Accounting Losses versus Profits and CEO Turnover

Aloke (Al) Ghosh

Stan Ross Department of Accountancy Baruch College, The City University of New York,

Box B12-225, One Bernard Baruch Way New York, NY 10010

E-mail address: Aloke.Ghosh@baruch.cuny.edu Phone: 646.312.3184

___________________ We benefited from our discussions with Ray Ball, Sudipta Basu, Jeremy Bertomeu, Scott Bronson, Ting Chen, Sonali Hazarika, John Elliott, Chris Hogan, Yinghua Li, Antonio Marra, Pietro Mazzola, Christina Mashruwala, Serena Morricone, Daniel Oyon, Annalisa Prencipe, Bill Ruland, Sasson Bar-Yosef and the participants at 2011 AAA Annual Meeting in Denver (Colorado), University of Lauzanne (UNIL), Bocconi University, and Yonsei University. Our special thanks to Masako Darrough, Val Dimitrov, Carol Marquardt, and Terry Shevlin for their many comments and suggestions on some of our prior versions.

Accounting Losses versus Profits and CEO Turnover

ABSTRACT

Relying on a linear specification, several studies examine the importance of accounting and stock performance measures for CEO turnover. We suggest that accounting losses reflect managerial effort and quality that are not fully captured in the prior performance measures including profits. Using a non-linear specification around losses, we find a statistically and economically significant relationship between accounting losses and subsequent CEO turnover. Further, the magnitude of the loss also increases the likelihood of CEO turnover. A crucial finding is that once we include losses, accounting performance is no longer incrementally important in explaining CEO turnover. We additionally hypothesize and find that: (1) the impact of losses on CEO turnover depends on whether other firms in the industry also report losses, (2) CEO turnover following losses leads to more outside CEO appointments, and (3) the sensitivity of CEO turnover to losses is affected by the strength of the board and the level of growth opportunities. Collectively, our results suggest that CEOs are penalized for losses and that boards consider other factors along with losses to arrive at CEO retention decisions. Keywords: accounting losses; accounting performance; CEO turnover; managerial quality. Data Availability: All data used in the paper are available from publicly available sources

noted in the text.

2

Accounting Losses versus Profits and CEO Turnover

I. INTRODUCTION

Relying on accounting (e.g., earnings levels and changes, industry-adjusted earnings) and

market measures (e.g., stock returns) as metrics for poor performance, several studies document

a positive relationship between CEO turnover and poor performance (see Hermalin and

Weisbach 2003). Almost all prior studies, with the exception of Jenter and Lewellen (2010),

assume a linear relationship between turnover and performance.1 A key criticism of prior studies

has been that the economic significance of performance is “arguably quite small” (Brickley 2003,

p. 228). One possible explanation for the weak association is that it becomes challenging for

researchers to develop a uniform metric for poor performance. In this study, we investigate

whether accounting loss is one such indicator of poor performance and whether losses dominate

other accounting measures of performance including accounting profits in determining CEO

turnover. Our fundamental contention is that a non-linear specification around losses is a more

parsimonious representation of turnover-performance relationship.

The classification of accounting performance into profits and losses presents a simple yet

powerful device for assessing managerial performance. Reporting of losses might act as a

heuristic for boards to calibrate managerial competence, which ultimately affects CEO retention

decisions (Pinnuck and Lillis 2007; Watts 2003). The idea that CEOs are penalized for losses

has been referenced by researchers and practitioners. Watts (2003) concludes that “managers

have incentives to hide losses to avoid being fired before their tenure is over.” Based on a survey

of key executives, Graham et al. (2005) find that three-fourths of the survey respondents believe

1For instance, Engel et al. (2003) find that a decline in accounting performance by 10% increases the likelihood of CEO departure by 0.31%. Farrell and Whidbee (2003) find that the implied probability of CEO turnover increases by 1.14% when ROA declines by one standard deviation. DeFond and Park (1999) find that a one standard deviation decrease in earnings increases the probability of CEO turnover by 1.3%. Kaplan and Minton (2006) conclude that the relationship is stronger in recent years.

3

that losses might lead to job losses. One recent anecdotal example is the removal of Howard

Stringer as the CEO of Sony Corp following losses (Wall Street Journal, February 2, 2012).

There are many reasons for losses serving as a barometer for failed performance and,

therefore, leading to more frequent CEO turnover. First, because boards are unable to observe

CEOs’ innate ability, they must rely on public signals, e.g., accounting earnings, and private

signals to assess managerial performance (Taylor 2010; Hermalin and Weisbach 1998; Huson et

al. 2004; Murphy and Zimmerman 1993). Losses are expected to be more informative in judging

managerial performance than profits because, under accounting conservatism, reported income

includes current and future losses while reported income is precluded from anticipating future

profits (Basu 1997; Watts 2003; Givoly and Hayn 2000). Because losses include future negative

payoffs, losses are more likely to indicate whether a CEO invested in negative net present value

(NPV) projects than profits (Ball and Shivakumar 2005; Watts 2003). Hence, the association

between CEO turnover and accounting performance is expected to be stronger for loss firms than

for profit firms.

Second, a loss indicates that a CEO failed in his/her stewardship role to manage the assets

of the firm thereby triggering questions about competence and ability. Consequently, some board

members might feel compelled to learn more about the CEO’s true ability following a loss

(Bowen 2008). Contrary to a routine annual evaluation, a loss might precipitate more critical

evaluation of the incumbent CEO. A critical review, or an escalation in the intensity with which

boards monitor CEOs, increases the likelihood of boards collecting adverse information (i.e.,

private information) which might lead to more frequent CEO turnover (Hermalin and Weisbach

1998).2 Finally, board members might also feel the pressure to hold CEOs accountable for losses

2In the extreme scenario, a board member might consider losses as sufficient justification for

removing a CEO from office. Degeorge et al. (1999) discuss three psychological effects for the zero-

4

because dissident shareholders often wage proxy fights to replace directors when boards fail to

initiate management changes following losses (DeAngelo 1988).

Based on a comprehensive sample of CEO turnovers from S&P 1500 firms between 1997

and 2007, we find, as in prior studies, a strong negative relationship between CEO turnover and

firm performance using both accounting and stock performance measures. Consistent with

concerns that the economic magnitude of accounting performance on CEO turnover is small, our

estimates suggest that a 10% increase in industry-adjusted return on assets decreases the

probability of CEO turnover by 2%. More importantly, when we additionally include an

indicator variable Loss for firms with negative net income while controlling for the other

determinants of CEO turnover, we find that the coefficient on Loss is positive and highly

significant. Additionally, when we examine whether the loss-size might be an added factor

determining CEO turnover (Klein and Marquardt 2006) by including Magnitude (decile ranks of

the absolute value of net income to book value equity) and interaction between Loss and

Magnitude, we find that the coefficients on Loss and the interaction term are positive and

significant. While losses increase the overall probability of a CEO turnover, the likelihood of

CEO turnover further increases with the magnitude of the loss.

In sharp contrast to the results from prior studies, we document that the economic

magnitude of losses on CEO turnover is large. Holding the other variables constant, some

estimates suggest that the odds of a CEO losing a job within two years are 54% higher for a firm

with losses compared to one with profits which suggests that losses substantially increase the

likelihood of the termination of CEO contracts. A crucial finding is that, once we include Loss,

earnings threshold acting as a heuristic or decision rule: (1) there is something fundamental between positive and non-positive numbers in the human thought process, (2) prospect theory predicts that individuals evaluate outcomes as changes from a reference point and losses can serve as one reference point, and (3) heuristics or a ‘rule of thumb’ can reduce transactions costs.

5

accounting performance is no longer incrementally important in explaining CEO turnover

regardless of how we measure accounting performance (e.g., earnings-to-total assets, industry-

adjusted earnings-to-total assets or change in earnings). In contrast, stock performance continues

to be incrementally important in explaining CEO turnover decisions. Thus, losses appear to

dominate as a metric for judging managerial competence while small or declining profits do not

appear to be useful for CEO retention decisions.3

We also develop several related hypotheses on the cross-sectional variations in the loss-

turnover relationship. The added tests are aimed to increase the confidence in our conclusion that

boards consider losses along with other factors to arrive at CEO retention decisions. First, while

losses from continuing operations are expected to lead to higher turnover, we also conjecture that

losses from non-recurring operations as a result of plant closings, rearrangement and shedding of

lines of businesses lead to more frequent CEO turnover because it suggests that a CEO made

poor investment decisions in the past that are currently being reversed (Elliott and Hanna 1996).

Decomposing net losses into recurring and non-recurring components, where non-recurring

losses include losses from discontinued operations and extraordinary items (e.g., Gaver and

Gaver 1998; Dechow et al. 1994), we find that both recurring and non-recurring losses are

incrementally important in explaining CEO turnover.

Second, boards are expected to hold CEOs accountable for poor performance when losses

result from mismanagement or incompetence (i.e., losses are idiosyncratic to the firm), but they

are expected to shelter CEOs from losses that are the outcome of industry-wide shocks affecting

many firms (Bushman et al. 2010; Gibbons and Murphy 1990; Sloan 1993). Therefore, we posit

3In a regression of CEO turnover on stock returns, Jenter and Lewellen (2010) allow the slope

coefficient of stock returns to vary by including quintile ranks instead of using a continuous variable. When we use a similar specification and include quintile ranks of accounting return, our results and conclusions remain unchanged.

6

that the relative frequency of industry-wide losses affects the sensitivity of CEO turnover to

losses. We find that the likelihood of a CEO turnover increases subsequent to losses when losses

are idiosyncratic or unique to the firm. However, as losses become more symptomatic of the

industry, the sensitivity of CEO turnover to losses is muted.

Third, we conjecture that losses increase the likelihood that an outsider would replace an

incumbent CEO. Parrino (1997) claims that outside candidates are more suitable for the top

position when boards want changes in the direction of the firm than when they want to maintain

status quo because “outside candidates…by virtue of their employment at other firms often have

a broader exposure to, and experience with, alternative ways of running a firm” (p. 167).

Because losses might suggest that a firm is in trouble (DeAngelo 1988), and if board members

feel that the current problems are an outcome of failed business policies, boards are more likely

to consider outsiders than insiders when replacing the incumbent CEO for poor performance.

Concentrating on the sub-sample with CEO turnovers, we find that turnovers following losses

lead to more frequent outside CEO appointments than those following profits.

Fourth, because stronger boards are more effective in disciplining top management for

poor performance (e.g., Huson et al. 2001; Goyal and Park 2002; Core et al. 1999; Hermalin and

Weisbach 1998; Byrd and Hickman 1992), we posit that the board strength affects the sensitivity

of turnover to losses. Our results indicate that the strength of the relationship between CEO

turnover and losses is stronger for firms with more effective boards, i.e., stronger boards are

more likely to hold CEOs accountable for reporting losses.

Fifth, while reported earnings might serve as reliable signals of current and past

managerial performance, accounting measures are less likely to capture future investment

opportunities. Therefore, boards might consider future growth opportunities along with current

7

performance to determine CEO retention decisions. Accordingly, we conjecture that the

sensitivity of CEO turnover to accounting losses is less pronounced for firms with large

investment opportunities. We find that the CEO turnover-loss relationship is weaker for firms

with large investment opportunities, implying that boards are willing to shield CEOs from losses

for growing firms to encourage long-term investments.4

Overall, our results suggest that boards rely on accounting losses for CEO retention

decisions. We also hypothesize and find that other firm- and industry-specific attributes affect

the loss-turnover relationship. 5 Prior studies provide persuasive evidence that accounting

recognition of losses is valuable to lenders and that it has a positive impact on debt contract

efficiency. Extending this line of literature, we provide evidence that accounting recognition of

losses is also valued by boards to monitor CEOs.

Our study also improves our understanding of three distinct areas of research. Relying on

a non-linear specification around losses, we document that once we control for losses,

accounting performance is no longer statistically significant in explaining CEO turnover. Also,

prior studies generally conclude that CEOs manage earnings to avoid losses.6 Our results suggest

that protecting jobs might be a key consideration for CEOs managing earnings to avoid losses.

Finally, prior studies conclude that the stock market reacts to profits, but not losses, because

4Additionally, similar to the studies examining the rewards from sustained growth for positive

earnings (Elliott et al. 2010; Ghosh et al. 2005; Barth et al. 1999), we analyze and find that CEO turnover likelihood increases with the frequency of annual losses.

5Our analyses assume that accounting loss is a pre-determined variable. Because prior research

suggests that accounting loss might be endogenously determined (Klein and Marquardt 2006; Joos and Plesko 2005), we also use a two-stage least squares estimation procedure to address endogeneity concerns and continue to find unusually high frequency of CEO turnover for firms reporting losses.

6 The discontinuity in the frequency of firm-years around zero earnings (e.g., Hayn 1995;

Burgstahler and Dichev 1997) is widely cited as evidence of earnings management to avoid reporting losses. Similarly, Roychowdhury (2006) provides evidence consistent with the premise that managers manipulate operating (‘real’) activities to avoid reporting losses.

8

CEOs are expected to exercise the ‘liquidation or abandonment option’ when losses are

persistent (Collins et al. 1999; Burgstahler and Dichev 1997; Hayn 1995; Berger and Ofek 1996).

Our results suggest that the enhanced threat of a job loss provides incumbent CEOs with strong

incentives to abandon operations when a firm reports a loss.

The rest of the paper is organized as follows. Section II develops the hypotheses, Section

III outlines our research design to test our hypotheses, and Section IV describes the sample

selection procedure and the data. Section V reports the empirical results, Section VI discusses

sensitivity analyses, and finally Section VII concludes the paper.

II. THE RELATIONSHIP BETWEEN ACCOUNTING LOSSES AND CEO TURNOVER

In a survey and interview of 400 key executives directly involved in the financial

reporting process, Graham et al. (2005) find that 78% of the executives admit to sacrificing long-

term growth to report immediate profits rather than a loss. Why are CEOs so concerned about

reporting accounting losses and why would they go to such lengths to manage earnings so as not

to report losses? We hypothesize that a key reason for CEOs avoiding losses is related to career

concerns.

Losses versus Profits and CEO Turnover

The board of directors is primarily charged with the responsibility of monitoring,

evaluating, and rewarding management and ultimately firing a CEO for poor performance (e.g.,

Hermalin and Weisbach 1998). Because board members cannot directly observe the ability of a

CEO, they must rely on various performance measures (public signals) and inside information

(private signals) to evaluate the performance of a CEO (Taylor 2010). Prior studies document

that CEO turnover is higher for firms performing poorly which suggests that boards rely on

9

public signals, along with their private information, to make CEO turnover decisions. 7 We

propose a new bright-line test that may be used in CEO retention decisions—whether a firm

reports a loss or a profit.

Researchers tend to agree that reported earnings in the U.S., and those around the world,

follow the ‘accounting conservatism principle’ (e.g., Ball et al. 2000). One interpretation of this

principle is that the income statement anticipates all losses (current and future) but not future

profits (Basu 1997). Lower of the cost or market for inventories, immediate recognition of future

losses on long-term contracts, recognition of future losses for operations designated as

discontinued, and asset impairments are some examples of reporting conservatism. A key

implication of accounting conservatism principle is that losses are more timely and reliable

signals of deteriorating managerial performance than small or declining profits. Therefore, the

relationship between CEO turnover and accounting performance is stronger for losses than for

profits.

Additionally, losses might act as a heuristic for ultimate failure (Pinnuck and Lillis

2007).8 Accounting losses are a signal that the underlying business model has failed under the

present leadership. Consistent with the premise of failed leadership, Graham et al. (2005) find

that three-fourths of the survey respondents believe that their inability to avoid losses is seen as a

“managerial failure” by the executive labor market and by corporate boards. According to one of

the surveyed executives, “if I miss the target, I’m out of a job.” One such target includes

7Most analytical CEO turnover models assume that over time the board learns more about the

CEO’s ability based on firm performance and other private information. The board decides to replace the incumbent manager whenever a CEO’s ability falls below a threshold which is less than the ability of a replacement manager (Jenter and Kanaan 2011).

8Some board members might favor CEO turnover for loss firms because of a loss aversion. If a

board member is more sensitive to losses than profits because a loss serves as a heuristic for failure (Barberis and Huang 2001; Kahneman and Tversky 1979), he/she might conclude that a loss is sufficient justification of a change.

10

avoiding losses; failure to report a profit may be seen as a sign of failed policies. Losses might

also indicate that a firm is experiencing serious difficulties and these problems are unlikely to be

resolved under the current leadership. These arguments suggest that boards might hold CEOs

accountable for losses.

In a related study, Watts (2003, 213) posits that “managers have incentives to hide losses

to avoid being fired before their tenure is over.” Admitting to losses might indicate that the CEO

invested in negative NPV projects and that the possibility of future profits under the incumbent

management might be remote. Therefore, losses provide board of directors with a signal to

investigate the reasons for those losses, to better understand why previously set goals and

objectives were not met, and to reevaluate the CEO’s future goals (Bowen 2008). A heightened

scrutiny of the CEO following a loss is more likely to lead to CEO turnover for at least two

reasons. First, holding other factors constant, an in-depth evaluation increases the chances of the

board uncovering more detrimental information about the CEO’s ability than through a more

routine annual evaluation. Second, if boards do not have the assurance that a change in business

strategy is likely even when confronted with losses, they are expected to replace the incumbent

CEO.9

Finally, shareholders might expect the board to dismiss the CEO when a firm reports a

loss because of erosion in equity value and the board might be acting to placate shareholders

(Watts 2003). For instance, DeAngelo (1988, 15) reports that dissident shareholders waging a

proxy contest “tend to emphasize losses as necessitating a management change.” Therefore,

boards might be willing to remove a CEO from office to placate dissident shareholders.

9For instance, instead of abandoning loss-making projects, CEOs may continue to operate their

pet projects by subsidizing the losses with the profits from other segments. Similarly, entrenched and powerful CEOs may be unwilling to discontinue projects with losses either because they are reluctant to acknowledge their mistakes or because of some personal benefits from managing a larger firm.

11

Hypothesis 1: Accounting losses increase the likelihood of CEO turnover. In the subsequent sub-sections, we theorize how the loss-turnover relationship is expected

to vary with several firm- and industry-specific attributes.

Recurring versus Non-Recurring Losses

Accounting performance includes recurring and non-recurring operations. A reasonable

question is whether boards discriminate between losses from recurring and non-recurring

operations. If CEOs are held accountable for losses, it seems plausible to assume that CEOs are

more likely to be held accountable for losses from continuing operations because it is indicative

of failed managerial performance. We conjecture that losses from discontinued operations also

provide useful additional information about the CEO’s competence. For instance, losses arising

from discontinued operations suggest that the CEO is likely to have made poor investment

decisions in the past that are currently being reversed. Therefore, we posit that

Hypothesis 2: Accounting losses from recurring and non-recurring operations both increase the likelihood of CEO turnover.

Idiosyncratic versus Systematic Losses

Boards are expected to hold CEOs accountable for poor performance when losses result

from mismanagement of the firm, lack of leadership or incompetence, or CEO’s inability to take

optimal decisions. In contrast, boards are expected to shelter CEOs from losses when poor

performance is systematic in nature because of industry- or economy-wide shocks which affect

many firms (Bushman et al. 2010; Gibbons and Murphy 1990; Jenter and Kanaan 2011).

Hypothesis 3: The sensitivity of CEO turnover to accounting losses declines as more firms from the same industry report losses.

Losses and Outside Replacement

The decision to fire a poorly performing CEO benefits shareholders only when the board

appoints a more capable successor. CEOs who are appointed from outside the firm are more

12

likely to change pre-existing firm policies that have resulted in losses. Consistent with this

perspective, Borokhovich et al. (1996) find that the stock market views the appointment of an

outside CEO more favorably than the appointment of an insider, especially when the incumbent

CEO is forced to resign. Therefore, accounting losses increase the likelihood that the board

might prefer an outside replacement to send a strong signal to investors that the CEO is

committed to turning around the firm.

Hypothesis 4: CEO turnover following losses leads to more outside CEO appointments. Board Strength

Several studies find that CEO turnover is more frequent following poor performance

when boards are more effective and independent (e.g., Faleye et al. 2011; Huson et al. 2001;

Goyal and Park 2002; Core et al. 1999; Hermalin and Weisbach 1998; Byrd and Hickman 1992).

This suggests that more effective and independent boards are more likely to hold CEOs

responsible for poor performance as measured by losses. Therefore,

Hypothesis 5: The sensitivity of CEO turnover to accounting losses increases with stronger boards.

III. RESEARCH DESIGN

Based on the extant literature on CEO turnover (e.g., Bushman et al. 2010; Fich and

Shivdasani 2006; Engel et al. 2003; Desai et al. 2006; Farrell and Whidbee 2003; Huson et al.

2001), we test the relationship between CEO turnover and accounting losses using the following

logistic regression.

Turnover = 0 + 1Loss + 2Accounting-return + 3ΔAccounting-return + 4Stock-return + 5Stock-volatility + 6Earnings-volatility + 7Forecast-error + 8Concentration + 9Size + 10Growth + 11Restructure + 12Restatement + 13Age + 14Tenure + Industry/Year Fixed effects + (1)

13

Where Turnover is an indicator variable that equals 1 when there is a change in the CEO in the

two years subsequent to the current fiscal year and 0 otherwise.10 Our main independent variable

is Loss, which is also an indicator variable with a value of 1 when net income is negative for the

current year and 0 otherwise. The predicted sign of the coefficient on Loss is positive; CEO

turnover for firms with losses is expected to be higher than firms with profits.

We also include the following performance and control variables. Accounting-return is

industry-adjusted return on assets measured as the difference between the firm-specific and the

industry-median income before extraordinary items deflated by total assets at the beginning of

the year. Accounting-return is the difference between current period income before

extraordinary items and the corresponding number in the prior year deflated by total assets at the

beginning of the year. Stock-return is the difference between the raw returns and the value-

weighted CRSP market returns over a twelve-month fiscal period. Stock-volatility is the standard

deviation of Stock-return based on prior twenty-four monthly returns. Earnings-volatility is the

standard deviation of Accounting-return over the previous five years. Forecast-error is the

difference between reported annual EPS and the mean forecast EPS deflated by stock price at the

beginning of the year. Concentration is the industry level Herfindahl index. Size is the

logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of

the market value of equity and the book value of debt scaled by the book value of total assets.

Restructure is an indicator variable that equals 1 if special items as a percentage of total assets

are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to

1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is

10Some studies measure performance and CEO turnover concurrently. Because we are interested in a causal relationship, it becomes critical that turnover is measured subsequent to losses. Also, if incumbent CEOs tend to take a “big bath” in the first year of their tenure, CEO turnover and losses should not be measured concurrently to avoid a mechanical relationship. Our results are similar when we limit changes in the CEO in the year subsequent to the current fiscal year.

14

an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise.

Tenure is the number of years that the CEO has been in office as of the fiscal year-end.

We include two accounting and one market measures of performance (Accounting-return,

Accounting-return, and Stock-return) because prior studies find that CEO turnover is related to

stock and accounting performance (e.g., Bushman et al. 2010; Fich and Shivdasani 2006; Farrell

and Whidbee 2003; DeFond and Park 1999; Murphy and Zimmerman 1993; Weisbach 1988).

We include two measures of volatility, one market (Stock-volatility) and another accounting

(Earnings-volatility), because firms with higher volatility are more prone to severe shocks that

lead to more frequent CEO turnovers (Bushman et al. 2010; Engel et al. 2003; DeFond and Park

1999). We control for analysts’ forecast errors because Farrell and Whidbee (2003) find that firm

performance expectations affect CEO turnover. We control for industry concentration because

CEO turnover is greater in highly concentrated industries than in less concentrated industries

(DeFond and Park 1999). We control for firm size (Size) and investment opportunity (Growth)

because larger firms and growing firms have a greater demand for high quality CEOs (Smith and

Watts 1992). We include indicator variables for restructuring activities (Restructure) and

financial restatements (Restatement) because firms with structural or reporting problems are

more likely to be associated with CEO turnovers (Desai et al. 2006; Pourciau 1993). Because not

all CEO turnovers are performance related, as in DeFond and Park (1999) and Desai et al.

(2006), we include an indicator variable for CEOs who are 60 years or older (Age) and a tenure

variable to measure the number of years in office (Tenure). Finally, we include fixed effects for

years and industry to control for variations in CEO turnover over time and across industries.

We also estimate an augmented equation that includes governance variables in addition to

the control variables included in Equation (1) because prior studies find that CEO turnover is

15

associated with board characteristics (e.g., Faleye et al. 2011; Dahya et al. 2002; Weisbach 1988;

Goyal and Park 2002). We include the number of directors on the board (Board-size), the

percentage of independent directors on the board (Board-independence), indicator variables

when a firm has a separate CEO and board chair (Separate-chair) and when a firm has separate

audit, nominating, and compensation committees (Separate-committees), and the percentage of

common stock held by the CEO (Ownership).

IV. DATA AND DESCRIPTIVE STATISTICS

Data and Sample Selection

Our sample consists of Standard and Poor’s (S&P) 1500 firms from Compustat’s

ExecuComp files during the period 1997 to 2007. Included in the ExecuComp files are the names

of the top executives in the firm, a CEOANN variable indicating which of the executives has the

title of a CEO, and the starting date of the CEO. Our CEO turnover indicator variable is

constructed from the information contained in ExecuComp files. If the name of the executive

listed as a firm’s CEO for the current year is different from the one listed as the CEO for the

prior year, we conclude that there is a change in the CEO, or a new CEO is hired, for the current

year. Because we define Turnover as one when there is a change in a CEO for the subsequent

two years, and our sample period ends with 2007, we consider accounting loss from 1997 to

2005.

Ideally, our sample would only consist of involuntary or forced turnovers. However, it is

often difficult to categorize CEO turnovers into voluntary and involuntary turnovers by reading

press articles (Engel et al. 2003). For instance, prior studies discuss the unreliable nature of the

press articles and how press releases often present involuntary turnovers as retirements (DeFond

and Park 1999; Warner et al. 1988). Therefore, as in prior studies we control for voluntary

16

turnovers by including a separate indicator variable for retirement age in our regression analyses

(e.g., Engel et al. 2003).

We also obtain CEO ownership, age and tenure data from the ExecuComp files. The data

on earnings and other firm characteristics are obtained from Compustat annual files. Stock return

data are obtained from CRSP files. We obtain analyst earnings forecast data from the IBES

summary files and board characteristics (size, composition, and structure) from the RiskMetrics

database (also previously known as IRRC). We construct one combined sample by merging the

CEO, accounting, stock return, forecast, and governance data. To remove the effect of outliers,

we winsorize the top or bottom 1 percent of the observations for Accounting-return,

Accounting-return, Stock-return, Earnings-volatility, Concentration, and Growth. 11 This

sample selection procedure results in 11,031 firm-year observations over fiscal years 1997

through 2005 with information about CEO turnover included up to 2007. However, when we

include analysts’ forecasts, the sample size reduces to 9,459.

Descriptive Statistics

Table 1 reports the descriptive statistics for the variables included in Equation (1). CEO

turnover levels are higher than those typically reported by prior studies; the frequency of CEO

turnover is 24.6% over the entire sample period. Losses are fairly common; of all the firm years,

17.6% report negative net income. The mean (median) industry-adjusted return on assets

(Accounting-return) and changes in income before extraordinary items deflated by total assets

(Accounting-return) are 5.1% (2.9%) and 0.9% (0.6%), respectively. The mean (median)

cumulative market-adjusted stock returns (Stock-return) are 8% (1.5%). The mean (median)

11Our results are not sensitive to other outlier identification methods and they remain qualitatively

unchanged when we remove the top and/or bottom 0.5 or 1 percent of observations or even retain all the outliers.

17

return volatility (Stock-volatility) is 0.115 (0.104), whereas the mean (median) earnings volatility

(Earnings-volatility) is 0.057 (0.029). The mean (median) analysts’ earnings forecast errors

(Forecast-error) are -0.007 (-0.000). The Herfindahl index (Concentration) has a median of

0.041. The mean fiscal-year end market value of equity (Market-equity) is $8.03 billion, while

the median number is much smaller ($1.52 billion). The mean (median) market-to-book ratio

(Growth) is 1.68 (1.20). 8.6% of firm years report special items less than or equal to -5 percent of

total assets and 8.2% of firm years are involved with restatements in the current or prior year.

The mean and median values of CEO age are very close at 56 years, whereas the mean (median)

CEO tenure (Tenure) is 8 (6) years.

Table 2 presents the relative frequency of CEO turnover for firms reporting losses and

those reporting profits. Consistent with our expectations that accounting losses are more likely to

lead to a CEO turnover, Turnover in Panel A is higher among loss firms than among profit firms.

More specifically, the frequency of a CEO turnover in the subsequent two years is 34% when

firms report negative net income in the current year while the corresponding number is 23%

when firms report a non-negative number as net income. The difference in frequency of turnover

between the two groups of firms is statistically significant at the 1 percent level. The preliminary

results indicate that firms with losses have higher CEO turnover than firms with profits.

Panel B of Table 2 reports the frequency of CEO turnover for loss and profit firms from

1997 to 2005. The frequency of CEO turnover for firms reporting profits appears to be constant

around 20% over the sample years. On the other hand, the frequency of CEO turnover for firms

reporting losses fluctuates over time but is always higher than that for profit firms. The

difference in the frequency of CEO turnover between loss and profit firms is statistically

significant for each of the sample years indicating that our hypothesis that losses lead to higher

18

CEO turnover is statistically reliable across each of the years.

V. EMPIRICAL RESULTS

CEO Turnover and Accounting Losses

Table 3 presents the logistic regression results for Equation (1). The dependent variable

Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two

years subsequent to the current year and 0 otherwise. Our interest is in the sign and magnitude of

the coefficient on Loss. In regression (1), we first replicate prior studies by including accounting

and stock performance measures but without including Loss. Consistent with prior research, we

find that accounting and stock performance measures are significantly negatively associated with

turnover. The coefficients on Accounting-return and Stock-return are -1.083 (2=15.57) and -

0.398 (2=53.48), respectively. As in prior studies, we find that the economic significance of

accounting performance on CEO turnover is small. Holding the values of the other explanatory

variables at their mean values, a 10% increase in industry-adjusted return on assets decreases the

probability of CEO turnover by 1.9%. Also, as in Huson et al. (2001) and Weisbach (1988), the

coefficient on Accounting-return is negative but insignificant (-0.114, 2=0.15).

More importantly, when we additionally include Loss in regression (2), we find a

statistically significant relationship between CEO turnover and accounting losses. The

coefficient on Loss is 0.435 (2=31.22). The economic magnitude of the coefficient is large.

Holding the other variables constant, the odds of a CEO losing his/her job within two years of

reporting a loss are about 54 percent higher than the odds of CEO turnover of reporting a profit.

Another key empirical result is that the coefficient on Accounting-return becomes insignificant (-

0.497, 2=2.86) when we include Loss as an additional explanatory variable. In sharp contrast,

19

the coefficient on Stock-return remains significant (-0.375, 2=47.46). The coefficient on

Accounting-return continues to be insignificant (-0.003, 2=0.01).

When we include all the control variables in regression (3), the results are very similar to

those in regression (2). Although the magnitude of the coefficient on Loss becomes smaller

(0.349, 2=14.25), we find it reassuring that the results continue to be significant even with the

reduced sample size from additionally including analysts’ forecasts errors (Forecast-error). Our

results in regressions (2) and (3) indicate that while accounting and stock performance is

important in CEO performance evaluations, accounting losses as a metric for judging CEO

competence dominate other accounting performance measures.

The results of the control variables are generally consistent with our expectations and

similar to those reported in prior studies (e.g., Bushman et al. 2010; Desai et al. 2006; Farrell and

Whidbee 2003; Engel et al. 2003; Huson et al. 2001; DeFond and Park 1999). In regression (3),

the coefficient estimate on Forecast-error is negative and highly significant, which indicates that

poor earnings performance relative to analyst forecasts significantly increases the likelihood of

CEO turnover. The coefficient estimates on Stock-volatility, Size, Restructure, Restatement, Age,

and Tenure are all positive and significant. The results suggest that the likelihood of CEO

turnover is higher for firms with higher stock return volatility, bigger firms, firms with

restructuring activities, restating firms, older firms, and firms with longer CEO-tenure. The

coefficients on Earnings-volatility, Concentration, and Growth are insignificant.

One concern with Table 3 is that our regression specification excludes governance

measures for various agency problems which might impact CEO turnover (e.g., Faleye et al.

2011; Fich and Shivdasani 2006; Dahya et al. 2002; Weisbach 1988; Goyal and Park 2002). For

instance, a CEO with a higher equity ownership has more power which might affect CEO

20

turnover decisions. Accordingly, we additionally include the size, composition, and structure of

the board, and CEO ownership in Equation (1). Specifically, we include the number of members

on the board (Board-size), the percentage of independent directors on the board (Board-

independence), the separation of CEO and board chair positions (Separate-chair), the presence

of separate standing sub-committees (Separate-committees), and the percentage of common

stock held by the CEO (Ownership). The additional data requirement reduces our sample to

8,077 and 7,233 firm-year observations, respectively, without and with controlling for Forecast-

error.

The results in Table 4 show that the inclusion of the additional board and ownership

variables does not alter the relation between CEO turnover and accounting losses. Consistent

with the results in Table 3, the positive relation between CEO turnover and losses continues to

hold. The coefficient on Loss is 0.383 (2=15.80) and 0.289 (2=6.81), respectively, without and

with the inclusion of Forecast-error in regressions (1) and (2). The parameter estimate in

regression (2) suggests that the odds of a CEO turnover are about 34 percent higher for loss firms

than for profit firms. We also find that, consistent with the findings in prior studies (e.g., Goyal

and Park 2002), the coefficient on Separate-chair is positive and significant, suggesting that

firms with separate CEO-Chair positions have higher turnover than firms with combined

positions. The coefficient on Separate-committees is also positive and significant, which

suggests that CEO turnover is more frequent when the board structure includes separate sub-

committees. The coefficient on Ownership is negative and significant, indicating that high equity

ownership by the CEO makes it more difficult for a board to remove a CEO. The coefficients on

Board-size and Board-independence are insignificant.

21

We use an indicator variable for losses to examine the relation between accounting

losses and CEO turnover. The underlying assumption that the impact of losses on CEO turnover

does not depend on the magnitude of accounting losses may be too restrictive. Therefore, we

relax this assumption by examining whether the strength of the relation between losses and

turnover varies with the size of losses by adding the magnitude of the loss (Magnitude).

Magnitude represents decile ranks of the absolute value of net income deflated by book value of

equity at the beginning of the year.12

Table 5 reports the results after including LossMagnitude to estimate how Magnitude

affects the sensitivity of top executive turnover to losses. We find that the coefficient on Loss

remains positive and significant in regression (1). More importantly, the sensitivity of CEO

turnover to accounting losses becomes larger as the magnitude of accounting losses increases.

The coefficient on LossMagnitude is 0.052 (2=4.31) in regression (1). When we include

governance variables in regression (2), the results are very similar to those in regression (1). The

coefficient on LossMagnitude is 0.045 (2=4.02). Our results suggest that boards take into

account both incidence and size of accounting losses in holding CEOs accountable for poor

performance.

Overall, the results from Tables 3 to 5 suggest that CEOs reporting losses are more likely

to lose their jobs within the two-year period compared to CEOs reporting profits.

Decomposition of Losses

Our analysis in the previous section partitions firms into profit and loss groups based on

net income, which is the bottom-line measure of accounting performance. We further examine

12We deflate loss by book value of equity because it indicates the degree of erosion in shareholder

equity, which aids the economic interpretation. We use deciles instead of including the continuous variable to minimize the effect of outliers (firms can report extremely large losses). The results are very similar when we interact Loss with Accounting-return.

22

whether the loss findings vary depending on whether income is recurring or non-recurring. By

partitioning net income, we are able to test whether losses from recurring or non-recurring

earnings components are treated differently when evaluating CEO performance. Similar to Gaver

and Gaver (1998), and Dechow et al. (1994), we decompose net income into: (1) recurring items

defined as income before extraordinary items and discontinued operations, and (2) non-recurring

items defined as the difference between net income and income before extraordinary items and

discontinued operations. Accordingly, we decompose Loss into two indicator variables measured

based on each earnings element. LossIB equals 1 when income before extraordinary items and

discontinued operations is negative. LossNR equals 1 when the difference between net income and

income before extraordinary items and discontinued operations is negative.

Table 6 reports the regression results of CEO turnover on the loss components. We find

that both LossIB and LossNR are significantly related to CEO turnover. In regression (1) when we

include all the variables except Forecast-error, the coefficients on LossIB and LossNR are 0.294

(2=8.01) and 0.211 (2=8.57), respectively. In regression (2) when we additionally include

Forecast-error, the coefficients on LossIB and LossNR continue to be significant (0.203, 2=5.86;

0.196, 2=6.30). The coefficients on recurring losses (11) and non-recurring losses (12) are

statistically indistinguishable; 11−12 is 0.083 (2=0.41) and 0.007 (2=0.01), respectively, in

regressions (1) and (2). Our results indicate that both recurring and non-recurring losses are

incrementally important in explaining CEO turnover.

Part of the earnings decomposition result might be mechanical. We find that 90% of the

firm-years with losses for net income also have negative recurring income (the correlation

between Loss and LossIB is also 0.923). Therefore, it is not surprising that losses based on

recurring earnings also lead to higher turnover as losses for net income. However, the correlation

23

between Loss and LossNR and that between LossIB and LossNR are low (0.148 and 0.042,

respectively). Low correlations imply that recurring income and non-recurring income can

provide distinct insights into the CEO’s ability. For instance, discontinued operations with losses

might indicate to boards that the CEO invested in negative NPV projects (Elliott and Hanna 1996)

which might be useful in assessing the CEO’s competence.

Idiosyncratic versus Systematic Losses

A related but important question is whether CEOs are held responsible for reporting

losses when other firms in the industry also report losses, i.e., losses are the outcome of

systematic negative shocks to the industry. Following Heflin and Hsu (2008), we construct a

variable, Industry-loss, which measures the proportion of firms reporting losses in each industry

for a given year where industry is defined using two-digit Standard Industry Classification (SIC)

codes. Our measure is constructed to capture industry-wide shocks. For instance, a negative

shock to the industry affecting many firms in that industry is likely to result in a significant

number of firms with losses. Therefore, as Industry-loss increases, a firm-specific loss is more

likely to be the outcome of industry-wide effects (systematic loss). In contrast, a smaller value

for Industry-loss indicates that a firm-specific loss is more likely to be the result of firm specific

factors rather than industry-wide effects (idiosyncratic loss).

Accordingly, in Table 7, we also examine whether Industry-loss affects the sensitivity of

turnover to losses by estimating an augmented logistic regressions after additionally including

LossIndustry-loss and Industry-loss along with the other variables. In regression (1) when we

include all control variables except Forecast-error, we find that the coefficient on Loss is

positive and significant (0.798, 2=40.08) while that on LossIndustry-loss is negative and

significant (-1.402, 2=11.97). Our results suggest that industry wide losses reduce the impact of

24

firm specific losses on CEO turnover. We get similar results from regression (2) when we

additionally include governance variables and Forecast-error.

The results from Table 7 provide direct evidence that boards are discriminating in

holding CEOs responsible for reporting losses. When losses are the outcome of poor managerial

performance, i.e., idiosyncratic in nature, the likelihood of CEO turnover is high following losses.

However, when losses are more systematic to the industry, boards are less likely to hold CEOs

accountable for losses.13

Outside Replacement and Accounting Losses

We also examine whether accounting losses increase the likelihood of an outside

replacement. We hand collect data to establish whether the successor CEO is an outsider by

reading press releases, 10-K reports, and associated proxy statements. The sample to examine the

impact of losses on outside appointments consists of 1,379 CEO turnover observations. Table 8

presents the regression results on the relationship between accounting losses and the likelihood

of outside succession, conditional on CEO turnover. We use a dichotomous dependent variable

that equals 1 when the incumbent CEO is replaced with a successor CEO from outside the firm

and 0 if the replacement CEO is appointed from within the firm. The independent variables are

based on two prior studies examining outside CEO turnover, Farrell and Whidbee (2003) and

Borokhovich et al. (1996).

We find that the coefficient on Loss is positive and significant in regressions (1) and (2).

Our results suggest that accounting losses lead to more frequent appointments of CEOs from

outside the firm. We also find that the coefficient on Separate-committees is positive and

13When we also use an alternative specification of Industry-loss by taking the natural logarithm of

the value plus 1, we find that our results remain unchanged. For example, the coefficients on Loss, LossIndustry-loss, and Industry-loss are 0.848 (2=38.80), -1.866 (2=12.85), and 0.251 (2=0.72), respectively, in regression (1).

25

significant, which indicates that the likelihood of outside succession is higher for firms with

specialized committees on audit, appointment, and remuneration issues. Among the other

variables in regression (2), we find that the coefficients on Stock-return, Size, and Tenure are

negative and significant, implying that the boards of larger firms with higher stock performance

and longer tenured CEOs tend to hire an insider to replace the incumbent CEO.

Boards, CEO Turnover, and Reporting of Losses

We also examine whether board strength affects the sensitivity of turnover to losses by

augmenting the logistic regressions reported in Table 3 after additionally including an interaction

term between accounting losses and a composite measure of board strength. Drawing on prior

studies (e.g., Bertrand and Mullainathan 2001), we construct a measure for board strength based

on six individual measures of board characteristics that are transformed into standardized values.

Specifically, we demean each of the six variables and divide it by its standard deviation. Board-

strength is then computed by summing the following six transformed measures: (1) board size,

(2) board independence, (3) separate CEO-Chair, (4) separate committees, (5) CEO ownership,

and (6) CEO tenure. For board size, CEO ownership, and CEO tenure, we use negative of board

size, CEO ownership, and CEO tenure before transforming them into standardized values so that

increasing values of individual characteristics indicate stronger boards.

The results from including LossBoard-strength and Board-strength in the turnover

regressions are presented in Table 9. In regression (1), we find that the coefficient on Loss

continues to be positive and significant (0.330, 2=11.59), but the coefficient on LossBoard-

strength is also positive and significant (0.077, 2=6.41). The results indicate that the strength of

the relationship between CEO turnover and losses is stronger for firms with stronger and more

26

effective boards. To the extent that Board-strength measures effectiveness of boards, our results

suggest that stronger boards are more likely to hold CEOs accountable for reporting losses.

In regression (2) when we additionally include Forecast-error, we get similar results to

those in regression (1). The coefficients on Loss and LossBoard-strength are 0.239 (2=4.55)

and 0.074 (2=4.77), respectively. Thus, our results from Table 9 corroborate the results from

prior tables. We additionally find direct evidence that stronger boards are more proactive in

holding CEOs accountable for eroding shareholder’s equity.

Key Implications

Our first hypothesis has broader implications. Several valuation studies find that the

relation between returns and earnings is weaker for loss firms than for profit firms (Collins et al.

1999; Burgstahler and Dichev 1997; Hayn 1995). The “liquidation/abandonment option” to

redeploy existing assets is often used as an explanation for the differential results between firm

values and earnings for profit and loss firms. Assuming that CEOs are willing to liquidate a firm

or to discontinue a segment when losses are expected to perpetuate, investors perceive losses as

being temporary. Therefore, the stock market reaction to losses is muted.

However, in the presence of agency problems, it is unclear whether incumbent CEOs

would exercise the liquidation/abandonment option when losses are expected to continue. For

example, Ofek (1993) finds that entrenched managers continue operations even when a firm is

distressed. Our study helps us better understand why losses are temporary. Because boards play a

proactive role in replacing entrenched CEOs that are unwilling or unable to change their failed

business strategies, losses are more likely to be temporary because new CEOs are more likely to

reverse the failed strategies of the predecessor CEO.

VI. SENSITIVITY ANALYSIS

27

CEO Turnover and the Frequency of Accounting Losses

Prior studies show that debt and equity markets reward firms with sustained earnings

growth because sustained earnings increases are indicative of the firms’ competitive advantages

and a higher probability of future earnings and cash flow growth (Elliott et al. 2010; Ghosh et al.

2005; Barth et al. 1999). Similar to the studies on the information content of sustained earnings

growth, we analyze whether the likelihood of a CEO turnover is affected by the frequency of

losses. Accordingly, we decompose Loss into 3 indicator variables depending on the frequency

of annual losses for the past five years including the current year. Loss1 equals 1 for firms with a

loss in the current year but not in the prior four years. Similarly, Loss2 (Loss3) equals 1 for firms

with two (three) years of losses in the past five years including the current year.14

Our results (not reported) show that the frequency of losses is incrementally important in

explaining CEO turnover. In the first regression when we include all the control variables except

Forecast-error, the coefficients on Loss1, Loss2, and Loss3 are all positive and statistically

significant (0.422, 2=16.71; 0.528, 2=25.06; 0.312, 2=6.29). In the second regression when

we additionally include Forecast-error, the coefficients on Loss1 and Loss2 continue to be

significant but that on Loss3 becomes insignificant (0.239, 2=2.60) probably because of lack of

power (i.e., there are few observations with CEO turnover and three annual losses).

Finally, when we include all the control and governance variables in the regression, the

results are very similar to those in the second regression. These results suggest that boards of

directors play a proactive role in replacing CEOs for reporting losses and, as the frequency of

14We do not include more than three years of annual losses in our analysis because very few firms

report losses more than three years over a five-year period. We get very similar results when we construct indicator loss variables to measure consecutive years of losses over a five year period including the current year.

28

losses increases, the likelihood of a CEO being replaced also increases. Our results provide a key

explanation why investors treat losses as being temporary.

CEO Turnover and Accounting Losses for Firms with High Growth Opportunities

For firms in emerging, high-tech, and high-growth industries, current earnings may not

serve as an adequate proxy for future earnings potential because of the mismatching of revenues

and expenses. Growing firms need to make large investments which are expected to yield future

revenues. However, R&D costs are expensed as incurred and growth in tangible/intangible assets

leads to depreciation/amortization expenses which could result in more frequent losses for

growing firms than non-growing firms (Hayn 1995). If CEOs are penalized for all losses

regardless of future growth considerations, they might reduce long-term investments to avoid

losses which would be detrimental for the long-run profitability of the firm. Therefore, boards of

firms with future growth opportunities are expected to shield CEO from losses.

As in prior studies, we use the market-to-book ratio (Growth) to measure future growth

opportunities. Similar to Magnitude, we include decile ranks of Growth to minimize the effect of

outliers. Our untabulated results indicate that the sensitivity of CEO turnover to losses becomes

smaller for firms with higher growth opportunities. In the first regression when we include all the

control variables except Forecast-error, the coefficient on Loss remains positive and significant

(0.570, 2=30.12). In contrast, the coefficient on the interaction between Loss and Growth is

negative and significant (-0.047, 2=4.53). When we additionally include Forecast-error in the

second regression, the results are very similar to those in the first regression. The coefficients on

Loss and interaction term are 0.500 (2=15.41) and -0.046 (2=4.09), respectively. Our results

suggest that boards of growing firms shield their CEOs from losses to encourage long-term

investments.

29

Non-linear Effects of Accounting Returns on CEO Turnover

Jenter and Lewellen (2010) find that the sensitivity of CEO turnover and stock price

performance increases substantially when they allow for non-linear effects of performance on

turnover. Specifically, instead of using a continuous variable, they include stock return quintiles

as separate explanatory variables in a CEO turnover regression. Similar to their approach, we

also allow for the Accounting-return slope to vary by including Accounting-return quintiles.

When we use Accounting-return quintile (each as an indicator variable) in our regression

specification, we find that our main conclusions remain unchanged. For instance, similar to

Table 3, regression (2) results, the coefficient on Loss remains positive and significant; it is 0.445

(2=28.40). On the other hand, treating top quintile Accounting-return as the benchmark, none of

the coefficients on Accounting-return quintiles are significant.

Endogeneity

A potential concern with our prior results is that accounting losses are likely to be

endogenously determined (Klein and Marquardt 2006; Joos and Plesko 2005). There are two

sources of endogeneity in our study. First, there might be a simultaneity problem; losses might

lead to CEO turnover, but CEO turnover might lead newly appointed CEOs to take a “big bath”

and report a loss. We avoid the simultaneity problem by examining CEO turnover subsequent to

the years following losses, i.e., we examine CEO turnover in years +1 and +2 where year 0

denotes a loss-year. Second, random shocks might affect the occurrence of losses which might

also affect CEO turnover decisions.

We address the second type of an endogeneity problem using a two-stage least squares

(2SLS) estimation procedure to obtain consistent and efficient estimates for losses. Specifically,

drawing on prior studies, we model losses in the first stage and then, in the second stage, we

30

regress CEO turnover on the probability of a firm reporting a loss obtained from the first stage

regression. We estimate the occurrence of losses by including four categories of variables: (1)

profitability measures including Cash-flow and Accrual (Joos and Plesko 2005), (2) Size because

of the strong link between losses and firm size (Hayn 1995), and Sales-growth because current

earnings may not capture future prospects of growing firms, (3) lag(Loss) to measure the

incidence of past losses; firms with prior losses are less likely to return to profitability in the

subsequent year, and (4) Dividend (Dividend-stop) because firms continuing to pay dividends

(firms stopping paying dividends) have a lower (higher) probability of incurring future losses

(Joos and Plesko 2005). We include governance and control variables in the first stage (Larcker

and Rusticus 2010). Because of the additional data requirements, the sample decreases to 6,906

observations in the first stage regression.

The results for the first stage estimation are presented in the first column of Table 10. The

coefficients on Cash-flow and Accrual are negative and significant, which suggests that firms

with higher cash flow from operations and larger accruals are less likely to report accounting

losses. In contrast, the coefficient on lag(Loss) is positive and significant, indicating that firms

with losses in the prior year are more likely to incur losses in the current year. The coefficients

on Sales-growth, Dividend, and Dividend-stop are not significant. The Nagelkerke R2 from the

first-stage regression which includes the instruments, governance and control variables is around

65% which is high indicating that our model explains substantial variations across firms

reporting losses. More importantly, we find that the partial Nagelkerke R2 after excluding the

governance and control variables is 30.43% and the partial likelihood ratio 2–statistic is

statistically significant at 2,452.66, which means that the instruments do a reasonably good job

of explaining cross sectional variations in losses.

31

In the second stage, we use the estimated values of losses (Pred-loss) from the first stage

as an instrumental variable and re-estimate Equation (1). After controlling for the endogeneity of

accounting losses, our results confirm the earlier findings on the positive relationship between

CEO turnover and losses. The coefficient on Pred-loss is 0.304 (2=4.23) when we include the

governance variables in the second column. Thus, our results once again confirm that the

likelihood of CEO turnover is higher for firms reporting accounting losses even after controlling

for issues related to endogeneity.

VII. CONCLUSIONS

Several studies examine the importance of earnings and stock returns as measures of firm

performance on CEO turnover considerations (e.g., Weisbach 1988; Murphy and Zimmerman

1993; Goyal and Park 2002; Bushman et al. 2010; Jenter and Kanaan 2011). We suggest that

accounting losses reflect managerial effort and quality that are not fully captured in the

traditional measures of firm performance. In this paper, we investigate whether accounting losses

provide incremental information that can be used to assess CEO retention/dismissal decision.

Specifically, we examine how accounting losses affect subsequent top executive turnover in

relation to the other measures of performance.

Based on a comprehensive sample of CEO turnover between 1997 and 2007, we find

that compared to profit firms, the likelihood of CEO turnover is significantly higher for loss

firms. Controlling for the other determinants of CEO turnover including accounting and stock

performance measures, the relative odds of a CEO losing a job within two years are more than 50

percent higher for firms with losses than with profits. More importantly, when we include losses,

accounting performance measures (earnings levels, earnings changes, or industry-adjusted

earnings) are no longer incrementally significant in explaining CEO turnover. Our results

32

suggest that only losses result in a higher likelihood of a subsequent CEO turnover and variations

in profit levels do not impact CEO turnover decisions.

Additionally, we find the following key results: (1) when we decompose losses (defined

using net income) into losses before extraordinary items and discontinued operations and losses

from extraordinary items and discontinued operations, we find that both measures affect CEO

turnover, (2) the magnitude of the loss also affects CEO turnover, (3) the impact of firm-specific

losses on CEO turnover is significant when losses tend to be confined to the firm but is muted

when losses are more systematic to the industry, (4) CEO turnover following losses leads to

more outside CEO appointments, and (5) the sensitivity of CEO turnover to losses is affected by

the strength of the board and the level of growth opportunities. Collectively, the results suggest

that CEOs are penalized for losses and that boards consider other factors along with losses to

arrive at CEO retention decisions.

Our results suggest that boards view losses as an additional indicator of management

failure and are more likely to penalize CEOs that report losses. Additionally, prior studies often

presume that effective managers exercise the liquidation/abandonment option when losses are

expected to persist. Our results suggest that board actions that threaten higher turnover following

losses increase CEO focus on the liquidation/abandonment option. Finally, our results also

provide one explanation why firms manage earnings to avoid reporting losses.

33

REFERENCES

Ball, R., S. P. Kothari, and A. Robin. 2000. The effect of international institutional factors on properties of accounting earnings. Journal of Accounting and Economics 29: 1-51.

Ball, R., and L. Shivakumar. 2005. Earnings quality in U.K. private firms: Comparative loss recognition timeliness. Journal of Accounting and Economics 39: 83-128.

Barberis, N., and M. Huang. 2001. Mental accounting, loss aversion, and individual stock returns. Journal of Finance 56: 1247-1292.

Barth, M. E., J. A. Elliott, and M. A. Finn. 1999. Market rewards associated with patterns of increasing earnings. Journal of Accounting Research 37: 387-413.

Basu, S. 1997. The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics 24: 3-37.

Berger, P. G., and E. Ofek. 1996 . Bustup takeovers of value-destroying diversified firms. Journal of Finance 51: 1175-200.

Bertrand, M., and S. Mullainathan. 2001. Are CEOs rewarded for luck? The ones without principals are. The Quarterly Journal of Economics 116: 901-932.

Borokhovich, K. A., R. Parrino, and T. Trapani. 1996. Outside directors and CEO selection. Journal of Financial and Quantitative Analysis 31: 337-355.

Bowen, W. G. 2008. The Board Book: An Insider’s Guide for Directors and Trustees. New York, NY: W. W. Norton and Company.

Brickley, J. A. 2003. Empirical research on CEO turnover and firm-performance: A discussion. Journal of Accounting and Economics 36: 227-233.

Burgstahler, D., and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses. Journal of Accounting and Economics 24: 99-126.

Bushman, R., Z. Dai, and X. Wang. 2010. Risk and CEO turnover. Journal of Financial Economics 96: 381-398.

Byrd, J. W., and K. A. Hickman. 1992. Do outside directors monitor managers? Evidence from tender offer bids. Journal of Financial Economics 32: 195-221.

Collins, D. W., M. Pincus, and H. Xie. 1999. Equity valuation and negative earnings: The role of book value of equity. The Accounting Review 74: 29-62.

Core, J. E., R. W. Holthausen, and D. F. Larcker. 1999. Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics 51: 371-406.

Dahya, J., J. J. McConnell, and N. G. Travlos. 2002. The Cadbury committee, corporate performance, and top management turnover. Journal of Finance 57: 461-483.

DeAngelo, L. 1988. Managerial competition, information costs, and corporate governance. Journal of Accounting and Economics 10: 3-36.

Dechow, P., M. Huson, and R. Sloan. 1994. The effect of restructuring charges on executives' cash compensation. The Accounting Review 69: 138-156.

DeFond, M. L., and C. W. Park. 1999. The effect of competition on CEO turnover. Journal of Accounting and Economics 27: 35-56.

Degeorge, F., J. Patel, and R. Zeckhauser. 1999. Earnings management to exceed thresholds. Journal of Business 72: 1-33.

Desai, H., C. E. Hogan, and M. S. Wilkins. 2006. The reputational penalty for aggressive accounting: Earnings restatements and management turnover. The Accounting Review 81: 83-112.

Elliott, J. A., A. Ghosh, and D. Moon. 2010. Asymmetric valuation of sustained growth by bond- and equity-holders. Review of Accounting Studies 15: 833-878.

Elliott, J. A., and J. D. Hanna. 1996. Repeated accounting write offs and the information content of earnings. Journal of Accounting Research 34: 135-155.

Engel, E., R. M. Hayes, and X. Wang. 2003. CEO turnover and properties of accounting information. Journal of Accounting and Economics 36: 197-226.

34

Faleye, O., R. Hoitash, and U. Hoitash. 2011. The costs of intense board monitoring. Journal of Financial Economics 101: 160-181.

Farrell, K. A., and D. A. Whidbee. 2003. Impact of firm performance expectations on CEO turnover and replacement decisions. Journal of Accounting and Economics 36: 165-196.

Fich, E. M., and A. Shivdasani. 2006. Are busy boards effective monitors? Journal of Finance 61: 689-724.

Gaver, J. J., and K. M. Gaver. 1998. The relation between nonrecurring accounting transactions and CEO cash compensation. The Accounting Review 73: 235-253.

Ghosh, A., Z. Gu, and P. Jain. 2005. Sustained earnings and revenue growth, earnings quality, and earnings response coefficients. Review of Accounting Studies 10: 33-57.

Gibbons, R., and K. J. Murphy. 1990. Relative performance evaluation for chief executive officers. Industrial and Labor Relations Review 43: 30S-51S.

Givoly, D., and C. Hayn. 2000. The changing time-series properties of earnings, cash flows and accruals: Has financial reporting become more conservative? Journal of Accounting and Economics 29: 287-320.

Goyal, V., and C. W. Park. 2002. Board leadership structure and CEO turnover. Journal of Corporate Finance 8: 49-66.

Graham, J. R., C. R. Harvey, and S. Rajgopal. 2005. The economic implications of corporate financial reporting. Journal of Accounting and Economics 40: 3-73.

Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20: 125-154. Heflin, F., and C. Hsu. 2008. The impact of the SEC’s regulation of non-GAAP disclosures. Journal of

Accounting and Economics 46: 349-365. Hermalin, B. E., and M. S. Weisbach. 1998. Endogenously chosen boards of directors and their

monitoring of the CEO. American Economic Review 88: 96-117. Hermalin, B. E., and M. S. Weisbach. 2003. Boards of directors as an endogenously determined

institution: A survey of the economic literature. Economic Policy Review 9: 7-26. Huson, M. R., P. H. Malatesta, and R. Parrino. 2004. Managerial succession and firm performance.

Journal of Financial Economics 74: 237-275. Huson, M. R., R. Parrino, and L. T. Starks. 2001. Internal monitoring mechanisms and CEO turnover: A

long-term perspective. Journal of Finance 55: 2265-2297. Jenter, D., and F. Kanaan. 2011. CEO Turnover and relative performance evaluation. Journal of Finance:

Forthcoming. Jenter, D., and K. Lewellen. 2010. Performance-induced CEO Turnover. Working paper, Stanford

University. Joos, P., and G. Plesko. 2005. Valuing loss firms. The Accounting Review 80: 847-870. Kahneman, D., and A. Tversky. 1979. Prospect theory: An analysis of decision under risk. Econometrica

47: 263-291. Kaplan, S., and B. A. Minton. 2006. How has CEO turnover changed? Increasingly performance sensitive

boards and increasingly uneasy CEOs, Working paper, University of Chicago. Klein, A., and C. Marquardt. 2006. Fundamentals of accounting losses. The Accounting Review 81: 179-

206. Larcker, D. F., and T. O. Rusticus. 2010. On the use of instrumental variables in accounting research.

Journal of Accounting and Economics 49: 186-205. Murphy, K. J., and J. L. Zimmerman. 1993. Financial performance surrounding CEO turnover. Journal of

Accounting and Economics 16: 273-315. Ofek, E. 1993. Capital structure and firm response to poor performance. Journal of Financial Economics

34: 3-30. Parrino, R. 1997. CEO turnover and outside succession: A cross-sectional analysis. Journal of Financial

Economics 46: 165-197. Pinnuck, M., and A. Lillis. 2007. Profits versus losses: Does reporting an accounting loss act as a

heuristic trigger to exercise the abandonment option and divest employees?’ The Accounting Review

35

82: 1031-1053. Pourciau, S. 1993. Earnings management and nonroutine executive change. Journal of Accounting and

Economics 16: 317-336. Roychowdhury, S. 2006. Earnings management through real activities manipulation. Journal of

Accounting and Economics 42: 335-370. Sloan, R. 1993. Accounting earnings and top executive compensation. Journal of Accounting and

Economics 16: 55-100. Smith, C. W., and R. L. Watts. 1992. The investment opportunity set and corporate financing, dividend,

and financing policies. Journal of Financial Economics 32: 262-292. Taylor, L. A. 2010. Why are CEOs rarely fired? Evidence from structural estimation. Journal of Finance

65: 2051-2087. Warner, J. B., R. L. Watts, and K. H. Wruck. 1988. Stock prices and top management changes. Journal of

Financial Economics 20: 461-492. Watts, R. L. 2003. Conservatism in accounting, Part I: Explanations and implications. Accounting

Horizons 17: 207-221. Weisbach, M. 1988. Outside directors and CEO turnover. Journal of Financial Economics 20: 431- 460.

36

TABLE 1 Descriptive Statistics

Mean

First quartile

Median

Third quartile Standard deviation

Turnover 0.246 0.000 0.000 0.000 0.431 Loss 0.176 0.000 0.000 0.000 0.381

Accounting-return 0.051 -0.002 0.029 0.098 0.117 Accounting-return 0.009 -0.010 0.006 0.028 0.088 Stock-return 0.080 -0.231 0.015 0.287 0.504

Stock-volatility 0.115 0.074 0.104 0.146 0.055 Earnings-volatility 0.057 0.013 0.029 0.066 0.077 Forecast-error -0.007 -0.012 -0.000 0.004 0.037 Concentration 0.057 0.027 0.041 0.067 0.049 Market-equity 8.033 0.566 1.520 5.121 25.938 Growth 1.683 0.828 1.195 1.958 1.481 Restructure 0.086 0.000 0.000 0.000 0.280 Restatement 0.082 0.000 0.000 0.000 0.274 Age 56.151 51.000 56.000 61.000 7.509 Tenure 7.994 3.000 6.000 11.000 7.566

Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. Loss is an indicator variable with a value of 1 when net income is negative for the current year and 0 otherwise. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Market-equity is the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is the age of the CEO in years (for the outgoing CEO in turnover firms) as of the fiscal-year end. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. Descriptive statistics are based on a sample with 11,031 firm-year observations between 1997 and 2005 for all the variables other than Forecast-error which has 9,459 observations.

37

TABLE 2 CEO Turnover for Loss and Profit Firms

Firms with Profits Losses Differences Panel A: Full Sample

0.226 0.339 -0.113**

Panel B: By Fiscal Year

1997 0.216 0.383 -0.167**

1998 0.243 0.363 -0.120**

1999 0.256 0.457 -0.201**

2000 0.227 0.315 -0.088** 2001 0.194 0.280 -0.086** 2002 0.194 0.293 -0.099** 2003 0.214 0.314 -0.100** 2004 0.220 0.360 -0.140** 2005 0.293 0.473 -0.180**

Firms with losses have negative net income for the current year and the rest of the firms are classified as profit firms. Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. The significance test of differences in means between profit and loss firms is based on the t-tests. ** denotes statistical significance at the 1 percent level based on a two-tailed test.

38

TABLE 3 CEO Turnover and Losses

Dependent variable: Turnover

(1) (2) (3)

Intercept -1.875 (105.09)** -1.940 (111.98)** -2.046 (98.46)**

Performance measures

Loss 0.435 (31.22)** 0.349 (14.25)**

Accounting-return -1.083 (15.57)** -0.497 (2.86) -0.507 (2.24)

Accounting-return -0.114 (0.15) -0.003 (0.01) 0.348 (0.92)

Stock-return -0.398 (53.48)** -0.375 (47.46)** -0.357 (32.23)**

Control variables

Stock-volatility 2.745 (18.14)** 2.073 (9.96)** 1.614 (4.68)*

Earnings-volatility -0.092 (0.05) -0.187 (0.21) -0.539 (1.30)

Forecast-error -2.731 (13.13)**

Concentration 0.220 (0.15) 0.264 (0.22) -0.482 (0.57)

Size 0.070 (18.47)** 0.077 (22.18)** 0.086 (20.61)**

Growth -0.025 (1.31) -0.027 (1.58) -0.003 (0.02)

Restructure 0.321 (13.52)** 0.200 (4.95)* 0.217 (4.55)*

Restatement 0.263 (10.60)** 0.249 (9.46)** 0.292 (11.53)**

Age 0.642 (156.63)** 0.653 (161.07)** 0.712 (162.59)**

Tenure 0.007 (4.93)* 0.007 (5.43)* 0.007 (5.31)*

Fixed effects

Industry Yes Yes Yes Year Yes Yes Yes

Observations 11,031 11,031 9,459 Nagelkerke R2 5.88% 6.27% 6.64%

The dependent variable Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. The independent variables are as follows. Loss is an indicator variable with a value of 1 when net income is negative for the current year and 0 otherwise. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.

39

TABLE 4 CEO Turnover and Losses: Inclusion of Additional Governance Variables

Dependent variable: Turnover

(1) (2)

Intercept -3.171 (141.55)** -3.522 (145.52)**

Performance measures Loss 0.383 (15.80)** 0.289 (6.81)**

Accounting-return -0.438 (1.24) -0.321 (0.53)

Accounting-return 0.085 (0.04) 0.079 (0.03)

Stock-return -0.380 (28.76)** -0.356 (20.52)**

Governance variables

Board-size 0.008 (0.33) 0.017 (1.40)

Board-independence 0.004 (3.47) 0.004 (3.49)

Separate-chair 0.708 (136.69)** 0.759 (136.76)**

Separate-committees 0.222 (7.05)** 0.242 (7.01)**

Ownership -0.021 (11.10)** -0.016 (5.80)*

Control variables

Stock-volatility 1.961 (5.58)* 1.685 (3.45)

Earnings-volatility -0.678 (1.49) -0.697 (1.26)

Forecast-error -2.231 (4.70)*

Concentration 0.286 (0.18) -0.270 (0.13)

Size 0.087 (13.58)** 0.097 (14.10)**

Growth -0.001 (0.01) 0.010 (0.11)

Restructure 0.257 (5.13)* 0.277 (4.93)*

Restatement 0.319 (11.85)** 0.314 (10.38)**

Age 0.688 (126.62)** 0.749 (132.61)**

Tenure 0.030 (53.89)** 0.028 (42.96)**

Fixed effects

Industry Yes Yes Year Yes Yes

Observations 8,077 7,233 Nagelkerke R2 8.80% 9.60%

The dependent variable Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. The independent variables are as follows. Loss is an indicator variable with a value of 1 when net income is negative for the current year and 0 otherwise. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Board-size is the number of directors on the board. Board-independence is the percentage of independent directors on the board. Separate-chair and Separate-committees are indicator variables set to 1 when a firm has a separate CEO and board chair and when a firm has separate audit, nominating, and compensation committees, respectively. Ownership is the percentage of common stock held by the CEO. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.

40

TABLE 5 CEO Turnover and the Magnitude of Losses

Dependent variable: Turnover

(1) (2)

Intercept -2.237 (128.33)** -3.512 (142.87)**

Performance measures

Loss 0.228 (4.33)* 0.211 (4.26)*

LossMagnitude 0.052 (4.31)* 0.045 (4.02)*

Magnitude -0.007 (0.31) 0.004 (0.08)

Accounting-return -0.049 (0.02) -0.108 (0.05)

Accounting-return 0.362 (0.99) 0.077 (0.03)

Stock-return -0.339 (29.47)** -0.356 (20.46)**

Governance variables

Board-size 0.017 (1.39)

Board-independence 0.005 (3.68)

Separate-chair 0.759 (136.29)**

Separate-committees 0.244 (7.15)**

Ownership -0.016 (5.98)*

Control variables

Stock-volatility 1.146 (2.36) 1.604 (3.09)

Earnings-volatility -0.539 (1.29) -0.844 (1.80)

Forecast-error -2.420 (10.00)** -2.075 (3.93)*

Concentration -0.412 (0.41) -0.281 (0.14)

Size 0.091 (22.95)** 0.094 (13.01)**

Growth -0.015 (0.40) 0.002 (0.01)

Restructure 0.158 (2.32) 0.242 (3.63)

Restatement 0.337 (15.84)** 0.314 (10.35)**

Age 0.716 (164.30)** 0.749 (132.68)**

Tenure 0.007 (4.79)* 0.028 (43.19)**

Fixed effects Industry Yes Yes Year Yes Yes

Observations 9,459 7,233 Nagelkerke R2 6.72% 9.65%

The dependent variable Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. The independent variables are as follows. Loss is an indicator variable with a value of 1 when net income is negative for the current year and 0 otherwise. Magnitude represents decile ranks of the absolute value of net income deflated by book value of equity at the beginning of the year. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Board-size is the number of directors on the board. Board-independence is the percentage of independent directors on the board. Separate-chair and Separate-committees are indicator variables set to 1 when a firm has a separate CEO and board chair and when a firm has separate audit, nominating, and compensation committees, respectively. Ownership is the percentage of common stock held by the CEO. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.

41

TABLE 6 CEO Turnover and Loss: Decomposition of Losses

Dependent variable: Turnover

(1) (2)

Intercept -3.132 (138.11)** -3.497 (143.35)**

Performance measures

LossIB (11) 0.294 (8.01)** 0.203 (5.86)*

LossNR (12) 0.211 (8.57)** 0.196 (6.30)*

Test: 11 12 0 0.083 (0.41) 0.007 (0.01)

Accounting-return -0.548 (1.92) -0.417 (0.89)

Accounting-return 0.020 (0.01) 0.027 (0.01)

Stock-return -0.383 (29.19)** -0.358 (20.67)**

Governance variables

Board-size 0.007 (0.28) 0.017 (1.35)

Board-independence 0.003 (3.26) 0.004 (3.79)

Separate-chair 0.712 (138.20)** 0.764 (138.17)**

Separate-committees 0.223 (7.14)** 0.243 (7.09)**

Ownership -0.021 (11.18)** -0.016 (5.78)*

Control variables

Stock-volatility 1.957 (5.54)* 1.667 (3.36)

Earnings-volatility -0.604 (1.18) -0.626 (1.01)

Forecast-error -2.390 (5.32)*

Concentration 0.253 (0.14) -0.314 (0.18)

Size 0.080 (11.64)** 0.092 (12.64)**

Growth 0.004 (0.02) 0.015 (0.25)

Restructure 0.273 (5.70)* 0.296 (5.53)*

Restatement 0.306 (10.86)** 0.302 (9.57)**

Age 0.685 (125.52)** 0.746 (131.68)**

Tenure 0.030 (54.84)** 0.029 (43.88)**

Fixed effects

Industry Yes Yes Year Yes Yes

Observations 8,077 7,233 Nagelkerke R2 8.84% 9.65%

The dependent variable Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. The independent variables are as follows. LossIB is an indicator variable with a value of 1 when income before extraordinary items and discontinued operations is negative for the current year and 0 otherwise. LossNR is an indicator variable with a value of 1 when the difference between net income and income before extraordinary items and discontinued operations is negative for the current year and 0 otherwise. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Board-size is the number of directors on the board. Board-independence is the percentage of independent directors on the board. Separate-chair and Separate-committees are indicator variables set to 1 when a firm has a separate CEO and board chair and when a firm has separate audit, nominating, and compensation committees, respectively. Ownership is the percentage of common stock held by the CEO. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.

42

TABLE 7 CEO Turnover and the Impact of Industry Losses

Dependent variable: Turnover

(1) (2)

Intercept -2.001 (117.11)** -3.512 (142.95)**

Performance measures

Loss 0.798 (40.08)** 0.492 (5.34)* LossIndustry-loss -1.402 (11.97)** -0.561 (4.26)*

Industry-loss 0.144 (0.34) -0.236 (0.57)

Accounting-return -0.508 (2.88) -0.239 (0.28)

Accounting-return -0.006 (0.01) 0.037 (0.01)

Stock-return -0.379 (48.18)** -0.360 (20.88)**

Governance variables

Board-size 0.017 (1.41)

Board-independence 0.005 (3.40)

Separate-chair 0.758 (136.12)**

Separate-committees 0.243 (7.08)**

Ownership -0.016 (5.89)*

Control variables

Stock-volatility 2.110 (10.26)** 1.768 (3.77)

Earnings-volatility -0.104 (0.07) -0.650 (1.09)

Forecast-error -2.254 (4.74)*

Concentration 0.303 (0.29) -0.310 (0.17)

Size 0.080 (23.64)** 0.097 (14.19)**

Growth -0.030 (2.01) 0.007 (0.06)

Restructure 0.214 (5.66)* 0.282 (5.11)*

Restatement 0.245 (9.09)** 0.311 (10.15)**

Age 0.654 (161.51)** 0.749 (132.83)**

Tenure 0.007 (5.97)* 0.028 (42.23)**

Fixed effects Industry Yes Yes Year Yes Yes

Observations 11,031 7,233 Nagelkerke R2 6.45% 9.62%

The dependent variable Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. The independent variables are as follows. Loss is an indicator variable with a value of 1 when net income is negative for the current year and 0 otherwise. Industry-loss is the proportion of firms reporting losses in each industry for a given year when industry is defined using SIC 2-digit codes. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Board-size is the number of directors on the board. Board-independence is the percentage of independent directors on the board. Separate-chair and Separate-committees are indicator variables set to 1 when a firm has a separate CEO and board chair and when a firm has separate audit, nominating, and compensation committees, respectively. Ownership is the percentage of common stock held by the CEO. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.

43

TABLE 8 Outside CEO Appointment Following Losses

Dependent variable: Outsider replacement

(1) (2)

Intercept 0.099 (0.03) -0.164 (0.03)

Performance measures

Loss 0.490 (4.38)* 0.612 (3.98)*

Accounting-return 0.199 (0.08) 0.743 (0.57)

Accounting-return -0.411 (0.36) 0.370 (0.10)

Stock-return -0.292 (3.56) -0.541 (4.51)*

Governance variables

Board-size -0.018 (0.17)

Board-independence 0.003 (0.27)

Separate-chair -0.104 (0.34)

Separate-committees 0.623 (4.41)*

Ownership -0.015 (0.57)

Control variables

lag(Stock-return) -0.264 (3.62) -0.293 (2.12)

Stock-volatility -0.593 (0.10) 2.207 (0.73)

Earnings-volatility 1.231 (1.42) 0.337 (0.04)

Forecast-error 0.376 (0.03)

Concentration -0.971 (0.30) -0.333 (0.02)

Size -0.132 (7.87)** -0.147 (3.87)*

Growth 0.070 (1.31) -0.001 (0.01)

Restructure 0.067 (0.10) 0.146 (0.24)

Restatement -0.020 (0.01) -0.264 (1.30)

Tenure -0.026 (7.52)** -0.026 (3.89)*

Fixed effects Industry Yes Yes Year Yes Yes

Observations 1,379 886 Nagelkerke R2 10.23% 13.10%

The dichotomous dependent variable Outsider replacement equals 1 when the incumbent CEO is replaced with a successor CEO from outside the firm and 0 if the replacement CEO is appointed from within the firm. The independent variables are as follows. Loss is an indicator variable with a value of 1 when net income is negative for the current year and 0 otherwise. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Board-size is the number of directors on the board. Board-independence is the percentage of independent directors on the board. Separate-chair and Separate-committees are indicator variables set to 1 when a firm has a separate CEO and board chair and when a firm has separate audit, nominating, and compensation committees, respectively. Ownership is the percentage of common stock held by the CEO. lag(Stock-return) is the one-year lagged value of Stock-return. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.

44

TABLE 9 CEO Turnover and Losses: The Impact of Board Strength

Dependent variable: Turnover

(1) (2)

Intercept -1.927 (79.48)** -2.106 (80.30)**

Performance measures

Loss 0.330 (11.59)** 0.239 (4.55)*

LossBoard-strength 0.077 (6.41)* 0.074 (4.77)*

Board-strength 0.032 (5.82)* 0.039 (7.51)**

Accounting-return -0.573 (2.21) -0.532 (1.52)

Accounting-return 0.103 (0.06) 0.111 (0.05)

Stock-return -0.389 (30.31)** -0.370 (22.33)**

Control variables

Stock-volatility 1.960 (5.64)* 1.613 (3.20)

Earnings-volatility -0.900 (2.65) -0.924 (2.23)

Forecast-error -2.339 (5.25)*

Concentration 0.123 (0.03) -0.468 (0.40)

Size 0.079 (15.76)** 0.094 (18.29)**

Growth 0.010 (0.14) 0.022 (0.60)

Restructure 0.257 (5.20)* 0.269 (4.71)*

Restatement 0.274 (8.74)** 0.270 (7.69)**

Age 0.739 (152.34)** 0.802 (159.26)**

Fixed effects Industry Yes Yes Year Yes Yes

Observations 8,077 7,233 Nagelkerke R2 6.20% 6.78%

The dependent variable Turnover is an indicator variable that equals 1 when there is a change in the CEO in the two years subsequent to the current fiscal year and 0 otherwise. The independent variables are as follows. Loss is an indicator variable with a value of 1 when net income is negative for the current year and 0 otherwise. To measure Board-strength, six individual measures of board strength are transformed into standardized values that are derived from measuring the distance from the average and dividing by the standard deviation. Board-strength is computed by summing the six transformed measures and is constructed such that a higher value indicates greater board strength. The individual measures are board size, board independence, separate CEO-Chair, separate committees, CEO ownership, and CEO tenure. -1 is multiplied to board size, CEO ownership, and CEO tenure before they are transformed into standardized values. Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.

45

TABLE 10 CEO Turnover and Losses: Two Stage Least Squares Estimation

First-stage Second-stage

Dependent variables = Loss Turnover

Intercept -4.724 (71.87)** -3.437 (131.78)** Performance measures

Pred-loss 0.304 (4.23)*

Accounting-return -0.548 (1.48)

Accounting-return -0.066 (0.02)

Stock-return -0.360 (20.52)**

Instruments Cash-flow -6.754 (69.67)** Accrual -4.667 (30.25)** Sales-growth -0.002 (0.99) lag(Loss) 1.195 (51.48)** Dividend -0.157 (1.18) Dividend-stop -0.241 (0.41)

Governance variables Board-size 0.023 (0.52) 0.030 (3.51)

Board-independence 0.001 (0.06) 0.005 (6.16)*

Separate-chair 0.085 (0.49) 0.739 (124.99)**

Separate-committees -0.117 (0.55) 0.232 (6.17)*

Ownership -0.011 (0.78) -0.018 (6.49)*

Control variables

Stock-volatility 12.428 (65.56)** 1.668 (3.04)

Earnings-volatility 5.279 (32.63)** -0.884 (1.89)

Forecast-error -37.339 (377.79)** -1.848 (1.94)

Concentration -3.362 (3.54) -0.661 (0.74)

Size 0.042 (0.73) 0.086 (10.32)**

Growth -0.327 (29.79)** 0.014 (0.21)

Restructure 3.361 (466.38)** 0.248 (2.29)

Restatement 0.502 (9.18)** 0.318 (10.37)**

Age -0.301 (4.74)* 0.713 (115.11)**

Tenure 0.017 (3.99)* 0.029 (41.99)**

Fixed effects Industry Yes Yes

Year Yes Yes

Observations 6,906 6,906 Nagelkerke R2 64.79% 9.33%

In the first stage regression, the dichotomous dependent variable Loss equals 1 when reported net income is negative in the current year and 0 otherwise. The instruments are: Cash-flow is cash flows from operations divided by total assets at the beginning of the year, Accrual is total accruals (defined as the difference between net income and cash flows from operations) divided by total assets at the beginning of the year, Sales-growth is the percentage growth in sales, lag(Loss) is the one-year lagged value of Loss, Dividend equals 1 when the firm pays dividends and 0 otherwise, and Dividend-stop is an indicator variable equal to 1 if the firm stopped paying dividends and 0 otherwise. All the instrumental variables in the first stage regression are measured at the end of the previous fiscal year. In the second stage, we regress the dichotomous dependent variable Turnover which equals 1 for firms with CEO turnover in the subsequent two years on the estimated value of loss from the first stage regression (Pred-loss). Accounting-return is the industry-adjusted return on assets measured as the difference between the firm-specific and the industry-median income before extraordinary items deflated by total assets at the beginning of the year. Accounting-return is the difference between current period income before extraordinary items and the corresponding number in the prior year deflated by total assets at the beginning of the year. Stock-return is the difference between the raw returns and the value-weighted CRSP market returns over a twelve-month fiscal period. Board-size is the number of directors on the board. Board-independence is the percentage of independent directors on the board. Separate-chair and Separate-committees are indicator variables set to 1 when a firm has a separate CEO and board chair and when a firm has separate audit, nominating, and

46

compensation committees, respectively. Ownership is the percentage of common stock held by the CEO. Stock-volatility (Earnings-volatility) is the standard deviation of Stock-return (Accounting-return) based on prior twenty-four monthly (five year) data. Forecast-error is the difference between reported annual EPS and the mean forecast EPS deflated by stock price at the beginning of the year. Concentration is the industry level Herfindahl index. Size is the logarithmic transformation of the fiscal year-end market value of equity. Growth is the sum of the market value of equity and the book value of debt scaled by the book value of total assets. Restructure is an indicator variable that equals 1 if special items as a percentage of total assets are less than or equal to -5 percent and 0 otherwise. Restatement is an indicator variable equal to 1 when a firm restates its financial statement in the current or prior year and 0 otherwise. Age is an indicator variable equal to 1 when the CEO is over the age of 60 years and 0 otherwise. Tenure is the number of years that the CEO has been in office as of the fiscal year-end. We report the estimated coefficients from a logistic regression and the corresponding 2–statistics in parenthesis. ** and * denote statistical significance at the 1 percent and 5 percent levels, respectively, based on a two-tailed test.