Sie sind auf Seite 1von 41

Overconfidence, CEO Awards, and Corporate Tax Aggressiveness

Thomas R. Kubick
Accounting and Information Systems
University of Kansas
Lawrence, KS 66045-7601
(785) 864-7318 (office)
tkubick@ku.edu

G. Brandon Lockhart
Department of Finance
Clemson University
Clemson, SC 29634
(864) 656-1135 (office)
blockha@clemson.edu

January 16, 2017

Forthcoming in Journal of Business Finance and Accounting

Abstract
Theory and prior research suggests that overconfidence leads managers to overestimate their
own ability to generate returns, leading to riskier corporate policies. We use a novel dataset
of media awards as an exogenous shock to overconfidence to test whether award-winning
CEOs adopt more aggressive corporate tax policies. Using propensity score matching and a
difference-in-difference design, we find strong evidence that firms with an award-winning
CEO exhibit significantly greater tax aggressiveness following the award. Overall, our results
suggest that CEO overconfidence has a meaningful impact on corporate tax policy.

JEL Codes: H2, H25, M4, M40.

Keywords: CEOs; Superstar CEOs; business awards; firm risk; tax avoidance; tax
aggressiveness; CEO overconfidence; CEO hubris; CEO optimism

* Corresponding author.

We gratefully acknowledge helpful comments and suggestions from Qi Chen (editor), an anonymous reviewer,
Tom Omer, John Robinson, Terry Shevlin, Casey Schwab (discussant), and participants at the 2013 AAA
annual meeting.

This article has been accepted for publication and undergone full peer review but has not been
through the copyediting, typesetting, pagination and proofreading process, which may lead to
differences between this version and the Version of Record. Please cite this article as doi:
10.1111/jbfa.12237.

This article is protected by copyright. All rights reserved.


1. INTRODUCTION

The CEO is often considered to have the most economically important role in shaping

corporate strategy and policy and, as a result, there is substantial interest among investors,

researchers, regulators, policymakers, and the media, in understanding the importance of

CEO characteristics and abilities for corporate outcomes. In the context of corporate taxes,

Dyreng et al. (2010) assert that CEOs can influence tax policy through their “tone at the top,”

by changing functional areas of the firm, by directing resource allocations, and by setting the

compensation of the tax director. However, the literature has yet to make substantial

empirical progress linking CEO characteristics to corporate tax outcomes. Our focus is on

overconfidence, an important executive characteristic, as overconfident CEOs tend to

overestimate their ability to generate favorable outcomes from their decisions. In this paper,

we utilize a shock to the status of the CEO to establish a robust link between CEO

overconfidence and the firm‟s tax policy.

Corporate taxes provide an interesting setting in which to examine the effect of

overconfidence on managerial behavior, as taxes represent a significant cost to the firm and

tax aggressive policies can increase internal cash flows, easing investment funding

constraints (Edwards, Schwab, and Shevlin, 2016). This is particularly important, as

overconfident CEOs often believe their firms are undervalued and, as a result, view external

finance as costly (Malmendier and Tate, 2005; 2008). Further, corporate taxes affect nearly

every transaction within the firm, making taxes an important strategic policy issue for firms

and CEOs.1

1
In the context of investment, the strategic importance of taxes is underscored in Foley, Hartzell, Titman, and
Twite (2007) and Hanlon, Lester, and Verdi (2015) who argue that repatriation tax burdens (on earnings
repatriated from low-tax foreign jurisdictions) can cause investment distortions through higher cash holdings
and less valuable foreign acquisitions, respectively. Similarly, Faulkender and Petersen (2012) find an increase
in domestic investment of multinational firms who repatriated foreign earnings pursuant to the tax holiday in the
American Jobs Creation Act of 2002.

This article is protected by copyright. All rights reserved.


2
Malmendier and Tate (2009), hereafter “MT,” find evidence consistent with

behavioral/policy changes for award-winning CEOs and the firms they manage, following the

CEO‟s receipt of prestigious recognition from a major business media outlet. MT suggest that

CEO awards can influence firm outcomes by cultivating hubris.2 Accordingly, we utilize

CEO awards conferred by major media outlets as a shock to CEO overconfidence and

investigate the extent to which the receipt of these prestigious media awards impacts future

corporate taxes, which can be significantly influenced by the structure and nature of certain

transactions, and can be managed to minimize the reported cost to the firm.3

We argue that overconfidence can affect tax aggressiveness in at least two ways. First,

overconfident CEOs underestimate the risks of their actions by setting narrow confidence

bounds (e.g., Larwood and Whittaker, 1977; Kidd, 1970; Malmendier and Tate, 2005; Moore,

1977), and overestimating the likelihood of a favorable outcome. Overconfidence can lead

CEOs to overestimate the returns to tax planning by underestimating the uncertainty of

realizing the benefits from tax planning. For example, an overconfident CEO may

underestimate the risk that the firm will be audited or that an uncertain tax position will be

successfully challenged by tax authorities conditional on audit. Second, overconfident CEOs

believe they have greater control over the outcome of their decisions (e.g., March and

Shapira, 1987; Langer, 1975; Malmendier and Tate, 2005). Hence, overconfident CEOs may

believe that they have a greater ability to select and motivate the best tax planners, or that

their firm is more able to identify and execute valuable tax planning opportunities.

An important empirical complication to this stream of research is that overconfidence

is an unobservable characteristic. Hence, empirical measures of overconfidence rely on

2
There are several labels in the literature for this construct (e.g., hubris, overconfidence, over-optimism). The
distinctions between these labels are not our focus. Mostly for expositional purposes, we consider hubris to be
synonymous with overconfidence or over-optimism.
3
See Graham and Tucker (2006), Wilson (2009), and Hanlon and Heitzman (2010) for specific examples of
these transactions.

This article is protected by copyright. All rights reserved.


3
observable manifestations, often using firm and compensation-based characteristics. While

researchers can control for observable characteristics that are likely correlated with both

overconfidence and corporate outcomes, there remains potentially significant unobserved

heterogeneity that could account for the observed empirical patterns. Thus, in the absence of

an exogenous shock to overconfidence, it is difficult to empirically disentangle the

importance of overconfidence from other executive and firm characteristics. Although our set

of award-winning CEOs is relatively small in number, the rarity of winning such a

prestigious media award is a source of the strength of the shock to overconfidence, which is

useful for analyzing a link to an important corporate policy such as corporate taxes. Using

this hand-collected dataset of CEO awards as a shock to overconfidence, our results make an

important contribution to the literature by showing that a shock to CEO status, generated

from outside the firm, can have a significant impact on corporate tax policy.

In our first set of results, we replicate and expand the dataset used by MT and

estimate the likelihood of winning an award. Then, at the time of the award, we match each

firm led by an award-winning CEO to a firm that is not led by an award-winning CEO but

has the highest propensity to be selected for an award. Matching on propensity scores

mitigates concerns that our treatment effect (winning an award) is confounded by selection

effects. Our prediction model closely mirrors the model used in MT and displays substantial

discriminant ability. After confirming that our matched pairs have good covariate balance at

the time of the match, we estimate difference-in-difference regressions with firm fixed effects

and observe a significant increase in subsequent tax aggressiveness among firms led by an

award-winning CEO. Specifically, in one specification, we find that firms led by award-

winning CEOs exhibit 1.1% greater discretionary book-tax differences, a common measure

This article is protected by copyright. All rights reserved.


4
for tax aggressiveness, in years after the CEO wins an award, an effect representing an

economically significant change in corporate tax policy.4

Throughout our analysis, we carefully consider possible alternative explanations.

First, we employ propensity score matching to construct a control sample of observationally

similar firms without an award-winning CEO, and we confirm that award-winning CEOs lead

firms exhibiting a clear shift toward a more aggressive tax policy. In all regressions, we

control for the sensitivity of the CEO‟s and other executives‟ option portfolios to changes in

stock price (delta) and stock volatility (vega), which are commonly used measures of equity-

based incentives.5 We also use firm fixed effects to control for latent, time-invariant, firm

characteristics (e.g., tax aggressiveness technology, firm culture).

In additional tests, we consider the potential confounding effects of managerial ability

and CEO board connections and confirm our results hold after controlling for these factors. In

addition, we find that firms led by award-winning CEOs have a higher likelihood of reporting

an increase in foreign earnings, and a corresponding decrease in domestic earnings, after the

receipt of a prestigious media award. These results indicate a change in transfer pricing,

which is an important tax policy. We also perform validation tests to confirm that CEOs are

more likely to be classified as overconfident after winning an award. Finally, we confirm that

our results are not moderated, or driven by, an increase in financial reporting aggressiveness.

Our paper is related to at least two important areas of research. First, our study

contributes to the growing and influential stream of research investigating causes and

4
In a concurrent study, Chyz, Gaertner, Kausar, and Watson (2014) report some evidence of a positive
association between measures of CEO overconfidence and tax avoidance. However, their results are based on
data limited to fiscal years 2007-2012, and their empirical design lacks a source of exogenous variation in CEO
overconfidence. We believe our study overcomes these limitations through our use of a longer time series and
our use of CEO awards as an arguably exogenous shock to overconfidence.
5
Delta captures the sensitivity of the CEO‟s option portfolio to a 1% change in the underlying stock price. Vega
captures the sensitivity of the CEO‟s option portfolio to a 0.01 unit change in the underlying stock volatility.
Prior research suggests that CEO options portfolios with greater Vega encourage CEO risk-taking (Coles,
Daniel, and Naveen, 2006; Brockman, Martin, and Unlu, 2010).

This article is protected by copyright. All rights reserved.


5
consequences of tax aggressiveness. There is considerable research examining the cross-

sectional determinants of tax aggressiveness (e.g., Mills, Erickson, and Maydew,1998; Rego,

2003; Desai and Dharmapala, 2006; Frank et al., 2009; Robinson, Sikes, and Weaver, 2010;

Chen, Chen, Cheng, and Shevlin, 2010; Cheng, Huang, Li, and Stanfield, 2012; Dyreng,

Mayew, and Williams, 2012), and the consequences of tax aggressiveness (e.g., Lev and

Nissim, 2004; Hanlon, 2005; Hanlon and Slemrod, 2009; Kim, Li, and Zhang, 2011). Within

this literature, our results contribute to the emerging line of research examining the influence

of individual executives on corporate tax policy (e.g., Dyreng et al., 2010) by showing that a

shock to CEO status can embolden CEOs to adopt riskier tax policies. In doing so, we also

address a specific request for additional research examining the “manager effect” on

corporate tax policy (Hanlon and Heitzman, 2010).

Our paper also contributes to the stream of research examining the impact of CEO

confidence on corporate outcomes (e.g., Roll, 1986; Malmendier and Tate, 2005, 2008;

Malmendier, Tate, and Yan, 2012; Hirshleifer et al., 2012; Ahmed and Duellman, 2013;

Hsieh, Bedard, and Johnstone, 2014). MT suggest that prestigious business awards, reflecting

a shift in status, may cultivate CEO hubris. We provide evidence supporting this claim, by

demonstrating that CEO awards are a significant predictor of CEO overconfidence.

The rest of this paper is organized as follows. Section 2 discusses the background

literature and motivates our hypothesis, Section 3 describes our empirical methodology,

Section 4 presents our main results, Section 5 discusses additional tests, and Section 6

provides a brief conclusion.

2. BACKGROUND LITERATURE AND HYPOTHESIS DEVELOPMENT

This article is protected by copyright. All rights reserved.


6
In this section, we review and connect the literatures on corporate tax aggressiveness

and CEO characteristics and influence, motivate our use of CEO awards as an exogenous

shock to CEO confidence, and develop our hypothesis.

(i) CEOs and Firm Policies

Prior research suggests that executives have a discernible impact on corporate

policies. For example, Bertrand and Schoar (2003) use a dataset of manager-firm pairs – with

a portion of managers working for multiple firms over time – to estimate both manager fixed

effects and firm fixed effects in regressions of a number of corporate variables related to

investment, financing, strategy, and performance. They find that executive fixed effects are

empirically important in many of their regressions, suggesting that executives have a certain

“style” that manifests in the financial performance of their firms, and that this style follows

executives to future employers. Bamber, Jiang, and Wang (2010) follow the Bertrand and

Schoar (2003) empirical strategy and find evidence of “executive effects” that can explain

differences in disclosure policies (e.g., forecast precision, forecast news, forecast bias, and

forecast errors) across firms. Similarly, Dyreng et al. (2010) examine the movement of 908

executives (CEOs, CFOs, and others) across firms and find strong evidence that book and

cash effective tax rates are influenced by the movement of these executives, suggesting some

executives have a stronger preference than others for lower effective tax rates (i.e., greater tax

avoidance), and that this “manager effect” is generally stronger for the CEO.

The economic importance of the CEO‟s influence is underscored in Dyreng et al.

(2010), who specifically assert that CEOs can influence tax policy through their “tone at the

top” by changing the functional areas of the firm, resource allocations, or setting the

compensation of the tax director.6

6
Armstrong, Blouin, and Larcker (2012) find a negative relationship between the GAAP effective tax rate and
the incentive compensation of the tax director. Further, the structure of the CEO‟s compensation can moderate

This article is protected by copyright. All rights reserved.


7
(ii) CEO awards and Overconfidence

While CEOs are usually the central figures of their firms, not all CEOs have the same

visibility. Media coverage has cultivated a celebrity-type awareness of many CEOs, and the

ex-post consequences of this elevated status are largely unknown. MT find that the operating

performance and market valuation of firms with award-winning CEOs declines following the

receipt of a prestigious media award. In addition to the relative underperformance, the

increase in CEO compensation following the award is attendant an increase in earnings

management and the number of distractions (e.g., increased number of books authored,

number of board seats held, lower golf handicaps), especially among firms with weaker

corporate governance characteristics.

MT suggest that CEO awards can influence firm outcomes by cultivating CEO hubris,

a hypothesis they do not test but one that has potentially important implications, as CEOs

with comparatively high confidence overestimate their ability to generate returns

(Malmendier and Tate, 2005, 2008). Consequently, overconfidence can lead to suboptimal

investment by assuming greater risk than shareholders would prefer. In fact, extant research

shows that CEOs with comparatively high levels of confidence engage in more value-

destroying mergers (Malmendier and Tate, 2008), pursue activities that increase firm risk

(Hirshleifer et al., 2012), and adopt less conservative financial reporting policies (Schrand

and Zechman, 2011; Ahmed and Duellman, 2013; Hsieh, Bedard, and Johnstone, 2014). CEO

awards can impact corporate policies similarly by increasing award-winning CEOs‟

confidence in their own ability, emboldening them to pursue activities and adopt policies that

increase firm risk. In other words, CEO awards can arguably provide an exogenous shock to

the CEO‟s influence on corporate policies. For example, compensation packages can be constructed to induce
greater (Coles, Daniel, and Naveen, 2006) or less investment risk (Cassell, Huang, Sanchez, and Stuart, 2012),
and Brockman, Martin, and Unlu (2010) find that lenders adjust financing contracts (e.g., the amount and
maturity of the debt) to manage the credit risk attendant this CEO compensation-induced investment risk.

This article is protected by copyright. All rights reserved.


8
the CEO‟s confidence in their own ability to generate returns, and therefore influence

decision-making.

(iii) Corporate Tax Aggressiveness

Addressing the voids in the literature described in Shackelford and Shevlin (2001),

Shevlin (2007), and Hanlon and Heitzman (2010), researchers have investigated the cross-

sectional determinants of tax aggressiveness (e.g., Gupta and Newberry 1997; Mills, Erickson,

and Maydew 1998; Shackelford and Shevlin 2001; Rego 2003; Chen, Chen, Cheng, and Shevlin

2010; McGuire, Omer, and Wang 2012; Cheng, Huang, Li, and Stanfield 2012; Graham et al.

2014), the consequences of tax aggressiveness (e.g., Lev and Nissim, 2004; Hanlon, 2005;

Desai and Dharmapala 2009; Hanlon and Slemrod 2009; Kim, Li, and Zhang 2011) and,

recently, the impact of executives on corporate tax aggressiveness (e.g., Dyreng, Hanlon, and

Maydew 2010; Law and Mills 2013). Hanlon and Heitzman (2010) suggest that an increase in

the understanding of a potential tax aggressiveness “manager-effect” is an important question

for future research. Indeed, prior research contends that tax aggressiveness is a risky activity

(e.g., Desai and Dharmapala, 2006; Chen et al., 2010; Kim et al., 2011; Hasan et al., 2014),

and that executives can be incentivized to engage in greater tax aggressiveness (Rego and

Wilson, 2012). Hence, an open question is whether CEO characteristics, such as

overconfidence, influence tax aggressiveness.

Prior research considers tax aggressiveness to be an inherently risky activity (e.g.,

Desai and Dharmapala, 2006; Chen et al., 2010; Kim et al., 2011; Rego and Wilson, 2012;

Hasan et al., 2014). Hasan et al. (2014) argues that tax aggressiveness can generate three

sources of risk: information risk, agency risk, and IRS audit risk. First, tax aggressiveness

creates information risk by reducing information quality through complex tax disclosures

(Balakrishnan, Blouin, and Guay, 2012; Hope, Ma, and Thomas, 2013). Second, tax

aggressiveness creates agency risk by allowing managers the potential opportunity to use the

This article is protected by copyright. All rights reserved.


9
complexity and opacity of tax structures to mask rent diversion (Desai and Dharmapala,

2006) or delay bad news recognition (Kim et al., 2011). Finally, tax aggressiveness creates

IRS audit risk by raising the likelihood of an audit and the assessment of deficiencies and

penalties conditional on audit. Indeed, Mills (1998) and Mills and Sansing (2000) find that

tax aggressive firms are more likely to be audited by the IRS, and Wilson (2009) documents

substantial penalties associated with participation in tax shelters deemed to be illegal by the

IRS. These risks, particularly IRS audit risk, are pertinent to an overconfident CEO.

(iv) Hypothesis

Overconfidence can affect tax aggressiveness in several ways. First, overconfident

CEOs underestimate the risks of their actions by setting narrow confidence bounds (e.g.,

Larwood and Whittaker, 1977; Kidd, 1970; Malmendier and Tate, 2005; Moore, 1977), and

overestimating the likelihood of a favorable outcome. Overconfidence can lead CEOs to

overestimate the returns to tax planning by underestimating the uncertainty of realizing the

benefits from tax planning. For example, an overconfident CEO may underestimate the risk

that the firm will be audited or that an uncertain tax position will be successfully challenged

by tax authorities conditional on audit. Second, overconfident CEOs believe they have greater

control over the outcome of their decisions (e.g., March and Shapira, 1987; Langer, 1975;

Malmendier and Tate, 2005). Overconfident CEOs may believe that they have a greater

ability to select and motivate the best tax planners or that their firm can better identify

valuable tax planning opportunities. Thus, overconfidence can lead to suboptimal investment

in tax aggressiveness. This leads us to our hypothesis stated in alternative form.

H: Prestigious CEO awards are associated with greater subsequent tax

aggressiveness.

3. EMPIRICAL METHODOLOGY

This article is protected by copyright. All rights reserved.


10
In this section, we describe the procedures used to hand-collect our dataset of CEO

awards, our empirical measures of tax aggressiveness, and our methodological design.

(i) CEO Awards

Following MT, we hand-collect data on award-winning CEOs from major media

outlets such as Time, Forbes, BusinessWeek, Financial World, Morningstar.com, Industry

Week, Chief Executive, and Ernst & Young.7 Although our dataset is similar to that used by

Malmendier and Tate (2009), our dataset extends through calendar year 2010 and includes

CEO awards conferred by Barron's (circulation 300,000).8 Anecdotally, we observe variation

in award winners across publications, potentially reflecting competition among media outlets

or timing differences in the award-issue publication dates. Our review of the published

criteria, as well as the model specified in MT, suggests that the primary driver of CEO

awards is past performance. For example, in a list of “World‟s Best CEOs”, Barron’s

emphasizes that “shares of nearly every company on our list have outpaced S&P 500 firms

during their CEO tenure” (Bary, 2014).

(ii) Tax Aggressiveness

Although several empirical measures of tax avoidance (or aggressiveness) have been

used in the literature, Hanlon and Heitzman (2010) suggest researchers carefully choose the

measures most appropriate based on theory and the empirical setting. Accordingly, as our

hypothesis relates to risk-taking, we choose two conventional measures of tax avoidance that

prior research considers to reflect decisions and activities that lie on the aggressive-end of the

spectrum of tax avoidance (McGuire, Omer, and Wang, 2012; Armstrong, Blouin, and

Larcker, 2013), and these measures have been empirically validated using samples of actual

7
Our sources of awards data differ from Malmendier and Tate (2009) in that we do not include awards
conferred by Electronic Business Magazine because no information could be obtained for this source.
8
The dataset used in Malmendier and Tate (2009) ends with 2002.

This article is protected by copyright. All rights reserved.


11
tax shelter firms (Frank et al., 2009; Wilson, 2009). Hence, we use the term tax

aggressiveness to refer to our construct of interest.

Our first measure of tax aggressiveness is the level of discretionary tax planning

(DTAX) following Frank et al. (2009). Specifically, DTAX is the residual resulting from a

regression of permanent book-tax differences on state income tax expense, unconsolidated

earnings, non-controlling interest in earnings, intangibles, existence of a net operating loss

and lagged permanent book-tax differences.9 Frank et al. (2009) validate this measure using a

sample of actual tax sheltering cases, by showing that DTAX outperforms alternative

measures, such as effective tax rates, in explaining tax sheltering activity.

Our second measure of tax aggressiveness is the tax sheltering prediction score from

Wilson (2009). Specifically, we define SHELTER as -4.86 + 5.20*BTD + 4.08*|ACC| -

1.41*LEV + 0.76*SIZE + 3.51*ROA + 1.72*FI + 2.42*R&D. We follow prior research (e.g.,

Kim, Li, and Zhang, 2011; Armstrong, Blouin, and Larcker, 2013) and specifically Wilson

(2009, p. 988) in specifying the variables used to construct SHELTER. An advantage of using

this measure is that it is derived from a sample of actual tax sheltering firms.10 Hence, for

these reasons, both DTAX and SHELTER have been regarded as having greater construct

validity than other measures of tax aggressiveness. Higher values of DTAX and SHELTER

reflect greater tax aggressiveness.

(iii) Sample

Our main sample is constructed from the intersection of the Compustat and

Execucomp databases, spanning fiscal years 1994-2011. We construct our dataset beginning

in fiscal year 1994 for several reasons. First, we want our sample selection to be consistent

9
Following Frank et al. (2009), we estimate DTAX for each industry-year, requiring industries (two-digit SIC)
to have at least 15 observations each year.
10
Our results are robust to using the empirical estimates from the other logit models in Wilson (2009).
Specifically, we use the following alternative specifications: (1) SHELTER = -4.30 + 6.63*BTD - 1.72*LEV +
0.66*SIZE + 2.26*ROA + 1.62*FI + 1.56*R&D; (2) SHELTER = -4.29 + 8.49*BTD – 0.76*LEV + 0.51*SIZE +
4.59*ROA + 1.28*FI + 5.24*R&D.

This article is protected by copyright. All rights reserved.


12
with recent empirical work and extant research contends the book-tax gap began sometime in

the early 1990s (Frank et al., 2009). Second, income tax accounting under U.S. GAAP

changed with the enactment of SFAS No. 109 in 1992. Third, because we lag our

compensation-related variables in our multivariate models and because the Execucomp

database did not become representative until 1993, we are limited to examining fiscal years

after 1993.

We exclude financial firms and utilities from our sample due to important regulatory

differences for these industries. For similar reasons, we restrict our sample to domestically-

incorporated U.S. firms. However, we control for foreign income in all of our multivariate

specifications, as some sample firms have foreign affiliates. We also omit firms that have

reported negative pretax income after special items (Compustat PIi,t - Compustat SPIi,t),

negative total tax expense (Compustat TXTi,t) or negative cash taxes paid (Compustat

TXPDi,t) as these firms are in an inherently different tax planning position relative to other,

more profitable firms.

Table 1 reports the distributions of award-winning CEOs (Award = 1) and non-award-

winning CEOs (Award = 0) for S&P 1500 firms during our sample period. The column

reflecting non-award-winning CEOs (Award = 0) captures all firm-years with non-missing

data for our primary variables of interest. Our propensity score matching design, discussed

later, limits this subsample to the closest match for each award observation. Regardless, this

table provides a sense for the distribution of award-winning CEOs across time (Panel A) and

industry (Panel B). Panel A indicates that at least three percent of firms each year are led by

recent award-winning CEOs, with a larger number appearing in 1995 and 1996.11 Panel B

reports some clustering within the food, tobacco, textiles, paper and chemicals industries

11
MT acknowledge a similar pattern, primarily due to one media outlet conferring a large number of awards
during these years. In untabulated tests, we find our results are robust to excluding these award years.

This article is protected by copyright. All rights reserved.


13
(one-digit SIC=2) as well as the manufacturing, machinery and electronics industries (one-

digit SIC=3). However, we observe that award winners and non-award winners exhibit

broadly similar distributions across industries. We control for firm and fiscal year fixed

effects in our multivariate regressions.

(iv) Multivariate Setting

Our empirical design consists of two parts. First, we estimate the likelihood of

winning an award and use propensity score matching to match firms led by award-winning

CEOs to firms led by non-award-winning CEOs with the closest probability of winning an

award. Second, we use the receipt of an award to identify a shock to CEO overconfidence,

and examine tax aggressiveness following the award using our propensity score matched

sample. Our tests are described in more detail below.

Determinants of CEO awards

One potential concern with our setting is that a pooled design in which we broadly

examine whether CEO awards impact tax aggressiveness might not adequately isolate the

effects of CEO awards from selection effects. MT acknowledge this concern, and construct a

control sample based on observable characteristics the media likely use when making the

award decisions. We follow MT and estimate their logit regression using the same

specification that is reported in their study.12 We then match firms led by award-winning

CEOs to firms that are led by non-award winning CEOs that have the closest predicted

probability of winning an award. This construction helps dispel concerns that any association

we observe between CEO awards and subsequent tax aggressiveness is not primarily driven

12
To ease comparability with MT, we follow their variable construction and use their terminology. Market
capitalization equals the stock price multiplied by the number of shares outstanding two months prior to the
award month. The book-to-market ratio equals the book value of equity divided by market capitalization at the
end of the fiscal year prior to (but within six months before) the award month. Returns_x_y equals the buy-and-
hold return compounded from y months prior to the award to x months prior to the award. CEO female equals
one if the CEO is a female. CEO age is the CEO‟s age at the time of the award and CEO tenure is the number of
years the CEO has been employed by the firm.

This article is protected by copyright. All rights reserved.


14
by innate differences in the probability of winning an award (selection effects). Importantly,

we match firms led by award-winning CEOs to firms that are led by CEOs that have never

won an award but have the closest predicted probability of winning an award. In this way, we

have a „clean‟ control sample as this procedure ensures that we are not inadvertently

matching an award-winning firm to a firm that is led by a future award-winning firm.

The association between CEO awards and tax aggressiveness

Next, we test the extent to which CEO awards are associated with a shift toward a

more aggressive tax policy by estimating the following regression (firm and time subscripts

are omitted for brevity and all variable definitions are provided in the appendix):

TAX = α0 + α1POST + Controls + firm and year fixed effects (1)

This difference-in-difference specification, particularly the inclusion of firm fixed effects, is

in the spirit of Bertrand, Duflo, and Mullainathan (2004), Chen, Chen, Schipper, Xu, and Xue

(2012), and Balakrishnan, Billings, Kelly, and Ljungqvist (2014).13 Our choice of control

variables is adapted from Chen et al. (2010), and has been used with some variation

extensively by other researchers (e.g., McGuire et al., 2012; Cheng et al., 2012, Omer et al.,

2012). The idea behind Equation (1) is to isolate the effect of the variable of interest (in our

case POST) while controlling for the indirect effects on tax aggressiveness from foreign

operations, intangibles, net operating losses, growth opportunities, and other factors. TAX

represents one of two measures of tax aggressiveness (DTAX or SHELTER) described above.

POST is an indicator variable equal to one for fiscal years after the CEO wins a prestigious

media award and operates as our difference-in-difference estimator.

A large set of control variables is included to isolate the effect of a shock to

overconfidence (through awards) on tax aggressiveness. First, we control for the equity

13
Balakrishnan et al. (2014), in particular, construct a matched sample of control firms to mitigate concerns that
their treatment effect (analyst coverage terminations) may be confounded by selection effects. Given our
matched sample design, our difference-in-difference design is most similar to this study.

This article is protected by copyright. All rights reserved.


15
incentives of the CEO, as well as other named executive officers (NEOs), by including option

Delta and Vega in the regression. Delta and Vega are computed following Core and Guay

(2002).14 Malmendier and Tate (2009) show that award-winning CEOs receive higher

subsequent compensation packages, and Rego and Wilson (2012) show that equity incentives

are positively associated with tax aggressiveness, suggesting these are important control

variables. We lag these compensation-related variables in our multivariate regressions and we

log transform them to mitigate the influence of outliers. Time subscripts are also included in

all of the tables to make it clear when we are using lagged and contemporaneous variables.

We also include controls found in other tax research to control for tax effects arising

from foreign operations, intangibles, net operating losses, growth opportunities and other

items. Operating performance is controlled for with pretax return on assets (ROA), measured

as pretax book income (Compustat PIi,t) divided by lagged total assets (Compustat ATi,t-1).

We include a measure of performance-matched pretax discretionary accruals (ACC)

following the procedures in Frank et al. (2009).15 The natural logarithm of lagged total assets

(Compustat ATi,t-1) controls for firm size (SIZE). Foreign income (FI) is measured as pretax

foreign income for the year (Compustat PIFOi,t) divided by lagged total assets (Compustat

ATi,t-1) to control for tax effects arising from foreign operations. EQINC is equal to one if

equity in earnings (Compustat ESUBi,t) is positive in order to capture differences arising from

unconsolidated earnings. We include a control for intangibles (INTAN) computed by dividing

reported intangibles (Compustat INTANi,t) by lagged total assets (Compustat ATi,t-1) to

capture the tax effects from differences in tax and book treatment of intangible assets. We

include a measure of capital intensity defined as property, plant and equipment (Compustat

14
We follow the Core and Guay (2002) “one-year approximation” method detailed in their Section 2. Brockman
et al (2010) also provide a clear summary in their appendix. For years after 2005, Execucomp altered their data
organization, and availability, corresponding to the change in U.S. GAAP accounting for stock options (SFAS
123R). For post-2005 observations we follow the procedures outlined in Hayes, Lemmon, and Qiu (2012).
15
Following Frank et al. (2009), we require at least 10 observations for each industry and year in order to
estimate ACC.

This article is protected by copyright. All rights reserved.


16
PPENTi,t) divided by lagged total assets (Compustat ATi,t-1). We control for the existence of a

positive net operating loss carryforward (Compustat TLCFi,t) with an indicator variable

(NOL), and with the change in NOL during the fiscal year (ΔNOL) scaled by lagged total

assets (Compustat ATi,t-1).

The market-to-book ratio (MTB), measured as the market value of equity at the

beginning of the year (Compustat PRCC_Fi,t-1*Compustat CSHOi,t-1) scaled by book value of

equity at the beginning of the year (Compustat CEQi,t-1), and research and development

activity (R&D) measured as research and development expense (Compustat XRDi,t) divided

by lagged total assets (Compustat ATi,t-1) are included to control for differences in growth

opportunities. We also control for differences in debt use with a measure of leverage (LEV),

equal to long-term debt (Compustat DLTTi,t) divided by lagged total assets (Compustat ATi,t-

1). A measure of free cash flow (FCF) computed as net operating cash flow minus capital

expenditures (Compustat OANCFi,t - Compustat CAPXi,t) divided by lagged total assets

(Compustat ATi,t-1) captures differences in corporate cash holdings for firms (Dhaliwal et al.,

2011).

In some of our regressions we include controls for the ability and connectedness of

the CEO. MT report that CEOs tend to serve on more boards after winning an award, and

Brown (2011) and Brown and Drake (2014) find that board connections can serve as an

important information channel through which tax planning knowledge might be transferred.

Our controls for ability (Demerjian et al., 2012 and Demerjian et al., 2013) are to control for

the possibility that the award reflects improved ability and/or learning that has occurred over

time. Finally, we include firm fixed effects to isolate the within-firm change in tax

aggressiveness after the CEO wins an award, and we include year fixed effects to control for

changes in tax aggressiveness across time.

This article is protected by copyright. All rights reserved.


17
4. EMPIRICAL RESULTS

(i) Figures

Figure 1 depicts mean tax aggressiveness before and after the award year, but for

treatment and control groups separately. The patterns in Figure 1 are consistent with

expectations. Specifically, mean DTAX for years before (after) the CEO wins an award is -

0.013 (0.010), mean SHELTER is 3.541 (4.828), and these differences are statistically

significant (p-value < 0.01). Moreover, DTAX for the control group firms moves in the

opposite direction, from 0.004 to 0.002. Overall, the univariate results for our measures of tax

aggressiveness suggest that there is a clear response to the CEO‟s award in a pattern

consistent with the predictions of a shock to CEO confidence. Next we turn to the

multivariate tests of our hypothesis.

(ii) Propensity Score Matching

Determinants of CEO awards

Our replication of MT‟s logit regression predicting CEO awards for our sample of

winners is reported in Table 2. This regression predicts the likelihood of winning the award

for each month during which an award is granted. Results in Table 2 are consistent with the

patterns reported in Malmendier and Tate (2009, Table II).16

It is important to highlight that, despite a relatively small number of awards in the

sample, the logistic regression demonstrates substantial discriminant ability, as the area under

the ROC curve is 0.951. Next, we use the logit model to construct a control sample of

nearest-neighbor firms led by CEOs who are observationally similar but have never won an

award. Specifically, we match without replacement and use the Abadie and Imbens (2006,

2011) bias-corrected matching estimator to form matched pairs.

16
Despite our longer time series, our sample has fewer observations than used by MT. Unlike MT, we exclude
financial firms and expand the award series through 2010.

This article is protected by copyright. All rights reserved.


18
Covariate balance

We require a one-to-one matching, and both treatment and control firms must have at

least one fiscal year of data available before and after the award. Our resulting propensity

score matching sample has 152 matched pairs.17 Table 3 reports the covariate balance of the

matched sample. With the exception of firm size (Market capitalization), treatment and

control firms are similar on all observable characteristics, suggesting that we have obtained a

valid match.18 Next, we merge this sample to our original dataset and re-estimate Equation

(1) using only the matched pairs resulting in a sample of 1,935 firm years.

Difference-in-difference regressions

Table 4 reports summary statistics and difference-in-difference regression results

from estimating Equation (1). Panel A reports descriptive statistics for the variables used in

the primary analyses. Mean discretionary permanent book-tax differences (DTAX) comprise

approximately 0.2% of lagged total assets, and mean SHELTER is 4.004. Means and medians

of our CEO overconfidence measures (CAPEX, OVERINVEST) are consistent with related

research (e.g., Malmendier and Tate, 2005; Malmendier and Tate, 2008; Ben-David et al.

2010; Schrand and Zechman, 2011; Ahmed and Duellman, 2013). Finally, means and

medians of our control variables are generally consistent with related tax research.

Regression results are reported in Panel B. All regressions include firm and year fixed

effects. Regressions (1) and (2) represent baseline estimations of Equation (1). Regressions

(3) and (4) control for managerial ability, and Regressions (5) and (6) control for the presence

of CEO board connections to other firms. In all specifications, our variable of interest is

POST. A positive coefficient estimate on POST suggests that firms led by award-winning

17
We ensure that firms led by a CEO who has not won an award in the previous year, but has won an award at
some other point in the sample, are not used as a potential control observation.
18
We estimate our logit model with industry and fiscal year dummies in order to be consistent with MT.
However, it is possible that a treatment firm could be matched to a control firm in a different year or industry. In
untabulated tests, we confirm that the results are robust to reducing the caliper width between the treatment and
control group so that the groups are also balanced by fiscal year and industry.

This article is protected by copyright. All rights reserved.


19
CEOs have a significant shift in tax aggressiveness during years after the CEO wins an

award.

Results confirm our hypothesis that firms led by award-winning CEOs experience an

increase in tax aggressiveness following the award. Specifically, the coefficient estimate in

the DTAX regression (Estimate = 0.0111, p-value < 0.05), and the estimate in the SHETLER

regression (Estimate = 0.1442, p-value < 0.05), suggest that firms led by CEOs who have

won an award in the previous fiscal year engage in significantly greater tax aggressiveness in

subsequent years. These results are also consistent with the univariate results depicted in

Figure 1. Specifically, the shift in tax aggressiveness is significantly greater, and in the

predicted direction, among firms led by award-winning CEOs. Using the coefficient estimate

of 0.0111 in the DTAX regression (Regression 1), for example, suggests these firms exhibit

on average $220 million greater discretionary book-tax differences (which is approximately

8.4% of pretax earnings) in years after the CEO wins an award. Overall, we interpret these

results as consistent with the hypothesis that prestigious CEO awards, acting as a shock to

overconfidence, encourage greater tax aggressiveness.

It is possible that the award could reflect improved ability of the CEO because of

learning that has occurred over time. In this way, ability could be time-varying in a manner

similar to overconfidence, thus confounding the effect of the award on tax outcomes. We

control for this possibility by estimating our regressions with a control for managerial ability,

using the innovative measure of managerial ability, described in Demerjian et al. (2012) and

Demerjian et al. (2013), to identify how efficiently managers are able to generate revenues

from a given set of inputs (MgrAbility).19 Higher values of MgrAbility reflect greater

managerial ability. Regressions (3) and (4) report the results.

19
Specifically, Demerjian et al. (2012, 2013) use data envelopment analysis, which is an advanced optimization
technique, in order to estimate the degree to which a firm is efficiently generating revenues from a given set of

This article is protected by copyright. All rights reserved.


20
Results suggest that our inferences hold after including a control for managerial

ability. Specifically, the coefficient estimate on POST in the DTAX regression (Estimate =

0.0111, p-value < 0.05) is identical to the estimate reported in Regression (1). In contrast, the

coefficient estimate on POST in the SHETLER regression is slightly attenuated but remains

statistically significant (Estimate = 0.1368, p-value < 0.05). Overall, we conclude that our

results are not driven by cross-sectional differences in managerial ability.

MT report that CEOs tend to serve on more boards after winning a prestigious award,

reflecting an increase in board connections (and information channels) of the CEO. This is

important for our setting, as Brown (2011) and Brown and Drake (2014) find that board

connections can serve as an important information medium through which tax planning

knowledge can be transferred. Thus, it is possible that the award may increase the number of

board connections of the CEO, thereby confounding the effect of the award on tax outcomes.

We address this possibility by re-estimating our regressions with a control for the number of

board connections of the CEO, and we use the BoardEx database to count the number of

boards on which the CEO serves each year (CEO Connections). To preserve data, we also

include a dummy variable (Missing CEO Connection) that equals one if we were unable to

match the CEO to the BoardEx database. Regressions (5) and (6) report the results.

Results suggest that our inferences hold when controlling for the board connections of

the CEO. Specifically, we observe a positive and significant coefficient estimate on POST in

the DTAX regression (Estimate = 0.0118, p-value < 0.05) and a positive, and a significant

coefficient estimate in the SHELTER regression (Estimate = 0.1290, p-value < 0.05). Overall,

inputs (cost of goods sold, selling, general and administrative expenses, property, plant and equipment,
operating leases, R&D, goodwill, and other intangibles). Firms are assigned a value between zero and one,
where one reflects the greatest level of efficiency. The residual from a Tobit regression of efficiency on size,
market share, positive free cash flow, firm age, segment concentration, and foreign operations and year
dummies reflects the efficiency attributable to the management team. The authors perform a number of validity
checks to ensure that their measure reflects manager characteristics. We thank Peter Demerjian for sharing his
data.

This article is protected by copyright. All rights reserved.


21
we conclude that our results are not driven by an increase in CEO board connections to other

firms.

Evidence of Changes in Tax Policies

Our results thus far reveal an increase in tax aggressiveness among firms led by

award-winning CEOs in a manner consistent with prestigious media awards acting as a shock

to overconfidence. In this section, we dig deeper by identifying evidence of tax policy

changes. Specifically, we examine the changes in foreign income relative to domestic income

which may be suggestive of a change in transfer pricing policy.20 We construct a new

dependent variable, ↑For↓Dom, which equals one if foreign income increases and domestic

income decreases relative to the previous year, thus identifying instances in which firms are

attempting to shift earnings to low-tax foreign jurisdictions. We include all of the

independent variables from Equation (1) in our regressions in order to control for important

firm characteristics that could be associated with these changes, such as firm size and the

level foreign income. Table 5 reports the results.

Table 5 reports results from two regressions: a linear prediction model using OLS and

a logit regression. Results reveal a higher likelihood of a firm reporting an increase in foreign

income and a corresponding decrease in domestic income after a CEO wins a prestigious

media award. Specifically, the coefficient estimate on POST is positive and significant in

both the linear prediction model (Estimate = 0.0959, p-value < 0.01) and the logit model

(Estimate = 0.4859, p-value < 0.05). However, we acknowledge these results are suggestive,

but not definitive, evidence of a change in transfer pricing and we interpret these results with

caution.

5. ADDITIONAL ANALYSES

20
We thank an anonymous referee for this suggestion.

This article is protected by copyright. All rights reserved.


22
In this section, we discuss additional tests, including validation tests confirming that

award-winning CEOs are more likely to be subsequently classified as overconfident, earnings

management interactions, and tests omitting large award years granted during the 1990s.

(i) The association between CEO awards and CEO overconfidence – Validation tests

In this section, we explicitly test the hypothesis offered by Malmendier and Tate

(2009), that CEO awards lead to greater hubris, to validate our use of CEO awards as a shock

to CEO confidence. Specifically, we examine whether CEO overconfidence increases

following the receipt of a major media award. We estimate these regressions over the sample

of firms led by award-winning CEOs in order to capture the within-firm change in

overconfidence by CEOs that have won an award.

We use two common measures of CEO overconfidence obtained from prior research.

Our first measure of overconfidence, CAPEX, equals one if the firm‟s capital expenditures,

scaled by lagged capital expenditures, exceeds the industry median for that fiscal year, and

zero otherwise (Ben-David, Graham, and Harvey, 2010; Ahmed and Duellman, 2013). Our

second measure of overconfidence, OVERINVEST, is equal to one if the residual from a

regression of total asset growth on sales growth is greater than zero.21 Intuitively, CEO

overconfidence is more likely to manifest in firms with asset growth outpacing that of sales

(Schrand and Zechman, 2011; Ahmed and Duellman, 2013). A firm‟s CEO is assumed to be

overconfident if these measures (CAPEX, OVERINVEST) equal one.

We follow Banerjee, Humphrey-Jenner, and Nanda (2015) and construct our

overconfidence measures on a yearly basis, allowing for time-variation consistent with

theoretical (Gervais and Odean, 2001) and empirical evidence (Billett and Qian, 2008; Hilary

and Menzly, 2006) that overconfidence can vary over time. Indeed, many of the papers in this

21
We follow Schrand and Zechman (2011) and Ahmed and Duellman (2013) and estimate this regression by
industry and fiscal year.

This article is protected by copyright. All rights reserved.


23
stream of research construct at least one (primary) measure of overconfidence that can vary

over time for the CEO (e.g., Ahmed and Duellman, 2013; Banerjee, Humphrey-Jenner,

Nanda, 2015; Malmendier and Tate, 2005; 2008; Schrand and Zechman, 2012).

In this specification, POST equals one for all years after the CEO wins an award. To

isolate the effect of media awards on overconfidence, we include a number of controls that

could also be correlated with overconfidence such as equity incentives (log of CEO Delta and

CEO Vega), sales (Log(Sales)), profitability (ROA), growth opportunities (MTB), annual firm

stock return (AnnRet), leverage (TotLeverage), and free cash flow (FCF), as well as firm

fixed effects. A positive coefficient on POST is consistent with CEO awards predicting CEO

overconfidence, and thus would validate our use of media awards as a shock to CEO

overconfidence. Table 6 reports the results.

Table 6 presents results where the dependent variable is one of two measures of CEO

overconfidence (CAPEX, OVERINVEST), and our independent variable of interest is POST.

The results suggest that CEO awards are positively associated with CEO overconfidence in

the following year, as we observe positive and significant coefficients in each of the

regressions. Coefficients on most of the control variables are statistically significant. Overall,

results in Table 6 support MT‟s conjecture that CEO awards are associated with hubris, and

provide validation for our use of awards as a shock to CEO overconfidence in the preceding

regressions.

(ii) Verification Tests of CEO Overconfidence during Years after Winning an Award

We have used CEO awards as an exogenous shock to overconfidence. Although we

have confirmed in the previous section that prior year award-winners are significantly more

likely to be classified as overconfident, in this section we examine the distribution of

overconfident CEOs during years before and after winning an award. This important

verification test provides some assurance that the awards are reflecting the correct construct

This article is protected by copyright. All rights reserved.


24
of overconfidence. In our main sample, we find only two such instances where the CEO is

not classified as overconfident in years after winning the award. Overall, this verification test

suggests that CEO awards are impacting CEO confidence, consistent with our expectations.

(iii) Earnings Management Interactions

MT find that award-winning CEOs engage in greater earnings management during

years following the award. Although we control for financial reporting aggressiveness (ACC)

in our regressions, one concern is that our results might be a byproduct of firms aggressively

managing book income upward during years after the CEO wins an award. We confront this

possibility by re-estimating all of our difference-in-difference regressions with ACC

interactions. Notably, we fail to find any evidence suggesting that the shift in tax

aggressiveness that we analyze is driven by an increase in financial reporting aggressiveness.

(iv) Omitting large award years

The time distribution of award-winning CEOs in Table 1 (Panel A) shows a couple of

years where a large number of CEO awards were granted. To address the possibility that our

results are affected by these large award years, which also coincide with a period of time

when tax shelters proliferated, we re-estimate our regressions after dropping observations

from these award years. Although we have a reduced sample size we find our results

generally continue to hold suggesting that CEO awards, acting as a shock to overconfidence,

encourage CEOs to adopt more aggressive tax policies during years following the award.

6. CONCLUSION

A growing and influential stream of literature examines whether CEO characteristics

impact corporate policies. CEO overconfidence, in particular, has been theoretically and

empirically linked to corporate outcomes such as mergers (Malmendier and Tate, 2008),

investment (Ben-David et al., 2010), firm risk (Hirshleifer et al., 2012) and financial

This article is protected by copyright. All rights reserved.


25
reporting policies (Schrand and Zechman, 2011; Ahmed and Duellman, 2013). However,

research examining CEO characteristics in general, and CEO overconfidence in particular,

suffers from an identification problem in that, in the absence of a shock, it is difficult to

separate the CEO characteristics from firm characteristics and incentives. We address this

issue by using CEO awards conferred by major media outlets as an exogenous shock to CEO

overconfidence. Given the economic importance of corporate tax policy, we examine whether

this shock to overconfidence leads to a more aggressive tax policy. We find strong evidence

in favor of this hypothesis.

This paper is related to several important lines of research. First, by providing robust

evidence of a CEO characteristic that impacts tax policy, we contribute to the growing

literature examining the cross-sectional determinants of corporate tax aggressiveness.

Second, we contribute to the stream of research investigating the impact of individual

executives on corporate outcomes. In this regard, like Dyreng et al. (2010) who document

that managers have an impact on corporate tax policy, we advance the discussion by

providing evidence that a shock to CEO confidence can result in more aggressive tax policies

going forward. Overall, our results provide interesting insights into the relationship between

executives and the management of corporate taxes, and suggest a potential unintended

consequence of award-winning status is an increase in hubris.

This article is protected by copyright. All rights reserved.


26
REFERENCES
Abadie, A., Imbens, G., 2006. Large sample properties of matching estimators for average
treatment effects. Econometrica 74, 235-267.

Abadie, A., Imbens, G., 2011. Bias-corrected matching estimators for average treatment
effects. Journal of Business and Economic Statistics 29, 1-11.

Ahmed, A., and S. Duellman. 2013. Managerial Overconfidence and Accounting


Conservatism. Journal of Accounting Research 51: 1-29.

Armstrong, C., J. Blouin, and D. Larcker. 2012. The Incentives for Tax Planning. Journal of
Accounting and Economics 53: 391-411.

Balakrishnan, K., J. Blouin, and W. Guay, 2012. Does Tax Aggressiveness Reduce Financial
Reporting Transparency? Unpublished working paper, University of Pennsylvania.

Bamber, L., J. Jiang, and I. Wang. 2010. What‟s My Style? The Influence of Top Managers
on Voluntary Corporate Disclosure. The Accounting Review 85: 1131-1162.

Banerjee, S., M. Humphrey-Jenner, and V. Nanda, 2015. Restraining Overconfident CEOs


through Improved Governance: Evidence from the Sarbanes-Oxley Act. Review of
Financial Studies, in press.

Bary, A. 2014. World‟s Best CEOs. Barron’s. Vol XCIV, No. 12. March 24, 2014.

Ben-David, I., J. Graham, and C. Harvey. 2010. Managerial Miscalibration. Unpublished


working paper. Duke University.

Bertrand, M., and A. Schoar. 2003. Managing with Style: The Effect of Managers on Firm
Policies. Quarterly Journal of Economics 118: 1169-1208.

Billett, M. and Y. Qian, 2008. Are Overconfident CEOs Born or Made? Evidence of Self-
Attribution Bias from Frequent Acquirers. Management Science 54, 1037-1051.

Brockman, P., X. Martin, and E. Unlu. 2010. Executive Compensation and the Maturity
Structure of Corporate Debt. Journal of Finance 65: 1123-1161.

Brown, J., 2011. The Spread of Aggressive Corporate Tax Reporting: A Detailed
Examination of the Corporate-Owned Life Insurance Shelter. The Accounting Review
86, 23-57.

Brown, J. and K. Drake, 2014. Network Ties Among Low-Tax Firms. The Accounting
Review 89, 483-510.

Campbell, T., M. Gallmeyer, S. Johnson, J. Rutherford, and B. Stanley. 2011. CEO Optimism
and Forced Turnover. Journal of Financial Economics 101: 695-712.

This article is protected by copyright. All rights reserved.


27
Cassell, C., S. Huang, J. Sanchez, and M. Stuart. 2012. Seeking Safety: The Relation between
CEO Inside Debt Holdings and the Riskiness of Firm Investment and Financial
Policies. Journal of Financial Economics 103: 588-610.

Chen, C., X. Chen, Q. Cheng, and T. Shevlin. 2010. Are Family Firms More Tax Aggressive
than Non-Family Firms? Journal of Financial Economics 95: 41-61.

Cheng, C., H. Huang, Y. Li, and J. Stanfield. 2012. The Effect of Hedge Fund Activism on
Corporate Tax Avoidance. The Accounting Review 87: 1493-1526.

Chyz, J., F. Gaertner, A. Kausar, and L. Watson, 2014. Overconfidence and aggressive
corporate tax policy. Unpublished working paper. University of Tennessee, University
of Wisconsin-Madison, Nanyang Technological University, University of Florida.

Coles, J., D. Naveen, and L. Naveen. 2006. Managerial incentives and risk-taking. Journal of
Financial Economics 79: 431-468.

Core, J., and J. Guay. 2002. Estimating the Value of Employee Stock Option Portfolios and
their Sensitivities to Price and Volatility. Journal of Accounting Research 40: 613-
630.

Demerjian, P., B. Lev, M. Lewis, and S. McVay, 2013. Managerial Ability and Earnings
Quality. The Accounting Review 88, 463-498.

Demerjian, P., B. Lev, and S. McVay, 2012. Quantifying Managerial Ability: A New
Measure and Validity Tests. Management Science 58, 1229-1248.

Desai, M., and D. Dharmapala. 2006. Corporate Tax Avoidance and High-Powered
Incentives. Journal of Financial Economics 79: 145-179.

Dhaliwal, D., S. Huang, S. Moser, and R. Pereira. 2011. Corporate Tax Avoidance and the
Level and Valuation of Firm Cash Holdings. Unpublished working paper. University
of Arizona, University of Arkansas, University of Missouri at Columbia. Available at
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1905076.

Dyreng, S., M. Hanlon, and E. Maydew. 2008. Long-Run Corporate Tax Avoidance. The
Accounting Review 83: 61-81.

Dyreng, S., M. Hanlon, and E. Maydew. 2010. The Effects of Executives on Corporate Tax
Avoidance. The Accounting Review 85: 1163-1189.

Dyreng, S., W. Mayew, and C. Williams. 2012. Religious Social Norms and Corporate
Financial Reporting. Journal of Business Finance & Accounting 39, 845-875.

Edwards, A., C. Schwab, and T. Shevlin, 2016. Financial Constraints and Cash Tax Savings.
The Accounting Review 91 (3): 859-881.

Faulkender, M. and M. Petersen, 2012. Investment and Capital Constraints: Repatriations


Under the American Jobs Creation Act. Review of Financial Studies, 3351-3388.

This article is protected by copyright. All rights reserved.


28
Foley, C., J. Hartzell, S. Titman, and G. Twite, 2007. Why Do Firms Hold So Much Cash? A
Tax-Based Explanation. Journal of Financial Economics, 579-607.

Frank, M., L. Lynch, and S. Rego. 2009. Tax Reporting Aggressiveness and its Relation to
Aggressive Financial Reporting. The Accounting Review 64: 467-496.

Gervais, S. and T. Odean, 2001. Learning to be Overconfident. Review of Financial Studies


14, 1-27.

Graham, J., and A. Tucker. 2006. Tax Shelters and Corporate Debt Policy. Journal of
Financial Economics 81: 563-594.

Graham, J., M. Hanlon, T. Shevlin, and N. Shroff. 2012. Incentives for Tax Planning and
Avoidance: Evidence from the Field. Unpublished working paper. Duke University,
MIT, University of California at Irvine. Available at
http://papers.ssrn.com/sol3/papers.cfm? abstract_id=2148407.

Hall, B., and K. Murphy, 2002. Stock options for undiversified executives, Journal of
Accounting and Economics 33, 3-42.

Hanlon, M. 2005. The Persistence and Pricing of Earnings, Accruals, and Cash Flows when
Firms Have Large Book-Tax Differences. The Accounting Review 80: 137-166.

Hanlon, M., and S. Heitzman. 2010. A Review of Tax Research. Journal of Accounting and
Economics 50: 127-178.

Hanlon, M., R. Lester, and R. Verdi, 2015. The Effect of Repatriation Tax Costs on U.S.
Multinational Investment. Journal of Financial Economics, in press.

Hanlon, M., and J. Slemrod. 2009. What does tax aggressiveness signal? Evidence from stock
price reactions to news about tax shelter involvement. Journal of Public Economics
93: 126-141.

Hasan, I., C. Hoi, Q. Wu, and H. Zhang, 2014. Beauty is in the Eye of the Beholder: The
Effect of Corporate Tax Avoidance on the Cost of Bank Loans. Journal of Financial
Economics 113, 109-130.

Hayes, R., M. Lemmon, and M. Qui. 2012. Stock Options and Managerial Incentives for
Risk-Taking: Evidence from FAS 123R. Journal of Financial Economics 105: 174-
190.

Higgens, D., T. Omer, and J. Phillips. 2013. The Influence of a Firm‟s Business Strategy on
its Tax Aggressiveness. Unpublished working paper, University of Connecticut,
University of Nebraska-Lincoln. Available at
http://papers.ssrn.com/sol3/papers.cfm?abstract_id= 1727592.

Hilary, G. and L. Menzly, 2006. Does Past Success lead Analysts to become Overconfident?
Management Science 52, 489-500.

This article is protected by copyright. All rights reserved.


29
Hirshleifer, D., A. Low, and S. Teoh. 2012. Are Overconfident CEOs Better Innovators?
Journal of Finance 67: 1457-1498.

Hope, O., M. Ma, and W. Thomas, 2013. Tax Avoidance and Geographic Earnings
Disclosure. Journal of Accounting and Economics 56, 170-189.

Hosmer, D. and S. Lemeshow. 2000. Applied Logistic Regression. Wiley, NewYork.

Hsieh, T., J. Bedard, K. Johnstone. 2014. CEO Overconfidence and Earnings Management
During Shifting Regulatory Regimes. Journal of Business Finance & Accounting 41,
1243-1268.

Kidd, J., 1970. The utilization of subjective probabilities in production planning. Acta
Psychologica 34, 338-347.

Kim, J., Y. Li, and L. Zhang. 2011. Corporate Tax Avoidance and Stock Price Crash Risk:
Firm-Level Analysis. Journal of Financial Economics 100: 639-662.

Langer, E., 1975. The illusion of control. Journal of Personality and Social Psychology 32,
311-328.

Larwood, L and W. Whittaker, 1977. Managerial myopia: Self-serving bias in organizational


planning. Journal of Applied Psychology 62, 94-198.

Law, K. and L. Mills, 2013. Following the Rules? Corporate Tax Reporting by CEOs with
Military Experience. Unpublished working paper, Tilbery University and the
University of Texas at Austin.

Lev, B. and D. Nissim. 2004. Taxable income, future earnings, and equity values. The
Accounting Review, 79: 1039-1074.

Malmendier, U. and G. Tate. 2005. CEO Overconfidence and Corporate Investment. Journal
of Finance 55: 2661-2700.

Malmendier, U. and G. Tate. 2008. Who Makes Acquisitions? CEO Overconfidence and the
Market‟s Reaction. Journal of Financial Economics 89: 20-43.

Malmendier, U. and G. Tate. 2009. Superstar CEOs. Quarterly Journal of Economics 124:
1593-1638.

Malmendier, U. G. Tate, and J. Yan. 2011. Overconfidence and Early-Life Experiences: The
Effect of Managerial Traits on Corporate Financial Policies. Journal of Finance 66:
1687-1733.

March, J. and Z. Shapira, 1987. Managerial perspectives on risk and risk-taking. Management
Science 33, 1404-1418.

This article is protected by copyright. All rights reserved.


30
McGuire, S., T. Omer, and D. Wang. 2012. Tax avoidance: Does tax-specific industry
expertise make a difference? The Accounting Review 87: 975-1003.

Mills, L., 1998. Book-Tax Differences and Internal Revenue Service Adjustments. Journal of
Accounting Research 36, 343-356.

Mills, L., M. Erickson, and E. Maydew. 1998. Investments in Tax Planning. Journal of
American Taxation Association 20: 1-20.

Mills, L. and R. Sansing, 2000. Strategic Tax and Financial Reporting Decisions: Theory and
Evidence. Contemporary Accounting Research 17, 85-106.

Moore, P., 1977. The manager‟s struggle with uncertainty. Journal of The Royal Statistical
Society Series A 149, 129-165.

Rego, S. 2003. Tax-Avoidance Activities of U.S. Multinational Corporations. Contemporary


Accounting Research 20: 805-833.

Rego, S. and R. Wilson. 2012. Equity Risk Incentives and Corporate Tax Aggressiveness.
Journal of Accounting Research 50: 775-809.

Robinson, J., S. Sikes, and C. Weaver. 2010. Performance Measurement of Corporate Tax
Departments. The Accounting Review 85: 1035-1064.

Roll, R. 1986. The hubris hypothesis of corporate takeovers. Journal of Business 59: 197-
216.

Shackelford, D. and T. Shevlin. 2001. Empirical Tax Research in Accounting. Journal of


Accounting and Economics 31: 321-387.

Schrand, C., and S. Zechman. 2011. Executive Overconfidence and the Slippery Slope to
Financial Misreporting. Journal of Accounting and Economics 53, 311-329.

Shevlin, T. 2007. The Future of Tax Research: From an Accounting Professor's Perspective.
Journal of American Taxation Association 29: 87-93.

Omer, T., C. Weaver, and J. Wilde. 2012. Investments in Tax Planning, Tax Avoidance and
the New Economy Business Model. Unpublished working paper, Texas A&M
University. Available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2001716.

Wilson, R. 2009. An Examination of Corporate Tax Shelter Participants. The Accounting


Review 84: 969-999.

This article is protected by copyright. All rights reserved.


31
Variable Appendix
Variable Source Definition
A.1. Dependent variables
DTAXt Compustat Computed as the residual from an industry-year regression of
permanent differences on intangibles, unconsolidated earnings,
non-controlling interest in earnings, current state income expense,
change in NOL, and lagged permanent differences. See
Frank et al. (2009, p. 473) for additional details.

SHELTERt Compustat Tax shelter score from Wilson (2009, p.988), defined as
-4.86 + 5.20*BTD + 4.08*|ACC| - 1.41*LEV + 0.76*SIZE +
3.51*ROA
+ 1.72*FI + 2.43*R&D.

↑For↓Domt Compustat Equals one if foreign income over sales increases and domestic income
over sales decreases relative to the previous year.

CAPEXt Compustat Equals one if the capital expenditures, scaled by lagged assets, exceed
the industry median for that year, zero otherwise.

Equals one if the residual from a regression of asset growth on sales


OVERINVESTt
Compustat growth is
positive, zero otherwise. Regressions are estimated by industry and
year.
A.2. Independent
variables
POST Various Equals one for all years after the CEO wins an award

CEO Deltat-1 Execucomp Dollar increase in the CEO's stock option portfolio given a 1%
change in the underlying stock price. Computed using the Core and
Guay (2002) "one-year approximation" method.

CEO Vegat-1 Execucomp Dollar increase in the CEO's stock option portfolio given a 1%
change in the underlying stock volatility Computed using the Core and
Guay (2002) "one-year approximation" method.

Log(CEO Delta)t-1 Execucomp Natural logarithm of CEO delta.

Log(CEO Vega)t-1 Execucomp Natural logarithm of CEO vega.

NEO Deltat-1 Execucomp Total dollar increase in the NEOs' stock option portfolio given a 1%
change in the underlying stock price. Computed using the Core and
Guay (2002) "one-year approximation" method.

NEO Vegat-1 Execucomp Total dollar increase in the NEOs' stock option portfolio given a 1%
change in the underlying stock volatility Computed using the Core and
Guay (2002) "one-year approximation" method.

Log(NEO Delta)t-1 Execucomp Natural logarithm of NEO delta.

Log(NEO Vega)t-1 Execucomp Natural logarithm of NEO vega.

This article is protected by copyright. All rights reserved.


32
ROAt Compustat Pretax book income (Compustat PI) divided by lagged total assets
(Compustat AT).

ACCt Compustat Pretax discretionary accruals. Computed following Frank et al. (2009).

Natural logarithm of lagged market capitalization (Compustat


SIZEt-1
Compustat PRCCF*CSHO).

FIt Compustat Pretax foreign income (Compustat PIFO) divided by lagged total assets
(Compustat AT).

EQINCt Compustat Equals one if unconsolidated earnings (Compustat ESUB) is positive,


zero otherwise.

INTANt Compustat Intangibles (Compustat INTAN) divided by lagged total assets


(Compustat AT).

PPEt Compustat Net property, plant, and equipment (Compustat PPENT) divided by
lagged total assets (Compustat AT).

NOLt Compustat Equals one if the firm reports a positive tax loss carryforward during
the year (Compustat TLCF); zero otherwise.

ΔNOLt Compustat Change in tax loss carryforward (Compustat TLCF) divided by


laggged total assets (Compustat AT).

MTBt-1 Compustat Lagged market-to-book value of equity ratio


(Compustat (PRCC_F*CSHO)/CEQ).

LEVt Compustat Long-term debt (Compustat DLTT) divided by lagged total assets
(Compustat AT).

FCFt Compustat Free cash flow divided by lagged total assets


((Compustat OANCF - CAPX)/AT).

R&Dt Compustat Research and development expense (Compustat XRD) divided by


lagged total assets (Compustat AT).

MgrAbilityt P. Demerjian Managerial ability score from Demerjian et al. (2012, 2013).

CEO Connectiont BoardEx Number of board seats held by CEO.

Log(Sales)t Compustat Natural logarithm of total sales (Compustat SALE).

SalesGrowtht Compustat Contemporaneous sales (Compustat SALE) divided by lagged sales.

Market capitalization plus dividends minus lagged market


AnnRett
Compustat capitalization,
scaled by lagged market capitalization.

Total leverage (Compustat DLC + DLTT) divided by assets


TotLeveraget
Compustat (Compustat AT).

This article is protected by copyright. All rights reserved.


33
Figure 1
Mean tax aggressiveness prior to (Pre) and after (Post) winning an award
This figure depicts mean measures of tax aggressiveness of firms led by CEOs before (Pre) and after (Post)
winning a prestigious media award. The sample used to construct these figures contains 1,935 observations and
is obtained using propensity score matching to construct a matched sample of firms led by award-winning CEOs
(Treatment) and firms led by CEOs who did not win an award but otherwise appear similar on observable
characteristics that explain award (Controls). Untabulated t-tests of differences in means confirm that the
differences are significant (p-value < 0.01).

DTAX
0.015

0.010

0.005

0.000
Pre Post
-0.005

-0.010

-0.015

Treatment Controls

SHELTER
6.000

5.000

4.000

3.000

2.000

1.000

0.000
Pre Post

Treatment Controls

This article is protected by copyright. All rights reserved.


34
Table 1
Sample composition.
Panel A: Time distribution of full (unmatched) sample
Fiscal year Award = 1 % Award = 0 %
1994 10 3.14 701 4.42
1995 35 11.01 829 5.37
1996 44 13.84 882 5.76
1997 13 4.09 922 5.82
1998 12 3.77 875 5.52
1999 11 3.46 936 5.89
2000 13 4.09 954 6.02
2001 8 2.52 763 4.80
2002 20 6.29 750 4.79
2003 16 5.03 781 4.96
2004 14 4.40 941 5.94
2005 27 8.49 934 5.98
2006 12 3.77 959 6.04
2007 16 5.03 942 5.96
2008 21 6.60 893 5.69
2009 17 5.35 833 5.29
2010 15 4.72 948 5.99
2011 14 4.40 915 5.78

Panel B: Industry distribution of full (unmatched) sample


One-digit SIC Award=1 % Award=0 %
0-1 (Agriculture, mining, oil and construction) 10 3.14% 772 4.90%
2 (Food, tobacco, textiles, paper and chemicals) 65 20.44% 3,523 22.36%
3 (Manufacturing, machinery and electronics) 120 37.74% 5,282 33.52%
4 (Transportation and communications) 26 8.18% 855 5.43%
5 (Wholesale and retail) 46 14.47% 2,591 16.44%
7 (Services) 32 10.06% 1,949 12.37%
8-9 (Health, legal and educational services and other) 19 5.97% 786 4.99%

This article is protected by copyright. All rights reserved.


35
Table 2
Determinants of CEO awards
This table reports results from a logistic regression of CEO award on variables prior research has shown to
impact award-winning status. Market capitalization equals the stock price multiplied by the number of shares
outstanding two months prior to the award month. The book-to-market ratio equals the book value of equity
divided by market capitalization at the end of the fiscal year prior to (but within six months before) the award
month. Returns_x_y equals the buy-and-hold return compounded from y months prior to the award to x months
prior to the award. CEO female equals one if the CEO is a female. CEO age is the CEO‟s age at the time of the
award and CEO tenure is the number of years the CEO has been employed by the firm. For brevity, the
intercept, and fiscal year, industry, and award month fixed effects are not reported. Robust z-statistics are
reported below each coefficient estimate, and all p-values are two-tailed. ***, **, * denotes significance at the
1%, 5%, and 10% level, respectively.

Variable Award
Market capitalization 1.1480***
(24.70)
Book-to-market ratio -0.6863**
(-2.13)
Returns_2_3 1.1466**
(2.31)
Returns_4_6 1.8907***
(4.70)
Returns_7_12 1.0746***
(5.02)
Returns_13_36 0.2504***
(3.06)
CEO female (dummy) 1.1652**
(2.37)
CEO age -0.0191
(-1.61)
CEO tenure 0.0393***
(4.53)

Observations 60,592
Psuedo R-squared 0.353
Area under the ROC curve 0.951

This article is protected by copyright. All rights reserved.


36
Table 3
Covariate balance between treatment and control groups.
This table reports the covariate balance for the matched pairs obtained from the logistic regression model in
Table 2. Firms led by recent award-winning CEOs (Treatment group) are matched without replacement to firms
led by CEOs who have never won an award but otherwise appear similar on the characteristics in Table 2
(Control group). The final column reports two-sided p-values from t-tests of differences in means across groups.
(1) (2)
Treatment group Control group
Variable N Mean N Mean p-value
Market capitalization 152 17.510 152 17.254 0.045
Book-to-market ratio 152 0.262 152 0.259 0.836
Returns_2_3 152 0.040 152 0.038 0.896
Returns_4_6 152 0.068 152 0.050 0.358
Returns_7_12 152 0.184 152 0.138 0.145
Returns_13_36 152 0.583 152 0.528 0.530
CEO age 152 57.020 152 56.724 0.722
CEO tenure 152 9.257 152 8.336 0.338

This article is protected by copyright. All rights reserved.


37
Table 4
The association between CEO awards and subsequent tax aggressiveness
This table reports descriptive statistics (Panel A) and results from difference-in-difference regressions (Panel B)
where the dependent variables are measures of tax aggressiveness (DTAX, SHELTER) and the variable of
interest (POST) equals one for all years after the CEO wins an award. The sample used to estimate these
regressions is obtained from the propensity score matching model described in Tables 2 and 3 and further
discussed in the text. For brevity, the intercept, and fiscal year and firm fixed effects are not reported. Robust t-
statistics are reported below each coefficient estimate, and all p-values are two-tailed. ***, **, * denotes
significance at the 1%, 5%, and 10% level, respectively.

Panel A: Descriptive statistics


Variable N Mean Std dev 10th Pctl 50th Pctl 90th Pctl
DTAXt 1,935 0.002 0.074 -0.069 -0.002 0.078
SHELTERt 1,935 4.004 1.531 1.800 4.170 5.959
HOLDER67t 1,935 0.401 0.490 0.000 0.000 1.000
CAPEXt 1,935 0.678 0.467 0.000 1.000 1.000
OVERINVESTt 1,935 0.312 0.463 0.000 0.000 1.000
Log(CEO Delta)t-1 1,935 6.678 1.596 4.927 6.725 8.817
Log(CEO Vega)t-1 1,935 4.749 2.095 0.000 5.276 6.912
Log(NEO Delta)t-1 1,935 6.621 1.308 5.033 6.711 8.137
Log(NEO Vega)t-1 1,935 5.307 1.674 3.324 5.555 7.077
ROAt 1,935 0.161 0.093 0.054 0.145 0.284
ACCt 1,935 -0.019 0.054 -0.083 -0.012 0.037
SIZEt-1 1,935 9.731 1.317 7.986 9.708 11.628
FIt 1,935 0.045 0.052 0.000 0.024 0.124
EQINCt 1,935 0.234 0.423 0.000 0.000 1.000
INTANt 1,935 0.182 0.213 0.000 0.096 0.492
PPEt 1,935 0.324 0.216 0.099 0.270 0.645
NOLt 1,935 0.310 0.462 0.000 0.000 1.000
ΔNOLt 1,935 0.002 0.018 -0.002 0.000 0.007
MTBt-1 1,935 4.946 3.669 1.866 3.861 9.186
LEVt 1,935 0.183 0.148 0.000 0.168 0.376
FCFt 1,935 0.091 0.082 0.000 0.086 0.190
R&Dt 1,935 0.042 0.055 0.000 0.018 0.119
Log(Sales)t 1,935 9.162 1.267 7.351 9.276 10.759
SalesGrowtht 1,935 1.129 0.187 0.977 1.097 1.311
AnnRett 1,935 0.201 0.452 -0.229 0.136 0.701
TotLeveraget 1,935 0.209 0.145 0.008 0.198 0.405

This article is protected by copyright. All rights reserved.


38
Table 4 (continued)
The association between CEO awards and subsequent tax aggressiveness
Panel B: Regressions
(1) (2) (3) (4) (5) (6)
Variable DTAXt SHELTERt DTAXt SHELTERt DTAXt SHELTERt
POST 0.0111** 0.1442** 0.0111** 0.1368** 0.0118** 0.1290**
(1.98) (2.48) (1.99) (2.38) (2.10) (2.22)
Log(CEO Delta)t-1 0.0002 0.0448*** 0.0002 0.0406*** -0.0000 0.0471***
(0.09) (2.87) (0.11) (2.61) (-0.01) (2.99)
Log(CEO Vega)t-1 0.0017 -0.0225 0.0017 -0.0203 0.0018 -0.0265*
(1.23) (-1.58) (1.25) (-1.44) (1.34) (-1.85)
Log(NEO Delta)t-1 0.0037 0.0970*** 0.0037 0.0835*** 0.0038 0.0960***
(1.55) (4.19) (1.53) (3.67) (1.56) (4.16)
Log(NEO Vega)t-1 -0.0015 -0.0112 -0.0016 -0.0068 -0.0015 -0.0104
(-0.78) (-0.61) (-0.85) (-0.37) (-0.79) (-0.56)
ROAt -0.0826* -0.0859* -0.0796*
(-1.87) (-1.88) (-1.78)
ACCt 0.0803 0.0778 0.0790
(1.48) (1.44) (1.45)
SIZEt-1 -0.0071* -0.0078* -0.0072*
(-1.73) (-1.92) (-1.75)
FIt 0.1453* 0.1516* 0.1461*
(1.88) (1.96) (1.89)
EQINCt -0.0126* -0.0122* -0.0125*
(-1.76) (-1.71) (-1.74)
INTANt 0.0897*** 0.0912*** 0.0904***
(4.76) (4.83) (4.78)
PPEt -0.0054 -0.0017 -0.0046
(-0.21) (-0.07) (-0.18)
NOLt 0.0040 0.0038 0.0039
(0.75) (0.72) (0.72)
ΔNOLt 0.3118** 0.3136** 0.3132**
(2.44) (2.45) (2.45)
MTBt-1 -0.0006 -0.0004 -0.0006
(-0.72) (-0.48) (-0.74)
LEVt -0.0325 -0.0399* -0.0335
(-1.54) (-1.93) (-1.58)
FCFt 0.0250 0.0233 0.0240
(0.56) (0.53) (0.54)
R&Dt 0.0321 -0.0034 0.0303
(0.29) (-0.03) (0.28)
MgrAbilityt -0.0017 1.0967***
(-0.07) (5.45)
CEO Connectiont -0.0007 0.0391**
(-0.40) (2.44)
Missing CEO Connectiont 0.0117 -0.1566
(1.18) (-1.37)
Observations 1,935 1,935 1,933 1,933 1,935 1,935
R-squared 0.237 0.845 0.239 0.848 0.238 0.846

This article is protected by copyright. All rights reserved.


39
Table 5
Changes in foreign income relative to changes in domestic income
This table reports results from regressions where the dependent variable equals one if foreign income increased
and domestic income decreased, compared to the previous year (↑For↓Dom). Regression (1) is estimated as a
linear prediction model using OLS. Regression (2) reports results from a fixed effects logit regression. Our
variable of interest (POST) equals one for all years after the CEO wins an award. The sample used to estimate
these regressions is obtained from the propensity score matching model described in Tables 2 and 3, and further
discussed in the text. For brevity, the intercept, and fiscal year and firm fixed effects are not reported. Robust
t/z-statistics are reported below each coefficient estimate, and all p-values are two-tailed. ***, **, * denotes
significance at the 1%, 5%, and 10% level, respectively.
(1) (2)
Estimation: OLS Logit
Variable ↑For↓Domt ↑For↓Domt
POST 0.0959*** 0.4859**
(2.59) (2.26)
Log(CEO Delta)t-1 0.0262** 0.1439**
(2.20) (2.17)
Log(CEO Vega)t-1 -0.0093 -0.0685
(-0.86) (-1.16)
Log(NEO Delta)t-1 0.0074 0.0168
(0.46) (0.19)
Log(NEO Vega)t-1 0.0023 0.0332
(0.16) (0.44)
ROAt -1.1184*** -6.6770***
(-4.62) (-4.79)
ACCt -0.2403 -0.9722
(-0.93) (-0.67)
SIZEt-1 -0.0045 0.0563
(-0.17) (0.38)
FIt 3.4348*** 20.1153***
(9.05) (7.82)
EQINCt 0.0445 0.2393
(1.07) (1.01)
INTANt 0.0247 0.0471
(0.30) (0.09)
PPEt 0.2625* 1.6399*
(1.67) (1.88)
NOLt 0.0112 0.0286
(0.31) (0.14)
ΔNOLt 0.3344 1.2749
(0.51) (0.41)
MTBt-1 0.0036 0.0223
(0.76) (0.80)
LEVt 0.0956 0.6017
(0.93) (0.87)
FCFt -0.0596 -0.3371
(-0.24) (-0.25)
R&Dt 0.1358 3.5551
(0.26) (1.15)
Observations 1,935 1,935
R-squared 0.296
Log likelihood -776.019

This article is protected by copyright. All rights reserved.


40
Table 6
The association between CEO awards and subsequent overconfidence
This table reports results from regressions where the dependent variable is a measure of overconfidence
(CAPEX, OVERINVEST) and the independent variable of interest is POST which equals one for all years after
the CEO wins an award. For brevity, the intercept and firm fixed effects are not reported. Robust z-statistics are
reported below each coefficient estimate, and all p-values are two-tailed. ***, **, * denotes significance at the
1%, 5%, and 10% level, respectively.
(1) (2)
Variable CAPEXt OVERINVESTt
POST 0.9296*** 0.9406***
(2.74) (3.83)
Log(CEO Delta)t-1 0.1754** 0.0771
(2.22) (1.18)
Log(CEO Vega)t-1 -0.0637 -0.0082
(-0.99) (-0.16)
Log(Sales)t -0.5060* -0.4081**
(-1.93) (-2.02)
ROAt 7.9836*** 2.0941
(3.81) (1.39)
MTBt-1 0.0756* 0.0082
(1.69) (0.27)
SalesGrowtht 0.9031 0.8484
(1.33) (1.62)
AnnRett -0.0507 0.0767
(-0.22) (0.42)
TotLeveraget -0.4878 3.5770***
(-0.35) (3.38)
FCFt -8.8505*** 0.8235
(-4.16) (0.58)

Observations 777 1,006


Log likelihood -268.971 -457.762

This article is protected by copyright. All rights reserved.


41

Das könnte Ihnen auch gefallen