Restraining Overconfident CEOs ∗
Suman Banerjee†
Nanyang Technological University
Mark Humphery-Jenner‡
University of New South Wales
Vikram Nanda§
Georgia Institute of Technology
This version: 2nd May, 2013
Abstract
Prior literature posits that while some CEO overconfidence may benefit shareholders, high levels ofoverconfidence do not. We investigate whether improvements in governance can help to mitigatethe adverse effects of overconfidence while preserving its positive aspects. We use the passage ofthe Sarbanes-Oxley (SOX) Act as a natural experiment to examine whether improvements in regu-lation and governance help to mitigate investment distortions and moderate risk-taking tendenciesof the more overconfident CEOs. We conduct tests using options-based proxies for CEO overcon-fidence. The results indicate that, after SOX, overconfident CEOs reduced investment, improvedperformance and market value, reduced their risk-exposure, increased dividends and substantiallyimproved long-term performance following acquisitions. We also find that these SOX-related benefitsare concentrated in the firms that were SOX non-compliant prior to its passage. While the beneficialaspects of SOX in restraining overconfident CEOs may have been an unintended consequence, themessage of our paper is simple: CEO over-confidence can be monitored and regulated – just like anyother CEO attribute.
JEL Classification Code: G23, G32, G34Keyword: CEO Over-Confidence, Over-investment, Risk-taking, Quality of Investment, SOX
∗We thank the seminar participants at Nanyang Technological University, University of New South Wales,and University of Technology Sydney. The paper also benefited from comments from Itzhak Ben-David, OlegChuprinin, Jarrad Harford, Russell Jame, Jon Karpoff, Asad Kausar, Andy Kim, Jaehoon Lee, Angie Low,Thomas Noe, Jianfeng Shen, Peter Swan, Gloria Tian, Robert Tumarkin, Terry Walter, and Emma Zhang.†Nanyang Business School, Nanyang Technological University, Singapore. Tel: +65-67906237. E-mail:
[email protected]‡Australian School of Business, University of New South Wales, Australia. Tel: +61 2 9385 5853. E-mail:
[email protected]§Scheller School of Business, Georgia Institute of Technology. Tel: 404-385-8156. E-mail:
1 Introduction
Prior literature suggests that firm value may be non-monotonic in CEO overconfidence: while some
overconfidence could benefit shareholders, higher levels of overconfidence might lead to poor in-
vestment decisions and outcomes (Geravis et al., 2011; Goel and Thakor, 2008). We ask whether
improvements in governance can moderate the actions of overconfident CEOs: preserving the bene-
fits of overconfidence, while mitigating its negative aspects. In particular, we use the passage of the
Sarbanes-Oxley (SOX) Act of 2002 as a natural experiment in which there is an exogenous improve-
ment in governance. We explore the effect of SOX on the investment and risk-taking behavior of
overconfident CEOs. Our results are striking. SOX appears to improve corporate policies of firms
with overconfident CEOs in a variety of ways: For instance, there is less over-investment (relative to
industry), there is an improvement in performance and firm value, risk-taking tendencies are curbed
and acquisition outcomes are better. Furthermore, these benefits are concentrated in the firms that
were impacted by the passage of the law i.e., firms that were SOX non-compliant prior to its passage.
Confidence is essential for success in myriad domains, including business (Johnson and Fowler,
2011; Puri and Robinson, 2007).5 Overconfidence can make managers more willing to take risk.
This can be important in innovative industries, where it can encourage CEOs to invest in high-
risk innovations (see eg. Galasso and Simcoe, 2011; Hirshleifer et al., 2012; Simsek et al., 2010).
However, overconfidence can cause faulty assessments of investment value and risk, resulting in
suboptimal decision making (Dittrich et al., 2005). Indeed, Ben-David et al. (Forthcoming) indicate
that managers often miscalibrate the risk-return relationship of investments. Such overconfidence can
induce over-investment (Malmendier and Tate, 2005), poor acquisitions (Ferris et al., Forthcoming;
Kolasinski and Li, Forthcoming; Malmendier and Tate, 2008) and low dividends (Deshmukh et al.,
Forthcoming). These investment-distortions could potentially lead to accounting practices that are
‘less conservative’ (Ahmed and Duellman, 2013), and possibly earnings management (Burg et al.,
2013; Yu, 2013) and financial misreporting (Schrand and Zechman, 2012).
Our objective is to explore whether improved governance can help correct some of the undesirable
consequences of CEO overconfidence. With the passage of SOX we have an event that, for many firms,
5Johnson and Fowler (2011) argue that overconfidence, and the investment and risk-taking associated withit, can create the (potentially false) signal of corporate profitability, which itself can deter competitors andimprove the company’s competitive position.
2
is arguably associated with an exogenous improvement in governance, disclosure, and monitoring (see
e.g. Coates, 2007).6 Using SOX as a natural experiment allows us to avoid the typical endogeneity-
concerns that are associated with examining corporate governance. We argue that because SOX
enhances internal governance, and induces a climate of increased disclosure, it can help to restrain
some of the negative aspects of overconfident CEOs. Further, we compare the impact of SOX on
firms that had previously complied with SOX’s requirements and those that had not.
Similar to Malmendier and Tate (2005), we construct measures of overconfidence based upon the
CEO’s decision to voluntarily expose her self to additional firm-specific risk by retaining vested deep-
in-the-money options. Usually CEOs receive large option grants as well as restricted stocks as a part
of their compensation package. In most cases, they are not allowed to trade their options/stocks
for at least some pre-specified period of time (known as vesting period); nor can they short-sell
company stock to hedge risk exposure. At the same time, a CEO’s human capital (or her labor-
generated income) is intimately linked to the firm’s performance. Hence, this potentially serious
under-diversification should prompt any unbiased (or normal) CEO to exercise her options and/or sell
her stock at the earliest opportunity, given a sufficiently high stock price (deep in-the-moneyness).7
Following the existing literature, we argue that a CEO is overconfident if she holds any of these
exercisable options that are in the money. We measure managerial overconfidence by examining the
value of the CEO’s exercisable (but not-yet-exercised) options, reflecting the CEO’s willingness to
keep more of her personal wealth tied to the company. We classify managers as highly overconfident
if their confidence value is in the top quartile of the set of Execucomp firms.8 We also examine
alternative measures of overconfidence, including press-based measures. We apply these measures to
a panel data set on the options and stock holdings of CEOs of Execucomp companies between 1992
and 2011. We also examine the acquisition decisions of these companies.
We have several important findings. Our results indicate that, prior to SOX, overconfident CEOs
tend to employ more capital relative to other CEOs in their industry. However, after the passage of
SOX, overconfident CEOs reduce capital expenditures quite drastically and bring it closer to other
6We briefly discuss the literature surrounding SOX in Section 2.7See, for example, Lambert et al. (1991), Meulbroek (2001), Hall and Murphy (2000) and Hall and Murphy
(2002) for more detailed discussions.8This is analogous to the measure of overconfidence used in (Malmendier and Tate, 2005, 2008), which is
based upon proprietary data to which we do not have access.
3
CEOs in their industry. In these regressions we control for firm/industry and year fixed effects along
with various other firm and CEO characteristics. The results are similar whether we use an indicator
for high overconfidence or a continuous measure of overconfidence.
We also examine Sales, General and Administrative expenses (SG&A) for overconfident and other
CEOs. In this, we follow Chen et al. (2013), who argue that overconfident CEOs are less likely to
downward-adjust SG&A, reflecting their beliefs about future growth prospects and SG&A needs.
Our results indicate that the passage of SOX is associated with a substantial drop in SG&A for
overconfident CEOs. The findings are similar for asset growth and, after SOX, there is a significant
reduction in the asset growth for firms with overconfident CEOs, bringing them closer to growth
policies of other firms.
We next investigate the effect of SOX on the sensitivity of investment to cash flows. As Mal-
mendier and Tate (2005) have shown, overconfident CEOs spend more of their cash flows on capital
expenditures – reflecting their greater propensity to invest available internal funds. We find evidence
that, after SOX, overconfident CEOs’ investment-sensitivity-to-cash-flows decreases.
The next issue is how overconfident CEOs influence the firm’s exposure to both systematic
and firm-specific risk. We conjecture that overconfident CEOs are likely to under-estimate the
risk associated with investments; and thus, are likely to engage in additional risk taking.9 Our
results indicate that post-SOX there is a significant reduction in the relationship between CEO
overconfidence and both systematic and firm-specific risk.
The next question is whether this reduction in investment and risk-taking works to the benefit
or shareholders. In other words, we test whether SOX results in curbing value-destroying tenden-
cies of overconfident CEOs or whether SOX results in curbing what were previously value-creating
tendencies. For our tests, we use several measures of firm performance. We use both market based
and accounting based measures of firm performance, namely Tobin’s Q, Earnings Before Interest &
Tax (EBIT), and Standard & Poor’s Earning Quality (EQ) measure. We also examine the impact of
overconfidence on the value of R&D and CAPEX. Our results are quite clear – despite the reduction
in investment-expenditure and risk , overconfident CEOs create more shareholder value post-SOX.
We next examine the performance of overconfident CEOs in the context of acquisitions. Mal-
9We do not argue that this is destructive of shareholder value, since other managers may well be excessivelycautious.
4
mendier and Tate (2008) find that overconfident CEOs tend to make acquisitions that create sig-
nificantly less shareholder wealth. However, we find that after the passage of SOX, overconfident
CEOs do takeovers that create relatively more long-term shareholder wealth than they did before
(or destroy significantly less shareholder wealth than they did before).
The foregoing analysis begs the question of how CEOs use the cash the they would otherwise
spend on investments. Specifically, we examine whether, after SOX, there is a reversal in overcon-
fident CEOs’ tendency to reduce dividends (see e.g. Deshmukh et al., Forthcoming). We find that
after SOX, overconfident CEOs pay significantly higher dividends. This suggests that, in conjunc-
tion with the reduction in expenditures, SOX encourages overconfident CEOs to distribute cash to
shareholders.
We also show that these SOX-related benefits are concentrated in the companies that were not
previously compliant with SOX. This result indicates that it was indeed the governance mandates
in SOX that were responsible for the documented changes. This is because the firms that complied
with the requirements of SOX before its passage, already had the governance necessary to constrain
overconfident CEOs.
We take steps to ensure the results are robust to various econometric issues. We control for
other firm , CEO , and governance characteristics, and ensure robustness to firm, year, and industry
effects. Also, we take steps to mitigate endogeneity (reverse-causality) concerns that may arise from
CEO overconfidence being affected by firm performance. This is unlikely to be a concern since,
in general, overconfident CEOs tend to perform worse, suggesting that performance does not lead
to overconfidence. Nonetheless, we confirm that overconfidence tends to be ‘sticky’ over time (as
Malmendier and Tate (2005) have also shown), suggesting that it is a behavioral characteristic rather
than a function of contemporaneous performance. The results are robust to using a press-based
measure of overconfidence. Our results are also robust to replacing our main CEO-overconfidence
measure with an indicator for whether the CEO is overconfident in any prior period, capturing the
behavioral aspect of overconfidence. Further the results are robust to using measures representing
the number of not-yet-exercised but exercisable options, or the value thereof (as per Schrand and
Zechman, 2012). Additionally, we find that the results hold when examining CEOs with “deep in the
money” options (as per the ‘Holder67’ variable in Malmendier and Tate, 2005) for whom performance
5
will have only limited impact on option value; and thus, for whom endogeneity is unlikely to be an
issue.
Our results contribute to the literature on managerial overconfidence and the literature on mar-
ket regulation. We confirm that overconfidence can lead to excessive risk-taking and expenditure.
The results provide support for exogenously mandated improvements in some governance practices.
Certainly, SOX appears to have had a positive effect in mitigating significant value-destruction at the
hands of overconfident CEOs. We believe that the paper provides novel evidence on the benefits of
SOX – and its moderating impact on CEO behavior. Though it may not extrapolate to other types
of broad governance changes that may have been proposed or enacted, in the specific case of SOX, it
appears that it acted as a restraint on CEO excesses and may have been value-enhancing regulation
from the point of view of the shareholders (and social welfare) in general.10 The message of the
paper is quite clear: overconfidence can be influenced, regulated and monitored – just like any other
managerial attribute. How best to harness overconfidence remains an interesting and open question.
There is, of course, unlikely to be a definitive answer to the question of how design regulation to
optimally balance the benefits and costs of overconfidence.
The remainder of this paper is structured as follows: Section 2 reviews the literature and develops
the hypotheses. Section 3 describes the sample selection process, variable definition, and summary
statistics of the data. Section 4 presents the results on the relations between various measures of
investment, quality of investment, and risk-taking with our main variables of interest (i.e., measure
of overconfidence, SOX and the interaction term of SOX with overconfidence). Section 5 presents
the results of robustness tests and Section 6 concludes the paper.
2 Hypotheses
Over-confident CEOs, by definition, value investments more positively than other CEOs do. These
overconfident CEOs would assign a greater expected value to their investments. Thus, this over-
confidence can induce hubris and over-investment. For example, hubris is one potential driver of
10Such evidence is consistent with prior literature that suggests that SOX prevents insiders from expropri-ating from minority shareholders (as in Duarte et al. (Forthcoming)), and is associated with improvements indisclosure and governance (see e.g. Arping and Sautner (2013); Ashbaugh-Sakife et al. (2009))).
6
over-bidding in takeovers (e.g., Roll (1986) and Hayward and Hambrick (1997)). Malmendier and
Tate (2008) argue that overconfident CEOs are more likely to undertake hubristic takeovers. By
parity of reasoning, overconfident CEOs would be more likely to spend more resources internally (i.e.
in CAPEX or asset growth). Thus, prior studies show that overconfident CEOs tend to over invest
on internal expenditures (Chen et al., 2013; Malmendier and Tate, 2005) . Similarly, there is related
evidence that CEOs who engage in increased personal risk-taking (and are presumably more prone
to overconfidence) tend to engage in increased corporate risk-taking (see e.g. Cain and McKeon,
2013). Because these overconfident CEOs over-estimate the expected value of their investments, and
under-estimate the downside risk, overconfident CEOs will be more likely to increase corporate risk
than other CEOs will.
The Sarbanes-Oxley Act of 2002 (SOX) aims to restrict managerial excesses, increase trans-
parency, and improve corporate governance. Several provisions would enable such an improvement
in governance (for a complete summary see Coates (2007)). These include having an independent
audit committee (Section 301), executive certification of financial reports (Section 302), disclosure of
managerial assessment of internal controls (Section 404), and a code of ethics for senior financial of-
ficers (Section 406). SOX also prevented accounting firms from providing auditing and non-auditing
services to the one firm and increased penalties for corporate fraud. A side-effect of these provisions
is an overall improvement in internal corporate governance . Put together, the increased environ-
ment of disclosure , and especially of monitoring, could help to restrain managerial excesses. Thus,
we hypothesize that SOX would help to constraint overconfident CEOs by increasing both internal
oversight and external scrutiny.
The evidence tends to suggest that SOX is associated with improvements in governance. While
prior evidence indicates that SOX might impose significant costs on companies (see e.g. Iliev (2010);
Leuz et al. (2008)). However, despite the potential costs of SOX, there is some evidence that it enables
better protection for minority shareholders against extraction of value by insiders (see Duarte et al.
(Forthcoming)), improvements in disclosure and governance (see e.g. Arping and Sautner (2013);
Ashbaugh-Sakife et al. (2009)), and increases in market value (see Switzer (2007)). Duarte et al.
(Forthcoming) find evidence that the way to reconcile the competing literatures (i.e. that SOX is
costly on the one hand versus the benefits of SOX on the other) is SOX imposes costs to insiders
7
by making it more difficult to extract value from minority shareholders; however, SOX creates value
for minority shareholders by enhancing governance and disclosure. Overall, the literature suggests
that SOX is generally associated with better governance and disclosure. Given that we expect that
overconfident CEOs will be likely to overinvest, and assume more risk, and that improved governance
should restrict this, we make the following hypotheses.
Hypothesis 1. SOX reduces the impact of CEO overconfidence on the amount of corporate invest-
ment.
Hypothesis 2. SOX weakens the impact of CEO overconfidence on the exposure to systematic as
well as unsystematic corporate risk.
The prediction that SOX reduces excessive risk-taking and eliminates wasteful expenditure of
overconfident CEOs implies that SOX should positively impact their firms’ operating performance
and other measures of firm valuation. We capture these predictions in the following hypotheses.
Hypothesis 3. SOX enhances the impact of CEO overconfidence on firm-value.
Hypothesis 4. SOX enhances the impact of CEO overconfidence on firm’s operating performance.
Since SOX curbs wasteful expenditure and excessive risk-taking tendencies of overconfident
CEOs, it follows that SOX helps to increases the value of the investments that they do make. For
example, Harford (1999) argues that the availability of excess cash holdings enables self-interested
CEOs to engage in self-interested, or hubristic, takeovers. SOX might help to curb such wasteful
expenditure by increasing monitoring and disclosure. Thus, we expect that SOX will enhance the
impact of CEO overconfidence on the value of major corporate investments like R&D and capital
expenditures. Subsequently, we state the following hypothesis.
Hypothesis 5. SOX enhances the value of CAPEX and R&D investment in firm managed overcon-
fident CEOs.
The impact of SOX in moderating CEO overconfidence should encourage better takeover deci-
sions. Managerial overconfidence can induce over-bidding and value-destruction in acquisitions (Kim,
Forthcoming; Kolasinski and Li, Forthcoming; Malmendier and Tate, 2008). Additionally, poor cor-
porate governance appears to facilitate such acquisitions. For example, entrenched CEOs appear
8
to make acquisitions that destroy more corporate value, implying overpayment in acquisitions (e.g.,
Masulis et al. (2007) and Harford et al. (2012)). Subsequently, given that SOX improves corporate
governance, we expect that it will reduce over-bidding in acquisitions and encourage CEOs to en-
gage in greater long-term value-creation. Kolasinski and Li (Forthcoming) provide some consistent
evidence, suggesting that a strong independent board reduces the likelihood that an overconfident
manager undertakes an acquisition. From an empirical stand-point, we are most interested in long-
term value-creation (c.f. short-run market returns) given the evidence that the market can take some
time to impound the value-implications of takeovers (e.g., Schijven and Hitt (2012) and Masulis et al.
(2013)). Thus, we make the following hypothesis.
Hypothesis 6. SOX improves the impact of CEO overconfidence on long-term value-creation in
acquisitions.
The above hypotheses suggest that SOX will help to mitigate excessive investments by overcon-
fident CEOs. Malmendier et al. (2011) argue that overconfident CEOs consider their firms under-
valued and hence, prefer not to use external financing. Thus, overconfident CEOs would prefer to
retain earnings to finance investments and as a result pay lower dividends (Deshmukh et al., Forth-
coming). We anticipate that SOX curbs over investment and other wasteful expenditures, thereby
freeing more cash for companies to pay as dividends. Subsequently, we make the following hypothesis.
Hypothesis 7. SOX will encourage overconfident CEOs to increase or initiate dividend payments.
3 Data
This study contains three-separate data-sets: a firm-year panel and an acquisition data-set. We
start with the set of companies in the Execucomp database, which we use to calculate proxies for
CEO optimism. We construct a continuous ‘CEO confidence’ variable. To do this we start with
the value-per-vested-option. We then divide this variable by the company’s stock price at the end
of the relevant fiscal year (as given in CRSP/Compustat). This gives an indication of the extent
to which the CEO retains in-the-money options that are vested. This is analogous to the variables
in Malmendier and Tate (2008). The variable differs slightly from that in Malmendier and Tate
9
(2008) because the Execucomp database does not provide the same set of variables as is available in
their database. In the main reported tests we allow the managerial overconfidence measure to vary
over time due to prior evidence that overconfidence can vary over time based upon past experience
and performance (see e.g. Billett and Qian (2008); Hilary and Menzly (2006)). We further create an
indicator variable for whether the CEO’s ‘confidence’ measure is in the top quartile of all firms in that
year. We examine a continuous variable, in addition to the indicator variable, due to prior evidence
(in Ben-David et al., Forthcoming) that many executives miscalibrate the risk/return distribution,
implying that there is a continuum of miscalibration and overconfidence. In robustness tests, we
ensure that the results are robust to various different definitions of overconfidence, including press-
based measures of overconfidence. We also use the Execucomp database to obtain other governance
variables that might influence corporate performance, including CEO tenure, CEO age, the ratio of
bonus-compensation to fixed-salary, and the CEO’s percentage ownership.
We use the firm-year panel to estimate the impact of SOX and overconfidence on firm-value,
expenditure (i.e. CAPEX and asset growth), corporate risk (beta, daily stock-return variance, and
mean squared error), and the moderating effect of SOX on the impact of overconfidence on the value
of cash holdings, CAPEX, and R&D. The firm-year panel contains all CRSP/Compustat firms for
which we have the necessary data. We start with the set of all Compustat firms and merge-in the
set of firms for which we have Execucomp data. We use Compustat to obtain fundamental firm-level
characteristics for use as control variables as appropriate in the models. We also obtain additional
data on the percentage holdings of all institutional investors from the Thomson 13f filing database.
The acquisition data-set starts with all acquisition-announcements in SDC, which we then merge
with fundamental data from Compustat, managerial optimism data (from Execucomp) and insti-
tutional ownership data (from the Thomson 13f filings). To construct this dataset we identify the
acquirer in an acquisition. We then obtain the relevant explanatory variables for the acquiring
company, including a set of control variables that are standard in the acquisition literature.
We report the sample composition by year in Table 1 and report the summary statistics in Table 2.
The statistics in Table 1 indicate that overconfidence is relatively stable over time. This is consistent
with the idea that CEO overconfidence is a behavioral trait (rather than a transient reflection of the
corporation’s position). The summary statistics in Table 2 provide some indication of the nature of
10
the sample. Panel A presents figures for the panel data sample, and Panel B presents figures for the
M&A sample. The figures in Panel B are broadly consistent with those reported in prior literature.
In particular, acquirer CARs are close to zero (for CAR(-10,10)) or slightly negative (for CAR(-42,
126)), which is consistent with prior literature (see e.g. Harford et al. (2012); Masulis et al. (2007);
Moeller et al. (2004)). The mean level of managerial confidence for the acquirers (0.38) is higher
than that for the general sample (0.31), which is consistent with prior evidence that managers who
are more confident tend to undertake more acquisitions (see e.g. Malmendier and Tate (2008)). The
following sections use these samples to conduct a multivariate analysis of effect of SOX on the impact
of managerial overconfidence.
4 Analysis
4.1 Does SOX restrain overconfident CEOs’ over-investment?
We first hypothesize that SOX prevents overconfident CEOs from over-investing. We analyze this
by examining how SOX moderates the impact of overconfidence on both CAPEX and different types
of asset growth.
4.1.1 Does SOX restrain overconfident CEOs’ capital expenditure?
We hypothesize that after the passage of SOX overconfident CEOs reduce internal corporate expen-
diture. We capture this by examining the relationship between SOX, CEO-confidence, and CAPEX.
The models have the following firm:
CAPEX/Assetsi, t+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + λj(i) + φt + εi,t
(1)
CAPEX/Assetsi, t+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + ηi + φt + εi,t
(2)
11
where, X represents a set of control variables, and φt, λj(i), and ηi are year, industry, and firm
dummies, respectively. We estimate the models using OLS regression that cluster standard errors
by firm.
The results from the regression are stated in Table 3. First we run the regression using industry
dummies, λj(i) and year dummies, φt as stated in equation (1). The regression results support
our hypothesis – we find that the coefficient on ‘Confidence’ in Column M1 of Table 3 is positive
(i.e., β(2) =+1.883) and is significant at less than 1% ; whereas, the coefficient associated with the
interaction term, ‘Confidence × SOX’ is negative (i.e., β(3) = −1.401) and is also significant at less
than 1% . Thus, these results support the view that before SOX, overconfident CEOs tend to employ
more capital relative to other CEOs in their industry and after the passage of SOX, overconfident
CEOs reduce capital expenditures quite drastically and bring it much closer to other CEOs in their
industry (= +0.483 = 1.883−1.401). Hence, we can conclude that SOX has a significant moderating
effect on capital expenditures of overconfident CEOs. We hypothesize that SOX helps to create an
atmosphere that promotes divergent opinions and candid discussions among board members and as
a result board decisions on capital expenditures more closer to the optimal and less influenced by
CEO’s behavioral biases.11
The results from models with firm dummies, ηi and year dummies, φt , as stated in equation
(1), also provide more support to our hypothesis. From Column M3 of Table 3, we find that the
coefficient on ‘Confidence’ is also positive (i.e., β(2) = 1.450) and the coefficient on ‘Confidence ×
SOX’ is also negative (i.e., β(3) = −0.912), showing that overconfident CEOs tend to employ more
capital prior to the passage of SOX but significantly reduce capital employed after the passage of
SOX (β(2) + β(3) = 1.450− 0.912 = 0.538). All these coefficients are highly significant.
We also obtain similar results when we replace our continuous measure of CEO overconfidence,
i.e., the variable ‘Confidence’, with the binary measure of the CEO’s overconfidence, the variable
‘ConfidenceTopQ’. ‘ConfidenceTopQ’ equals one if the CEO’s confidence-measure is in the top 25%
of the sample for that year; otherwise, ‘ConfidenceTopQ’ is zero. We document a set of results
using ‘ConfidenceTopQ’ in M2 and M4 of Table 3 and find results that are qualitatively similar and
consistent with the expectation that SOX restrains over-investment by overconfident CEOs.
11Lower capital expenditures is not necessarily a good thing for a firm, unless eliminated expenditures arewasteful in nature. We document the value of effect of changed CAPEX in our subsequent analysis.
12
The next issue is whether SOX induces overconfident CEOs to reduce expenses. We examine
Selling and General Administrative expenses (SG&A) following Chen et al. (2013), who argue that
overconfident CEOs can often over-spend on SG&A account. The logic being that overconfident
CEOs tend to have over-positive views about future investment-prospects, so are less likely to reduce
SG&A. The regression equations and the results from the regression in Subsection 4.1.2 below.
4.1.2 Does SOX helps to reduce overconfident CEOs’ SG&A expenses?
Chen et al. (2013) argue that overconfident CEOs can often over-spend on SG&A, reflecting their
overconfidence about the level of future demand for their products. We conjecture that SOX should
also be able to moderate such over-spending tendencies of overconfident CEOs. We capture this in
the following model:
SG&A/Salesi,t+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + λj(i) + φt + εi,t
(3)
SG&A/Salesi,t+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + ηi + φt + εi,t
(4)
where, SG&A/Sales is the firm’s sales, general, and administrative expense scaled by its total sales, X
is a vector of control variables, and φt, λj(i), and ηi are year, industry, and firm dummies, respectively.
We estimate the model using OLS and clustering standard errors by firm.
The results are stated in Table 4 and they are consistent with our conjectures. The main finding is
that the overconfident CEOs do not necessarily over-spend on SG&A – we find that the coefficients on
‘Confidence’ and ‘ConfidenceTopQ’ are not statistically significant in columns M1, M2, M3 and M4 of
Table 4. Although, the coefficients on the interaction term ‘Confidence × SOX’ and ‘ConfidenceTopQ
× SOX’ are negative, −0.013 and −0.015 for the industry fixed effect models and −0.009 and −0.007
for the firm fixed effect models respectively and are statistically significant. This suggests that SOX
makes these overconfident CEOs to try harder in their efforts to reduce costs, like SG&A. This result
is consistent both SOX restraining CEO over-spend and with the view that the passage of SOX
preventing overconfident CEOs from acting on their excessive positive view of future demand by
13
over-spending on marketing and advertising.12
Next, we try to provide some more evidences in favor of overconfident CEO’s excesses. In
particular, we look at four things: 1) growth in total asset between year t and t + 1; 2) growth in
property, plant and equipment (PPE) between year t and t+1; 3) growth in total asset between year
t+ 1 and t+ 2; and 4) growth in property, plant and equipment (PPE) between year t+ 1 and t+ 2.
We argue that overconfident CEOs can often instigate undesirable growth in total assets and as well
as undesirable growth in property, plant and equipment.13 These excesses may look innocuous but
may lead to increase in agency problems and thereby, may decrease firm value. We conjecture that
SOX helps to moderate undesirable growth in total assets as well as undesirable growth in property,
plant and equipment. We document the design and the results from these regressions in Subsection
4.1.3 below.
4.1.3 Does SOX restrain overconfident CEOs’ asset growth?
We hypothesize that after the passage of SOX overconfident CEOs reduce corporate expenditure.
We capture this by constructing a firm-year panel of data and by examining the growth rate in
assets and property plant and equipment (PPE), both between year t and t + 1 and between year
t+ 1 and year t+ 2. For brevity, we only report models that include year and firm fixed effects. In
un-tabulated models, we find that the results are robust to replacing firm fixed effects with industry
fixed effects. The models for asset growth have the following form
Asset Growthi,(t,t+1) = ln
[Assett+1
Assett
]and Asset Growthi,(t+1, t+2) = ln
[Assett+2
Assett+1
](5)
Asset Growthi,(t,t+1) = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)+
Xi,tθ + ηi + φt + εi,t,
(6)
Asset Growthi,(t+1,t+2) = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)+
Xi,tθ + ηi + φt + εi,t.
(7)
12The R-squared in the models is high. This arises because we control for lagged SG&A and we know fromChen et al. (2013) that SG&A is sticky. The R-squared are low in models that use firm dummies because firmdummies consumes a lot more degrees of freedom compared to the regressions with industry dummies.
13The reason we look at growth rates rather than on the scaled Asset or scaled PPE is because we thinkwe lack appropriate scaler for these dependent variable. Hence, instead we use growth rates of these twodependent variables.
14
Similarly, the models for property, plant and equipment growth have the following form
PPE Growthi,(t,t+1) = ln
[PPEt+1
PPEt
]and PPE Growthi,(t+1, t+2) = ln
[PPEt+2
PPEt+1
](8)
PPE Growthi,(t,t+1) = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + ηi + φt + εi,t
(9)
PPE Growthi,(t+1,t+2) = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + ηi + φt + εi,t
(10)
We estimate a panel regression using firm dummies, year dummies and we cluster standard errors
by firm. We report the regression results in Table 5. In all cases we find that the coefficient
associated with the interaction terms ‘Confidence × SOX’ (−0.087 and −0.069 , −0.042, and −0.034)
as well as the coefficients associated with the interaction term ‘ConfidenceTopQ × SOX’ (−0.042,
−0.021, −0.019 and −0.001) are all negative in sign and in all cases except Column M8 we find them
statistically significant. Thus, in the pre-SOX era overconfident CEOs tend to do excesses in growing
assets (all types of assets including property, plant and equipment) but after the passage of SOX
we find that these overconfident CEOs tend to control their tendencies to grow assets considerably.
Thus, after the passing of SOX, overconfident CEOs tend to grow assets in close harmony with other
CEOs. Hence, SOX has significant moderating effect on overconfident CEO’s ability to grow total
assets as well as investments in property, plant and equipment.
4.1.4 Sensitivity of investment to cash flows
We next examine how SOX impacts on the sensitivity of investment to cash flows. Malmendier and
Tate (2005) find that overconfident CEOs spend more of their cash flows on capital expenditures.
However, we expect SOX will restrain overconfident CEOs from over-spending. Thus, we expect
that after SOX, investment by overconfident CEOs will be less sensitive to cash flows. That is,
we expect that after SOX, overconfident CEOs will spend less of their available cash. We examine
the sensitivity of expenditure in year t to cash flow in year t within a framework similar to that in
Malmendier and Tate (2005). This type of sensitivity-model is relatively standard in the literature
(see e.g. Agca and Mozumdar (2008); Almeida et al. (2004); Brown and Petersen (2009); Fazzari
15
et al. (1988, 2000); Hovakimian (2009)).14 Specifically, we run regressions of the following form:
CAPEX/Assetsi,t = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ SOXi,t × Cash Flowi,tβ(4) + Confidencei,t × Cash Flowi,tβ
(5)
+ SOXi,t × Confidencei,t × Cash Flowi,tβ(6) + Xi,tθ + λj(i) + φt + εi,t
(11)
CAPEX/Assetsi,t = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ SOXi,t × Cash Flowi,tβ(4) + Confidencei,t × Cash Flowi,tβ
(5)
+ SOXi,t × Confidencei,t × Cash Flowi,tβ(6) + Xi,tθ + ηi + φt + εi,t
(12)
Where, ‘Cash Flow’ represents one of two measures of cash flows: EBIT/Assets and OCF/Assets,
X is a vector of control variables, and φt, λj(i), and ηi represent year, industry, and firm fixed effects,
respectively. We anticipate a negative sign on β(6), which would suggest that SOX attenuates the
tendency of overconfident CEOs to pay-out cash flows.
The results are in Table 6. Consistent with Malmendier and Tate (2005), we find a that overconfi-
dent CEOs do in deed spend more of their cash flows (i.e. we find a positive value for β(5)). However,
the coefficient on the triple interaction term, β(6), is negative and statistically highly significant in
M1, M2, M3, and M4, and is negative and mostly significant in M5-M8. This result suggests that
SOX attenuates the tendency of overconfident CEOs to payout cash flows.
4.2 Does SOX reduce overconfident CEO’ excess risk-taking ten-
dencies?
The next issue is firms’ exposure to risk – both systematic or market risk and unsystematic or firm-
specific risk – under overconfident CEOs and does SOX help to mitigate effects of overconfidence
on levels of risk exposure. We conjecture that these overconfident CEOs, as they over estimate
internal rate of returns of investment projects, are likely candidates to under-estimate the levels of
14The investment cash flow sensitivity models have received some criticism as measures of financial con-straints. However, do not use the model to measure financial constraints (see e.g. Chen and Chen (2012);Kaplan and Zingales (1997)). We use the model to analyze the amount of tendency of overconfident CEOs tospend available cash flows as per Malmendier and Tate (2005).
16
risk associated with these projects. Thus, the firms under their control is expected to load more
than desirable(or sub-optimal) amount of risk. We argue that after the passage of SOX a relatively
more independent board and/or relative more independent audit committee and/or higher mandate
for more disclosures lead to an environment where these overconfident CEOs are constrained from
completely influencing firm’s choice of risk level. Hence, we conjecture that SOX helps to mitigate
excessive risk-taking tendencies of these overconfident CEOs.
We expect that the reduction in over-investment will be associated with a reduction in risk-taking
(Hypothesis 2). We examine two different types of risk: exposure to market risk (measured by beta)
and the level of idiosyncratic/firm-specific risk (as per Low (2009)). We estimate beta by running a
single-index model over the course of the year using daily data. The proxy for idiosyncratic risk is
the mean squared error (MSE) from that single-index-model. When examining variance and MSE
we take logs in order to adjust for non-linearities. The model is of the following form
Betat+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + λj(i) + φt + εi,t
(13)
ln(MSE)t+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ Xi,tθ + λj(i) + φt + εi,t
(14)
where, X represents a set of control variables, and φt, λj(i), and ηi are year, industry, and firm
dummies, respectively. We cluster standard errors by firm.
The results are stated in Table 7. In Columns M1 to M4 we state the results for market risk
(beta) and in Columns M5 to M8 we state the results for log MSE. We find that coefficients associated
with the variable ‘Confidence’ as well as the variable ‘ConfidenceTopQ’ are all positive and highly
significant. This implies that overconfident CEOs tend to expose their firms to relatively more market
risk when compared with their industry peers and when compared to their firm’s averages. But after
the passage of SOX we find that these same overconfident CEOs tend to reduce the level of risk
exposure considerably. For example, in models M2 and M4 where we use firm and year dummies, the
coefficient associated with the interactions terms ‘Confidence ×SOX’ and ‘ConfidenceTopQ × SOX’
are both negative and statistically significant at less than one percent level (−0.176 and −0.106).
In models M1 and M3 where we use industry and year dummies, the coefficient associated with the
17
interactions term ‘Confidence ×SOX’ and ‘ConfidenceTopQ × SOX’ are again both negative and
statistically significant at less than one percent level (−0.331 and −0.180).
These results are true for even firm-specific risk. For example, in models M6 and M8 where we
use firm and year dummies, the coefficient associated with the interactions term ‘Confidence ×SOX’
and ‘ConfidenceTopQ × SOX’ are both negative and statistically significant at less than one percent
level (−0.059 and −0.038). In models M5 and M7 where we use replace firm dummies with two-digit
SIC code industry dummies, the coefficient associated with the interactions term ‘Confidence ×SOX’
and ‘ConfidenceTopQ × SOX’ are again both negative and statistically significant at less than one
percent level (−0.042 and −0.031). Hence, SOX has significant moderating effects on the risk-taking
tendencies of these overconfident CEOs.
The coefficients control variables have similar signs and significance-levels to expectations. For
example, higher Tobin’s Q values are associated with higher risk (as in Low (2009)), consistent with
the idea that Tobin’s Q reflects future growth prospects. Further, high EBIT/Assets is associated
with lower risk (as in Dutt and Humphery-Jenner (2013)), which is consistent with the idea that stable
companies with higher earnings also have lower risk. Older CEOs tend to be associated with lower
risk, consistent with prior evidence that younger CEOs tend to be more willing to undertake more
risky investments, such as acquisitions (see e.g. Yim (Forthcoming)). Unsurprisingly, institutional
ownership is associated with (weakly) significantly higher betas, reflecting the argued adherence to
market-benchmarks by institutional investors (as per Baker et al. (2011)). Illiquid shares (as proxied
by the proportion of no-return dais) tend to have lower betas and higher MSEs, consistent with the
idea that these companies tend to have higher unsystematic risk.
4.3 Does SOX encourage overconfident CEOs to make better in-
vestments?
Our analysis so far is restricted to mitigating effects of SOX on the effects of overconfidence on
the levels (as well as rates) of investments and on the levels of risk exposure of these firms. The
question that naturally arises: does SOX help overconfident CEOs to improve the quality of their
investments? We conjecture that the increased discipline associated with SOX will induce CEOs to
focus on value-creating investments. Subsequently, in this section we test the overarching hypothesis
18
that SOX causes overconfident CEOs to make better quality investments. We use several measures
of firm performance. First, we use a market based measure of firm performance, namely Tobin’s
Q. Next, we use an accounting measure of performance, Earnings Before Interest & Taxes (EBIT).
We also examine industry-adjusted measures of performance. Lastly, we use the Standard & Poor’s
Index of Earning Quality. We document our research design and our regression results in Subsections
4.3.1 , 4.3.2 and 4.3.3 below.
4.3.1 Is SOX associated with value-increases in overconfident CEOs’ companies?
We expect that SOX encourages overconfident CEOs to create more corporate value (Hypothesis
3), improve operating performance (as measured by earnings), and enhance the quality of those
earnings (Hypothesis 4). We use Tobin’s Q and industry-adjusted Tobin’s Q as proxies of firm-value
(as per Bebchuk et al. (2009); Gompers et al. (2003)). The proxies for operating performance are
the firm’s EBIT/Assets and industry-adjusted EBIT/Assets (following Powell and Stark (2005)).
The industry-adjusted Q (or EBIT/Assets) is the firm’s Q (or EBIT/Assets) less the average Q (or
EBIT/Assets) for all firms in its two-digit industry and year.
We examine the impact of SOX and overconfidence on firm-value by constructing a firm-year
panel of all companies in Compustat that have the necessary data. We run models with two-digit
SIC code based industry dummies and year dummies and also models with firm and year dummies.
We cluster standard errors by firm. We also examine the moderating role of SOX on the impact of
overconfidence on earnings quality. We obtain each company’s S&P earnings quality variable from
Compustat (Compustat code: spcsrc). The earnings quality variable ranks the firm’s quality from
‘A+’ through to ‘D’, with ‘A+’ being the highest. We re-code the original earnings quality variable to
be a numerically ordered variable from 1 through to 8 and run an ordered logit model as in Equation
19
(17). The model are of the following form
Performancei, t+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ xi,tθ + λj(i) + φt + εi,t
(15)
Performancei, t+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ xi,tθ + ηi + φt + εi,t
(16)
Earnings Qualityt+1 = α+ SOXi,tβ(1) + Confidencei,tβ
(2) + SOXi,t × Confidencei,tβ(3)
+ xi,tθ + λj(i) + φt + εi,t
(17)
where, ‘Performance’ refers to either (a) the firm’s Tobin’Q or industry-adjusted Tobin’s Q, or (b) the
firm’s EBIT/Assets or industry-adjusted EBIT/Assets, and ‘Earnings Quality’ refers to the firm’s
S&P earnings quality rating. Equation (15) is an OLS regression with industry and year dummies.
Equation (16) is an OLS regression with firm and year dummies. Equation (17) is an ordered logit
model with industry dummies and year dummies. All models cluster standard errors by firm.
The results for models that examine Tobin’s Q are in Table 8. What we find? CEO overconfidence
has no effect on Tobin’s Q prior to the passage of SOX. The coefficient associated with the variable
‘Confidence’ as well as ‘ConfidenceTopQ’ are not statistically different from zero.15 But after the
passage SOX CEO overconfidence starts influencing firm performance for the better. We find that
the coefficients associated with the interaction terms ‘Confidence ×SOX’ and ‘ConfidenceTopQ ×
SOX’ are all positive and statistically significant. For example, after the passage of SOX, in the
model with firm dummies and year dummies we obtain coefficient of ‘Confidence’ equal to +0.239 =
0.094 + 0.145 when we consider just Tobin’s Q (i.e., Column M2). When we consider industry-
adjusted Tobin’s Q (i.e., Column M4), we obtain that the coefficient associated with ‘Confidence’ is
equal to +0.218 = 0.062 + 0.156. We get similar effects when we replace our continuous measure of
overconfidence with the discrete measure, ‘ConfidenceTopQ’.
The results for the EBIT models are in Table 9 and are consistent with our hypotheses. We
find that CEO overconfidence per se do not influence earnings. All the coefficients, columns M1-M8,
associated with our ‘Confidence’ and ‘ConfidenceTopQ’ are not statistically different from zero. But
15There is only one exception – in Column M2 where the coefficient β(2) is positive and slightly significant.
20
after the passage of SOX we find that CEO overconfidence starts influencing earning in a positive
way. We find that the coefficients associated with the interaction terms ‘Confidence ×SOX’ and
‘ConfidenceTopQ × SOX’ are all positive and highly statistically significant. For example, after
the passage of SOX, in the model with firm dummies and year dummies we obtain coefficient of
‘Confidence’ equal to +0.021 = 0.022 − 0.001 when we consider EBIT over Assets (i.e., Column
M2). When we consider industry-adjusted EBIT over Assets (i.e., Column M4), we obtain that the
coefficient associated with ‘Confidence’ is equal to +0.022 = 0.021+0.001. We get similar effects when
we replace our continuous measure of overconfidence with the discrete measure, ‘ConfidenceTopQ’.
This suggests that SOX helps to redirect overconfident CEOs towards investments that create more
shareholder wealth.
The earnings-equality models in Table 10 support the results in Table 9. The dependent variable is
a discrete ordered variable that represents the company’s S&P earnings quality. A positive coefficient
on a variable indicates that it is associated with higher earnings quality. What do we find? CEO
overconfidence weakly but negatively affects earning quality. After the passage of SOX, overconfident
CEOs tend to use their overconfidence to improve the earnings quality of their firm. For example, in
both models M1 and M2, the coefficient associated with the interaction terms, ‘Confidence ×SOX’
and ‘ConfidenceTopQ × SOX’ are 0.831 and 0.368 respectively – both positive and highly significant.
The results suggest that SOX positively moderates the impact of CEO confidence on earnings quality.
Thus, results from tables 9 and 10 support the view that SOX encourages overconfident CEOs to
both increase earnings and increase the quality of those earnings.
The coefficients on the control variables are largely consistent with expectations. Size signifi-
cantly reduces Q and EBIT/Assets, suggesting the potential presence of reduced growth prospects
or diseconomies of scale (as per Moeller et al. (2004)). Stock return standard deviation is associated
with lower market values (i.e. Q) and earnings (consistent with the ‘low volatility effect’ in Baker
et al. (2011) and Dutt and Humphery-Jenner (2013)). The CEO’s Bonus/Salary is positively re-
lated to Q and EBIT, suggesting that incentive compensation can help encourage CEOs to improve
corporate performance (see e.g. Jensen and Murphy (1990)). R&D is positively associated with Q
but negatively associated with EBIT, highlighting that high-R&D firms tend to be smaller and more
innovative, leading to higher growth prospects (as impounded in Q) but potentially lower earnings.
21
Next, we study the impact of CEO overconfidence after the passage of SOX on values of R&D
and CAPEX. This is important given our result that post SOX overconfident CEOs reduce CAPEX
significantly. Hence, the natural question that arises: does SOX help to eliminate relatively wasteful
expenditures? We do this analysis using a triple interaction term ‘Confidenc×SOX’ times either
CAPEX/Sales or R&D/Sales. Also, we run separate regressions for the pre-SOX and the post-SOX
periods and observe the sign and significance of the double interaction term between ‘Confidence’
and either CAPEX/Sales or R&D/Sales. We document our research design and our regression results
in Subsection 4.3.2 below.
4.3.2 Does SOX positively influence the impact of overconfidence on the value
of R&D and value of CAPEX?
We expect that SOX will positively moderate the impact of managerial overconfidence on the value
of CAPEX and R&D (Hypothesis 5). These models are similar to Equation (15) and Equation (16)
but with the addition of a triple interaction term capturing the interaction of ‘SOX’ and ‘Confidence’
with either CAPEX/Sales or R&D/Sales. For brevity, we only report models that include firm fixed
effects and year fixed effects and cluster standard errors by firm (rather than models that include
industry fixed effects).
The results in Table 11 support the hypothesis that SOX positively moderates the impact of
confidence on the value of R&D (in Panel A) and CAPEX (in Panel B). The triple interaction terms
‘Confidence × SOX × R&D/Sales’ and ‘ConfidenceTopQ × SOX × R&D/Sales’ are 1.910 and 1.941
respectively – both positive and statistically significant at less than 1% level. Further, there are
economically significant differences between the impact of CEO confidence on the value of R&D
between the pre-SOX and post-SOX periods. For example, coefficients of ‘Confidence× R&D/Sales’
is negative (−2.126) and highly significant in the pre-SOX period whereas the same coefficient is
positive though insignificant in the post-SOX period. Overall, this suggests that SOX significantly
moderates the impact of overconfidence on the value of R&D.
The results in Table 11 (Panel B) indicate that SOX positively moderates the impact of CEO
confidence on the value of CAPEX. As with Panel A, the key variables of interest are the triple
interaction terms in Columns M1 and M4. Both triple interaction terms are positive and significant
22
at less than 3% level. Further, whereas CEO confidence significantly negatively moderates the value
of CAPEX in the pre-sox period (i.e. Columns M2 and M5) it insignificantly positively moderates
its value in the post-SOX period (i.e. Columns M3 and M6). These results are consistent with the
idea that SOX encourages overconfident CEO’s to focus on value-creating capital expenditures.
Next, we look at acquisitions by overconfident CEOs. We look at announcement day return,
both short and long windows and we also look at value to 1-year buy-and-hold strategy. Further, we
look at industry-adjusted performance measures like Tobin’s Q and EBIT over assets. We document
our research design and our regression results in Subsection 4.3.3 below.
4.3.3 Does SOX increase the value-created by overconfident CEOs’ acquisitions?
We expect that SOX will encourage overconfident CEOs to create more value (or, at least, destroy
less value) in acquisitions (Hypothesis 6). We examine this by collecting data on all acquisitions
made by all firms for which we have the necessary data on executive compensation and governance.
The acquisition must occur between 1992 and 2011 to appear in the sample.
We start with an examination of the cumulative abnormal returns (CARs) from 60 days before the
acquisition to 128 days after the acquisition. The CARs are based an OLS estimation of the market
model from 125 days to 375 days before the acquisition announcement. Figure 1 plots the CARs. The
figure reveals that there is a significant negative pre-announcement run-up and post-announcement
decline for acquisitions by overconfident CEOs. The decline is less negative for overconfident CEOs
after SOX than it is before SOX. The pre-announcement and post-announcement returns are largely
consistent with prior studies, documenting a relatively low cumulative abnormal return for acquirers
on average (as per Duchin and Schmidt, 2013; Humphery-Jenner and Powell, 2011). We obtain
similar results if we look at the sub-set of acquisitions of public targets (see Figure 2) and of non-
public targets (see Figure 3), with acquisitions of publicly listed targets generally performing worse
(as per Chang (1998); Fuller et al. (2002)). This figure does not control for other firm-level and
deal-level factors that might drive acquisition performance, which leads us to conduct multivariate
tests.
We examine the long-run post-acquisition performance, as proxied by post-acquisition BHAR,
and industry-adjusted Tobin’s Q and industry-adjusted EBIT/Assets, and their industry-adjusted
23
counter-parts (as in Healy et al. (1992), Powell and Stark (2005), and Harford et al. (2012)).16 The
industry-adjusted values are the firm’s value less the value for all companies in the firm’s two-digit
SIC industry and year. We control for factors that are standard in the literature for models that
examine long-run post-takeover performance. For an acquisition that is announced in year t, we run
a model of the following form:
Performancet+1 = α+ SOXi,tβ(1) + Confidencei,t−1β
(2) + SOXi,t × Confidencei,t−1β(3)
+ Xi,tθ + λj(i) + φt + εi,t
(18)
where, the vector of controls, Xi,t contains a set of standard control variables, which are lagged as
appropriate to ensure that they pre-date the acquisition announcement. We include year dummies
and industry dummies, to account for the documented industry and time-effects in mergers (e.g.,
Powell and Yawson (2007), Powell and Yawson (2005), Harford (2005), Ovtchinnikov (2013)) and
cluster standard errors by firm.
The results for short-horizon windows are in models M1 and M2 in Table 12. The dependent
variable is the acquirer’s abnormal return on announcing the takeover. As in Malmendier and Tate
(2005), CEO overconfidence is negatively related to acquisition-performance. The results do not
suggest that SOX significantly changes the impact of overconfidence on acquisition returns. However,
given that Figure 1 suggests that the negative returns manifest themselves over a longer time horizon
in our sample, this is not surprising.
The results in relation to long-horizon market-performance are in models M3-M6 of Table 12
are consistent with the idea that SOX increases the value created by overconfident CEOs’ acquisi-
tions. CEO overconfidence is negatively related to long-term value creation, as proxied by BHARs.
However, SOX positively moderates the relationship between CEO overconfidence and long-term
value-creation from acquisitions. The results in relation to long-run operating performance (models
M7 and M8) and long-run value (models M9 and M10) support the BHARs-results. They indi-
cate that while overconfident CEOs are associated with significantly lower post-acquisition operating
returns and market values, SOX helps to mitigate the impact of CEO overconfidence.
16We focus on long-term performance due to evidence that it can take some time for the market to fullyimpound the value created by a takeover. See, for example, Masulis et al. (2013) for more details.
24
The foregoing analysis points to the fact that, after the passage of SOX, firms run by overconfident
CEOs cut their investments (CAPEX, PPE, etc.) and at the same time improve the quality of their
investments. Hence, these firms are likely to accumulate relatively high free cash flows. This begs
the question of whether these overconfident CEOs hold on to these liquid assets or they disburse it
to the rightful owners in the form of dividends. Next, we analyze the effects of SOX on dividend
policies of the overconfident CEOs. We document our research design and our regression results in
Subsection 4.4 below.
4.4 Does SOX cause overconfident CEOs to increase dividend pay-
out?
We expect that if overconfident CEOs reduce expenditure and also due to the additional discipline
imposed by SOX on CEOs, we expect these overconfident CEOs to payout the excess cash as div-
idends (Hypothesis 7). This follows from prior evidence that overconfident CEOs tend to reduce
dividends because they overinvest, and want to maintain the flexibility to continue to overinvest
(Deshmukh et al., Forthcoming). We test this hypothesis in the following framework:
Dividendsi,t+1 = α+ SOXi,tβ(1) + Confidencei,t−1β
(2) + SOXi,t × Confidencei,t−1β(3)
+ Xi,tθ + λj(i) + φt + εi,t
(19)
Dividendsi,t+1 = α+ SOXi,tβ(1) + Confidencei,t−1β
(2) + SOXi,t × Confidencei,t−1β(3)
+ Xi,tθ + ηi + φt + εi,t
(20)
where, ‘Dividends’ represents the firm’s total dividend payout scaled by its assets, X is the set of
control variables as in the prior regressions, and ηi, λj(i), and φt represent firm, industry, and year
dummies, respectively.
The results are in Table 13 and are consistent with our conjectures. We find that the coeffi-
cients on the overconfidence measures, ‘Confidence’ and ‘ConfidenceTopQ’ are −0.232 and −0.124
in Columns M2 and M4 which depicts models with firm and year dummies. This is consistent with
the findings documented in Deshmukh et al. (Forthcoming) that overconfident CEOs prefer to limit
25
dividends payments. But we also find that the coefficients associated with the interaction terms
‘Confidence × SOX’ and ‘ConfidenceTopQ × SOX’ are 0.284 and 0.147 in models where we control
for firm fixed effects. The results are similar when we replace the firm dummies with two-digit SIC
code based industry dummies. These results suggest that after the passage of SOX overconfident
CEOs pay significantly higher dividends.
The coefficients on the controls are largely as expected. Large firms and firms with higher earnings
pay higher dividends (as in Allen et al. (2012); Li and Zhao (2008); Mansi and Wald (2011)). Riskier
firms (i.e. those with a higher return standard deviation) pay lower dividends (as per Li and Zhao
(2008); Mansi and Wald (2011)). Unsurprisingly, firms with higher proportion of intangibles-to-assets
pay lower dividends (as in Allen et al. (2012)), likely reflecting the possibility that such firms reinvest
more cash flow in growth, leaving less cash to pay to shareholders.
It is important to note one caveat with these results: In 2001, the U.S. government reduced the
personal tax payable on dividends, potentially making dividends a more favorable way for companies
to return cash to shareholders. We address this caveat in Section 5.3 .
5 Additional robustness tests and alternative explana-
tions
We take steps to address alternative explanations and to mitigate concerns about the robustness of
the models.
5.1 Voluntary Compliance, SOX and Overconfidence
First, we consider an important experiment. We argue that SOX mandates more independent think-
ing within the boardroom as well as within the audit committee. This independent thinking helps to
provide alternative contingencies to future planning and, such alternatives helps to reduce errors and
biases in the firm’s decision making processes. Thus, SOX leads to improvement in the performance
of firms run by overconfident CEOs.
If our chain of argument is true, it should also imply that the set of companies who voluntarily
complied with the requirements of SOX, even before SOX was even perceived by the policymakers,
26
should not experience the same level of moderation as well as performance enhancements as the
non-compliant firms. That is, the relevant interaction terms should not be statistically significant for
the set of companies that were voluntarily compliant with the requirements of SOX in the pre-SOX
period.
We test this by examining two definitions of compliance: (1) having at lest 50% independent
directors; and (2) having a completely independent audit committee. We then create four categories
for firms that satisfy (1) and (2) individually, either (1) or (2), or both (1) and (2), in all of 1998-
2001.17 We re-run the tests for this set of companies as a ‘placebo test’.
The results are in Table 15. For brevity, we do not report all results. We report models that
use ‘ConfidenceTopQ’ as the measure for CEO overconfidence (however, the results are qualitatively
similar if we use ‘Confidence’ instead). The key finding is that the relevant interaction terms, i.e.,
ConfidenceTopQ×SOX, is statistically insignificant in almost all models that we re-run. This implies
that SOX did not moderate the impact of overconfidence in the set of voluntarily compliant firms.
This result provides some confidence that the SOX-effect in the main models does capture the impact
of SOX rather than a spurious contemporaneous change.
5.2 Placebo test - the impact of SOX on less confident managers
We conduct placebo tests to examine the impact of SOX on less overconfident managers. We do this
by creating an indictor for whether the ‘Confidence(t)’ measure is in the bottom quartile of all firms
in that year, which we denote ‘ConfidenceBottomQ(t)’. We then re-run the models by replacing
the confidence measures with the ‘ConfidenceBottomQ(t)’. We also run tests in which we further
exclude firms with confidence in the top quartile from the control sample (in order to ensure that we
are not just looking at the mirror image of highly overconfident managers). The results (unreported
for brevity) indicate that SOX is not associated with a reduction in investment or risk taking, and is
not associated with improvements in value, acquisition performance, earnings, or dividends. That is,
the SOX-effect observed for high-confidence managers is not observed for low-confidence managers.
17We require compliance in all four years because firms who become compliant in 2001 likely do so inanticipation of SOX.
27
5.3 Dividends and Bush’s Tax Cut
The dividend-related results are unlikely to attributable to contemporaneous tax-changes. In 2001,
Bush’s government lowered the tax rate that shareholders paid on dividends. This could arguably
induce low-dividend companies to increase dividends, thereby raising the concern that our reported
results merely reflect the impact of the tax-cut rather than of SOX. There are two responses to this
issue: First, the aforementioned tests that examine the impact of SOX on overconfident CEOs in
compliant firms suggest that SOX only impacted non-compliant firms (as expected), implying that
the results are likely to reflect the impact of SOX and are unlikely to merely reflect contemporaneous
tax cuts. Second, the dividends-related results are qualitatively similar if we restrict the sample to
the set of firms whose dividends were in the top 50% of the sample in 2001 – i.e., firms who were
likely to already be catering to dividend-demanders; and thus, who were less likely to respond to the
tax cut.
5.4 Alternative Measures of CEO Overconfidence
Our results are not sensitive to the definition of CEO overconfidence. The results are robust to
using a press-based measure of overconfidence. Prior studies have used press-based measures of
overconfidence, usually based on a comparison of the number of articles that report a CEO as being
confident with those that report the CEO as being non-overconfident (see e.g. Hirshleifer et al., 2012;
Shu et al., 2013). We follow a similar method and construct a ‘Net News’ measure, which is equal
to the number of articles that report the CEO as confident less the number that report the CEO as
non-overconfident. We obtain news articles by conducting a Factiva keyword search (as in Hirshleifer
et al., 2012) of articles in the New York Times, US Today, Business Week, and Wall Street Journal.
We have this data for the years 2000, 2004, 2006. We run the analysis between the years 2000-2006
and back-fill data for the years in which we are missing news-articles. We report results for a subset
of our results in Table 14; for brevity we do not report all results. The important finding is that the
results are qualitatively similar to those in the main model, and indicate that the results are robust
to alternative measures over overconfidence.
We also examine several other ways of constructing overconfidence measures. The results hold if
we use cut-offs other than the top quartile to identify the highly overconfident CEOs. That is, if we
28
examine CEOs with confidence measures in the top 50% through to top 10%. Further, the results
are robust to an alternative definition of overconfidence being the value of all the CEO’s unexercised
exercisable options scaled by his/her salary or total compensation. The results are also robust to
using a dummy variable that equals one if the CEO’s confidence-measure exceeds 67%, which is
analogous to the ‘Holder67’ variable in Malmendier and Tate (2005). Additionally, the results are
robust to replacing the ‘Confidence’ measure with the natural log of the total value of the CEO’s
unexercised, but exercised, options (as per Schrand and Zechman (2012) and similar to Li et al.
(2012)).
5.5 Addressing the issue of the impact of market returns
One concern is that the reported overconfidence measures are based on option prices; and thus, are
sensitive to the performance of the market. We take steps to address this in several ways. First, the
reported models also use the ‘ConfidenceTopQ(t)’ indictor, which indicates whether the confidence
measure is in the top quartile in year t. Given that this measure ranks firms within each year and
that all firms are exposed to market forces, this variables helps to partially mitigate the concern
that the results merely reflect changes in the value of options due to changes in market conditions.
Second, as indicated above, the results are robust to measuring CEO confidence by using the number
of vested, but unexercised, options (rather than its value). Third, the reported results control for the
firm’s stock market performance. Third, the reported models include year dummies, which control
for market-wide effects.
5.6 Robustness to changes in sample composition
We take steps to mitigate concerns that the possibility that the results reflect changes in sample
composition and/or improvements in Execucomp’s report. We do this by examining the sub-sample
of observations from 1998-2006 for which the company is in the database for all of 1998-2006 (i.e. a
sample that does not change over the tight window surrounding SOX). For these tests, the results
are qualitatively similar the the reported results.
29
5.7 Addressing the Issue of Endogeneity
We take some steps to mitigate endogeneity concerns. One issue is that the variable, as constructed,
reflects how in-the-money the CEO’s options are. This would be related to strong future prospects.
This potential source of endogeneity would actually bias against finding the results that we obtain.
This is because the core impact of overconfidence on corporate value is negative. Whereas, if en-
dogeneity were an issue, one would expect a positive relationship between the CEO overconfidence
measure and firm-value. Nonetheless, we address endogeneity issues in several ways:
1. We construct the aforementioned Holder67-type variable, which equals one if the CEO’s con-
fidence exceeds 67%. Such levels of confidence are associated with deep in the money options,
for which an increase in risk will lead to only a minimal increase in option value. Thus, this
set of CEOs is unlikely to engage in risk-taking purely for the reason of increasing their option
value. The results hold if we replace our confidence-measures with this Holder67-type measure,
suggesting that endogeneity does not drive the results.
2. The results hold if we construct dummy variables that equal one if the CEO’s overconfident
measure, ‘ConfidenceTopQ’, equaled one in any prior year, which we call ‘PriorTopQ’. The
interpretation of ‘PriorTopQ’ is that it reflects overconfidence as a behavioral trait of the CEO.
We also construct an variable ‘AnytimeTopQ’, which equals one if the CEO’s ‘ConfidenceTopQ’
variable equals one in any year (either before or after the present year). This variable works on
the assumption that overconfidence is a behavioral attribute that can manifest itself after the
present year, even if the CEO does not currently appear to be overconfident. When using these
results, we focus on models that include year fixed effects and industry fixed effects (rather
than firm fixed effects) due to the firm-time-invariant nature of ‘AnytimeTopQ’. The results
hold if we replace the confidence variables with ‘AnytimeTopQ’ or ‘PriorTopQ’.
3. The results are robust to using further lags of the overconfidence measure (in the reported
results, the overconfidence measures date from year t while the outcome measures date from
year t+ 1).
4. The results are also robust to using the natural log of the number of unexercised exercisable
options (rather than their value), which would arguably be less subject to endogeneity concerns.
30
5.8 CEO age
One issue is that there can be a significant relationship between age and risk-taking (see e.g. Kim,
2013; Waelchi and Zeller, 2013). The reported models control for CEO age. However, we also
find that the results are qualitatively robust to splitting the sample into halves based on CEO age,
suggesting that mere CEO age does not drive the results.
5.9 Entrenchment
The results are robust to controlling for managerial entrenchment. Prior literature indicates that
managerial entrenchment, as proxied by having a preponderance of anti-takeover provisions, can
insulate managers from the market for corporate control and enable managers to make self-interested,
or hubristic, investments (see e.g. Bebchuk et al. (2009); Gompers et al. (2003); Harford et al. (2012);
Masulis et al. (2007)). However, other evidence suggests that there might be endogeneity between
entrenchment and performance (Core et al. (2006); Lehn et al. (2007); O’Connor and Rafferty (2012)).
Further, requiring data on managerial entrenchment (from IRRC/Risk Metrics) significantly reduces
the sample-size. Subsequently, in robustness tests we examine models where we control for the state-
average Bebchuk et al. (2009) entrenchment-index and the firm’s entrenchment index . We examine
the state-average values on grounds that the degree of entrenchment that is allowed depends on
the laws of the firm’s state-of-incorporation, and there prior literature suggests that there might be
a race-to-the-bottom in corporate governance (McCahery and Vermeulen (2005)). The results are
qualitatively robust to controlling for entrenchment and are untabulated for brevity.
5.10 Time from IPO
The results are robust to excluding companies that recently IPO-ed. One concern is that CEOs in
newly public often obtain options at the IPO issue-price rather than the first-day-close price (see
e.g. Lowry and Murphy (2007)). This would enable them to benefit from underpricing, and would
the presence of deeply in the money options and the appearance of overconfidence. This situation
should not undermine the reported results as such a CEO would rationally exercise the deep in the
money options just as would a CEO of an established firm. Thus, retaining these options would still
31
connote overconfidence. Nonetheless the results, which are untabulated for brevity, are qualitatively
consistent with the reported results.
5.11 Other robustness tests
The results are robust to industry and time effects. In the reported models, we include two-digit
SIC dummies and year dummies, or firm and year dummies. All models cluster standard errors by
firm. The results are robust to using any level of industry dummy (or none at all), to excluding
year-dummies, or to clustering by industry or year instead of by firm.
The results are not due to any one time period. We include year fixed effects, which capture
time effects. However, the results hold if we look at smaller windows around SOX in 2002, holding
when we restrict the sample to 1998-2004 or 1999-2003. The results are also qualatively similar if we
omit the tech-crash years of 2000 and 2001. The results are also robust to omitting high-tech firms
from the sample. Additionally, the results are robust to omitting the financial crisis years (of 2007
onwards).
The results are robust to excluding firms who experience a CEO changeover around SOX in 2002
(i.e. for whom the CEO in 2001 is different from the CEO in 2003). Thus, the results do not merely
reflect a mechanical change in overconfidence owing to a change in CEO.
The M&A results are robust to controlling for the presence of ‘large loss’ and ‘large gain’ deals.
Outlying ‘large loss’ deals can account for a significant portion of value-destruction in acquisitions
(see Moeller et al. (2005)). Conversely, ‘large gain’ deals account for a significant portion of value-
creation in acquisitions (see Fich et al. (2012)). We address the presence of such outlying deals
by ensuring that the results hold in quantile regressions (i.e. regressions through the median) and
robust regressions and to omitting from the sample acquisitions whose ‘performance’ variable (i.e.
CAR, BHAR, Q, or EBIT/Assets, as applicable) is in the top 1% or bottom 1% of the sample.
6 Conclusion
Two contentious issues in the literature are how to restrain CEO overconfidence and whether SOX
has conveyed tangible governance benefits. CEO overconfidence can convey benefits and disbenefits.
32
The benefits of overconfidence include improvements in innovation (see e.g. Hirshleifer et al. (2012)).
Conversely, CEO overconfidence can be associated with over-investment and risk-taking (see e.g.
(Malmendier and Tate, 2005, 2008)). We hypothesize that, improving internal governance and dis-
closure can help to constrain overconfident CEOs. We use SOX as a natural experiment in which
there is an improvement in governance and disclosure. We find that SOX reduces over-investment
and risk-taking by overconfident CEOs. Further, SOX positively moderates the impact of overconfi-
dence on firm-value, earnings, earnings-quality, the value of R&D, and the value of CAPEX. Further,
after SOX, overconfident CEOs’ acquisitions create significantly more value (or at least destroy sig-
nificantly less value. We also find evidence that SOX is associated with increases in dividends by
overconfident CEOs. Overall, these results suggest that improvements in governance and disclosure,
as mandated by SOX, help to mitigate the potentially harmful aspects of CEO overconfidence.
References
Agca, S., Mozumdar, A., 2008. The impact of capital market imperfections on investment–cashflow sensitivity. Journal of Banking and Finance 32 (2), 207–216.
Ahmed, A. S., Duellman, S., 2013. Managerial overconfidence and accounting conservatism.Journal of Accounting Research 51 (1), 1–30.
Allen, L., Gottesman, A., Saunders, A., Tang, Y., 2012. The role of banks in dividend policy.Financial Management 41 (3), 591–613.
Almeida, H., Campello, M., Weisbach, M. S., 2004. The cash flow sensitivity of cash. Journalof Finance 59 (4), 1777–1804.
Arping, S., Sautner, Z., 2013. Did the Sarbanes-Oxley Act of 2002 make firms less opaque?evidence from analyst earnings forecasts. Contemporary Accounting Research , Forthcom-ing.
Ashbaugh-Sakife, H., Collins, D. W., Kinney, W. R., Lafond, R., 2009. The Effect of SOXInternal Control Deficiencies on Firm Risk and Cost of Equity. Journal of AccountingResearch 47 (1), 1–43.
Baker, M., Bradley, B., Wurgler, J., 2011. Benchmarks as limits to arbitrage: Understandingthe low-volatility anomaly. Financial Analysts Journal 67 (1), 40–54.
Bebchuk, L., Cohen, A., Ferrell, A., 2009. What Matters in Corporate Governance? Reviewof Financial Studies 22 (2), 783–827.
Ben-David, I., Graham, J. R., Harvey, C. R., Forthcoming. Managerial miscalibration. Quar-terly Journal of Economics.
33
Billett, M. T., Qian, Y., 2008. Are Overconfident CEOs Born or Made? Evidence of Self-Attribution Bias from Frequent Acquirers. Management Science 54 (6), 1037–1051.
Brown, J. R., Petersen, B. C., 2009. Why has the investment-cash flow sensitivity declinedso sharply? rising r&d and equity market developments. Journal of Banking and Finance33 (5), 971–984.
Burg, V., Pierk, J., Scheinert, T., 2013. Managerial Overconfidence and Accounting Behaviorfollowing CEO Turnover. Working Paper, Humboldt University of Berlin.URL http://dx.doi.org/10.2139/ssrn.2244870
Cain, M. D., McKeon, S. B., 2013. CEO Personal Risk-Taking and Corporate Policies. WorkingPaper.URL http://dx.doi.org/10.2139/ssrn.1785413
Chang, S., 1998. Takeovers of privately held targets, method of payment, and bidder returns.Journal of Finance 53 (2), 773–784.
Chen, C. X., Gores, T., Nasev, J., 2013. Managerial overconfidence and cost stickiness. Work-ing Paper, University of Illinois - UC.
Chen, H., Chen, S., 2012. Investment-cash flow sensitivity cannot be a good measure offinancial constraints: Evidence from the time series. Journal of Financial Economics 103 (2),393–410.
Coates, J. C., 2007. The Goals and Promise of the Sarbanes-Oxley Act. Journal of EconomicPerspectives 21 (1), 91–116.
Core, J. E., Guay, W. R., Rusticus, T. O., 2006. Does weak governance cause weak stockreturns? an examination of firm operating performance and investor’s expectations. Journalof Finance 61 (2), 655–687.
Deshmukh, S., Goel, A. M., Howe, K. M., Forthcoming. CEO Overconfidence and DividendPolicy. Journal of Financial Intermediation.
Dittrich, D. A. V., Guth, W., Maciejovsky, B., 2005. Overconfidence in investment decisions:An experimental approach. European Journal of Finance 11 (6), 471–491.
Duarte, J., Kong, K., Young, L. A., Siegel, S., Forthcoming. The Impact of the Sarbanes-Oxley Act on Shareholders and Managers of Foreign Firms. Review of Finance.URL http://dx.doi.org/10.2139/ssrn.1062641
Duchin, R., Schmidt, B., 2013. Riding the merger wave: Uncertainty, reduced monitoring,and bad acquisitions. Journal of Financial Economics 107 (1), 69–88.
Dutt, T., Humphery-Jenner, M., 2013. Journal of Banking and Finance 37 (3), 999–1017.
Fazzari, S., Hubbard, R., Petersen, B., 1988. Financing constraints and corporate investment.Brookings Papers on Economic Activity 1, 141–195.
34
Fazzari, S., Hubbard, R., Petersen, B., 2000. Investment-cash flow sensitivities are useful: acomment on Kaplan and Zingales. Quarterly Journal of Economics 115, 695–705.
Ferris, S. P., Jayaraman, N., Sabherwal, S., Forthcoming. CEO Overconfidence and Interna-tional Merger and Acquisition Activity. Journal of Financial and Quantitative Analysis.
Fich, E. M., Nguyen, T., Officer, M., 2012. Large wealth creation in mergers and acquisitions.Working Paper.
Fuller, K., Netter, J., Stegemoller, M., 2002. What do returns to acquiring firms tell us?evidence from firms that make many acquisitions. Journal of Finance 57 (4), 1763–1794.
Galasso, A., Simcoe, T. S., 2011. CEO overconfidence and innovation. Management Science57 (8), 1469–1484.
Geravis, S., Heaton, J., Odean, T., 2011. Overconfidence, Compensation Contracts, and Cap-ital Budgeting. Journal of Finance 66 (5), 1735–1777.
Goel, A. M., Thakor, A. V., 2008. Overconfidence, ceo selection, and corporate governance.Journal of Finance 63 (6), 2737–2784.
Gompers, P., Ishii, J., Metrick, A., 2003. Corporate Governance and Equity Prices. QuarterlyJournal of Economics 118 (1), 107–155.
Hall, B., Murphy, K., 2000. Optimal exercise prices for executive stock options. AmericanEconomic Review 90, 209–214.
Hall, B., Murphy, K., 2002. Stock options for undiversified executives. Journal of Accountingand Economics 33, 3–42.
Harford, J., 1999. Corporate cash reserves and acquisitions. Journal of Finance 54 (6), 1969–1997.
Harford, J., 2005. What drives merger waves? Journal of Financial Economics 77, 529–560.
Harford, J., Humphery-Jenner, M. L., Powell, R. G., 2012. The sources of value destructionin acquisitions by entrenched managers. Journal of Financial Economics 106 (2), 247–261.
Hayward, M., Hambrick, D., 1997. Explaining the premiums paid for large acquisitions: Evi-dence of ceo hubris. Administrative Science Quarterly 42 (1), 103–127.
Healy, P., Palepu, K., Ruback, R., 1992. Does corporate performance improve after mergers?Journal of Financial Economics 31, 135–175.
Hilary, G., Menzly, L., 2006. Does Past Success Lead Analysts to Become Overconfident?Management Science 52 (4), 489–500.
Hirshleifer, D., Low, A., Teoh, S. H., 2012. Are Overconfident CEOs Better Innovators?Journal of Finance 67 (4), 1457–1498.
35
Hovakimian, G., 2009. Determinants of investment cash flow sensitivity. Financial Manage-ment 38 (1), 161–183.
Humphery-Jenner, M. L., Powell, R. G., 2011. Firm size, takeover profitability, and the effec-tiveness of the market for corporate control: Does the absence of anti-takeover provisionsmake a difference? Journal of Corporate Finance 17 (3), 418–437.
Iliev, P., 2010. The Effect of SOX Section 404: Costs, Earnings Quality, and Stock Prices.Journal of Finance 65 (3), 1163–1196.
Jensen, M. C., Murphy, K. J., 1990. Performance pay and top-management incentives. Journalof Political Economy 98, 225–264.
Johnson, D., Fowler, J., 2011. The evolution of overconfidence. The Nature 477, 317–320.
Kaplan, S., Zingales, L., 1997. Do investment-cash flow sensitivities provide useful measuresof financing constraints? Quarterly Journal of Economics 112, 169–215.
Kim, S., April 2013. The acquisitiveness of youth: Ceo age and acquisition behavior. Journalof Financial Economics 108 (1), 250–273.
Kim, Y. H., Forthcoming. Self Attribution Bias of the CEO: Evidence from CEO interviewson CNBC. Journal of Banking and Finance.
Kolasinski, A. C., Li, X., Forthcoming. Do Strong Boards and Trading in Their Own Firm’sStock Help CEOs Make Better Decisions? Evidence from Corporate Acquisitions by Over-confident CEOs. Journal of Financial and Quantitative Analysis.URL http://dx.doi.org/10.2139/ssrn.1573395
Lambert, R., Larcker, D., Verrecchia, R., 1991. Portfolio considerations in valuing executivecompensation. Journal of Accounting Research 29, 129–149.
Lehn, K., Patro, S., Zhao, M., 2007. Governance indexes and valuation: Which causes which?Journal of Corporate Finance 13 (5), 907–928.
Leuz, C., Triantis, A., Yue Wang, T., 2008. Why do firms go dark? Causes and economicconsequences of voluntary SEC deregistrations. Journal of Accounting and Economics 45 (2-3), 181–208.
Li, F., Minnis, M., Nagar, V., Rajan, M. V., 2012. Knowledge, compensation, and firm value:An empirical analysis of firm communication. Working Paper.
Li, K., Zhao, X., 2008. Asymmetric information and dividend policy. Financial Management37 (4), 673–694.
Low, A., 2009. Managerial risk-taking behavior and equity-based compensation. Journal ofFinancial Economics 92 (3), 470–490.
Lowry, M., Murphy, K. J., 2007. Executive stock options and IPO underpricing. Journal ofFinancial Economics 85 (1), 39–65.
36
Malmendier, U., Tate, G., 2005. CEO overconfidence and corporate investment. Journal ofFinancial Economics 60 (6), 2661–2700.
Malmendier, U., Tate, G., 2008. Who makes acquisitions? CEO overconfidence and the mar-ket’s reaction. Journal of Financial Economics 89, 20–43.
Malmendier, U., Tate, G., Yan, J., 2011. Overconfidence and early-life experiences: The effectof managerial traits on corporate financial policies. Journal of Finance 66 (5), 1687–1733.
Mansi, S. A., Wald, J. K., 2011. Payout policy with legal restrictions. Financial Management40 (3), 701–732.
Masulis, R., Swan, P., Tobiansky, B., 2013. Do wealth creating mergers and acquisitions reallyhurt acquirer shareholders? University of New South Wales, Working Paper Series.
Masulis, R., Wang, C., Xie, F., 2007. Corporate Governance and Acquirer Returns. Journalof Finance 62 (4), 1851–1889.
McCahery, J. A., Vermeulen, E. P. M., 2005. Does the european company prevent the ’delawareeffect’? European Law Journal 11 (6), 785–801.
Meulbroek, L., 2001. The efficiency of equity-linked compensation: Understanding the fullcost of awarding executive stock options. Financial Management 30, 5–30.
Moeller, S. B., Schlingemann, F. P., Stulz, R. M., 2004. Firm size and the gains from acqui-sitions. Journal of Financial Economics 73 (2), 201–228.
Moeller, S. B., Schlingemann, F. P., Stulz, R. M., 2005. Wealth destruction on a massivescale? a study of acquiring-firm returns in the recent merger wave. Journal of Finance60 (2), 757–782.
O’Connor, M., Rafferty, M., 2012. Corporate governance and innovation. Journal of Financialand Quantitative Analysis 47 (2), 397–413.
Ovtchinnikov, A. V., 2013. Merger waves following industry deregulation. Journal of CorporateFinance 21 (1), 51–76.
Powell, R. G., Stark, A. W., 2005. Does operating performance increase post-takeover for UKtakeovers? A comparison of performance measures and benchmarks. Journal of CorporateFinance 11 (1-2), 293–317.
Powell, R. G., Yawson, A., 2005. Industry aspects of takeovers and divestitures: evidence fromthe UK. Journal of Banking & Finance 29, 3015–3040.
Powell, R. G., Yawson, A., 2007. Are Corporate Restructuring Events Driven by CommonFactors? Implications for Takeover Prediction. Journal of Business Finance & Accounting34 (7-8), 1169–1192.
Puri, M., Robinson, D. T., 2007. Optimism and economic choice. Journal of Financial Eco-nomics 86 (1), 71–99.
37
Roll, R., 1986. The hubris hypothesis of corporate takeovers. The Journal of Business 59 (2),197–216.
Schijven, M., Hitt, M., 2012. The vicarious wisdom of crowds: toward a behavioral perspectiveon investor reactions to acquisition announcements. Strategic Management Journal 33 (11),1247–1268.
Schrand, C. M., Zechman, S. L. C., 2012. Executive overconfidence and the slippery slope tofinancial misreporting. Journal of Accounting and Economics 53 (1-2), 311–329.
Shu, P.-G., Yeh, Y.-H., Chiang, T.-L., Hung, J.-Y., 2013. Managerial overconfidence and sharerepurchases. International Review of Finance 13 (1), 39–65.
Simsek, Z., Heavy, C., Veiga, J. F., 2010. The impact of CEO core self-evaluation on the firm’sentrepreneurial orientation. Strategic Management Journal 31 (1), 110–119.
Switzer, L. N., 2007. Corporate governance, Sarbanes-Oxley, and small-cap firm performance.Quarterly Review of Economics and Finance 47 (5), 651–666.
Waelchi, U., Zeller, J., 2013. Old captains at the helm: Chairman age and firm performance.Journal of Banking and Finance 37 (5), 1612–1628.
Yim, S., Forthcoming. The acquisitiveness of youth: CEO age and acquisition behavior. Jour-nal of Financial Economics.
Yu, C.-F., 2013. CEO Overconfidence, CEO Compensation, and Earnings Manipulation.Working Paper, Monash University.
Appendix 1: Definition of Variables
Table A1: Variable Definitions
Variable Definition
Overconfidence Measures
Confidence A measure of how in-the-money the CEO’s vested stock options are. First, we obtain the totalvalue-per option of the ITM options by dividing the value of all unexercised exercisable options (Ex-ecucomp: opt unex exer est val) by the number of options (Execucomp: opt unex exer num). Nextwe scale this ‘value-per-option’ by the price at the end of the fiscal year as reported in (Compustat:prcc f)
SOX Measure
SOX An indicator that equals one if the observation occurs in 2002 or later and equals zero otherwise
Performance Measures
MTB The firm’s market-to-book ratio, being its market value at the end of the fiscal year(CRSP/Compustat: prcc f × csho) divided by its book assets (Compustat: at)
Ind Adj MTB The firm’s industry adjusted Tobin’s Q, defined as its Tobin’s Q less the average Tobin’s Q for allfirms in its two-digit SIC industry and year.
EBIT/Assets The firm’s EBIT (Compustat: ebit) scaled by its book assets (Compustat: at)
38
Ind Adj EBIT/Assets The firm’s EBIT/Assets less the mean EBIT/Assets for all companies in the firm’s two-digit SICindustry and year
Risk Measures
Beta The firm’s beta as estimated using a single index model using daily stock returns over the course ofthe year with an CRSP equally weighted market index.
ln (Variance) The firm’s daily stock price variance over the course of the year.ln (MSE) The mean squared error from the estimation of the single index model (above) over the course of
that year.
Governance Variables
ln (CEO Tenure) The natural log of one plus the number of years that the CEO has been the CEO of the company.ln (CEO Age) The natural log of the CEO’s ageCEO Bonus/Salary The ratio of the CEO’s bonus payment as ratio of his/her fixed salaryCEO%Ownership The percentage of the firm that the CEO ownsInst%Ownership The percentage of the firm that institutional investors ownsBCF The Bebchuk et al. (2009) entrenchment index. We use this in robustness tests in Table ??. The
data is from IRRC/Risk Metrics.State Ave BCF The average Bebchuk et al. (2009) entrenchment index for all companies in the subject-company’s
state and year. We use this in robustness tests in Table ??. The data is from IRRC/Risk Metrics.
Corporate Variables
Cash/Assets The firm’s cash holdings (Compustat: ch) divided by its book assets (Compustat: at)R&D/Sales The firm’s R&D expenditure (Compustat: xrd) divided by its sales (Compustat: sale)CAPEX/Assets The firm’s capital expenditures (Compustat: capx) scaled by its assets (Compustat:at)CAPEX/Sales The firm’s capital expenditure (Compustat: capx) divided by its sales (Compustat: sale)Ln (Assets) The natural log of the firm’s book assets (Compustat: at)Debt/Assets The firm’s long-term debt (Compustat: dltt) scaled by its assets (Compustat: at)Intangibles/ Assets The firm’s intangible assets (Compustat: intan) scaled by its total book assets (Compustat: at)Stock Return The firm’s cumulative daily stock return over year t. The data is from CRSPStock Std Dev The firm’s standard deviation of daily stock returns over year t. The data is from CRSPProp No Trade Days The proportion of days in year t on which there was no trade in the company’s stock
Acquisition Characteristics
CAR(-5,5) The acquirer’s cumulative abnormal return from five days before the acquisition announcement tofive days after the acquisition announcement. The cumulative abnormal return on day t is the firm’sraw return less the return predicted by a market model. We obtain the market model parametersusing an OLS estimation of the market model from 125 days before the acquisition announcementfor a period of 250 days.
CAR(-42,126) The acquirer’s cumulative abnormal return over the period 42 days before the acquisition announce-ment to 126 days after the acquisition announcement. The cumulative abnormal return on day tis the firm’s raw return less the return predicted by a market model. We obtain the market modelparameters using an OLS estimation of the market model from 125 days before the acquisitionannouncement for a period of 250 days.
BHAR(-5,250) The buy and hold abnormal return earned from holding the acquirer’s stock from five days before theacquisition announcement to 250 days after the acquisition announcement. The abnormal returnsare based on a market model with parameters estimated using OLS from 11 days to 210 days beforethe acquisition announcement.
Diversifying Deal A dummy variable that equals one if the bidder and target are in different SIC two-digit industriesand equals zero otherwise.
Run-up The acquirer’s cumulative abnormal return earned over the period 260 days to 11 days before theacquisition announcement.
Competed Deal A dummy variable that equals one if there is more than one bidder and equals zero otherwise.Tender offer A dummy variable that equals one if the deal was a tender offer and equals zero otherwise.Tender offer A dummy variable that equals one if the target is publicly listed and equals zero otherwise.Cash only A dummy variable that equals one if the method of payment was 100% cash and equals zero other-
wise.Rel Deal Size The ratio of the target’s market capitalization scaled by the acquirers assets.ln(Transaction Value) The natural log of the deal value
39
Tables
Table 1: Sample Composition by Year
This table contains the sample composition by year. Variable definitions are in the appendix. Figures are sample averages
Year Num Obs Confidence(t) Confidence(t)- Confidence
(t-1)Confidence(t) Median 25th Pctile 75th Pctile Std Dev
1992 198 0.329 0.301 0.153 0.459 0.2411993 633 0.348 0.325 0.137 0.525 0.258 -0.0171994 910 0.311 0.274 0.086 0.478 0.259 -0.0571995 944 0.337 0.319 0.126 0.499 0.252 0.0251996 998 0.356 0.337 0.132 0.550 0.258 0.0071997 1,049 0.411 0.418 0.200 0.601 0.278 0.0351998 1,090 0.365 0.364 0.105 0.584 0.283 -0.0761999 1,148 0.348 0.292 0.039 0.587 0.316 -0.0522000 1,190 0.355 0.319 0.043 0.582 0.370 -0.0122001 1,246 0.304 0.251 0.063 0.488 0.276 -0.0532002 1,374 0.220 0.151 0.004 0.368 0.232 -0.0952003 1,436 0.322 0.291 0.103 0.487 0.271 0.0802004 1,513 0.357 0.343 0.156 0.526 0.249 0.0162005 1,492 0.355 0.330 0.122 0.534 0.287 -0.0102006 1,510 0.380 0.364 0.165 0.554 0.268 0.0072007 1,597 0.337 0.300 0.082 0.537 0.298 -0.0472008 1,546 0.165 0.047 0.000 0.280 0.223 -0.1742009 1,525 0.201 0.121 0.000 0.338 0.226 0.0332010 1,468 0.257 0.202 0.050 0.409 0.242 0.0582011 1,253 0.244 0.179 0.032 0.407 0.242 -0.018
40
Table 2: Summary Statistics
This table shows the summary statistics of all the variables. We depict sample averages, median 25th and 75th percentile of allof our variables of interest as well as our control variables. These are averages over all years between 1992 and 2011.
Variable Mean Median 25th Pctile 75th Pctile Std Dev
Panel A: Statistics for the panel data sample
Confidence 0.309 0.268 0.062 0.496 0.277Beta 1.244 1.157 0.799 1.576 0.654MSE 0.024 0.021 0.015 0.030 0.013Variance 0.100 0.058 0.031 0.116 0.142Q 1.324 0.935 0.509 1.649 1.299Ind Adj Q -0.034 -0.195 -0.624 0.240 1.127EBIT/Assets 0.085 0.085 0.042 0.133 0.095Ind Adj EBIT/Assets -0.001 -0.001 -0.036 0.040 0.088Assets 8702 1593 528 5389 24983PPE/Assets 0.535 0.444 0.220 0.782 0.400LT Debt/ Assets 0.192 0.170 0.038 0.299 0.168R&D / Assets 0.042 0.000 0.000 0.033 0.100Intangibles/Assets 0.154 0.086 0.011 0.244 0.176CAPEX/Sales 0.076 0.038 0.020 0.076 0.124Cash/ Assets 0.093 0.050 0.016 0.131 0.109Bonus/Salary 0.726 0.359 0.000 1.002 1.163CEO Tenure 6.726 5.000 2.000 9.000 7.167CEO Age 55.379 55.000 51.000 60.000 7.225CEO%Own 0.020 0.003 0.001 0.012 0.048Inst%Own 0.575 0.654 0.399 0.813 0.319Dividends/ Assets 0.010 0.002 0.000 0.015 0.016SG&A/ Sales 0.252 0.216 0.119 0.339 0.177
Panel B: Statistics for the M&A sample
CAR(-10,10) 0.002 0.003 -0.055 0.064 0.111BHAR(-42,125) -0.106 -0.054 -0.314 0.166 0.471BHAR(-5,125) -0.080 -0.040 -0.255 0.144 0.385Confidence 0.383 0.364 0.148 0.575 0.274SOX 0.681 1.000 0.000 1.000 0.466Diversifying deal 0.439 0.000 0.000 1.000 0.496Run-up 0.003 0.015 -0.326 0.312 0.668Compted deal 0.014 0.000 0.000 0.000 0.119Tender Offer 0.058 0.000 0.000 0.000 0.233Public Target 0.195 0.000 0.000 0.000 0.396Cash Only 0.400 0.000 0.000 1.000 0.490Rel Deal Size 0.136 0.039 0.011 0.130 0.262ln(Transaction Value) 4.519 4.430 3.246 5.690 1.756ln(Assets) 7.792 7.613 6.558 8.945 1.705CEO Bonus/Salary 0.995 0.657 0.000 1.307 1.392ln(Tenure) 1.742 1.792 1.099 2.303 0.816ln(CEO Age) 3.993 4.007 3.912 4.094 0.132CEO%Own 0.016 0.003 0.001 0.011 0.040Inst%Own 0.668 0.706 0.559 0.828 0.239LT Debt/ Assets 0.178 0.155 0.035 0.276 0.157R&D/Sales 0.052 0.011 0.000 0.065 0.091EBIT/Assets 0.102 0.098 0.060 0.144 0.080Intangibles/Assets 0.208 0.162 0.041 0.334 0.188Q 1.666 1.202 0.749 2.046 1.504Cash/Assets 0.099 0.059 0.020 0.137 0.108CAPEX/Sales 0.072 0.038 0.022 0.068 0.119
41
Table 3: Capital Investments
This table contains regression models that examine the relationship between SOX, overconfidence, and CAPEX. The dependentvariable is the firm’s CAPEX in year t + 1 scaled by its assets in year t. The appendix contains the variable definitions. Thesignificance levels at the 1%, 5%, and 10% are denoted by ***, ** and *, respectively.
CAPEX (t+1)/Assets (t) × 100M1 M2 M3 M4
a: Confidence (t) 1.883*** 1.450***[0.000] [0.000]
b: ConfidenceTopQ (t) 0.902*** 0.614***[0.000] [0.000]
c: SOX 0.153 -0.238 0.061 -0.265[0.431] [0.173] [0.946] [0.768]
a × c -1.401*** -0.912***[0.000] [0.003]
b × c -0.674*** -0.350**[0.000] [0.043]
CEO Bonus/Salary 0.033 0.042* 0.059* 0.067**[0.192] [0.093] [0.052] [0.027]
ln(Tenure(t)) -0.009 0.005 0.015 0.028[0.769] [0.876] [0.762] [0.581]
ln(CEO Age (t)) -0.846*** -0.863*** -0.548 -0.575[0.000] [0.000] [0.200] [0.179]
CEO%Own(t) 1.331* 1.169* 3.172** 3.063**[0.057] [0.096] [0.039] [0.046]
Inst%Own (t) 0.212** 0.237*** 0.282 0.327*[0.018] [0.008] [0.152] [0.097]
ln(Assets(t)) -0.135*** -0.127*** -1.377*** -1.389***[0.000] [0.000] [0.000] [0.000]
Stock Return (t) 1.489*** 1.585*** 0.788*** 0.849***[0.000] [0.000] [0.000] [0.000]
Stock Std Dev (t) -11.804*** -12.010*** -12.767*** -12.907***[0.000] [0.000] [0.000] [0.000]
Prop No Trade Days (t) -1.214 -1.227 5.825 5.983[0.537] [0.563] [0.556] [0.550]
LT Debt/ Assets(t) -0.158 -0.178 -2.587*** -2.615***[0.450] [0.401] [0.000] [0.000]
R&D/Sales (t) -1.033*** -1.037*** 3.455*** 3.498***[0.009] [0.009] [0.001] [0.001]
EBIT/Assets (t) 2.263*** 2.394*** 6.752*** 6.960***[0.000] [0.000] [0.000] [0.000]
Intangibles/Assets (t) -1.082*** -1.048*** 1.829*** 1.889***[0.000] [0.000] [0.000] [0.000]
Q(t) 0.254*** 0.277*** 0.487*** 0.522***[0.000] [0.000] [0.000] [0.000]
CAPEX/Assets (t-1) 80.186*** 80.387*** 45.944*** 46.135***[0.000] [0.000] [0.000] [0.000]
Constant 5.092*** 5.455*** 14.724*** 15.179***[0.000] [0.000] [0.000] [0.000]
Year Fixed Effects Yes Yes Yes YesFirm Fixed Effects No No Yes YesIndustry Fixed Effects Yes Yes No No
Observations 19,349 19,349 19,349 19,349R-squared 72.10% 72.00% 35.30% 35.10%
42
Table 4: Selling & General Administrative Expenses
This table contains regressions that examine the moderating effect of SOX on SG&A. The dependent variable is the SG&Aexpense in year t + 1 scaled by sales in year t. The models are OLS models that include year, industry, and firm fixed effects asstated in the column. All models use standard errors clustered by firm. Brackets contain p-values and superscripts ***, **, and* denote significance at 1%, 5%, and 10%, respectively.
Dependent Variable SG&A/Sales (t+1)M1 M2 M3 M4
a: Confidence (t) 0.030*** 0.020***[0.000] [0.002]
b: ConfidenceTopQ (t) 0.015*** 0.007**[0.000] [0.039]
c: SOX -0.007 -0.010** 0.043*** 0.038***[0.179] [0.031] [0.001] [0.004]
a × c -0.013** -0.015**[0.035] [0.033]
b × c -0.009** -0.007*[0.018] [0.092]
CEO Bonus/Salary -0.001** -0.001* -0.002** -0.002**[0.030] [0.057] [0.019] [0.027]
ln(Tenure(t)) 0.003*** 0.004*** 0.003* 0.003**[0.000] [0.000] [0.052] [0.037]
ln(CEO Age (t)) -0.027*** -0.027*** -0.023* -0.024*[0.000] [0.000] [0.059] [0.055]
CEO%Own(t) 0.027 0.024 0.041 0.039[0.226] [0.274] [0.246] [0.269]
Inst%Own (t) 0 0 0 0[0.963] [0.883] [0.937] [0.961]
ln(Assets(t)) -0.002*** -0.002*** -0.023*** -0.023***[0.000] [0.000] [0.000] [0.000]
Stock Return (t) 0.022*** 0.023*** 0.016*** 0.017***[0.000] [0.000] [0.000] [0.000]
Stock Std Dev (t) -0.086 -0.093 -0.236*** -0.233***[0.259] [0.220] [0.008] [0.009]
Prop No Trade Days (t) 0.078 0.076 0.429 0.424[0.237] [0.224] [0.391] [0.396]
LT Debt/ Assets(t) 0.012** 0.012** 0.013 0.013[0.018] [0.027] [0.118] [0.124]
R&D/Sales (t) 0.105*** 0.105*** 0.163*** 0.163***[0.000] [0.000] [0.000] [0.000]
EBIT/Assets (t) 0.045*** 0.049*** 0.022 0.026[0.003] [0.001] [0.220] [0.154]
Intangibles/Assets (t) 0.033*** 0.034*** 0.032*** 0.033***[0.000] [0.000] [0.009] [0.007]
Mkt/Boook (t) 0.012*** 0.013*** 0.012*** 0.012***[0.000] [0.000] [0.000] [0.000]
CAPEX/Sales (t) 0.054*** 0.055*** 0.080*** 0.081***[0.000] [0.000] [0.000] [0.000]
SG&A/Sales (t) 1.034*** 1.033*** 0.809*** 0.809***[0.000] [0.000] [0.000] [0.000]
Constant 0.143*** 0.150*** 0.259*** 0.266***[0.000] [0.000] [0.000] [0.000]
Year Fixed Effects Yes Yes Yes YesFirm Fixed Effects No Yes No YesIndustry Fixed Effects Yes No Yes No
Observations 15,778 15,778 15,778 15,778R-squared 89.60% 89.50% 41.90% 41.90%
43
Tab
le5:
Ass
etan
dP
rop
erty
Gro
wth
Th
ista
ble
conta
ins
regre
ssio
ns
that
exam
ine
the
rela
tion
ship
bet
wee
nC
EO
over
con
fid
ence
,S
OX
an
dass
etgro
wth
.T
he
colu
mn
hea
der
state
sth
ed
epen
den
tvari
able
.T
he
mod
els
conta
infi
rman
dyea
rfi
xed
effec
ts,
an
du
sest
an
dard
erro
rscl
ust
ered
by
firm
.B
rack
ets
conta
inp
-valu
esan
dsu
per
scri
pts
***,
**,
an
d*
den
ote
sign
ifica
nce
at
1%
,5%
,an
d10%
,re
spec
tivel
y.D
epen
den
tvari
ab
leln
(PP
E(t
+1)/
PP
E(t
))ln
(Ass
ets(
t+1)/
Ass
ets(
t))
ln(P
PE
(t+
2)/
PP
E(t
+1))
ln(A
sset
s(t+
2)/
Ass
ets(
t+1))
M1
M2
M3
M4
M5
M6
M7
M8
a:
Con
fid
ence
(t)
0.0
99***
0.0
42**
0.0
36***
0.0
16
[0.0
00]
[0.0
13]
[0.0
08]
[0.2
19]
b:
Con
fid
ence
Top
Q(t
)0.0
44***
0.0
01
0.0
15*
0.0
04
[0.0
00]
[0.9
16]
[0.0
69]
[0.6
95]
c:
SO
X0.0
07
-0.0
17
0.1
41***
0.1
23***
-0.0
33
-0.0
43*
-0.0
5-0
.058*
[0.8
20]
[0.5
89]
[0.0
01]
[0.0
02]
[0.2
03]
[0.0
91]
[0.1
35]
[0.0
83]
a×
c-0
.087***
-0.0
69***
-0.0
42**
-0.0
34**
[0.0
00]
[0.0
00]
[0.0
13]
[0.0
35]
b×
c-0
.042***
-0.0
21*
-0.0
19**
-0.0
1[0
.000]
[0.0
53]
[0.0
48]
[0.3
40]
CE
OB
onu
s/S
ala
ry0.0
02
0.0
02
-0.0
01
00.0
06***
0.0
07***
-0.0
01
-0.0
01
[0.5
22]
[0.3
99]
[0.7
77]
[0.9
24]
[0.0
07]
[0.0
06]
[0.7
32]
[0.7
62]
ln(T
enu
re(t
))0.0
04
0.0
05
0.0
03
0.0
03
0.0
04
0.0
04
0.0
03
0.0
03
[0.3
15]
[0.2
41]
[0.4
16]
[0.3
61]
[0.2
14]
[0.1
92]
[0.3
77]
[0.3
78]
ln(C
EO
Age
(t))
-0.1
19***
-0.1
20***
-0.0
87***
-0.0
89***
-0.0
90***
-0.0
91***
-0.0
74**
-0.0
75**
[0.0
01]
[0.0
01]
[0.0
10]
[0.0
09]
[0.0
02]
[0.0
02]
[0.0
14]
[0.0
13]
CE
O%
Ow
n(t
)0.2
59***
0.2
54***
0.3
05***
0.3
11***
0.1
58**
0.1
57**
0.2
08**
0.2
12**
[0.0
03]
[0.0
04]
[0.0
02]
[0.0
01]
[0.0
29]
[0.0
31]
[0.0
13]
[0.0
12]
Inst
%O
wn
(t)
-0.0
13
-0.0
1-0
.050***
-0.0
49***
00.0
01
-0.0
31**
-0.0
31**
[0.4
36]
[0.5
58]
[0.0
01]
[0.0
01]
[0.9
97]
[0.9
30]
[0.0
28]
[0.0
31]
LT
Deb
t/A
sset
s(t)
-0.1
07***
-0.1
09***
-0.1
25***
-0.1
24***
-0.1
21***
-0.1
21***
-0.1
28***
-0.1
27***
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
R&
D/S
ale
s(t
)0.0
17
0.0
19
-0.0
69
-0.0
70.0
21
0.0
21
0.1
41
0.1
4[0
.849]
[0.8
34]
[0.4
71]
[0.4
65]
[0.8
58]
[0.8
56]
[0.1
55]
[0.1
58]
EB
IT/A
sset
s(t
)0.3
30***
0.3
43***
0.3
53***
0.3
57***
0.1
03*
0.1
07*
0.1
04**
0.1
02**
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
70]
[0.0
64]
[0.0
17]
[0.0
18]
Inta
ngib
les/
Ass
ets
(t)
0.0
21
0.0
22
-0.2
92***
-0.2
91***
-0.0
46
-0.0
46
-0.3
02***
-0.3
03***
[0.5
48]
[0.5
16]
[0.0
00]
[0.0
00]
[0.1
05]
[0.1
08]
[0.0
00]
[0.0
00]
Q(t
)0.0
30***
0.0
32***
0.0
57***
0.0
60***
0.0
24***
0.0
25***
0.0
17***
0.0
18***
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
CA
PE
X/S
ale
s(t
)0.1
22***
0.1
27***
0.0
86*
0.0
91*
-0.0
38
-0.0
37
-0.1
07*
-0.1
07*
[0.0
07]
[0.0
05]
[0.0
91]
[0.0
77]
[0.3
97]
[0.4
03]
[0.0
59]
[0.0
58]
Sto
ckR
etu
rn(t
)0.0
25***
0.0
29***
0.0
60***
0.0
62***
0.0
22***
0.0
24***
0.0
21***
0.0
21***
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
Sto
ckS
tdD
ev(t
)-1
.101***
-1.1
00***
-1.5
86***
-1.5
57***
0.0
07
0.0
17
-0.4
19*
-0.3
99
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.9
82]
[0.9
56]
[0.1
00]
[0.1
20]
Pro
pN
oT
rad
eD
ays(
t)-0
.032
-0.0
19
0.0
10.0
15
0.9
80.9
81
1.0
22
1.0
23
[0.9
69]
[0.9
82]
[0.9
92]
[0.9
87]
[0.3
27]
[0.3
23]
[0.3
70]
[0.3
69]
Con
stant
0.5
06***
0.5
29***
0.3
63***
0.3
80***
0.4
27***
0.4
34***
0.4
82***
0.4
87***
[0.0
01]
[0.0
00]
[0.0
09]
[0.0
06]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
Ob
serv
ati
on
s18,1
45
18,1
45
19,3
80
19,3
80
16,0
78
16,0
78
17,1
04
17,1
04
R-s
qu
are
d9.4
0%
9.2
0%
17.0
0%
16.9
0%
5.1
0%
5.1
0%
7.5
0%
7.5
0%
44
Table 6: Sensitivity of Investment to Cash Flows
This table contains regressions that examine how SOX attenuates the senitivity of investment by overconfident managers to cash flows. Thedependent variable is the firm’s CAPEX/Assets in year t (as per Malmendier and Tate (2005)). Variable definitions are in appendix 1. The modelsare OLS models that include firm fixed effects, industry fixed effects, year fixed effects (as indicated in the Column footer) and cluster standarderrors by firm. Brackets contain p-values and superscripts ***, **, and * denote significance at 1%, 5%, and 10%, respectively.
Depenent Variable CAPEX/Assets (t)Column M1 M2 M3 M4 M5 M6 M7 M8a: Confidence (t) 0.013*** 0.005 0.009** 0
[0.000] [0.138] [0.020] [0.988]b: ConfidenceTopQ (t) 0.007*** 0.003 0.006** 0
[0.007] [0.252] [0.024] [0.848]c: SOX -0.017*** -0.018*** -0.018** -0.018** -0.014* -0.015* -0.018** -0.016**
[0.000] [0.000] [0.022] [0.020] [0.089] [0.069] [0.030] [0.039]d: EBIT/Assets (t) 0.012 0.029*** 0.034*** 0.056***
[0.319] [0.006] [0.010] [0.000]e: OCF/Assets 0.096*** 0.116*** 0.055*** 0.080***
[0.000] [0.000] [0.000] [0.000]
a × c 0.004 0.006 0.001 0.011**[0.347] [0.115] [0.758] [0.012]
a × d 0.057*** 0.067***[0.006] [0.005]
a × c × d -0.095*** -0.094***[0.002] [0.001]
b × c 0.003 0.002 -0.001 0.005*[0.366] [0.317] [0.840] [0.068]
b × d 0.026 0.022[0.104] [0.148]
b × c × d -0.053*** -0.038**[0.010] [0.028]
a × e 0.064*** 0.080***[0.003] [0.001]
a × c × e -0.055* -0.093***[0.094] [0.001]
b × e 0.022 0.028*[0.152] [0.067]
b × c × e -0.021 -0.041**[0.311] [0.015]
c × d 0.029** 0.011 0.024 0.001[0.025] [0.287] [0.106] [0.904]
c × e 0.007 -0.01 0.003 -0.020*[0.617] [0.385] [0.826] [0.091]
CEO Bonus/Salary -0.001 0 0 0 -0.001** -0.001* 0 0[0.302] [0.385] [0.936] [0.757] [0.043] [0.063] [0.612] [0.810]
ln(Tenure(t)) 0 0.001 0.001** 0.001** 0.001 0.001 0.001** 0.001***[0.433] [0.267] [0.023] [0.010] [0.236] [0.145] [0.012] [0.005]
ln(CEO Age (t)) -0.011** -0.011** -0.009** -0.009** -0.012*** -0.012*** -0.010** -0.011**[0.016] [0.012] [0.033] [0.023] [0.006] [0.005] [0.015] [0.010]
CEO%Own(t) 0.042*** 0.040*** 0.040*** 0.039*** 0.041*** 0.039*** 0.041*** 0.040***[0.002] [0.003] [0.003] [0.004] [0.002] [0.002] [0.002] [0.003]
Inst%Own (t) 0.009*** 0.009*** 0.006*** 0.006*** 0.008*** 0.008*** 0.006*** 0.006***[0.000] [0.000] [0.004] [0.003] [0.000] [0.000] [0.005] [0.003]
ln(Assets(t)) -0.001*** -0.001** -0.002* -0.002* -0.001** -0.001** -0.001 -0.001[0.009] [0.012] [0.068] [0.070] [0.025] [0.032] [0.183] [0.189]
Stock Return (t) -0.011*** -0.010*** -0.011*** -0.011*** -0.011*** -0.010*** -0.011*** -0.011***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Stock Std Dev (t) 0.107*** 0.098** 0.095*** 0.091*** 0.226*** 0.222*** 0.093*** 0.092***[0.009] [0.016] [0.004] [0.005] [0.000] [0.000] [0.007] [0.007]
Prop No Trade Days (t) -0.008 -0.01 0.03 0.029 -0.014 -0.016 0.036 0.034[0.827] [0.792] [0.530] [0.548] [0.731] [0.706] [0.455] [0.491]
LT Debt/ Assets(t) 0.012*** 0.011*** -0.006 -0.007 0.011** 0.011** -0.007 -0.007[0.007] [0.009] [0.146] [0.130] [0.016] [0.020] [0.132] [0.113]
R&D/Sales (t) -0.028*** -0.028*** 0.033*** 0.033*** 0.009 0.008 0.039*** 0.039***[0.000] [0.000] [0.002] [0.002] [0.191] [0.208] [0.000] [0.000]
Intangibles/Assets (t) -0.061*** -0.061*** -0.029*** -0.029*** -0.061*** -0.061*** -0.028*** -0.027***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Q(t) 0.004*** 0.004*** 0.003*** 0.004*** 0.001 0.001 0.003*** 0.003***[0.000] [0.000] [0.000] [0.000] [0.362] [0.207] [0.000] [0.000]
Constant 0.097*** 0.102*** 0.099*** 0.102*** 0.099*** 0.101*** 0.098*** 0.098***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesFirm Fixed Effects No No Yes Yes No No Yes YesIndustry Fixed Effects Yes Yes No No Yes Yes No No
Observations 21,714 21,714 21,714 21,714 21,267 21,267 21,267 21,267R-squared 48.70% 48.60% 15.90% 15.60% 50.00% 49.90% 16.60% 16.30%
45
Table 7: Risk
This table contains regressions that examine the relationship between CEO overconfidence, SOX, and risk taking. The columnheader states the dependent variable. The models contain fixed effects as stated in the column, and use standard errors clusteredby firm. Brackets contain p-values and superscripts ***, **, and * denote significance at 1%, 5%, and 10%, respectively.
Dependent variable Beta (t+1) ln(MSE(t+1))M1 M2 M3 M4 M5 M6 M7 M8
a: Confidence (t) 0.357*** 0.260*** 0.083*** 0.076***[0.000] [0.000] [0.000] [0.000]
b: ConfidenceTopQ (t) 0.181*** 0.129*** 0.068*** 0.058***[0.000] [0.000] [0.000] [0.000]
c: SOX 0.254** -0.042 0.174 -0.097 -0.136*** -0.249*** -0.146*** -0.262***[0.027] [0.732] [0.128] [0.424] [0.001] [0.000] [0.000] [0.000]
a × c -0.331*** -0.176*** -0.042** -0.059***[0.000] [0.003] [0.022] [0.007]
b × c -0.180*** -0.106*** -0.031*** -0.038***[0.000] [0.000] [0.002] [0.000]
ln(Assets (t)) 0.025*** 0.03 0.026*** 0.03 -0.057*** -0.058*** -0.056*** -0.059***[0.000] [0.137] [0.000] [0.136] [0.000] [0.000] [0.000] [0.000]
Stock Return (t) 0.041*** 0.004 0.058*** 0.015 -0.055*** -0.060*** -0.054*** -0.058***[0.004] [0.740] [0.000] [0.220] [0.000] [0.000] [0.000] [0.000]
Stock Std Dev (t) 24.338*** 18.611*** 24.305*** 18.528*** 18.336*** 11.247*** 18.272*** 11.220***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Prop No Trade Days(t) -8.466*** -6.401** -8.564*** -6.453** 1.012*** -0.674 1.001*** -0.68[0.004] [0.013] [0.004] [0.013] [0.001] [0.403] [0.001] [0.400]
CEO Bonus/Salary 0 -0.009* 0.002 -0.008 0.007*** 0.002 0.006*** 0.002[0.951] [0.080] [0.676] [0.130] [0.003] [0.502] [0.004] [0.531]
ln(Tenure(t)) 0.019*** 0.020** 0.022*** 0.022** 0.006** 0.004 0.006** 0.004[0.005] [0.023] [0.001] [0.011] [0.039] [0.289] [0.047] [0.288]
ln(CEO Age (t)) -0.163*** -0.250*** -0.164*** -0.252*** -0.100*** -0.085** -0.098*** -0.084**[0.000] [0.001] [0.000] [0.001] [0.000] [0.013] [0.000] [0.014]
CEO%Own(t) -0.063 0.181 -0.095 0.146 0.131** 0.095 0.119** 0.083[0.622] [0.367] [0.454] [0.465] [0.024] [0.274] [0.040] [0.339]
Inst%Own (t) 0.048** 0 0.053*** 0.007 -0.003 0.005 -0.002 0.007[0.012] [0.990] [0.006] [0.832] [0.725] [0.767] [0.776] [0.678]
LT Debt/ Assets(t) 0.041 0.045 0.037 0.037 0.098*** 0.114*** 0.097*** 0.113***[0.356] [0.462] [0.409] [0.552] [0.000] [0.000] [0.000] [0.000]
R&D/Sales (t) 0.239** -0.01 0.238** -0.001 0.053 -0.218*** 0.057* -0.216***[0.046] [0.959] [0.048] [0.998] [0.117] [0.006] [0.090] [0.005]
EBIT/Assets (t) -0.458*** -0.473*** -0.426*** -0.427*** -0.500*** -0.517*** -0.497*** -0.513***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Intangibles/Assets (t) -0.117*** -0.092 -0.112*** -0.086 -0.047** -0.019 -0.049** -0.02[0.002] [0.210] [0.004] [0.242] [0.015] [0.565] [0.012] [0.528]
Q (t) 0.038*** 0.095*** 0.042*** 0.100*** 0.005* 0.024*** 0.004 0.023***[0.000] [0.000] [0.000] [0.000] [0.092] [0.000] [0.189] [0.000]
CAPEX/Sales (t) 0.211*** 0.430*** 0.222*** 0.442*** 0.032 0.032 0.029 0.029[0.002] [0.000] [0.001] [0.000] [0.303] [0.386] [0.351] [0.439]
Constant 0.379* 1.208*** 0.454** 1.260*** -3.620*** -3.476*** -3.616*** -3.462***[0.081] [0.001] [0.036] [0.000] [0.000] [0.000] [0.000] [0.000]
Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesIndustry Fixed Effects Yes No Yes No Yes No Yes NoFirm Fixed Effects No Yes No Yes No Yes No Yes
Observations 17,110 17,110 17,110 17,110 17,110 17,110 17,110 17,110R-squared 49.40% 27.20% 49.10% 27.00% 74.30% 62.90% 74.40% 63.00%
46
Table 8: Overconfidence and Tobin’s Q
This table contains regressions that examine the relationship between CEO overconfidence, SOX, firm value, as proxied by thefirm’s Tobin’s Q or industry-adjusted Tobin’s Q ratio, where we define the firm’s Tobin’s Q as its market capitalization dividedby its book assets. The column header states the dependent variable. The models contain fixed effects as stated in the column,and use standard errors clustered by firm. Brackets contain p-values and superscripts ***, **, and * denote significance at 1%,5%, and 10%, respectively.
Dependent Variable Q (t+1) Ind Adj Q (t+1) Q (t+1) Ind Adj Q (t+1)Model M1 M2 M3 M4 M5 M6 M7 M8
a: Confidence (t) -0.073 0.094* -0.074 0.062[0.140] [0.062] [0.106] [0.188]
b: ConfidenceTopQ (t) -0.043 0.060* -0.041 0.047[0.144] [0.068] [0.135] [0.131]
c: SOX 0.029 0.077 0.078*** 0.305** 0.041* 0.081 0.092*** 0.316**[0.318] [0.498] [0.007] [0.019] [0.098] [0.474] [0.000] [0.016]
a × c 0.085 0.145** 0.096* 0.156***[0.110] [0.016] [0.054] [0.007]
b × c 0.069** 0.065* 0.074** 0.070**[0.032] [0.067] [0.016] [0.039]
ln(Assets (t)) -0.028*** -0.341*** -0.027*** -0.304*** -0.028*** -0.338*** -0.027*** -0.301***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Stock Return (t) -0.034* 0.039** -0.03 0.025 -0.039** 0.046*** -0.035* 0.030*[0.078] [0.027] [0.103] [0.138] [0.042] [0.009] [0.055] [0.066]
Stock Std Dev (t) -1.918*** 0.856 -2.097*** -0.068 -1.889*** 0.752 -2.082*** -0.158[0.001] [0.238] [0.000] [0.919] [0.001] [0.300] [0.000] [0.815]
Prop No Trade Days(t) -0.446 -0.281 -0.372 0.082 -0.452 -0.263 -0.378 0.096[0.257] [0.829] [0.322] [0.961] [0.236] [0.839] [0.297] [0.954]
CEO Bonus/Salary 0.016*** 0.020*** 0.015*** 0.016** 0.016*** 0.021*** 0.014*** 0.016**[0.004] [0.002] [0.006] [0.012] [0.005] [0.001] [0.008] [0.011]
ln(Tenure(t)) 0.000 0.000 -0.002 -0.006 -0.001 0.001 -0.003 -0.004[0.944] [0.964] [0.771] [0.581] [0.851] [0.894] [0.680] [0.685]
ln(CEO Age (t)) 0.017 0.04 0.005 0.035 0.018 0.035 0.006 0.031[0.708] [0.679] [0.903] [0.710] [0.703] [0.717] [0.897] [0.745]
CEO%Own(t) -0.079 0.463* -0.058 0.383 -0.077 0.434* -0.056 0.357[0.545] [0.069] [0.637] [0.117] [0.559] [0.088] [0.652] [0.144]
Inst%Own (t) 0.016 0.078* 0.01 0.064 0.015 0.080* 0.009 0.065[0.334] [0.062] [0.523] [0.134] [0.353] [0.057] [0.549] [0.128]
LT Debt/ Assets(t) -0.084 -0.167* -0.088* -0.104 -0.084 -0.178* -0.087* -0.114[0.119] [0.094] [0.088] [0.273] [0.123] [0.075] [0.092] [0.230]
R&D/Sales (t) 1.121*** 0.341 1.090*** 0.388 1.123*** 0.355 1.090*** 0.401[0.000] [0.222] [0.000] [0.145] [0.000] [0.204] [0.000] [0.133]
EBIT/Assets (t) 0.957*** 0.340** 0.886*** 0.299** 0.948*** 0.381** 0.876*** 0.333**[0.000] [0.031] [0.000] [0.044] [0.000] [0.015] [0.000] [0.024]
Intangibles/Assets (t) -0.169*** -0.122 -0.122*** -0.047 -0.171*** -0.117 -0.124*** -0.044[0.000] [0.182] [0.000] [0.578] [0.000] [0.198] [0.000] [0.603]
Q(t) 0.702*** 0.400*** 0.701*** 0.403***[0.000] [0.000] [0.000] [0.000]
Ind Adj Q (t) 0.712*** 0.410*** 0.711*** 0.412***[0.000] [0.000] [0.000] [0.000]
CAPEX/Sales (t) -0.103 -0.116 -0.083 -0.095 -0.107 -0.107 -0.087 -0.088[0.173] [0.430] [0.245] [0.504] [0.157] [0.466] [0.221] [0.534]
Constant 0.319* 2.887*** 0.149 1.804*** 0.306 2.897*** 0.134 1.809***[0.095] [0.000] [0.412] [0.000] [0.108] [0.000] [0.460] [0.000]
Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesIndustry Fixed Effects Yes No Yes No Yes No Yes NoFirm Fixed Effects No Yes No Yes No Yes No Yes
Observations 19,378 19,378 19,378 19,378 19,378 19,378 19,378 19,378R-squared 72.20% 37.90% 65.90% 29.90% 72.20% 37.90% 65.90% 29.90%
47
Table 9: EBIT
This table contains regressions that examine the relationship between CEO overconfidence, SOX, and operating performance, asproxied by the firm’s EBIT/Assets and industry-adjusted EBIT/Assets. The column header states the dependent variable. Themodels contain fixed effects as stated in the column, and use standard errors clustered by firm. Brackets contain p-values andsuperscripts ***, **, and * denote significance at 1%, 5%, and 10%, respectively.
Dependent Variable EBIT/Assets Ind Adj EBIT/Assets EBIT/Assets Ind Adj EBIT/Assets(t+1) (t+1) (t+1) (t+1)
M1 M2 M3 M4 M5 M6 M7 M8
a: Confidence (t) -0.004 -0.001 -0.001 0.001[0.124] [0.748] [0.654] [0.830]
b: ConfidenceTopQ (t) -0.001 -0.001 0.000 0.000[0.491] [0.611] [0.969] [0.934]
c: SOX -0.015* -0.014 -0.017* -0.009 -0.013 -0.012 -0.016 -0.007[0.092] [0.125] [0.076] [0.348] [0.144] [0.201] [0.103] [0.459]
a × c 0.011*** 0.022*** 0.010*** 0.021***[0.001] [0.000] [0.001] [0.000]
b × c 0.005*** 0.010*** 0.005** 0.008***[0.008] [0.000] [0.015] [0.000]
ln(Assets (t)) 0.000 -0.012*** 0.000 -0.009*** 0.000 -0.012*** 0.000 -0.009***[0.860] [0.000] [0.528] [0.000] [0.836] [0.000] [0.547] [0.000]
Stock Return (t) 0.023*** 0.022*** 0.018*** 0.016*** 0.023*** 0.022*** 0.018*** 0.017***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Stock Std Dev (t) -0.533*** -0.330*** -0.435*** -0.187*** -0.539*** -0.337*** -0.442*** -0.197***[0.000] [0.000] [0.000] [0.003] [0.000] [0.000] [0.000] [0.002]
Prop No Trade Days (t) 0.072** 0.264** 0.084** 0.282** 0.071** 0.260** 0.083** 0.278**[0.040] [0.015] [0.012] [0.027] [0.042] [0.019] [0.012] [0.032]
CEO Bonus/Salary 0.001* 0.002*** 0.001** 0.001*** 0.001* 0.002*** 0.001** 0.001***[0.058] [0.000] [0.045] [0.002] [0.068] [0.000] [0.044] [0.001]
ln(Tenure(t)) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000[0.480] [0.859] [0.490] [0.690] [0.469] [0.949] [0.547] [0.896]
ln(CEO Age (t)) 0.008** 0.01 0.008** 0.01 0.008** 0.009 0.007** 0.01[0.030] [0.160] [0.032] [0.131] [0.031] [0.192] [0.036] [0.160]
CEO%Own(t) 0.000 -0.008 0.002 -0.005 0.000 -0.009 0.001 -0.007[0.993] [0.777] [0.877] [0.819] [0.999] [0.731] [0.883] [0.763]
Inst%Own (t) 0.004*** 0.003 0.005*** 0.003 0.004*** 0.003 0.005*** 0.003[0.001] [0.323] [0.000] [0.297] [0.001] [0.328] [0.000] [0.297]
LT Debt/ Assets(t) 0.014*** 0.024*** 0.013*** 0.023*** 0.014*** 0.023*** 0.013*** 0.022***[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
R&D/Sales (t) -0.051*** -0.086*** -0.046*** -0.072*** -0.052*** -0.084*** -0.047*** -0.070***[0.000] [0.001] [0.000] [0.004] [0.000] [0.001] [0.000] [0.004]
EBIT/Assets (t) 0.728*** 0.442*** 0.727*** 0.447***[0.000] [0.000] [0.000] [0.000]
Ind Adj EBIT/Assets (t) 0.743*** 0.452*** 0.744*** 0.457***[0.000] [0.000] [0.000] [0.000]
Intangibles/Assets 0.005* 0.002 0.004 -0.003 0.005** 0.002 0.004 -0.003[0.050] [0.719] [0.150] [0.652] [0.048] [0.712] [0.137] [0.649]
Constant 0.015 0.110*** -0.021 0.028 0.013 0.110*** -0.021 0.028[0.375] [0.000] [0.199] [0.315] [0.408] [0.000] [0.205] [0.314]
Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesIndustry Fixed Effects Yes No Yes No Yes No Yes NoFirm Fixed Effects No Yes No Yes No Yes No Yes
Observations 19,366 19,366 19,366 19,366 19,366 19,366 19,366 19,366R-squared 68.60% 33.30% 65.90% 29.20% 68.60% 33.20% 65.90% 29.00%
48
Table 10: Earnings Quality
This table examines the relationship between SOX, overconfidence and earnings quality. We run a order-logit regression withboth industry and year fixed effects and cluster standard errors by firm. We use S&P’s earnings and dividend ranking as ourdependent variable. The earning quality rank goes from 1 to 8 with 8 representing the best quality (called A+ rating). Bracketscontain p-values and superscripts ***, **, and * denote significance at 1%, 5%, and 10%, respectively.
M1 M2
a: Confidence (t) -0.136[0.479]
b: ConfidenceTopQ (t) -0.190*[0.082]
c: SOX -0.371*** -0.255***[0.000] [0.000]
a × c 0.831***[0.000]
b × c 0.368***[0.002]
ln(Assets (t)) 0.376*** 0.373***[0.000] [0.000]
Stock Return (t) 0.021 0.074**[0.527] [0.020]
Prop No Trade Days (t) -3.181* -3.352*[0.055] [0.057]
CEO Bonus/Salary -0.001 0.005[0.986] [0.857]
ln(Tenure(t)) 0.016 0.029[0.684] [0.451]
ln(CEO Age (t)) 0.323 0.278[0.260] [0.331]
CEO%Own(t) -0.75 -0.706[0.348] [0.376]
Inst%Own (t) 0.116 0.127[0.359] [0.320]
LT Debt/ Assets(t) -1.288*** -1.323***[0.000] [0.000]
R&D/Sales (t) -1.374*** -1.336***[0.003] [0.004]
EBIT/Assets (t) 6.934*** 7.168***[0.000] [0.000]
Intangibles/Assets (t) 0.443* 0.477*[0.093] [0.071]
Constant Cut 1 -0.076 -0.206[0.956] [0.878]
Constant Cut 2 1.758 1.627[0.195] [0.225]
Constant Cut 3 3.271** 3.136**[0.016] [0.019]
Constant Cut 4 4.531*** 4.391***[0.001] [0.001]
Constant Cut 5 5.829*** 5.685***[0.000] [0.000]
Constant Cut 6 6.577*** 6.433***[0.000] [0.000]
Constant Cut 7 7.671*** 7.527***[0.000] [0.000]
Year Fixed Effects Yes YesIndustry Fixed Effects Yes Yes
Observations 17,021 17,021Pseudo R-squared 9.96% 9.85%
49
Table 11: Value of CAPEX and R&D
This table contains regressions that examine how SOX moderates the impact of overconfidence on the value of R&D (in Panel A) and CAPEX (inPanel B). The dependent variable is the firm’s industry-adjusted market-to-book ratio in year t + 1. Columns 1 and 4 examine the whole sampleof firms. Columns 2, 3, 5, and 6 examine sub-samples of pre-SOX and post-SOX firms. Variable definitions are in appendix 1. The models areOLS models that include firm fixed effects and year fixed effects and cluster standard errors by firm. The models also include the standard controlvariables from Table 3. Brackets contain p-values and superscripts ***, **, and * denote significance at 1%, 5%, and 10%, respectively.
Dependent Variable Ind Adj Q (t+1)Sample All Pre-SOX Post-SOX All Pre-SOX Post-SOX
M1 M2 M3 M4 M5 M6
Panel A: Value of R&D
a: Confidence (t) 0.161*** 0.143** 0.144***[0.004] [0.045] [0.003]
b: ConfidenceTopQ (t) 0.129*** 0.111*** 0.070***[0.000] [0.002] [0.000]
c: SOX 0.320** 0.306**[0.016] [0.019]
d: R&D/Sales (t) 1.124** 1.164 0.590* 0.866** 0.83 0.533*[0.016] [0.111] [0.075] [0.024] [0.233] [0.076]
a × c 0.056[0.379]
a × c × d 1.910***[0.008]
b × c -0.032[0.338]
b × c × d 1.941***[0.000]
a × d -2.126*** -2.376*** 0.148[0.003] [0.004] [0.836]
b × d -1.457*** -1.466*** 0.643[0.000] [0.004] [0.146]
c × d -0.544 -0.36[0.188] [0.262]
Control Variables Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes YesFirm Fixed Effects Yes Yes Yes Yes Yes Yes
Observations 19,378 7,423 11,955 19,378 7,423 11,955R-squared 30.20% 15.20% 21.40% 30.30% 15.20% 21.60%
Panel B: Value of CAPEX
a: Confidence (t) 0.162*** 0.166** 0.145***[0.004] [0.021] [0.007]
b: ConfidenceTopQ (t) 0.101*** 0.097** 0.082***[0.007] [0.023] [0.000]
c: SOX 0.319** 0.297**[0.013] [0.020]
e: CAPEX/Sales (t) 0.128 0.301 -0.15 -0.079 0.102 -0.182[0.452] [0.296] [0.252] [0.584] [0.678] [0.149]
a × c 0.053[0.447]
a × c × e 1.067**[0.015]
b × c 0.01[0.797]
b × c × e 0.614**[0.023]
a × e -1.048*** -1.122** 0.039[0.005] [0.021] [0.883]
b × e -0.563** -0.599** 0.137[0.018] [0.027] [0.374]
c × e 0.026 0.242**[0.871] [0.043]
Control Variables Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes YesFirm Fixed Effects Yes Yes Yes Yes Yes Yes
Observations 19,378 7,423 11,955 19,378 7,423 11,955R-squared 30.10% 14.70% 21.40% 30.00% 14.70% 21.50%
50
Figure 1: CARs around takeoversThis figure plots the cumulative abnormal returns from 50 days before the takeover to 50 days after the acquisition
announcement. The abnormal returns are based on an OLS estimation of the market model over the period 125 to 375 days
before the acquisition announcement.
-.08
-.06
-.04
-.02
0.0
2C
umat
ive
Abn
orm
al R
etur
n
-50 0 50Days from Acquisition Announcement
Overconfident Post-SOX Non-Overconfident Post-SOX Overconfident Pre-SOX Non-Overconfident Pre-SOX
Figure 2: CARs around takeovers of public targetsThis figure plots the cumulative abnormal returns from 50 days before the takeover to 50 days after the acquisition
announcement. The abnormal returns are based on an OLS estimation of the market model over the period 125 to 375 days
before the acquisition announcement.
-.06
-.04
-.02
0.0
2C
umat
ive
Abn
orm
al R
etur
n
-50 0 50Days from Acquisition Announcement
Overconfident Post-SOX Non-Overconfident Post-SOX Overconfident Pre-SOX Non-Overconfident Pre-SOX
51
Figure 3: CARs around takeovers of non-public targetsThis figure plots the cumulative abnormal returns from 50 days before the takeover to 50 days after the acquisition
announcement. The abnormal returns are based on an OLS estimation of the market model over the period 125 to 375 days
before the acquisition announcement.
-.1
-.05
0C
umat
ive
Abn
orm
al R
etur
n
-50 0 50Days from Acquisition Announcement
Overconfident Post-SOX Non-Overconfident Post-SOX Overconfident Pre-SOX Non-Overconfident Pre-SOX
52
Table
12:
Acquis
itio
ns
This
table
exam
ines
how
SO
Xm
odera
tes
the
perf
orm
ance
of
overc
onfi
dence
CE
Os’
acquis
itio
ns.
The
colu
mn
header
state
sth
edep
endent
vari
able
.If
acquis
itio
nis
inth
eyeart+
1,
then
we
use
all
expla
nato
ryvari
able
sfr
om
the
yeart.
The
models
are
OL
Sre
gre
ssio
nm
odels
that
inclu
de
year
and
indust
ryfi
xed
eff
ects
and
clu
ster
standard
err
ors
by
firm
.B
rackets
conta
inp-v
alu
es
and
sup
ers
cri
pts
***,
**,
and
*denote
signifi
cance
at
1%
,5%
,and
10%
,re
specti
vely
.D
ep
endent
Vari
able
CA
RB
HA
RB
HA
RIn
dA
dj
EB
IT/A
ssets
Ind
Adj
QT
ime
Win
dow
(-10,1
0)
(-42,1
25)
(-5,1
25)
(t+
1)
(t+
1)
Model
M1
M2
M3
M4
M5
M6
M7
M8
M9
M10
a:
Confi
dence
(t)
-0.0
23*
-0.2
99***
-0.2
13***
-0.0
09
-0.3
07***
[0.0
75]
[0.0
00]
[0.0
00]
[0.1
21]
[0.0
00]
b:
Confi
denceT
opQ
(t)
-0.0
01
-0.1
14***
-0.0
85***
-0.0
07**
-0.1
81***
[0.8
97]
[0.0
00]
[0.0
00]
[0.0
24]
[0.0
00]
c:
SO
X-0
.059
-0.0
43***
-0.1
89
-0.3
86***
-0.1
36
-0.2
71***
-0.0
07
-0.0
04
-0.0
20.2
91***
[0.2
47]
[0.0
08]
[0.2
81]
[0.0
00]
[0.3
53]
[0.0
00]
[0.7
61]
[0.8
37]
[0.8
22]
[0.0
01]
a×
c0.0
19
0.1
10**
0.0
54
0.0
21***
0.3
19***
[0.2
12]
[0.0
37]
[0.2
26]
[0.0
02]
[0.0
00]
b×
c0.0
03
0.0
51*
0.0
26
0.0
12***
0.2
31***
[0.7
16]
[0.0
91]
[0.2
94]
[0.0
01]
[0.0
00]
Div
ers
ifyin
gdeal
-0.0
08**
-0.0
08**
-0.0
22
-0.0
17
-0.0
14
-0.0
18
0.0
01
0.0
01
-0.0
33
-0.0
31
[0.0
45]
[0.0
39]
[0.1
23]
[0.2
17]
[0.2
19]
[0.1
13]
[0.6
06]
[0.5
70]
[0.1
33]
[0.1
58]
Run-u
p-0
.017***
-0.0
17***
-0.1
11***
-0.1
14***
-0.1
34***
-0.1
22***
0.0
08***
0.0
08***
0.1
11***
0.1
09***
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
Com
ple
ted
deal
-0.0
01
0.0
01
-0.0
05
0.0
11
-0.0
37
-0.0
24
0.0
06
0.0
04
0.0
74
0.0
9[0
.934]
[0.9
49]
[0.9
27]
[0.8
32]
[0.4
08]
[0.5
87]
[0.3
59]
[0.5
19]
[0.3
75]
[0.2
69]
Tender
Off
er
0.0
13
0.0
18*
-0.0
03
-0.0
10.0
24
0.0
26
-0.0
04
-0.0
05
-0.0
77
-0.0
83*
[0.1
51]
[0.0
52]
[0.9
23]
[0.7
51]
[0.3
70]
[0.3
29]
[0.3
50]
[0.2
13]
[0.1
21]
[0.0
87]
Public
Targ
et
-0.0
16***
-0.0
19***
-0.0
35*
-0.0
24
-0.0
44***
-0.0
39**
0.0
03
0.0
04
0.0
62*
0.0
67**
[0.0
05]
[0.0
01]
[0.0
84]
[0.2
27]
[0.0
10]
[0.0
20]
[0.1
95]
[0.1
23]
[0.0
51]
[0.0
32]
Cash
only
Off
er
0.0
06
0.0
06*
0.0
18
0.0
17
0.0
21*
0.0
19*
0.0
02
0.0
01
0.0
24
0.0
22
[0.1
33]
[0.0
93]
[0.1
77]
[0.1
95]
[0.0
60]
[0.0
91]
[0.3
48]
[0.3
94]
[0.2
53]
[0.2
90]
Rela
tive
deal
size
-0.0
09
-0.0
11
-0.0
71**
-0.0
65*
-0.0
31
-0.0
36
-0.0
29***
-0.0
29***
-0.3
40***
-0.3
51***
[0.3
59]
[0.2
80]
[0.0
36]
[0.0
53]
[0.2
73]
[0.2
04]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
Ln
(Tra
nsa
cti
on
valu
e)
0.0
02
0.0
01
0.0
15***
0.0
10*
0.0
09**
0.0
08*
00
0-0
.003
[0.2
92]
[0.3
86]
[0.0
08]
[0.0
67]
[0.0
48]
[0.0
76]
[0.5
82]
[0.5
06]
[0.9
56]
[0.7
09]
Ln
(Ass
et
(t))
-0.0
05***
-0.0
05***
-0.0
06
-0.0
02
-0.0
05
-0.0
04
0.0
01
0.0
01
-0.0
06
-0.0
05
[0.0
03]
[0.0
02]
[0.3
22]
[0.7
30]
[0.3
10]
[0.4
02]
[0.3
38]
[0.4
25]
[0.5
40]
[0.6
04]
CE
OB
onus
(t)
/Sala
ry(t
)0.0
02
0.0
02
0.0
02
0.0
01
0.0
02
0.0
02
00
0.0
04
0.0
02
[0.2
97]
[0.1
98]
[0.7
40]
[0.8
59]
[0.6
99]
[0.5
84]
[0.8
91]
[0.9
17]
[0.6
17]
[0.7
98]
Ln
(Tenure
)0
00.0
11
0.0
05
0.0
04
0-0
.003**
-0.0
03**
-0.0
15
-0.0
16
[0.9
94]
[0.9
79]
[0.2
13]
[0.5
85]
[0.5
99]
[0.9
91]
[0.0
18]
[0.0
14]
[0.2
70]
[0.2
43]
Ln
(age)
0.0
15
0.0
13
0.1
00*
0.1
10**
0.0
47
0.0
69
0.0
20***
0.0
19***
-0.0
8-0
.085
[0.3
23]
[0.4
00]
[0.0
66]
[0.0
42]
[0.3
05]
[0.1
24]
[0.0
06]
[0.0
05]
[0.3
51]
[0.3
12]
CE
O%
Ow
n0.0
02
0.0
08
0.0
23
0.1
36
-0.0
04
0.0
56
-0.0
22
-0.0
07
0.0
56
-0.0
04
[0.9
61]
[0.8
79]
[0.8
97]
[0.4
41]
[0.9
79]
[0.7
00]
[0.3
52]
[0.7
45]
[0.8
43]
[0.9
87]
Inst
%O
wn
-0.0
16*
-0.0
16*
0.0
01
-0.0
09
-0.0
35
-0.0
41*
-0.0
07*
-0.0
07*
-0.1
17**
-0.1
27***
[0.0
54]
[0.0
62]
[0.9
84]
[0.7
43]
[0.1
53]
[0.0
86]
[0.0
61]
[0.0
72]
[0.0
11]
[0.0
05]
LT
Debt
(t)
/A
sset
(t)
-0.0
07
-0.0
07
-0.0
83
-0.0
8-0
.073*
-0.0
72*
0.0
16**
0.0
15**
-0.1
53*
-0.1
34
[0.6
39]
[0.6
64]
[0.1
13]
[0.1
26]
[0.0
98]
[0.0
96]
[0.0
20]
[0.0
20]
[0.0
64]
[0.1
00]
R&
D(t
)/Sale
s(t)
-0.0
03
-0.0
08
0.0
64
0.1
06
0.0
19
0.0
78
-0.0
47***
-0.0
54***
0.8
27***
0.8
68***
[0.9
23]
[0.7
90]
[0.5
18]
[0.2
82]
[0.8
22]
[0.3
37]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
EB
IT(t
)/A
sset
(t)
0.0
46
0.0
42
0.7
65***
0.7
57***
0.5
86***
0.5
86***
0.5
71***
0.5
70***
1.2
02***
1.2
05***
[0.1
42]
[0.1
75]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
Inta
ngib
les/
Ass
et(
t)-0
.016
-0.0
19
-0.0
6-0
.082*
-0.0
3-0
.042
0.0
23***
0.0
23***
0.1
81**
0.1
70**
[0.2
06]
[0.1
30]
[0.1
75]
[0.0
57]
[0.4
13]
[0.2
41]
[0.0
00]
[0.0
00]
[0.0
10]
[0.0
14]
Q(t
)-0
.005**
-0.0
06***
-0.0
60***
-0.0
68***
-0.0
48***
-0.0
54***
0.0
08***
0.0
08***
0.5
20***
0.5
11***
[0.0
19]
[0.0
04]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
Cash
(t)/
Ass
et(
t)-0
.042*
-0.0
43*
-0.2
05***
-0.2
14***
-0.2
13***
-0.2
17***
-0.0
26**
-0.0
28***
0.0
30.0
57
[0.0
59]
[0.0
54]
[0.0
08]
[0.0
05]
[0.0
01]
[0.0
01]
[0.0
10]
[0.0
05]
[0.8
06]
[0.6
37]
CA
PE
X(t
)/Sale
s(t)
0.0
01
-0.0
08
-0.1
94**
-0.2
20***
-0.1
34**
-0.1
08*
-0.0
17*
-0.0
16
-0.1
19
-0.1
82
[0.9
59]
[0.7
26]
[0.0
11]
[0.0
04]
[0.0
37]
[0.0
87]
[0.0
86]
[0.1
02]
[0.3
15]
[0.1
18]
Const
ant
-0.0
31
-0.0
6-0
.571
-0.7
16
-0.3
05
-0.4
2-0
.254***
-0.2
49***
-1.6
85**
-1.6
74**
[0.8
39]
[0.6
78]
[0.2
81]
[0.1
50]
[0.4
93]
[0.3
09]
[0.0
00]
[0.0
00]
[0.0
25]
[0.0
24]
Obse
rvati
ons
5,2
62
5,2
62
5,2
62
5,2
62
5,2
62
5,2
62
4,7
04
4,7
04
4,7
07
4,7
07
R-s
quare
d18.6
0%
17.8
0%
30.8
0%
30.1
0%
48.8
0%
48.7
0%
43.7
0%
43.5
0%
61.7
0%
61.7
0%
53
Table 13: Dividend Payments
This table contains OLS regression models that examine the relationship between CEO overconfidence, SOX and dividendpayments. The dependent variable is the firm’s total dividend payment scaled by its assets in year t + 1. The models are OLSregression models that contain firm, industry, and year fixed effects as stated in the column, and use standard errors clusteredby firm. Variable definitions are in Appendix 1. Brackets contain p-values and superscripts ***, **, and * denote significance at1%, 5%, and 10%, respectively.
Dependent Variable 100 x Total Dividends/ Assets (t+1)M1 M2 M3 M4
a: Confidence (t) -0.145*** -0.232***[0.000] [0.000]
b: ConfidenceTopQ (t) -0.098*** -0.124***[0.000] [0.000]
c: SOX 0.084** -0.449*** 0.103*** -0.390***[0.018] [0.001] [0.002] [0.003]
a × c 0.081** 0.284***[0.038] [0.000]
b × c 0.063*** 0.147***[0.006] [0.000]
CEO Bonus/Salary -0.007 -0.028*** -0.007 -0.028***[0.277] [0.003] [0.267] [0.002]
ln(Tenure(t)) -0.001 0.006 -0.002 0.006[0.893] [0.638] [0.826] [0.659]
ln(CEO Age (t)) 0.111*** 0.111 0.110*** 0.109[0.007] [0.282] [0.008] [0.290]
CEO%Own(t) -0.157 -0.602 -0.143 -0.588[0.292] [0.134] [0.341] [0.145]
Inst%Own (t) -0.068*** -0.166*** -0.070*** -0.173***[0.000] [0.001] [0.000] [0.001]
ln(Assets(t)) 0.031*** 0.123*** 0.031*** 0.127***[0.000] [0.000] [0.000] [0.000]
Stock Return (t) 0.032** 0.034** 0.026* 0.029*[0.024] [0.025] [0.055] [0.055]
Stock Std Dev (t) -5.774*** -6.680*** -5.678*** -6.670***[0.000] [0.000] [0.000] [0.000]
Prop No Trade Days (t) 1.509 -0.073 1.509 -0.076[0.279] [0.962] [0.283] [0.962]
LT Debt/ Assets(t) -0.117** -0.330*** -0.114** -0.330***[0.014] [0.003] [0.016] [0.003]
R&D/Sales (t) 0.129 0.573** 0.127 0.574**[0.109] [0.015] [0.114] [0.015]
EBIT/Assets (t) 0.803*** 0.886*** 0.789*** 0.872***[0.000] [0.000] [0.000] [0.000]
Intangibles/Assets (t) -0.158*** -0.549*** -0.158*** -0.549***[0.000] [0.000] [0.000] [0.000]
Q(t) 0.034*** 0.006 0.034*** 0.005[0.000] [0.549] [0.000] [0.624]
CAPEX/Sales (t) -0.130*** -0.318*** -0.131*** -0.320***[0.001] [0.000] [0.001] [0.000]
100 x Total Dividends/ Assets (t) 0.828*** 0.503*** 0.829*** 0.505***[0.000] [0.000] [0.000] [0.000]
Constant -0.553*** -0.083 -0.574*** -0.149[0.001] [0.847] [0.001] [0.729]
Year Fixed Effects Yes Yes Yes YesFirm Fixed Effects No Yes No YesIndustry Fixed Effects Yes No Yes No
Observations 19,012 19,012 19,012 19,012R-squared 79.10% 30.30% 79.10% 30.30%
54
Table
14:
Pre
ss-b
ase
dm
easu
reof
overc
onfi
dence
This
table
conta
ins
regre
ssio
nm
odels
that
use
apre
ss-b
ase
dm
easu
reof
overc
onfi
dence,
‘Net
New
s’.
The
vari
able
‘Net
New
s’is
equal
toth
enum
ber
of
‘overc
onfi
dent’
art
icle
sin
year
tle
ssth
enum
ber
of
‘non-o
verc
onfi
dent’
art
icle
sin
yeart.
The
models
inclu
de
the
contr
ols
of
the
main
models
and
inclu
de
indust
ryfi
xed
eff
ects
.G
iven
the
lim
ited
tim
ep
eri
od
(2000-2
006)
we
om
ityear
dum
mie
sfr
om
these
models
infa
vor
of
reta
inin
gth
e‘S
OX
’in
dic
ato
r.T
he
app
endix
conta
ins
the
vari
able
defi
nit
ions.
The
signifi
cance
levels
at
the
1%
,5%
,and
10%
are
denote
dby
***,
**
and
*,
resp
ecti
vely
.
Dep
endent
Vari
able
CA
PX
SG
&A
PP
EA
ssets
EB
ITIn
dA
dj
EB
ITQ
Ind
Adj
QB
eta
MSE
QIn
dA
dj
Ind
Adj
QD
ivid
ends
Win
dow
/T
ime
(t+
1)
(t+
1)
(t,t
+1)
(t,t
+1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
Model
M1
M2
M3
M4
M5
M6
M7
M8
M9
M10
M11
M12
M13
M14
M15
a:
Net
New
s(t)
0.0
83*
0.0
02**
0.0
15***
0.0
14***
-.001**
-.001**
-.058***
-.049***
0.0
45***
-.003
-.038***
-.031***
-.033***
-.037***
-.007
[0.0
64]
[0.0
27]
[0.0
00]
[0.0
00]
[0.0
12]
[0.0
23]
[0.0
00]
[0.0
00]
[0.0
00]
[0.2
59]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.1
62]
b:
SO
X0.4
12**
0.0
06**
.027***
.051***
-0.0
01
-.013***
-0.0
06
-.264***
.394***
-.208***
0.0
26
-.235***
-0.0
02
-.197***
0.0
19
[0.0
10]
[0.0
23]
[0.0
09]
[0.0
00]
[0.7
63]
[0.0
00]
[0.8
13]
[0.0
00]
[0.0
00]
[0.0
00]
[0.3
56]
[0.0
00]
[0.9
23]
[0.0
00]
[0.5
01]
c:
CA
PX
/Sale
s25.4
77***
0.0
02
.208***
0.0
10
-0.1
03
-0.0
43
-0.0
18
0.0
26
0.6
94**
0.7
05**
-0.0
89
-0.0
50
-.210***
[0.0
00]
[0.8
74]
[0.0
00]
[0.8
70]
[0.4
46]
[0.7
46]
[0.8
70]
[0.5
85]
[0.0
19]
[0.0
34]
[0.4
87]
[0.6
96]
[0.0
02]
d:
R&
D/Sale
s-6
.36***
0.0
84***
-0.1
05
-.215***
-.056***
-.055***
0.8
85***
0.8
32***
0.3
56**
0.0
31
0.8
69***
0.8
19***
0.6
70*
0.5
74
0.1
42
[0.0
00]
[0.0
00]
[0.1
36]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
27]
[0.5
02]
[0.0
00]
[0.0
00]
[0.0
98]
[0.1
57]
[0.1
44]
a×
b-0
.058
-.003***
-.009***
-.010***
.003***
.002***
.079***
.064***
-.027***
0.0
00
0.0
65***
0.0
53***
0.0
56***
0.0
57***
0.0
15**
[0.1
71]
[0.0
00]
[0.0
01]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.8
71]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
00]
[0.0
28]
a×
c-.
214***
-.189***
[0.0
00]
[0.0
02]
a×
b×
c0.1
21*
0.0
89
[0.0
71]
[0.1
55]
a×
d-.
317***
-.295***
[0.0
00]
[0.0
01]
a×
b×
d0.2
67***
0.2
53***
[0.0
02]
[0.0
03]
b×
c-0
.294
-0.2
63
[0.3
13]
[0.3
73]
b×
d0.8
17**
0.8
88**
[0.0
27]
[0.0
20]
Contr
ols
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Indust
ryF
EY
es
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Obse
rvati
ons
7,3
41
6,0
77
6,9
03
7,3
51
7,3
39
7,3
39
7,3
51
7,3
51
6,9
94
6,9
94
7,3
51
7,3
51
7,3
51
7,3
51
7,2
72
R-s
quare
d63.9
0%
92.8
0%
12.8
0%
18.2
0%
71.8
0%
69.9
0%
75.1
0%
70.1
0%
46.2
0%
73.8
0%
75.2
0%
70.2
0%
75.8
0%
69.4
0%
77.4
0%
55
Table
15:
Exam
inin
gth
ese
tof
volu
nta
rily
com
pliant
com
panie
s
This
table
conta
ins
regre
ssio
ns
that
focus
on
the
sub-s
am
ple
of
firm
sth
at
were
com
pli
ant
in1998,
1999,
2000,
and
2001.
We
focus
on
two
defi
nit
ions
of
com
pliance:
(1)
havin
ga
100%
indep
endent
audit
com
mit
tee
and
(2)
havin
gm
ore
than
50%
indep
endent
dir
ecto
rson
the
board
.W
eth
en
const
ruct
sub-s
am
ple
sof
com
panie
sth
at
sati
sfy
(in
all
of
1998-2
001)
defi
nit
ion
1(P
anel
A),
defi
nit
ion
2(P
anel
B),
eit
her
defi
nit
ion
1or
defi
nit
ion
2(P
anel
C),
or
both
defi
nit
ion
1and
defi
nit
ion
2(P
anel
D).
The
dep
endent
vari
able
sare
:T
he
firm
’sC
AP
EX
inyeart
+1
scale
dby
its
ass
ets
inyeart
(M1),
SG
&A
/Sale
s(M
2),
ln(P
PEt
+1/P
PEt)
(M3),
ln(A
ssetst
+1/A
ssetst)
(M4),
Ind
Adj
EB
IT/A
ssets
(t+
1)
(M5),
Ind
Adj
Q(t
+1)
(M6,
M9,
M10),
Beta
(t+
1)
(M7),
ln(M
SE
(t+
1))
(M8),
and
Div
idends/
Ass
ets
(t+
1)
(M11).
All
models
conta
inth
econtr
ol
vari
able
sfr
om
the
main
regre
ssio
ns,
whic
hw
eom
itfo
rbre
vit
y.
All
models
inclu
de
firm
fixed
eff
ects
and
year
fixed
eff
ects
and
use
clu
stere
dst
andard
err
ors
as
inth
em
ain
regre
ssio
ns.
Bra
ckets
conta
inp-v
alu
es
and
sup
ers
cri
pts
***,
**,
and
*denote
signifi
cance
at
1%
,5%
,and
10%
,re
specti
vely
.
Dep
endent
Vari
able
CA
PE
XSG
&A
PP
EA
ssets
Ind
Adj
EB
IT(t
+1)
Ind
Adj
Q(t
+1)
Beta
MSE
Ind
Adj
QIn
dA
dj
QD
ivid
ends
Win
dow
/T
ime
(t+
1)
(t+
1)
(t+
1,t
+2)
(t+
1,t
+2)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
(t+
1)
Model
M1
M2
M3
M4
M5
M6
M7
M8
M9
M10
M11
Panel
A:
More
than
50%
indep
endent
dir
ecto
rsC
onfi
denceT
opQ
(t)
0.7
24**
0.0
02
0.0
16
0.0
09
0.0
03
0.1
11**
0.1
15***
0.0
71***
0.1
57**
0.1
73**
-0.1
14***
[0.0
25]
[0.5
55]
[0.1
59]
[0.5
04]
[0.3
05]
[0.0
36]
[0.0
00]
[0.0
00]
[0.0
13]
[0.0
31]
[0.0
07]
SO
X-1
.183**
-0.0
03
-0.0
28
0.0
14
0.0
11*
0.2
07**
0.1
50***
-0.0
52**
0.1
93**
0.4
51***
0.0
1[0
.020]
[0.6
43]
[0.1
76]
[0.5
48]
[0.0
96]
[0.0
10]
[0.0
01]
[0.0
43]
[0.0
16]
[0.0
00]
[0.9
22]
Confi
denceT
opQ
(t)
xSO
X-0
.144
-0.0
05
-0.0
16
-0.0
07
0.0
05
0.0
66
-0.0
38
-0.0
37*
-0.0
02
-0.0
30.1
11**
[0.6
87]
[0.1
38]
[0.2
08]
[0.6
79]
[0.2
26]
[0.2
23]
[0.2
63]
[0.0
53]
[0.9
81]
[0.7
07]
[0.0
32]
Confi
denceT
opQ
(t)
xC
AP
EX
/Sale
s(t)
-0.6
99
[0.2
35]
CA
PE
X/Sale
s(t)
xSO
X0.2
27
[0.2
78]
Confi
denceT
opQ
(t)
xSO
Xx
CA
PE
X/Sale
s(t)
0.9
12
[0.1
65]
Confi
denceT
opQ
(t)
xR
&D
/Sale
s(t)
-0.7
32
[0.6
26]
R&
D/Sale
s(t)
xSO
X0.6
32
[0.5
70]
Confi
denceT
opQ
(t)
xSO
Xx
R&
D/Sale
s(t)
1.9
15
[0.1
93]
Const
ant
18.1
70***
0.0
39
0.3
08**
0.4
26**
0.0
02
1.1
63*
0.9
74***
-3.8
43***
1.2
17**
1.0
89
-0.7
68
[0.0
00]
[0.4
26]
[0.0
38]
[0.0
12]
[0.9
68]
[0.0
60]
[0.0
07]
[0.0
00]
[0.0
49]
[0.1
35]
[0.2
81]
Obse
rvati
ons
4,4
96
3,6
04
4,1
37
4,1
99
4,4
97
4,4
97
4,2
08
4,2
08
4,4
97
4,4
97
4,4
60
R-s
quare
d33.1
0%
49.8
0%
6.2
0%
9.7
0%
41.6
0%
36.6
0%
35.5
0%
66.1
0%
36.8
0%
31.2
0%
48.0
0%
Panel
B:
Indep
endent
Audit
Com
mit
tee
Confi
denceT
opQ
(t)
0.4
70.0
01
0.0
21
0.0
16
0.0
02
0.1
86***
0.0
97***
0.0
62***
0.1
45*
0.2
10**
-0.1
26**
[0.1
56]
[0.7
61]
[0.1
54]
[0.2
90]
[0.6
02]
[0.0
08]
[0.0
02]
[0.0
00]
[0.0
78]
[0.0
23]
[0.0
20]
SO
Xx
Confi
denceT
opQ
(t)
-0.0
13
-0.0
08*
-0.0
09
0.0
01
0.0
07
0.0
36
-0.0
59
-0.0
53**
0.0
74
-0.0
45
0.1
16*
[0.9
74]
[0.0
93]
[0.5
72]
[0.9
79]
[0.1
73]
[0.5
96]
[0.1
49]
[0.0
20]
[0.3
83]
[0.6
12]
[0.0
79]
SO
X3.3
27
0.0
08
-0.0
06
-0.0
01
-0.0
05
-0.7
83
-0.2
97***
0.0
77**
-0.7
82
-0.5
12
-0.3
42*
[0.4
34]
[0.7
06]
[0.7
96]
[0.9
83]
[0.8
55]
[0.4
05]
[0.0
00]
[0.0
42]
[0.4
04]
[0.5
84]
[0.0
69]
Confi
denceT
opQ
(t)
xC
AP
EX
/Sale
s(t)
0.5
3[0
.385]
CA
PE
X/Sale
s(t)
xSO
X0.0
59
[0.7
81]
Confi
denceT
opQ
(t)
xSO
Xx
CA
PE
X/Sale
s(t)
-0.4
58
[0.4
36]
Confi
denceT
opQ
(t)
xR
&D
/Sale
s(t)
-0.0
43
[0.9
80]
R&
D/Sale
s(t)
xSO
X-0
.432
[0.7
82]
Confi
denceT
opQ
(t)
xSO
Xx
R&
D/Sale
s(t)
2.3
24
[0.2
34]
Const
ant
13.2
27**
-0.0
22
0.4
47**
0.5
18**
0.0
31
3.3
92**
0.9
31**
-3.7
30***
3.4
00**
3.5
66**
0.3
2[0
.036]
[0.6
68]
[0.0
17]
[0.0
14]
[0.5
40]
[0.0
11]
[0.0
39]
[0.0
00]
[0.0
11]
[0.0
10]
[0.7
20]
Obse
rvati
ons
2,9
64
2,4
22
2,7
02
2,7
63
2,9
66
2,9
66
2,7
71
2,7
71
2,9
66
2,9
66
2,9
40
R-s
quare
d35.0
0%
52.6
0%
6.9
0%
12.2
0%
44.3
0%
35.6
0%
34.4
0%
66.8
0%
35.6
0%
31.4
0%
48.5
0%
Panel
C:
More
than
50%
Indep
endent
Dir
ecto
rsor
Indep
endent
Audit
Com
mit
tee
56
Confi
denceT
opQ
(t)
0.7
56**
0.0
01
0.0
20*
0.0
09
0.0
03
0.1
14**
0.1
10***
0.0
71***
0.1
16*
0.1
44**
-0.1
29***
[0.0
12]
[0.6
67]
[0.0
89]
[0.4
54]
[0.3
01]
[0.0
24]
[0.0
00]
[0.0
00]
[0.0
50]
[0.0
48]
[0.0
01]
SO
Xx
Confi
denceT
opQ
(t)
-0.0
87
-0.0
05
-0.0
16
-0.0
08
0.0
06
0.0
69
-0.0
44
-0.0
44**
0.0
55
-0.0
01
0.1
38***
[0.7
97]
[0.1
19]
[0.1
77]
[0.5
88]
[0.1
22]
[0.1
68]
[0.1
71]
[0.0
16]
[0.3
82]
[0.9
92]
[0.0
05]
SO
X-1
.410***
-0.0
08
-0.0
29
0.0
12
0.0
12*
0.2
22***
-0.2
73***
0.1
04***
0.2
11***
0.4
68***
0.0
18
[0.0
06]
[0.2
26]
[0.1
31]
[0.6
00]
[0.0
73]
[0.0
05]
[0.0
00]
[0.0
00]
[0.0
07]
[0.0
00]
[0.8
47]
Confi
denceT
opQ
(t)
xC
AP
EX
/Sale
s(t)
-0.0
68
[0.8
89]
CA
PE
X/Sale
s(t)
xSO
X0.2
33
[0.2
15]
Confi
denceT
opQ
(t)
xSO
Xx
CA
PE
X/Sale
s(t)
0.1
56
[0.7
76]
Confi
denceT
opQ
(t)
xR
&D
/Sale
s(t)
-0.1
14
[0.9
33]
R&
D/Sale
s(t)
xSO
X0.5
24
[0.6
09]
Confi
denceT
opQ
(t)
xSO
Xx
R&
D/Sale
s(t)
1.6
39
[0.2
30]
Const
ant
18.6
05***
0.0
32
0.3
73***
0.5
13***
0.0
05
1.4
11**
0.9
54***
-3.7
43***
1.4
58**
1.3
80**
-0.9
93
[0.0
00]
[0.5
02]
[0.0
08]
[0.0
02]
[0.9
00]
[0.0
21]
[0.0
05]
[0.0
00]
[0.0
16]
[0.0
44]
[0.1
54]
Obse
rvati
ons
4,8
83
3,9
26
4,4
86
4,5
58
4,8
85
4,8
85
4,5
68
4,5
68
4,8
85
4,8
85
4,8
45
R-s
quare
d32.7
0%
49.9
0%
6.3
0%
9.6
0%
41.2
0%
36.9
0%
35.2
0%
65.7
0%
37.0
0%
31.5
0%
50.6
0%
Panel
D:
More
than
50%
Indep
endent
Dir
ecto
rsand
Indep
endent
Audit
Com
mit
tee
Confi
denceT
opQ
(t)
0.4
14
0.0
02
0.0
17
0.0
21
0.0
03
0.1
91**
0.0
98***
0.0
62***
0.2
20*
0.2
61**
-0.1
10*
[0.2
46]
[0.5
80]
[0.2
63]
[0.2
15]
[0.6
17]
[0.0
13]
[0.0
03]
[0.0
01]
[0.0
78]
[0.0
13]
[0.0
66]
SO
Xx
Confi
denceT
opQ
(t)
-0.1
47
-0.0
08
-0.0
08
00.0
06
0.0
2-0
.049
-0.0
45*
-0.0
24
-0.1
0.0
79
[0.7
31]
[0.1
13]
[0.6
26]
[0.9
92]
[0.3
52]
[0.7
87]
[0.2
68]
[0.0
65]
[0.8
56]
[0.3
20]
[0.2
86]
SO
X0.8
71
0.0
16
-0.0
29
0.0
06
-0.0
39
-0.9
90.5
42*
-0.1
35
-0.9
94
-0.8
76
-0.5
35***
[0.7
51]
[0.4
37]
[0.2
57]
[0.8
58]
[0.1
63]
[0.3
74]
[0.0
56]
[0.3
94]
[0.3
72]
[0.4
43]
[0.0
05]
Confi
denceT
opQ
(t)
xC
AP
EX
/Sale
s(t)
-0.4
15
[0.7
96]
CA
PE
X/Sale
s(t)
xSO
X0.0
26
[0.9
11]
Confi
denceT
opQ
(t)
xSO
Xx
CA
PE
X/Sale
s(t)
0.6
28
[0.6
94]
Confi
denceT
opQ
(t)
xR
&D
/Sale
s(t)
-0.8
83
[0.6
56]
R&
D/Sale
s(t)
xSO
X-0
.338
[0.8
51]
Confi
denceT
opQ
(t)
xSO
Xx
R&
D/Sale
s(t)
2.7
74
[0.1
99]
Const
ant
15.6
41***
-0.0
20.3
81*
0.4
61**
0.0
39
3.3
25**
0.1
84
-4.0
96***
3.3
26**
3.5
73**
0.6
7[0
.009]
[0.7
07]
[0.0
60]
[0.0
38]
[0.4
08]
[0.0
21]
[0.7
37]
[0.0
00]
[0.0
22]
[0.0
22]
[0.4
45]
Obse
rvati
ons
2,5
91
2,1
00
2,3
66
2,4
17
2,5
92
2,5
92
2,4
24
2,4
24
2,5
92
2,5
92
2,5
69
R-s
quare
d36.4
0%
52.5
0%
6.8
0%
12.9
0%
45.3
0%
35.1
0%
34.9
0%
67.7
0%
35.1
0%
30.9
0%
44.9
0%
57
Top Related