Big baths and CEO overconfidence

Unlock access to all the studying documents.

View Full Document

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rabr20

Accounting and Business Research

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rabr20

Big baths and CEO overconfidence

Jochen Pierk

To cite this article: Jochen Pierk (2020): Big baths and CEO overconfidence, Accounting and

Business Research, DOI: 10.1080/00014788.2020.1783634

To link to this article: https://doi.org/10.1080/00014788.2020.1783634

UK Limited, trading as Taylor & Francis

Group

Published online: 26 Jun 2020.

Submit your article to this journal

Article views: 152

View related articles

View Crossmark data

Big baths and CEO overconﬁdence

†

JOCHEN PIERK*

Erasmus University Rotterdam, Rotterdam, Netherlands

This paper empirically investigates the relationship between managerial overconﬁdence and

write-offs following CEO turnover. Incoming CEOs often engage in big bath accounting as

they dispose of poorly performing projects. Overconﬁdent managers overestimate their

abilities and consequently have upwardly biased expectations concerning future ﬁrm

performance. I hypothesise that overconﬁdent CEOs are less likely to engage in a big bath

following managerial change. The empirical results conﬁrm this hypothesis by showing that

big baths at CEO turnover are signiﬁcantly less frequent among overconﬁdent CEOs.

Keywords: Big bath; earnings management; managerial characteristics; overconﬁdence

1. Introduction

Overconﬁdence is a particular form of a biased managerial view. Overconﬁdent individuals over-

estimate their abilities and therefore have upwardly biased expectations related to their future per-

formance (e.g. Malmendier and Tate 2008, Ahmed and Duellman 2013). Overconﬁdent managers

are more optimistic about their ability to turn around poorly performing projects and are therefore

more likely to overestimate the likelihood and magnitude of projects that go well and underesti-

mate the likelihood and magnitude of those that do not perform well. For example, Ahmed and

Duellman (2013) argue that they may erroneously perceive a project as proﬁtable. In line with

these arguments, Ahmed and Duellman (2013) show that overconﬁdent managers are generally

less conservative as they tend to report losses later compared to non-overconﬁdent managers.

A speciﬁc form of accounting conservatism is large one-time write-offs, commonly known as

taking a big bath, highlighting the magnitude of these write-offs. This paper argues that CEOs

overconﬁdence determines the likelihood of taking a big bath after being hired, providing a poten-

tial channel through which the results of Ahmed and Duellman (2013) can arise. Recognition of

This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License

(http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium,

provided the original work is properly cited, and is not altered, transformed, or built upon in any way.

†

This paper is based on one part of my dissertation at Humboldt University Berlin. A previous version of this

paper was co-authored with Valentin Burg and Tobias Scheinert, to whom I am especially grateful. I

acknowledge comments and suggestions by Maria Correia (the editor) and two anonymous referees. I

further would like to thank Tim R. Adam, Ulf Brüggemann, Joachim Gassen, Igor Goncharov, Urska

Kosi, Beatriz García Osma, Caspar David Peter, Catherine Shakespeare, Jeroen Suijs, and seminar partici-

pants at Humboldt University Berlin, conference participants at the 2013 EAA Annual Meeting in Paris, and

workshop participants at the EAA 29th Doctoral Colloquium for their helpful comments.

*Email: pierk@ese.eur.nl

Accounting and Business Research, 2020

https://doi.org/10.1080/00014788.2020.1783634

losses is particularly relevant at CEO turnovers as empirical evidence indicates that they are more

frequent (Johnson et al. 2011) and more extreme in the turnover year (Strong and Meyer 1987).

When new CEOs step into ofﬁce they often recognise problems ignored by their predecessors

(Elliott and Shaw 1988), and managerial change induces restructuring (Strong and Meyer

1987). Consequently, managers engage in write-offs.

Overconﬁdent CEOs overestimate their ability relative to other managers and are more opti-

mistic about the company’s projects when these projects are managed by them. Thus, if CEOs

report in accordance with their overconﬁdent beliefs, this will result in a lower likelihood of enga-

ging in a big bath especially at CEO turnover. When overconﬁdent managers fail to report these

write-downs in the year of the turnover and postpone them to future periods, they do not ‘clear the

air’but ‘muddy the waters’and degrade the ﬁrm’s information environment (Haggard et al.

2015).

In the empirical analyses, I employ several measures of CEO overconﬁdence used in prior

literature based on the CEOs stock option portfolios (e.g. Malmendier and Tate 2008), on their

investment behaviour or on the magnitude of their capital expenditure in comparison to their

industry peers. Following Elliott and Shaw (1988) and in line with recent literature (e.g.

Haggard et al. 2015), I use the magnitude of write-offs in the form of special items to measure

big baths.

The results support the empirical prediction. I ﬁnd that overconﬁdent CEOs are about 6.3–

10.6 percent less likely to engage in a big bath in the turnover year than non-overconﬁdent

CEOs. Furthermore, I ﬁnd that this difference is only prevalent in the year of the turnover but

generally not in the years before or after the turnover.

An alternative explanation for the ﬁnding could be that there is self-selection of non-overcon-

ﬁdent managers into ﬁrms with higher potential for large write-offs in the turnover year. I address

potential endogeneity and omitted correlated variables concerns in several ways. First, I show that

the observed big bath choices are not driven by whether CEO turnover is forced. Big baths are

especially prevalent in forced turnovers (Pourciau 1993, Wells 2002). Second, I use entropy bal-

ancing so that the ﬁrst and second moments of all covariates in the year of the turnover are the

same between overconﬁdent and non-overconﬁdent managers. This mitigates concerns that

ﬁrm characteristics simultaneously explain the choice to hire a CEO of a certain behavioural

type and determine the predicted big bath pattern. Third, by controlling in big bath regressions

for the behavioural type of the outgoing CEO, I alleviate the concern that non-overconﬁdent

CEOs are selected to clean up bloated asset values left behind by overconﬁdent CEOs. Fourth,

I include ﬁrm ﬁxed effects to control for time-invariant ﬁrm characteristics. The results remain

qualitatively similar in all speciﬁcations.

This paper contributes to the literature by showing that differences in accounting conservatism

in the form of write-offs across overconﬁdent and non-overconﬁdent managers only arise in the

year of CEO turnover and not in subsequent years. This is in contrast to Ahmed and Duellman’s

ﬁndings (2013) that suggest that overconﬁdent managers are generally less conservative in their

accounting policies. Their market and accruals-based measures of conservatism do not explicitly

identify one-time write-offs or any other channels through which accounting conservatism can be

practiced. I add to their paper by showing that big baths at CEO turnovers are a potential channel

for their ﬁndings. Investors should be aware that when overconﬁdent managers fail to report these

write-downs, earnings are overstated and the ﬁnancial reporting does not accurately reﬂect the

underlying economics of the ﬁrm, thereby degrading the ﬁrm’s information environment. Further-

more, my results suggest that investors should especially focus on CEOs (non)overconﬁdence at

CEO turnover; in all other periods, this CEO characteristic is of less importance.

The remainder of this paper is organised as follows. Section 2 develops the empirical hypoth-

esis and section 3 introduces the research methodology. In section 4, I interpret the results and

2J. Pierk

section 5 mitigates endogeneity concerns. Section 6 presents robustness tests and section 7

concludes.

2. Literature review and hypothesis development

Recent literature has documented several links between overconﬁdence and more aggressive

accounting policies. For example, Schrand and Zechman (2012) report that overconﬁdence is

related to ﬁnancial misreporting and fraud, whereas Ahmed and Duellman (2013)ﬁnd a negative

relation between managerial overconﬁdence and accounting conservatism. Hribar and Yang

(2016), Libby and Rennekamp (2012) and Hilary and Hsu (2011) indicate that overconﬁdent

managers are more likely to engage in more speciﬁc and optimistic management forecasts, and

Davis et al. (2014) show that they use more positive language in conference calls. Hsieh et al.

(2014)ﬁnd that they are more likely to engage in income-increasing earnings management

even after the introduction of the Sarbanes-Oxley Act. Wong and Zhang (2014) show that

CEO optimism is a source of analyst forecast bias.

These results are not surprising as overconﬁdent managers systematically overestimate their

abilities and consequently the future cash ﬂows they are able to generate. In line with the literature

on CEO overconﬁdence and accounting outcomes (e.g. Ahmed and Duellman 2013), I argue that

overconﬁdent CEOs overestimate their ability relative to other managers. Consequently, they are

more optimistic about the company’s projects and they are more likely to overestimate the like-

lihood and magnitude of projects that go well and underestimate the likelihood and magnitude of

projects that do not go well. This overestimation has important implications for managers’

accounting decisions as they will tend to delay loss recognition (Ahmed and Duellman 2013).

When CEOs report in accordance with their overconﬁdent beliefs, this will result in a lower like-

lihood of engaging in a big bath. When new CEOs take the reins, they recognise problems ignored

by their predecessors (Elliott and Shaw 1988), and managerial change induces restructuring

(Strong and Meyer 1987). I expect that overconﬁdent managers are less likely to show this

pattern at CEO turnover.

HYPOTHESIS:Incoming overconﬁdent CEOs are less likely to engage in big baths compared

to incoming non-overconﬁdent CEOs in the turnover year.

3. Research methodology

3.1. Measurement of overconﬁdence

Following Malmendier and Tate (2008), I construct the overconﬁdence measures based on execu–

tive option holdings. CEOs are classiﬁed as overconﬁdent if they hold an option until maturity

which is at least 40 percent

in-the-money at the year-end prior to maturity (OC40). Thus, over-

conﬁdence is considered as an inherent, time-invariant personal characteristic of CEOs. The

rationale for relying on the option exercise behaviour to classify overconﬁdent and non-overcon-

ﬁdent managers is the following: CEOs face a trade-off between exercising their options and

retaining them for later use. By retaining their options, they maintain the right to purchase

company stock at potentially more favourable conditions in the future. The downside of this strat-

egy is that it involves substantial costs for the CEO in terms of exposure to idiosyncratic risk.

Executive stock options typically have a maturity of ten years and become vested after two to

four years. Furthermore, CEOs are legally prohibited from short-selling their company’s stock

in the US. Given the large proportion of personal wealth tied to their company, diversiﬁcation

abilities across alternative investments are also limited. Besides ﬁnancial exposure, substantial

human capital is also tied to the company (Malmendier and Tate 2008). Consequently, CEOs

Accounting and Business Research 3

can be considered as under-diversiﬁed investors who have large exposure to their company’s risk.

Thus, a non-overconﬁdent CEO should divest as soon as the option is sufﬁciently in-the-money

because the cost of delayed exercise typically exceeds its option value. In contrast, an overconﬁ–

dent CEO will not likely exercise stock options in this situation.

Considering CEO overconﬁdence as a ﬁxed effect allows me to exploit within-ﬁrm variation

in a CEO turnover setting. However, I acknowledge that there might also be a time-varying part of

an individual’s overconﬁdence that impacts his/her decisions. Therefore, I run the tests with

additional measures also to ensure that the respective managers’level of overconﬁdence is

measured at the time of CEO change.

Similar to Schrand and Zechman (2012), I use a measure of overconﬁdence based on the

dollar value of exercisable options. Managers are identiﬁed as overconﬁdent if the dollar value

of their exercisable options (measured as the difference between the current stock price and the

average exercise price of the options times the number of options held) exceeds the industry

median based on three-digit SIC codes in the respective year (SZ).

To further alleviate potential concerns associated with using option-based measures, I also

employ two investment-based measures of overconﬁdence. For example, Malmendier and Tate

(2005) show that overconﬁdence relates to ﬁrms’investment decisions. In line with Ahmed

and Duellman (2013), OVERINV is coded one if the residual of industry-year regressions of

total asset growth on sales growth is positive. Similarly, I classify CEOs as overconﬁdent

when capital expenditures deﬂated by lagged total assets exceed the industry-year median

(CAPEX).

3.2. Measurement of big baths

Following Elliott and Shaw (1988), I classify all ﬁrm-years with special items (SPI, Compustat

item #17) less than minus one percent of total assets as big bath years. Special items include

any non-recurring items, impairment of goodwill, non-recurring inventory write-downs, bad

debt expenses, restructuring costs, and provisions for doubtful accounts.

Special items could

occur predominantly in a speciﬁc year due to economic downturns or other exogenous shocks

(e.g. natural disasters). This should not have an impact in this setting as CEO changes are distrib-

uted over a period of 19 years for both groups (overconﬁdent vs. non-overconﬁdent) and both

groups are approximately equally distributed over time.

3.3. Control variables

Besides the main variable of interest, I control for the following variables which could inﬂuence

big bath behaviour.

Routine vs. Non-Routine CEO Turnover: Pourciau (1993) and Wells (2002) show that big

baths are especially pronounced after non-routine turnovers when negative outcomes can

easily be attributed to the CEO who has left the ﬁrm in discord. I hand-collect data on routine

and non-routine turnovers following Hazarika et al. (2012). A CEO turnover is classiﬁed as

non-routine (NONROUTINE)‘if (i) the CEO was ﬁred, forced out from the position, or departed

due to policy differences; or (ii) the departing CEO’s age is less than 60, and the announcement

does not report that the CEO died, left because of poor health, or accepted another position else-

where or within the ﬁrm; or (iii) the CEO ‘retires’but leaves the job within six months of the

‘retirement announcement’.’

Former CEO Type: If former CEOs were overconﬁdent, they might have pursued less conser-

vative accounting and depreciated less than necessary. Consequently, assets might be overvalued

4J. Pierk

and big baths would be a justiﬁed correction of inﬂated asset values. Therefore, I control for the

former CEO type.

Past Special Items: Elliott and Hanna (1996) show that ﬁrms repeatedly write-down assets,

indicating that past asset write-downs are correlated with future ones. Therefore, I include a

control variable (PRIORSPI) for the average special items of the last three years.

Conservatism: Ahmed and Duellman (2013) show that overconﬁdent managers use less con-

servative accounting which could lead to fewer big baths. I control for the level of accounting

conservatism prior to CEO turnover by including the average accruals over the three previous

years (PRIORACC). The measure is multiplied by minus one so that larger values denote

greater accounting conservatism.

Firm Performance: Prior research suggests that weak ﬁrm performance is related to more

aggressive earnings management. If current ﬁrm performance is poor, earnings are shifted

from the future to the current period (e.g. DeFond and Park 1997, Keating and Zimmerman

1999). Furthermore, performance could be mechanically linked to the magnitude of special

items since poor performance might trigger extraordinary write-offs. To control for ﬁrm perform-

ance, I include return on assets (ROA), which is income before extraordinary items and before

special items divided by total assets at the beginning of the year.

Firm Size: The size of the ﬁrm could also affect the earnings management behaviour of man-

agers. Skinner (1993), for example, shows that ﬁrm size increases the likelihood of income-

decreasing depreciation procedures. Further, big baths might be related to ﬁrm size as more

visible ﬁrms behave differently with respect to earnings manipulation. SIZE is measured as the

natural logarithm of total assets in millions of dollars.

Debt: The leverage ratio of a ﬁrm is related to debt covenant violations. Various papers show

that earnings are manipulated before and after debt covenant violations (e.g. Press and Weintrop

1990, DeFond and Jiambalvo 1994, Sweeney 1994). Covenant violations are most often triggered

by exceeding preset debt levels. Thus, I control for leverage in all regressions and deﬁne LEV as

total debt divided by total assets at the beginning of the year.

Corporate Governance: Weak internal control systems are often correlated with poor earnings

quality (Doyle et al. 2007). I include the Entrenchment Index (GOVINDEX) proposed by

Bebchuk et al. (2009) to account for the impact of corporate governance mechanisms on earnings

management and the Gompers et al. (2003) governance index (GINDEX) as a robustness check

(not tabulated).

Managerial Compensation: Earnings-based compensation of CEOs provides several incen-

tives to manipulate earnings. Holthausen et al. (1995) show that managers engage in income-

decreasing earnings management when bonus schemes are at their maximum. Bergstresser and

Philippon (2006)point out that earnings manipulation is especially prevalent if compensation

is closely tied to ﬁrm value. I collect information on CEO compensation (bonus and salary)

from ExecuComp. BONUS is deﬁned as the annual bonus payment divided by the sum of

bonus and salary.

Growth Opportunities: Missing earnings benchmarks such as analyst forecasts can be particu-

larly severe for high-growth ﬁrms (Skinner and Sloan 2002), which may incentivise them to

manipulate earnings. To control for growth opportunities, I include the market-to-book ratio

(MTB) and future growth in the regressions. MTB is equal to the market value of a company’s

assets (ﬁscal year closing price times common shares outstanding plus preferred stock plus

total liabilities divided by the book value of a company’s assets). GROWTH is an ex-post

measure of growth opportunities and is deﬁned as the relative increase in market value in the

next three years.

Accounting and Business Research 5

3.4. Big bath model

I use a logit model to test the paper’s hypothesis, using BIGBATH as the dependent variable.

BIGBATH is a dummy variable that is equal to one if special items are less than minus one

percent of total assets. OC is the respective overconﬁdent measure and is equal to one if the

CEO is classiﬁed as overconﬁdent and zero if classiﬁed as non-overconﬁdent. OC40 is the over-

conﬁdence classiﬁcation based on 40 percent in-the-money at the year-end prior to maturity. SZ

classiﬁes managers as overconﬁdent if the moneyness of their exercisable options exceeds the

industry median based on three-digit SIC codes. OVERINV is coded one if the residual of an indus-

try-year regression of total asset growth on sales growth is positive, and zero otherwise. CAPEX is

coded one if the capital expenditures exceed the industry-year median, and zero otherwise.

I include ten years surrounding the CEO turnover to investigate whether the different big bath

behaviour between overconﬁdent and non-overconﬁdent CEOs occurs only in the turnover year

or also before or in the subsequent years. I use indicators for each year that the new CEO is in

ofﬁce. Thus, I include six indicator variables, for the year of the turnover and the ﬁve years

after the turnover (e.g. YEAR0 is the year of the turnover and YEAR1 is the year thereafter).

Next, I interact each year with the respective overconﬁdence measure (e.g. YEAR0*OC,

YEAR1*OC, etc.). I expect a negative coefﬁcient of the interaction of the turnover year with

the overconﬁdence indicator (YEAR0*OC).

logit(p) =ln(p /(1 – p)) =

0+

j=1

j∗YEARit +

j=1

j∗YEARit∗OC +controlsit +FEit +1it

where pis the probability of engaging in big bath accounting.

6J. Pierk