustered Firm CEO Firm and CEOOLS Stand. Errors Stand. Errors Fixed Effects Fixed Effects Fixed Effects

Variables (1) (2) (3) (4) (5) (6)

InitialLev 0.024a

(4.7)

TDMt−1 -0.150a -0.150a -0.155a -0.457a -0.548a -0.574a

(-23.7) (-7.5) (-23.8) (-37.3) (-36.9) (-37.9)

IndustLevt−1 0.097a 0.097a 0.090a 0.055a 0.071a 0.069a

(10.8) (7.7) (10.0) (3.2) (3.6) (3.5)(

MB

)t−1

-0.002b -0.002a -0.002a -0.001b -0.002a -0.002a

(-5.0) (-2.2) (-4.7) (-2.5) (-3.3) (-3.4)

Colltrlt−1 0.014a 0.014c 0.012a 0.033b 0.030c 0.028(3.2) (1.8) (2.9) (2.4) (1.8) (1.6)

Profitt−1 -0.028a -0.028c -0.030a -0.034a -0.039a -0.042a

(-3.6) (-1.8) (-3.8) (-2.8) (-2.7) (-2.9)

Dividendt−1 -0.002 -0.002 -0.003 0.015a 0.016b 0.022a

(-1.3) (-0.8) (-1.4) (2.8) (2.5) (3.4)

Assetst−1 0.003a 0.003 0.003a 0.026a 0.028a 0.031a

(4.9) (1.6) (4.8) (11.1) (9.4) (9.7)

Constant -0.018a -0.018 -0.018a -0.122a -0.120a -0.133a

(-3.8) (-1.4) (-3.8) (-6.4) (-5.2) (-5.4)

Year Fixed effects Yes Yes Yes Yes Yes Yes

R2 −Adjusted 0.107 0.107 0.108 0.238 0.287 0.302

Observations 16130 16130 16130 16130 16130 16130

27

Table IIILeverage around CEO turnover

This table examines leverage changes around CEO turnovers. The sample consists of nonfinancial Com-pustat firms with data on CEO compensation in Execucomp. t-statistics reported are robust to het-eroscedasticity and within firm correlation. All regressions include manager fixed effects.

Forced Forced Forced ForcedAll categories excl. mergers Terminated All categories

(1) (2) (3) (4)

Normal turnover × Ipre 0.004 0.005 0.008a

(1.0) (1.4) (2.1)Normal turnover × IPost -0.004 -0.005 -0.004

(-1.1) (-1.4) (-1.0)Forced turnover × IPre 0.043a 0.046a 0.065a

(6.5) (6.5) (6.4)Forced turnover × IPost 0.008 0.011c 0.024b

(1.3) (1.7) (2.4)Outsider ×Normal × Ipre 0.006

(0.7)Outsider ×Normal × IPost -0.010

(-1.0)Insider ×Normal × IPre 0.004

(1.0)Insider ×Normal × IPost -0.004

(-1.1)Outsider × Forced× Ipre 0.039a

(4.0)Outsider × Forced× IPost -0.005

(-0.5)Insider × Forced× IPre 0.044a

(5.0)Insider × Forced× IPost 0.017b

(2.0)Constant -0.211a -0.209a -0.209a -0.214a

(-6.9) (-6.8) (-6.8) (-7.0)

Leverage Factors Yes Yes Yes YesFirm-Manager Fixed Effects Yes Yes Yes YesYear Fixed Effects Yes Yes Yes YesR2 −Adjusted 0.820 0.820 0.820 0.820Observations 16,132 16,132 16,132 16,132

28

Table IVAdjustment Speeds around CEO turnover

This table examines leverage adjustment speeds around CEO turnovers. The sample consists of nonfi-nancial Compustat firms with data on CEO compensation in Execucomp. t-statistics reported are robustto heteroscedasticity and within firm correlation. All regressions include manager fixed effects.

Forced Forced Forced ForcedAll categories excl. mergers Terminated All categories

(1) (2) (3) (4)

TDMt−1 -0.583a -0.583a -0.584a -0.583a

(-37.4) (-37.4) (-37.4) (-37.4)Normal × Ipre × TDMt−1 0.010 0.013 0.016

(0.8) (1.0) (1.4)Normal × IPost × TDMt−1 -0.007 -0.013 -0.007

(-0.5) (-1.0) (-0.6)Forced× IPre × TDMt−1 0.089a 0.101a 0.145a

(4.4) (4.5) (4.6)Forced× IPost × TDMt−1 0.007 0.022 0.033

(0.3) (1.0) (0.9)Outsider ×Normal × Ipre × TDMt−1 -0.007

(-0.2)Outsider ×Normal × IPost × TDMt−1 -0.006

(-0.2)Insider ×Normal × IPre × TDMt−1 0.016

(1.1)Insider ×Normal × IPost × TDMt−1 -0.008

(-0.5)Outsider × Forced× Ipre × TDMt−1 0.074b

(2.5)Outsider × Forced× IPost × TDMt−1 -0.004

(-0.1)Insider × Forced× IPre × TDMt−1 0.100a

(3.7)Insider × Forced× IPost × TDMt−1 0.019

(0.7)Constant -0.149a -0.147a -0.147a -0.149a

(-5.3) (-5.3) (-5.3) (-5.4)

Leverage Factors Yes Yes Yes Yes

Firm-Manager Fixed Effects Yes Yes Yes Yes

Year Fixed Effects Yes Yes Yes Yes

R2 −Adjusted 0.305 0.305 0.306 0.305

Observations 16,132 16,132 16,132 16,132

29

Table VIncentives and leverage

This table presents augmented leverage regressions with compensation and ownership. The sampleconsists of nonfinancial Compustat firms with data on CEO compensation in Execucomp. t-statisticsreported are robust to heteroscedasticity and within firm correlation. All regressions include firm andyear fixed effects.

(1) (2) (3) (4)

Stock Pay Performance Sensitivity 0.019(0.6)

Option Pay Performance Sensitivity -0.141a

(-4.4)

Total Pay Performance Sensitivity -0.066a

(-2.9)

Stock owned× Iown≤1% -1.829a

(-3.0)

Stock owned× I1<own≤5% 0.154(1.0)

Stock owned× Iown>5% 0.077a

(2.8)

Constant -0.177a -0.168a -0.165a -0.166a

(-6.2) (-6.0) (-5.9) (-7.0)

Leverage factors Yes Yes Yes Yes

R2 −Adjusted 0.785 0.786 0.786 0.775Observations 14135 14062 14062 15889

30

Table VIIncentives and leverage adjustment speeds

This table examines the effect of incentives on leverage adjustment speeds. The sample consists ofnonfinancial Compustat firms with data on CEO compensation in Execucomp. t-statistics reported arerobust to heteroscedasticity and within firm correlation. All regressions include firm and year fixedeffects.

(1) (2) (3) (4)

TDMt−1 -0.489a -0.485a -0.479a -0.454a

(-35.9) (-36.1) (-35.2) (-32.7)

Stock Pay Performance Sensitivity × TDMt−1 -0.231b

(-2.0)

Option Pay Performance Sensitivity × TDMt−1 -0.788a

(-6.4)

Total Pay Performance Sensitivity × TDMt−1 -0.499a

(-5.9)

Stock owned× TDMt−1 × Iown≤1% -1.194(-0.6)

Stock owned× TDMt−1 × I1<own≤5% -0.316(-0.5)

Stock owned× TDMt−1 × Iown>5% 0.057(0.8)

Constant -0.136a -0.132a -0.132a -0.127a

(-5.6) (-5.4) (-5.4) (-6.2)

Leverage Factors Yes Yes Yes Yes

Firm Fixed Effects Yes Yes Yes Yes

Year Fixed Effects Yes Yes Yes Yes

R2 −Adjusted 0.256 0.262 0.261 0.238

Observations 14,136 14,063 14,063 15,891

31

Tabl

eV

II:M

anag

erch

arac

teri

stic

sTh

ista

ble

prov

ides

aver

age

char

acte

rist

ics

ofC

EOs.

The

tabl

eal

sode

scri

bes

both

new

and

depa

rtin

gm

anag

ers

-re

plac

edin

forc

edan

dro

utin

gtu

rnov

ers.

App

endi

xB

prov

ides

ade

finit

ion

offo

rced

and

rout

ine

turn

over

s.

Rou

tine

Turn

over

sFo

rced

Turn

over

s

Insi

deA

ppoi

ntm

ents

Out

side

App

oint

men

tsIn

side

App

oint

men

tsO

utsi

deA

ppoi

ntm

ents

Att

ribu

tes

All

Dep

arti

ngIn

com

ing

Dep

arti

ngIn

com

ing

Dep

arti

ngIn

com

ing

Dep

arti

ngIn

com

ing

CEO

sC

EOC

EOC

EOC

EOC

EOC

EOC

EOC

EO

Age

,Ten

ure

and

Gen

der

Age

(in

year

s)55

.37

62.2

454

.54

60.6

955

.49

54.0

752

.93

54.1

052

.70

Tenu

reat

firm

(in

year

s)16

.47

24.9

214

.08

23.1

10.

6812

.96

11.9

815

.86

0.66

Tenu

reas

ceo

(in

year

s)7.

8311

.94

1.06

10.1

20.

987.

181.

727.

400.

98

Perc

ento

fFem

ale

CEO

s0.

010.

000.

020.

000.

010.

010.

030.

020.

01

Educ

atio

n

%of

CEO

sw

/MBA

0.37

0.32

0.40

0.29

0.52

0.40

0.33

0.35

0.44

%of

CEO

sw

/Fin

ance

edu

0.07

0.04

0.08

0.07

0.06

0.02

0.05

0.05

0.07

%of

CEO

sw

/Te

chn.

edu

0.39

0.43

0.38

0.62

0.45

0.44

0.43

0.45

0.48

%of

CEO

sw

/La

wed

u0.

100.

080.

090.

090.

040.

080.

110.

080.

10

32

Tabl

eV

II,C

onti

nued

Rou

tine

Turn

over

sFo

rced

Turn

over

s

Insi

deA

ppoi

ntm

ents

Out

side

App

oint

men

tsIn

side

App

oint

men

tsO

utsi

deA

ppoi

ntm

ents

Att

ribu

tes

All

Dep

arti

ngIn

com

ing

Dep

arti

ngIn

com

ing

Dep

arti

ngIn

com

ing

Dep

arti

ngIn

com

ing

CEO

sC

EOC

EOC

EOC

EOC

EOC

EOC

EOC

EO

Car

eer

%w

/pr

evio

usfin

.car

eer

0.22

0.20

0.29

0.16

0.28

0.26

0.28

0.28

0.20

%w

/pr

evio

usTe

ch.c

aree

r0.

150.

200.

110.

250.

130.

190.

140.

200.

20

Man

ager

ialc

ontr

acts

Sala

ry($

m)

588.

0864

7.60

571.

0857

5.81

525.

1358

5.93

511.

7956

7.62

489.

69

Bonu

s($

m)

571.

7666

2.65

552.

5236

8.87

602.

7532

9.73

399.

1229

8.73

520.

54

Opt

ions

:BS

Val

ue($

m)

3837

.27

3519

.81

4980

.49

2746

.34

5895

.49

2341

.63

4213

.39

2337

.50

6710

.03

Stoc

kPP

S(p

er$1

000)

25.9

429

.45

0.06

25.7

20.

0016

.65

0.10

13.0

30.

00

Opt

ion

PPS

(per

$100

0)7.

754.

841.

474.

040.

684.

182.

476.

241.

70

Tota

lPPS

($)

33.6

934

.38

1.53

29.7

60.

6821

.17

2.57

19.2

91.

70

Shar

esow

ned

(in

%)

2.97

2.96

0.76

2.18

0.20

1.66

1.70

1.27

1.23

33

Tabl

eV

III:

Man

ager

char

acte

rist

ics

and

leve

rage

This

tabl

eex

amin

esth

eef

fect

ofm

anag

erch

arac

teri

stic

son

leve

rage

.Th

esa

mpl

eco

nsis

tsof

nonfi

nanc

ial

Com

pust

atfir

ms

wit

hda

taon

CEO

com

pens

atio

nin

Exec

ucom

p.t-

stat

isti

csre

port

edar

ero

bust

tohe

tero

sced

asti

city

and

wit

hin

firm

corr

elat

ion.

All

regr

essi

ons

incl

ude

man

ager

fixed

effe

cts.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

CEO

olde

rth

an55

year

s0.

004

(1.5

)Lo

ngC

EOte

nure

-0.0

37a

(-13

.9)

#co

mpa

nies

wor

ked

prev

ious

ly0.

007a

(4.6

)Fe

mal

eC

EO-0

.019

(-1.

6)M

BAD

egre

e0.

015a

(4.0

)Fi

nanc

eEd

ucat

ion

-0.0

02(-

0.3)

Tech

nica

lEdu

cati

on-0

.007

c

(-1.

9)La

wed

ucat

ion

0.03

0a

(4.5

)Fi

nanc

ialC

aree

r0.

009c

(1.9

)Te

chni

calC

aree

r0.

020a

(3.9

)C

onst

ant

-0.0

74a

-0.0

76a

-0.0

57a

-0.0

71a

-0.0

40a

-0.0

37a

-0.0

34a

-0.0

36a

-0.0

27c

-0.0

27c

(-8.

5)(-

9.0)

(-3.

7)(-

8.3)

(-3.

1)(-

2.8)

(-2.

6)(-

2.8)

(-1.

9)(-

1.9)

Leve

rage

fact

ors

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

R2−

Adj

ust

ed0.

444

0.45

00.

493

0.44

40.

508

0.50

70.

508

0.51

00.

485

0.48

7O

bser

vati

ons

1613

016

130

4638

1613

064

8764

8764

8764

8759

0461

67

34

Tabl

eIX

:Man

ager

char

acte

rist

ics

and

leve

rage

adju

stm

ents

peed

sTh

ista

ble

exam

ines

the

effe

ctof

man

ager

char

acte

rist

ics

onle

vera

gead

just

men

tsp

eeds

.Th

esa

mpl

eco

nsis

tsof

nonfi

nanc

ial

Com

pust

atfir

ms

wit

hda

taon

CEO

com

pens

atio

nin

Exec

ucom

p.t-

stat

isti

csre

port

edar

ero

bust

tohe

tero

sced

asti

city

and

wit

hin

firm

corr

elat

ion.

All

regr

essi

ons

incl

ude

man

ager

fixed

effe

cts.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

TD

Mt−

1-0

.157

a-0

.145

a-0

.152

a-0

.150

a-0

.159

a-0

.160

a-0

.158

a-0

.167

a-0

.165

a-0

.161

a

(-20

.9)

(-21

.8)

(-12

.1)

(-23

.7)

(-16

.5)

(-17

.8)

(-16

.5)

(-17

.9)

(-16

.7)

(-16

.4)

CE

O>

55yea

rs×

TD

Mt−

10.

012c

(1.8

)L

ong

CE

Ote

nure×

TD

Mt−

1-0

.017

a

(-2.

7)#

com

pw

orked

pre

vio

usl

y×

TD

Mt−

1-0

.002

(-0.

7)F

emale

CE

O×

TD

Mt−

10.

024

(0.5

)M

BA×

TD

Mt−

1-0

.008

(-0.

8)F

inance

educa

tion×

TD

Mt−

1-0

.042

b

(-2.

1)T

echnic

aled

uca

tion×

TD

Mt−

1-0

.013

(-1.

3)L

aw×

TD

Mt−

10.

031a

(2.7

)F

inance

care

er×

TD

Mt−

10.

009

(0.8

)T

echnic

alca

reer×

TD

Mt−

10.

002

(0.2

)C

onst

ant

-0.0

18a

-0.0

19a

-0.0

18b

-0.0

18a

-0.0

21a

-0.0

22a

-0.0

21a

-0.0

21a

-0.0

17b

-0.0

15b

(-3.

7)(-

4.1)

(-2.

3)(-

3.8)

(-3.

3)(-

3.3)

(-3.

3)(-

3.2)

(-2.

4)(-

2.2)

Leve

rage

fact

ors

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

R2−

Adj

ust

ed0.

106

0.10

60.

118

0.10

50.

134

0.13

50.

134

0.13

50.

127

0.12

4O

bser

vati

ons

1613

016

130

4638

1613

064

8764

8764

8764

8759

0461

67

35

Tabl

eX

:Lev

erag

ean

dFi

rm,C

EO,a

ndC

FOfix

edef

fect

s

TD

M∆

TD

M

Firm

Firm

and

CEO

Firm

and

CFO

Firm

,CEO

,CFO

Firm

Firm

and

CEO

Firm

and

CFO

Firm

,CEO

,CFO

Fixe

def

fect

sFi

xed

effe

cts

Fixe

def

fect

sFi

xed

effe

cts

Fixe

def

fect

sFi

xed

effe

cts

Fixe

def

fect

sFi

xed

effe

cts

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

TD

Mt−

1-0

.482

a-0

.596

a-0

.610

a-0

.663

a

(-32

.6)

(-32

.4)

(-33

.0)

(-31

.4)

Indust

Lev

t−1

0.29

9a0.

239a

0.21

7a0.

176a

0.07

2a0.

080a

0.06

6b0.

051c

(11.

7)(8

.3)

(7.5

)(5

.5)

(3.3

)(3

.1)

(2.5

)(1

.7)

( M B

) t−1

-0.0

08a

-0.0

07a

-0.0

06a

-0.0

06a

-0.0

01c

-0.0

02a

-0.0

02a

-0.0

02a

(-9.

0)(-

8.1)

(-7.

1)(-

6.6)

(-1.

9)(-

3.0)

(-2.

6)(-

3.0)

Col

ltrl

t−1

0.07

9a0.

058a

0.10

2a0.

087a

0.04

5a0.

039c

0.06

9a0.

062a

(4.2

)(2

.6)

(4.5

)(3

.4)

(3.0

)(1

.9)

(3.3

)(2

.6)

Pro

fit

t−1

-0.1

81a

-0.1

63a

-0.1

15a

-0.1

05a

-0.0

38b

-0.0

42b

-0.0

09-0

.009

(-7.

4)(-

5.9)

(-5.

3)(-

4.1)

(-2.

3)(-

2.0)

(-0.

6)(-

0.5)

Div

iden

dt−

1-0

.009

0.01

20.

013

0.01

7c0.

015b

0.01

9b0.

023a

0.02

2b

(-1.

2)(1

.4)

(1.5

)(1

.7)

(2.4

)(2

.4)

(2.8

)(2

.4)

Ass

ets t−

10.

049a

0.05

1a0.

057a

0.05

5a0.

030a

0.03

3a0.

034a

0.03

4a

(13.

3)(1

1.3)

(11.

8)(9

.7)

(10.

0)(8

.0)

(8.0

)(6

.7)

Con

stant

-0.1

84a

-0.1

97a

-0.2

51a

-0.2

29a

-0.1

48a

-0.1

47a

-0.1

68a

-0.1

54a

(-6.

4)(-

5.7)

(-6.

8)(-

5.3)

(-6.

4)(-

4.7)

(-5.

2)(-

4.0)

Year

Fixe

def

fect

sYe

sYe

sYe

sYe

sYe

sYe

sYe

sYe

s

R2−

Adj

ust

ed0.

768

0.81

10.

821

0.83

60.

244

0.30

50.

334

0.36

4O

bser

vati

ons

1225

512

255

1225

512

255

1225

512

255

1225

512

255

36

Table XICEO CFO Combined Turnover and Leverage

Firm, CEO, CFO Firm, CEO, CFOFixed effect Fixed effect

Variables (1) (2)TDM ∆TDM

IPost -0.045c

(-2.3)TDMt−1 -0.771a

(-14.9)IPost × TDMt−1 -0.156a

(-3.5)IndustLevt−1 0.129 0.080

(1.5) (0.9)(MB

)t−1

-0.008 -0.006(-1.7) (-1.2)

Colltrlt−1 0.018 -0.007(0.3) (-0.1)

Profitt−1 -0.051 -0.039(-1.1) (-0.8)

Dividendt−1 0.023 0.022(1.2) (1.1)

Assetst−1 0.043b 0.033c

(3.1) (2.2)Constant -0.084 -0.054

(-0.7) (-0.5)

Fixed year effects Yes YesR2 −Adjusted 0.874 0.475N 2708 2708

37