board management: an analysis of investor optimism, accounting...
TRANSCRIPT
-
1
Board Management: An Analysis of Investor Optimism,
Accounting Expertise of Board of Directors, and Financial
Restatements
Shantaram Hegde
School of Business
University of Connecticut
2100 Hillside Road Unit-1041
Storrs, CT 06269, U.S.A.
Phone: +1 (860) 486-5135
Tingyu Zhou
John Molson School of Business
Concordia University
1455 de Maisonneuve Boulevard West
Montreal, Quebec H3G 1M8, Canada
Phone: +1 (514) 848-2424, ext. 2459
November 27, 2017
Abstract
We document that investor optimism about industry-wide business prospects is associated with
a significant decrease in the firm-level appointment of directors with accounting expertise and
an increase in the incidence of accounting irregularities. In addition, a decrease in the
accounting expertise of the board is associated with a higher likelihood of both accounting
irregularities and errors. These findings are robust to alternative measures of investor optimism
and the accounting composition of the board. Our results highlight opportunistic board
management by incumbent management and directors as a novel channel through which
investor optimism influences the accounting composition of corporate boards and fin
restatements. (102 words)
Keywords: investor optimism; accounting expertise; financial reporting; restatements
JEL classification: M41; G30; K30
___________________________ * We appreciate insightful comments from Staurt Gillan, Johan Sulaeman, Jie Zhu, Assaf Eisdorfer,
Efdal Misirli, John Clapp, Joseph Golec, John Glascock and seminar participants at University of
Connecticut and the CFA-JCF-Schulich Conference on Financial Market Misconduct. We are very
grateful to Andrew J. Leone who generously shared the General Accountability Office (GAO) data on
classification of errors and irregularities.
PRELIMINARY – PLEASE DONOT QUOTE OR CITE
-
2
Board Management: An Analysis of Investor Optimism, Accounting Expertise of Board
of Directors, and Financial Restatements
Abstract
We document that investor optimism about industry-wide business prospects is associated with
a significant decrease in the firm-level appointment of directors with accounting expertise and
an increase in the incidence of accounting irregularities. In addition, a decrease in the
accounting expertise of the board is associated with a higher likelihood of both accounting
irregularities and errors. These findings are robust to alternative measures of investor optimism
and the accounting composition of the board. Our results highlight opportunistic board
management by incumbent management and directors as a novel channel through which
investor optimism influences the accounting composition of corporate boards and fin
restatements. (102 words)
Keywords: investor optimism; accounting expertise; financial reporting; restatements
JEL classification: M41; G30; K30
-
3
1. Introduction
An important consequence of the separation of ownership and control in corporations is the
conflict of interest between stockholders and managers over who controls the nomination and
election of board of directors. Although outside stockholders have formal control over board
structure under the standard one share – one vote system, the incumbent management has
strong incentives to dominate the board in order to strengthen their power over business
strategy and extract private benefits. The extant literature on corporate governance shows that
in practice investors have little power over director appointment. Instead, the incumbent
management and directors enjoy effective control over the election of corporate boards (Aghion
and Bolton, 1992; Hermalin and Weisbach, 1998; Cai et al. 2009 and 2017; Adams, Hermalin
and Weisbach, 2010). Another strand of literature on behavioral biases indicates that while
investors monitor directors and managers more intensely when they are pessimistic about the
overall business environment, they slacken their oversight over insiders when they feel very
optimistic about business conditions, contributing at times to increased likelihood of corporate
fraud (Shleifer and Vishny, 1997; Povel, Singh, and Winton, 2007; Wang, Winton, and Yu,
2010).
Corporate boards are facing increasing financial reporting complexity, broadly defined as
managers’ willingness and ability to properly (and truthfully) apply generally accepted
accounting principles and communicate the overall financial performance and health of a
company (SEC 2008, p. 18). A stronger accounting profile is likely to endow the board with
comprehensive and deeper knowledge and expertise needed to scrutinize and challenge the
financial reporting choices of the management and improve financial reporting quality. Prior
studies show that the accounting expertise of directors is associated with a lower likelihood of
financial misreporting (see, for example, Krishnan and Visvanathan 2008; Dhaliwal et al., 2010;
-
4
Zhang et al., 2007; Carcello et al., 2009; Badolato et al., 2014).1 In light of the pivotal role
of accounting expertise of directors in determining the effectiveness of their monitoring and
advising roles, we focus on the how investor beliefs influence the accounting profile of the
board.2
It seems reasonable to expect that this environment creates opportunities for the incumbent
management and directors to manipulate the accounting composition of the board. Aware that
investors are monitoring the firm more closely during periods of low optimism, insiders may
strategically refrain from lowering the independence and accounting composition of the board.
However, when stockholders are very optimistic about macroeconomic conditions, they tend
to be less concerned with board performance in advising and monitoring of the management
over conflicts of interest. Insiders are more likely exploit this window of high investor
optimism to weaken the effectiveness of board monitoring by reducing the appointment of
independent directors with accounting expertise and manipulating financial statements.
Specifically, we investigate two hypotheses regarding the propensity of incumbent managers
and directors to opportunistically manage the accounting composition of corporate boards. The
Board Management Hypothesis posits that the appointment of directors with accounting
expertise decreases as investor optimism about the state of the economy increases. In addition,
the Accounting Misconduct Hypothesis claims that a decrease in the appointment of directors
with accounting expertise and an increase in investor optimism about business prospects are
both related positively to the incidence of accounting misconduct.
1 We use “accounting restatements” and “accounting misstatements” as interchangeable terms. Our sample includes some accounting-related fraudulent cases from SEC Accounting and Auditing Enforcement Releases (AAER) and Stanford Law
School’s Securities Class Action Clearinghouse (SCAC). As compared to GAO and AA databases, AAER and SCAC use a
fundamentally different way to identify accounting failures. Since we are interested in comparing intentional and unintentional
misstatements, we focus only on the GAO and AA datasets. We elaborate on our sample construction in Section 3, and thank
Stuart Gillan for his valuable comments on our sample construction.
2 A 2015 survey of U.S. corporations indicates that accounting/financial expertise is one of the top three backgrounds on their list of desired director attributes. About 24% of new board members in the S&P 500 are financial experts, although 54% of
boards report that they want new directors with financial expertise, https://www.spencerstuart.com/research-and-
insight/spencer-stuart-us-board-index-2015.
https://www.spencerstuart.com/research-and-insight/spencer-stuart-us-board-index-2015https://www.spencerstuart.com/research-and-insight/spencer-stuart-us-board-index-2015
-
5
We use annual industry median EPS growth forecasts of analysts(based on the Fama-French
(1987) classification) to proxy for investor optimism, firm-level appointment of directors with
accounting expertise to corporate boards, and financial restatements over 1996-2012 to test
these hypotheses. Our proxies for financial misconduct are intentional accounting
misstatements (irregularities) and unintentional misreporting (errors), which are driven by
different underlying forces (Hennes et al., 2008).3 Our first main finding is that annual
appointment of directors with accounting expertise is negatively related to investor optimism
regarding business prospects as proxied by industry median EPS growth forecast, consistent
with our hypothesis that high optimism allows insiders to opportunistically weaken the
accounting composition of the board. This result is robust to multiple measures of director
accounting expertise (the presence or absence, number and proportion of directors with
accounting expertise among independent directors, audit committee members and all directors)
and industry-level Tobin’s Q as an alternative measure of investor optimism.
In addition, we find a strong positive association between investor optimism and the
likelihood of accounting restatements and a negative relation between director accounting
expertise and accounting misconduct. This result is in line with our argument that more
optimistic beliefs about business conditions lead to a higher likelihood of financial misconduct
by weakening the effectiveness of board oversight.
It is important to note that we treat investor optimism about macroeconomic conditions as
an exogenous variable, not affected by the propensity of firms’ management to engage in
misconduct. This treatment is consistent with the fraud model of Povel et al. (2007) and its
empirical testing in the context of fraud in initial public offerings by Wang et al. (2010). We
further emphasize that we are not examining investor optimism about an individual firm’s
3 Intentional misrepresentations are necessary but not sufficient conditions for classifying misconduct as accounting fraud. That is, an accounting fraud is an irregularity, but an irregularity is not necessarily a fraud. Hence, an analysis of irregularities
allows us to study a broader class of accounting misconduct.
-
6
business prospects, which can be affected by misleading financial reports and earnings
management by the firm. The remaining two key variables – director accounting expertise and
financial misreporting are endogenous variables in our analysis.
It is plausible that accounting experts are less likely to work at firms with high financial
reporting risk, and firms that are more likely to commit accounting irregularities are less
inclined to select directors with accounting expertise. Further, our study of accounting
misconduct is subject to partial observability in that we cannot observe all irregularities and
errors but only those that are eventually disclosed (i.e., we do not have data on undetected
misstatements). These identification concerns indicate that our argument that stronger
monitoring by more accounting experts on the board is associated with lower likelihood of
financial misdeeds is subject to sample selection and endogeneity biases. We conduct a battery
of empirical tests including propensity-score matching, two-stage bivariate probit models and
instrumental variables to address concerns about observable and unobservable heterogeneity.
Results drawn from propensity-score matched samples based on investor optimism,
different types of board monitoring mechanisms and other observable firm attributes and
obtained from tests using binary, categorical and continuous treatments indicate that director
accounting expertise is associated with lower likelihood of both types of misconduct –
accounting irregularities and errors. Further, the likelihood of irregularities is positively related
to investor optimism about industry prospects but this relation rarely exists in errors. Our results
are robust to controls for executive compensation and other types of board monitoring
mechanisms such as board size, board independence, independence of audit committee, and
non-accounting financial expertise of directors.
To address the problem of partial observability of accounting misstatements, we examine a
bivariate probit model and find consistent results. Our results are also robust to two-stage
bivariate probit which uses predicted rather than actual measures of director accounting
-
7
expertise. Moreover, we conduct an instrumental variable (IV) analysis to mitigate endogeneity
concerns by exploiting the role of the local labor market in supplying directors. Specifically,
we use the top metropolitan statistical areas with the largest number of headquarters of firms
and the population of the county where a firm’s headquarter is located as instruments for
director accounting expertise (Knyazeva et al., 2013; Badolato et al., 2014; Nguyen et al., 2015).
Both the instrumental variables are expected to correlate positively with board accounting
expertise but not with financial misreporting. In all these specifications, we control for board
monitoring such as board size, board independence and independence of audit committee.
These results confirm our original findings that monitoring by accounting experts on the board
is correlated with a reduction in the likelihood of both irregularities and errors even after
controlling for the level of investor optimism.
In our S&P 1500 sample, about 29% of our sample firm-years do not have any director
with accounting expertise. Our analysis shows that, all else equal, a one-standard-deviation
increase from the mean investor optimism is associated with a decrease in the likelihood of the
presence of accounting experts among independent directors of 3%. The corresponding figures
for all directors and audit committee members are 3% and 7%, respectively. In turn, the
predicted probability of irregularities (roughly) doubles for firms without any director with
accounting expertise as compared to firms with at least one director with accounting expertise
(holding other variables at their means) and the predicted probability of errors increases by 20%
for firms without director accounting expertise. In addition, although the increase in investor
optimism has a significant (positive but non-linear) impact on the likelihood of accounting
irregularities, the net effect of the positive impact of investor optimism and the negative impact
of board accounting expertise on the incidence of misstatements is statistically insignificant.
These results imply that investors, corporate boards and regulators can counteract the increased
likelihood of accounting restatements due to investor optimism by appointing more directors
-
8
with accounting expertise.
This study makes the following noteworthy contributions First of all, this is the first paper,
to the best of our knowledge, to investigate whether investor optimism about the state of the
economy influences the probability of appointment of directors with accounting expertise to
the board, whether investor optimism increases the incidence of intentional misreporting, and
whether director accounting expertise mitigates the exacerbating effects of investor optimism
on the likelihood of financial misstatements.
The literature on earnings management indicates that firms may opportunistically shift
earnings from good to bad economic conditions, and report more negative discretionary
accruals during uncertain times (Stein and Wang, 2016). De Bondt and Thaler (1985) investors
tend to overreact to unexpected and dramatic news, driving stock valuations away from
fundementals. Benartzi and Thaler (1995) investors dislike losses more than they value gains.
Antoniou, Doukas, and Subrahmanyam (2016) indicate that high investor optimism lowers
investor monitoring intensity because it attracts unsophisticated (noise) investors, while
rational investors with greater monitoring effectiveness are more active in trading during
periods of investor pessimism. We contribute to the behavioral finance literature by showing
that investor overoptimism is associated with lower odds of appointment of directors with
accounting expertise, consistent with the notion that incumbent management and directors
attempt to biased corporate boards to be more ‘insider-friendly’ than ‘shareholder-friendly’.
Povel et al. (2007) predict that the probability of corporate fraud increases with investor
beliefs about the state of the economy because of weakened investor monitoring over
management (also see Shleifer and Vishny (1997), and Ball (2009)). But the incidence of fraud
drops when investors are over-optiomistic as managers are able to obtain unmonitored funding.
Supporting these predictions, Wang et al. (2010) find a hump-shaped relationship between
investor beliefs and firms’ propensity to commit fraud at initial public offerings. We extend this
-
9
line of research and show that high investor optimism is associated with lower odds of
appointment of directors with accounting expertise and the resulting decrease in board
monitoring is related to an increase in the likelihood of accounting misconduct.
Gillan, Hartzell, and Starks (2006) suggest that an independent board can act as a substitute
for an active takeover market. Song and Thakor (2006) predict that the intensity of monitoring
of projects and strategies by corporate boards is greater when directors are pessimistic about
business conditions but weaker when times are good, and CEOs prefer less talented boards
during good times with higher probability of good investment projects. Ferreira, Ferreira, and
Raposo (2011) focus on adverse selection issues (about management quality) and find that
stock price informativeness acts as a substitute for board independence, particularly strong for
firms with better external and internal governance mechanisms. We contribute to this line of
research by highlighting that high investor optimism makes investors indifferent to board
oversight and allows for board management by insiders. Our results suggest that higher odds
of diminished board oversight and increased accounting misconduct during periods of high
investor optimism can lower earnings informativeness.
Cai, Garner, and Walking (2009) and Cai, Nguyen, and Walking (2017) find that
shareholders do not have much influence in the nomination, appointment, or removal of
directors, yet fewer shareholder votes result in lower abnormal CEO compensation and a higher
probability of removing poison pills, classified boards, and CEOs. Gal-Or, Hoitash, and
Hoitash (2016) report that shareholders offer more voting support in elections to directors who
serve on the audit committee of the board and hold them accountable for the quality of financial
reporting. Our analysis highlights opportunistic board management as a new channel through
which investor optimism influences the accounting composition of corporate boards and
financial restatements. It suggests that regulatory requirement on the minimum proportion of
directors with accounting expertise is warranted in order to strengthen board oversight during
-
10
periods of high investor optimism.
We develop in Section 2 our hypotheses about the impact of investor optimism on the
proportion of directors with accounting-specific expertise and the effects of both of these
variables on the likelihood of irregularities and errors. Section 3 describes our sample. Sections
4 discusses our empirical tests and results, and Section 5 concludes the paper.
2. Hypotheses Development
A large body of literature on conflict of interest between stockholders and managers
highlights that incumbent management and directors have strong incentives to dominate board
nomination and election processes to gain the support of directors for their business strategy
and operating activities and dampen potential challenges to their attempts to extract private
benefits. Although in theory shareholders elect directors, there is widespread evidence that in
practice the director nomination and voting processes are biased in favor of insiders. (Aghion
and Bolton, 1992). Hermalin and Weisbach, 1998; Holmstrom and Tirole, 1993; Ferreira et al.,
2011; Adams, Hermalin and Weisbach, (2010) Cai et al. (2009 and 2017). Critics note that
many current board practices contribute to board entrenchment, lack of board refreshment,
inadequate director evaluation, term and age limits, succession planning, inadequate
managerial effort, empire building, private benefits extraction, lax board oversight, and lower
firm value (Manne, 1965; Jensen, 1988, 1993; Bebchuk and Cohen, 2005; Bebchuk, Cohen,
and Ferrell, 2009).
Another strand of studies highlights that investors intensify monitoring over management
when they are less optimistic about macroeconomic conditions but slacken it when they feel
very optimistic. PSW (2007)and WWY (2010) note that the probability of corporate fraud
increases initially with investor optimism but decreases when investors are overoptimistic
-
11
because investors are willing to provide unmonitored funding. De Bondt and Thaler (1985)
point out that investors tend to overreact to unexpected and dramatic news, driving stock
valuations away from fundementals. In Benartzi and Thaler (1995), investors dislike losses
more than they value gains. Song and Thakor (2006) predict that the intensity of monitoring of
projects and strategies by the board of directors is higher when directors are pessimistic about
business conditions but weaker when times are good and CEOs prefer less able boards during
economic upturns when there are many good investment projects to choose from.
Antoniou, Doukas, and Subrahmanyam (2016) focus on the effects of investor optimism
on the relation between systematic risk (beta) and stock returns. They suggest the monitoring
effects of stock prices is lower when investor optimism is high because overoptimism attracts
unsophisticated investors, leading to overvaluation and less informative stock prices. By
contrast, rational investors are more active in trading during periods of investor pessimism,
which increases monitoring effectiveness. IO provides the management with the opportunity
to manage board composition.
Against this backdrop, we focus on how investor optimism influences the accounting
composition of the board of directors. Financial statements have become increasingly complex
over time, and accounting misconduct is very costly to stockholders (Agrawal and Chadha,
2005; Kedia and Philippon, 2009; Bhattacharya, U. and C.D. Marshall, 2012).4 We expect that
less optimistic investors view board accounting expertise as a natural, first-order remedy for
curbing managerial propensity to manipulate financial reports. Similar to venture capitalists
and investment banks in the context of IPOs (WWY, 2010), a board equipped with more
specialized financial expertise, knowledge and experience in accounting and auditing should
face lower monitoring costs, have the courage to scrutinize and challenge the financial
reporting choices of top executives, monitor managers more effectively, prevent and detect
4 In addition, managers and directors face substantial loss of reputation and legal costs if the firm engages in fraudulent behavior (Srinivasan, 2005; Helland, 2006; Fich and Shivdasani, 2007).
-
12
accounting misconduct, and improve financial reporting quality (see Guner et al., 2008; Duchin
et al., 2010; Badolato et al. 2014). Thus, a higher proportion of directors with AE is likely to
strengthen the authority, competence and willingness of the board to confront managers over
questionable accounting practices and enhance the deterrence effect of their monitoring on the
likelihood of irregularities. This argument suggests that incumbent management and directors
have a strong propensity to dampen the accounting composition of the board in order to pursue
private benefits.5
As prior studies show, low investor optimism about business conditions intensifies
investor monitoring through a variety of channels including takeover threats, increased
informativeness of stock prices, activist campaigns for more independent directors, industry
expertise of directors, limits on board tenure, director age limits, more scrutiny of individual
director’s and board performance. Despite the fact that shareholders have limited influence
over director nomination and election, poorly performing firms may face mounting pressure
from activist investors to restructure their operations and are more likely concede a board seat
in order to avoid the distraction of a potential proxy fight.6 Even when proxy fights and
campaigns by activist and institutional investors fail to gain board seats in director elections,
they push the incumbent management to not only take a hard look at their existing strategies
and policies but also refrain from weakening board oversight. Therefore, we anticipate that
intense monitoring and mounting pressure for improved firm performance constrains insider
attempts to weaken the accounting profile of the board of directors during periods of low
investor optimism.
However, when investors are very optimistic, their indifference and lax monitoring allows
5 Studies on corporate governance discuss many other attributes that influence board advising and monitoriing effectiveness, such as, board independence, size, age limit, tenure, CEO-chairman duality, staggered boards, busy boards, and so on (Adams, et al. (2010). Wang, Xie, and Zhu (2016) find that relevant industry expertise improves independent directors’ monitoring effectiveness. 6 For example, faced with a sharp decline in stock price, General Electric Co. bowed to mounting pressure from activist investor Trian Fund Management and agreed to give the fund a seat on its board. http://www.foxbusiness.com/features/2017/10/10/ge-adds-activist-to-board-as-stock-slumps-wsj.html
-
13
insiders a window of opportunity to dampen the appointment of directors with accounting
expertise, turn the board around to be more ‘management-friendly’ and insulate themselves
from market discipline.
These arguments lead us to formulate the following hypothesis regarding the impact of
the exogenous investor optimism on the accounting profile of the board:
Board Management Hypothesis: The appointment of directors with accounting expertise decreases as investor optimism about the state of the economy increases.
xxxxxxxxxxxx
In early 2000, there was significant debate among regulators, investors, and researchers
whether only individuals with direct accounting knowledge should be considered as ‘‘financial
experts.’’ The SEC defined a financial expert as one who has direct accounting knowledge or
experience (SEC, 2002). Its final rule broadened the definition to allow persons with either
accounting (‘‘accounting financial expert’’) or non-accounting financial expertise
(‘‘supervisory financial expert’’) to be designated as a ‘‘financial expert” (SEC, 2003). Recent
literature suggests that the SEC 2003 definition does not capture the essential monitoring power
of accounting expertize as well as its original narrow definition (Carcello et al. 2009; Zhang et
al. 2007; Krishnan and Visvanathan, 2008). Defond et al. (2005) find positive market reaction
to the appointment of accounting financial experts but no reaction to non-accounting financial
experts assigned to audit committees. Krishnan and Visvanathan (2008) find that accounting
conservatism is positively related to accounting financial expertise but not related to non-
accounting financial expertise and nonfinancial expertise. Guided by these studies, we follow
the narrow definition in SEC (2002) and define accounting experts as ‘‘accounting financial
experts.’’
The SOX mandates that a firm must disclose whether at least one member of the audit
committee is a financial expert. If not, the firm has to explain why. Further, the listing stock
exchanges, such as NYSE and NASDAQ, require that at least one member of the audit
-
14
committee have accounting or related financial management expertise and/or experience
(Krishnan and Visvanathan, 2008). These rules suggest that investors may be able to influence
how many additional directors with accounting expertise, over and above the minimum set by
the regulatory and listing provisions, are appointed to the board depending on their optimistic
beliefs.
Prior studies point out that the presence of accounting experts on the board improves a
firm’s financial reporting quality. For example, Agrawal and Chadha (2005) and Carcello et al.
(2011) find that independent audit committees and audit committee financial experts are
generally effective in monitoring the financial reporting and auditing processes, thus lowering
the incidence of restatements. Badolato et al. (2014) find that audit committees with both
financial expertise and high status relative to top management lead to lower levels of earnings
management, as measured by accounting irregularities and abnormal accruals. As optimistic
investor beliefs about macroeconomic prospects dampen the likelihood of appointment of
directors with accounting expertise, we expect the combination of high optimism and lower
accounting expertise to increase the probability of accounting misconduct, leading us to
advance the following hypothesis:
Accounting Misconduct Hypothesis: A decrease in the appointment of directors with accounting expertise and an increase in investor optimism about business prospects are both related positively to the incidence of accounting misconduct.
As highlighted earlier, it is important to note that investor optimism about booms or busts
is an exogenous variable that cannot be affected by an individual firm’s actions (see in the
context of Povel et al. (2007) and WWY (2010)). By contrast, the accounting composition of
the board and financial misreporting are endogenous variables that the management can
influence.
3. Data and Sample Construction
3.1 S&P 1500 and COMPUSTAT-CRSP Merged Samples
Given our focus on the impact of investor optimism about business prospects on accounting
-
15
composition of the board and financial misconduct, we select a sample of financial restatements
from 1996 to 2012 from three different sources: General Accountability Office (GAO, 2002
and 2006), and Audit Analytics (AA).7 The GAO (2002, 2006) database contains accounting
restatements announced between January 1996 and September 2005. We extend this sample by
including additional accounting restatements identified by AA from October 2005 to December
2012.8 Since we only observe accounting mistakes that are committed and then detected, we
need to collect information on both the year of commission (the first misstated period) and the
year of detection (or announcement) for each restatement to address partial observability.
Restatements prior to 1996 are excluded to minimize the confounding effects of the passage of
the Private Securities Litigation Reform Act in 1995 which might affect a firm’s incentives to
engage in fraudulent behavior. Since the median length of the misstated period (from
commission to detection) in our sample is about two years, we exclude cases with commission
dates after 2010 to mitigate data truncation. We follow Hennes et al. (2008) and classify each
case of restatements into either an irregularity or an error.9 Our comprehensive (merged)
restatement sample consists of 868 cases of irregularity and 4,730 cases of errors spanning 15
years.
We hand-collect data on director accounting expertise from proxy statements. As manual
data collection is quite time-consuming, we focus on S&P1500 firms (for which data on board
characteristics are available in RiskMetrics) to gather a reasonable sample on the appointment
7 Earlier studies that use a combined sample of GAO and AA include Badertscher et al. (2011), Ettredge et al. (2012) and Files (2012). 8 We cross-check the overlapping cases in both GAO and AA as the latter (AA) database covers misstatements from January 2000. As one case might be associated with more than one event (restatement), we carefully investigate each case and keep
only the earliest event. 9 The GAO data on the classification of errors versus irregularities are generously provided by Professor Andrew J. Leone (http://sbaleone.bus.miami.edu/). In the AA database, there are two variables which help us distinguish irregularity from errors.
One is “Res_fraud”, which equals 1 if the restatement identified financial fraud, irregularities and misrepresentations. The
other is “Res_sec_investigation”, which equals 1 if the restatement disclosure identified that the SEC, PCAOB or another
regulatory body is investigating the registrant. Including “Res_sec_investigation” is consistent with the procedure to
distinguish irregularities from errors employed by Hennes, Leone and Miller (2008), where the first step is self-disclosure of
irregularity or fraud, the second step is SEC investigation and the third step is non-SEC investigation (Hennes et al. 2008,
p.1494). Badertscher (2011) applies a similar method to identify irregularities in AA.
http://sbaleone.bus.miami.edu/
-
16
and departure of directors with accounting expertise. This choice significantly reduces our
restatement sample to 193 irregularities and 922 errors for which firm-level data on director
accounting expertise are available as of the beginning date (commission) of misstatements. The
existing literature has documented a significant industry effect in the accounting expertise of
directors due to the complexity of accounting practice in some industries (Bills et al., 2013).
We exploit this evidence to relax the data constraint due to firm-level board accounting
expertise, estimate median values of board A accounting expertise for the Fama and French
(1997) 49 industry categories and use them in the bivariate Probit analysis. This yields a
broader restatement sample of 737 irregularities and 3,997 errors with non-missing values of
industry-level accounting expertise.10
Our control sample for testing the relation between director accounting expertise and
likelihood of accounting misstatements includes all firm-years in the COMPUSTAT-CRSP
merged database except firms that are in the restatement sample. Also we delete shell holding
companies and acquisition vehicles (SIC code 99) because the characteristics of these firms
change dramatically after acquisition. The initial control sample includes 96,667 firm-year
observations. After merging it with our measures of firm-level board accounting expertise, the
control sample drops to 20,907 observations. As a large set of control variables in regression
analyses imposes further reduction in sample size, we provide a detailed description on sample
construction in Appendix T.1B.
3.2 Measures of Investor Optimism and Director Accounting Expertise
As emphasized earlier, investor optimism about the state of the economy is an exogenous
variable in our setting. Following Wang et al. (2010), we use median annual earnings per share
(EPS) consensus growth forecasts of analysts for a firm’s industry (Indus. EPS Growth) based
10 Bivariate Probit analyses further reduce the sample size because of additional control variables and the requirement of data availability at both commission and disclosure of misconduct (see Appendix 1B for details on sample construction).
-
17
on Fama-French 49 industry categories as our primary proxy for investor optimism.11 Greater
accounting expertise of the board of directors should lower their own monitoring costs while
strengthening the confidence of investors in the delegated monitors’ capabilities. We define a
member of the board of directors as an accounting expert if the individual is a CPA and/or has
experience as a public accountant, auditor, principal or chief financial officer, controller, or
principal or chief accounting officer. This definition is the same as that of accounting financial
experts in the original SEC proposal in 2002 (SEC 2002). As emphasized earlier, recent
literature concludes that this measure is more effective in capturing monitoring effectiveness
than the revised definition in SEC’s final ruling (SEC 2003).12
To conduct a comprehensive analysis, we use six firm-level measures (three binary and
three continuous) and three industry-level measures (continuous) of accounting expertise of
directors summarized in Appendix T.1A. The six firm-level measures include (a) the presence
and the proportion of accounting experts among independent directors (Firm_AE_IndDir_D
and Firm %AE_IndDir), (b) the presence and the proportion of accounting experts among all
directors on the board (Firm_AE_AllDir_D and Firm %AE_AllDir), (c) the presence and the
proportion of accounting experts on the audit committee (Firm_AE_AuditDir_D and
Firm %AE_AuditDir). Firm_AE_IndDir_D is a dummy variable equal to one if there is at least
one director with accounting expertise among independent directors. Firm%AE_IndDir is the
number of independent directors with accounting expertise divided by the total number of
independent directors. Accounting expertise measures of all directors and audit committee
members are defined in an analogous fashion.
11 To mitigate endogeneity concerns, we exclude firms with misstatements when computing industry median proxies for investor optimism. 12 “Proposed Rule: Disclosure Required by 404, 406 and 407 of the Sarbanes-Oxley Act of 2002. U.S. Securities and Exchange Commission, 2002” (SEC 2002) originally proposed that accounting financial experts are individuals who have knowledge of GAAP that has been obtained through direct experience in accounting and/or auditing positions, which would have only
included individuals with experience as a public accountant, auditor, principal or chief financial officer, controller, or principal
or chief accounting office. In the “Final Rule: Disclosure Required by Sections 406 and 407 of the Sarbanes-Oxley Act of
2002. U.S. Securities and Exchange Commission, 2003” (SEC 2003), it relaxed the range and allowed for individuals who
have obtained knowledge of GAAP by supervising or otherwise monitoring the performance of others who are directly engaged
in accounting and/or auditing functions to qualify.
-
18
By far the most important accounting expertise measure we use is Firm%AE_IndDir as this
measure combines board independence as well as the depth of its accounting expertise, thus
making it a strong instrument of board monitoring. We also construct Firm%AE_AllDir
because boards may not have many independent accounting experts. Since previous studies
focus on the effects of audit committee on accounting misreporting, we include
Firm%AE_AuditDir. In bivariate probit analyses, we construct industry-level accounting
expertise measures by using the industry median %AE of firms in a given year. Similar to firm-
level accounting expertise measures, we construct Indus.%AE_IndDir, Indus.%AE_AllDir and
Indus.%AE_AuditDir.13
4. Empirical Results
4.1. Summary Statistics
Panel A of Table 1 presents the distribution of number of directors with accounting expertise
in our S&P 1500 sample over 1996-2010. Of the total of 22,022 firm-year observations, 6,362
firm-years (29% of total) have no directors with accounting expertise.14 In untabulated results
we find that the mean and median numbers of directors per firm are 9.47 and 9. Of these, the
mean and median numbers of independent directors are 6.51 and 6, and the corresponding data
for the audit committee members are 3.14 and 3. Roughly 46% of independent directors and
55% audit committees in the overall sample have no directors with accounting expertise. The
minimum and maximum number of directors with accounting expertise per board (or firm) are
0 and 7, respectively. The number of directors with accounting expertise per board varies from
1 to 3 for the majority (15,017 out of 22,022) of firm-years. Column (2) indicates that the
number of firms with no director with accounting expertise decreases from 705 in 1996 to 165
13 We construct only Indus. %AE because a dummy variable for the industry-level DAE has no practical meaning. 14 The entries in the “Total” column vary from 1,500 because (1) our sample is based on RiskMetrics, which includes a few additional large-cap companies in earlier years or (2) missing data on DAE. In addition, the gap reflects the additions and
deletions of firms from the index.
-
19
in 2010, while the next three columns show that the number of firms with 1 to 3 accounting
experts on the board has increased over our sample period, especially after SOX.
In column (1) of Panel B, we find that the average annual industry-level analyst EPS growth
forecast is 11.11%, varying from -24.06% to 16.38%. Column (2) shows our S&P 1500 sample
has 1.2092 accounting experts per firm on average, increasing from 0.6235 in 1996 to 1.7833
in 2010. The grand mean proportion of independent directors with accounting expertise among
all independent directors is 11.49%, which also increases from 5.72% to 20.38%, see the last
column.
Panel C shows the distribution of accounting expertise (# for numbers and % for percentage)
among independent directors, all directors, and directors serving on audit committees, sorted
by the deciles of investor optimism measured by industry median annual EPS growth forecast.
As shown in column (3), the average % and number (#) of independent DAE in the first decile
of investor optimism are 0.1626 and 1.1052, respectively, per firm. They drop to 0.1133 and
0.6548 in the last decile of investor optimism. The t- and z-statistics in the bottom rows show
that the mean decreases in both DAE measures are highly significant across all measures. The
negative correlation between investor optimism and DAE is confirmed by additional tests on
whether the mean and median DAE are lower for firm-years associated with high (above
median) investor beliefs relative to low optimism, see Appendix T.2. This preliminary evidence
offers strong support for our Board Management Hypothesis that investor optimism is
negatively related to the accounting makeup of the board.
Our first objective is to study how investor optimism about the state of the economy
contributes to this variation in the number of accounting experts serving on the board. In the
U.S. nominating committees appointed by the top management and incumbent boards put
together a roster of candidates for director election or reelection. It is typical to hold annual
election for all board members, but some firms have staggered (i.e., classified) boards in which
-
20
only a fraction of the board members is elected each year. For example, directors may serve
three-year terms, so only a third of the board is up for election each year. Classified boards are
unpopular with many institutional and activist investors, and a large majority of boards have
been declassified over time.15
To probe the time-series and cross-sectional variation in directors with accounting expertise,
we present the distribution of annual turnover among independent directors with accounting
expertise in Table 2. The sample is drawn from S&P 1500 firms over 1996-2010 (with complete
data on all control variables used in regression analysis in Table 4). Panel A shows the annual
change in the number of independent DAE ranges from -2 to +3. We define the annual change
as number of independent DAE in year t minus those in year t-1. For example, “0” in the Annual
Change row means no change in DAE from the past year. Entries within the table show the
number of firms experiencing the reported annual turnover by year. The last row of Panel A
indicates the net annual change in the number of independent DAE is zero for 5,386 out of
6,379 firm-years (84% of observations). Net annual additions of one-to-three directors occur
in 749 firm-years, while 244 firm-years are marked by net decreases of one-to-two directors.
Panel B (Panel C) tabulates the distribution of number of independent directors with accounting
expertise added to (i.e., appointment) and removed from (i.e., departures) corporate boards by
year. In Panel B, annual appointment of one-to three independent DAE occurs in 11% of firm-
years, while in Panel C about 9% of firm-years exhibit annual departure of one-to-three (or
more) independent DAE. Overall, these board renewal through annual elections patterns are
consistent with the widely documented evidence of low turnover among directors, but show
(considerable) entry and exit of DAE in 20% of our sample observations.16
15 Only about 50 of the S&P 500 companies still have staggered boards, one-sixth of the number just 15 years ago, http://www.wsj.com/articles/twitter-takeover-defenses-may-fly-out-the-window-1454463255 (last access: March 7, 2016). 16 Our estimates are in line with the continued low turnover in board seats. For instance, 255 boards included in the S&P 500 index elected 376 new independent directors during the 2015 proxy year (about 7% of total board members), with 91 boards
welcoming more than one director, see https://www.spencerstuart.com/research-and-insight/spencer-stuart-us-board-index-
2015 (last access: March 7, 2016).
http://www.wsj.com/articles/twitter-takeover-defenses-may-fly-out-the-window-1454463255https://www.spencerstuart.com/research-and-insight/spencer-stuart-us-board-index-2015https://www.spencerstuart.com/research-and-insight/spencer-stuart-us-board-index-2015
-
21
Columns (1) and (2) of Panel A of Table 3 report the annual frequency of accounting
irregularities and errors in the full (COMPUSTAT-CRSP merged) sample, both measured as of
the year of commission. Our sample covers in aggregate 868 accounting irregularities and
4,730 errors. There is a declining trend in the incidence of both irregularities and errors after
2000, but the drop-off appears more pronounced in intentional misstatements.
Turning to the influence of investor optimism about business conditions on managerial misconduct, we
find in the first row of Panel B of Table 3 that our COMPUSTAT-CRSP merged sample has 737 firm-year
observations on irregularities, 3,997 on errors and 96,667 firm-years in the control sample. The mean and
median industry EPS growth forecasts for the irregularities sample are 11% and 14%, respectively (column
(2)), and those for the error sample are 9% and 12% (column (3)). The univariate t statistics (t-stat) for mean
differences and Wilcoxon z statistics (z-stat) for median differences in the last two columns in Appendix T.3
show that accounting irregularities are typically associated with significantly higher levels of investor
optimism as compared to errors. Similarly, estimates in columns (1) and (2) and the related tests in Appendix
T.3 indicate that on average higher investor optimism accompanies irregularities relative to the control firm-
years with no reporting issues. These univariate tests are consistent with our Accounting Misconduct
Hypothesis regarding the positive relation between investor beliefs and accounting restatements. Moreover,
in line with the falsification of this argument, the estimates (and tests in Appendix T.3) reported in the
columns (1) and (3) suggest that the impact of investor optimism on errors (median of 12%) is not
significantly different from that on the control sample (median of 11%).
The remaining comparisons in Panel B indicate that, on average, firms reporting irregularities have
significantly smaller number and proportion of DAE relative to those in the no-restatement and error samples
in both the COMPUSTAT-CRSP merged sample and S&P 1500 sample. Further, firms committing errors
have significantly less DAE relative to the no-restatement sample. For example, in the S&P 1500 sample 36%
of firms in the irregularity sample have independent DAE, compared with 48% in the error sample and 54%
in the no-restatement sample. Mean percentage accounting expertise among independent directors is 8% in
the irregularity sample, compared with 11% in the error sample and 12% in the non-restatement sample.
Tests on the differences in means and medians between the sub-samples reported in Appendix T.3 are
supportive of our Accounting Misconduct Hypothesis that firms with more accounting experts on
-
22
corporate boards commit fewer accounting misstatements, especially irregularities.
Pairwise correlations between alternative measures of board accounting expertise and
investor optimism are negative and highly significant (see Appendix T.4). Similarly,
correlations between DAE and restatements (both irregularities and errors) are negative and
highly significant and that between investor optimism and irregularities is positive and
significant. All of these correlation estimates are supportive of our two hypotheses.
4.2. Investor Optimism and Board Accounting Expertise
The Board Management Hypothesis predicts that the proportion of directors with
accounting expertise decreases with optimistic investor beliefs. To test this prediction, we
estimate the following probit regressions on the presence of director accounting expertise and
ordinary least squares (OLS) regressions on the proportion of DAE:
Director Accounting Expertise = β0 + β1 Indus. EPS Growth + β2 Board Size + β3 Board
Independence + β4 Audit Committee Independence + β5 G Index + β6 Capital Intensity + β7
Prior ROA + β8 Leverage + β9 Earning Volatility + β10 Sales Growth + β11 Age + β12Log
Assets + β13 SOX + Industry/Firm fixed effect + ε (1)
The dependent variable is one of the six measures of firm-level board accounting expertise
discussed in Section 3.2 (three binary measures in the probit regressions in Models (1), (3) and
(5) and three continuous measures in OLS regressions in Models (2), (4) and (6)). The test
variable is the contemporaneous industry median EPS growth forecast, Indus. EPS Growth, our
proxy for investor optimism. In all models, we lag firm characteristics, control for industry or
firm fixed effect, and cluster standard errors at the firm level.17
Following prior studies, we add a number of regressors, as summarized in Appendix T.1A,
17 Coefficient estimates of Indus. EPS Growth in probit regressions with the firm-fixed effect are still statistically significant (at 5% for Firm AE_IndDir_D, 10% for Firm AE_IndDir_D and 1% for Firm AE_AuditDir_D). However, imposing the firm-
fixed effect in probit regressions reduces the number of observations to 2,319 with 261 firms due to collinearity. Alternatively,
coefficients estimates of Indus. EPS Growth using linear probability models with the firm-fixed effect are statistically similar.
-
23
to control for corporate governance and firm characteristics. We include Board Size because
larger boards are more likely to include a board member of any type. Previous studies (such as
Defond et al. 2005; Erkens and Booner 2013) document the impact of board independence and
audit committee independence on board structure, including the appointment of accounting
experts on the board. Following Beasley (1996), Klein (2002), and others, we use the
percentage of outside directors on the board and on the audit committee as proxies for Board
Independence and Audit Committee Independence, respectively. G Index, developed by
Gompers, Ishii and Metrick (2003), measures the strength of a firm’s governance system and
is constructed based on a simple counting of 24 corporate governance provisions. It proxies for
a firm’s exposure to the discipline of the market for corporate control (Ferreira et al., 2011).18
Since the data on G Index is not available after 2008, we use the 2008 index data for 2009 as
well 2010 – the last two years in our sample period. A low (high) G Index is associated with a
strong (weak) governance system. Previous literature (see, for example, DeFond, et al., 2005
and Krishnan and Visvanathan, 2008) suggests a negative relation between the appointment of
a financial expert and the G Index.19
Agrawal and Chadha (2005) predict that firms that are more capital intensive are likely to
have a greater need for financial expertise on the board. We measure Capital Intensity as total
assets scaled by number of employees. We include Leverage and Prior ROA because more
leveraged and poorly performing firms are likely to have a greater need for external financing
and boardroom financial expertise. Following Agrawal and Chadha (2005), and Krishnan and
Visvanathan (2008), we include Earnings Volatility, measured as the standard deviation of
earnings per share including extraordinary items for the past five years, Sales Growth,
measured as annual percentage change of sales, firm age (Age) and Log Assets (defined as the
18 Ferreira et al.(2011) find both %institutional ownership and institutional ownership insignificant in their tests on board independence. 19 We find similar results by replacing G Index with BCF Index, developed by Bebchuk et al. (2009).
-
24
logarithm of book value of total assets). To control for regulatory changes, we include the SOX
dummy (that equals one for misstatements committed in 2002 or after and zero otherwise).
The estimates in Panel A of Table 4 show a significantly negative relation between investor
optimism and all the six director accounting expertise measures. The associated marginal
effects indicate that, all else equal, a one-standard-deviation increase in investor optimism
(Indus. EPS Growth) from the mean is associated with a decrease in the likelihood of the
presence of director accounting expertise (Firm AE_IndDir_D, Firm AE_AllDir_D and Firm
AE_AuditDir_D) of 3%, 3% and 7% (respectively). Further, a one-standard-deviation increase
in Indus. EPS Growth is associated with a 10% decrease in the proportion of audit committee
members with accounting expertise (Firm %AE_AuditDir). These results seem economically
significant, especially in the case of audit committee accounting expertise, and indicate that
investor optimism regarding the state of the economy is one of the major drivers of the
accounting composition of the board. They offer strong support for our Board Management
Hypothesis that higher optimistic beliefs are associated with lower investor monitoring through
the appointment of accounting experts to the board. The coefficient estimates on control
variables indicate that board accounting expertise increases in general with firm size, firm age,
board size, board independence, better corporate governance, poor prior performance and
lower sales growth, although they are not significant in all the models. All proxies for
accounting expertise tend to increase significantly after SOX.
In Figure 1, we plot the median predicted values of the proportion of independent directors
with accounting expertise (based on Model (2) in Panel A of Table 4, shown on the vertical
axis) for the ten deciles of industry EPS growth forecast (horizontal axis). The solid line
connects the 10 predicted median values, and the dashed line is a fitted line. The observed trend
is consistent with our novel prediction that the predicted proportion of independent directors is
declining in investor optimism about business prospects.
-
25
To explore whether investor optimism affects other mechanisms (attributes) of board
monitoring, we conduct separate regressions (similar to the one in Panel A of Table 4) for each
of the following dependent variables: board independence, audit committee independence and
G Index. In unreported results, we find that the coefficient estimates on our test variable,
Industry median EPS growth (-0.062, -0.043, and 0.745, respectively, with p-values of 0.349,
0.675, and 0.392) are all insignificant. These results indicate that investor optimism has
negligible influence on other widely used board monitoring mechanisms.
Investor optimism decreases dramatically during the periods surrounding the recession in
2001 and the global financial crisis in 2008 in our sample period. On the other hand, director
accounting expertise increases almost monotonically, especially after the passage of SOX in
2002. A potential concern is that our results with respect to the impact of investor optimism on
board accounting expertise are driven by these events. To mitigate this concern, we exclude the
data over 2001-2002 (roughly the period of economic slowdown and recession as identified by
the National Bureau of Economic Research’s Business Cycle Dating Committe) and 2008-
2010 (covering global financial crisis and thereafter). In untabulated results, we find the
negative relation between EPS growth and direct accounting expertise remains intact. For
example, the coefficient estimates of EPS growth based on Model (1) in Panel A of Table 4 is
-0.158 with p-value of 0.036. The coefficient estimates of EPS growth based on Model (2) is -
0.038 with p-value of 0.049. The estimates for the remaining measures of board accounting
expertise are similar to those based on the full sample, suggesting that our results are not driven
only by these two major events.
To address potential concern that the time-series and cross-sectional variations in the
proportion of directors with accounting expertise are driven by the changes in board size
instead of the number of directors per se, we replace %AE with the absolute number of
accounting experts in the three continuous measures and still find negative coefficient estimates
-
26
of Indus. EPS Growth (except for the number of accounting experts among all directors). In
untabulated results, the coefficient estimates on the number of independent directors (number
of audit committees) with accounting expertise is -0.183 (-0.459) with a p-value of 0.078
(0.005).
The six measures of board accounting expertise used as dependent variables in Panel A are
in levels – either the presence (or absence) of accounting experts or the proportions of
accounting experts. Given the relatively low rate of director transition in our panel data, these
variables are persistent over time. Our first hypothesis posits that investors elect more directors
with accounting expertise to the board when they are less optimistic about macroeconomic
conditions. But investors do not care much about the accounting composition of the board when
they are optimistic. This indifference allows opportunistic managers to reduce the number of
accounting watchdogs on the board. To scrutinize this prediction, we present additional tests in
Panel B that focus on director turnover as dependent variable. We estimate director transition
as the annual change in the number and proportion of directors with accounting expertise
among independent directors, all directors and audit committee members. These six annual
board refreshments capture the election or exit of accounting experts. Five of the six estimates
of coefficients on Indus. EPS Growth are negative and significant at 10% or better. These
robustness tests based on turnover of accounting experts on the board provide more compelling
evidence that high investor optimism is associated with fewer accounting experts on the board.
4.3 Appointment and Departure of Directors with Accounting Expertise
The annual DAE turnover analysis presented in Panel B of Table 4 is based on net changes
(the difference between appointment and departure of DAE). Our arguments underlying the
Board Management Hypothesis indicate that the negative relation between investor optimism
and turnover of DAE is less likely come from the departure of DAE because it is unlikely that
investors as well as insiders would actively remove sitting directors with accounting expertise
-
27
regardless of high or low investor optimism. Therefore, we expect to find that investor
optimism plays an insignificant role in explaining the departure of DAE. On the other hand,
opportunistic management is more likely to appoint fewer DAE during times of high investor
optimism, but refrain from doing so when investors are pessimistic and more likely to question
board oversight over incumbent management. If is even possible that management would bow
to mounting pressure from activist shareholder campaigns and grant some board seats to avoid
the distraction form proxy fights, especially when poor firm performance coincides with lower
investor optimism about business prospects. These arguments suggest a negative relation
between appointment of DAE and high optimism, but no relation, or even a positive relation,
between appointment of DAE and low investor optimism.
To examine the above arguments, we specify a Poisson model to analyze the likelihood of
appointment and departure of DAE in Table 5. Our dependent variables include the number of
accounting experts added to (removed from) independent directors (in Model (1)), the number
of accounting experts added to (removed from) all directors on the board (in Model (2)), and
the number of accounting experts added to (removed from) audit committee (in Model (3)). In
Panel A, our test variable is high investor optimism that takes value of one if Indus. EPS Growth
forecast exceeds the 67th percentile of forecasts in a given year, and zero otherwise. We add
additional controls based on the existing literature (such as Hermalin and Weisback, 1988;
Farrell and Hersch, 2005). The likelihood of appointment of a DAE to a board should be higher,
the lower the existing representation of DAE. To control for this assertion, we include the lgged
percentage DAE among independent directors (Lagged %AE) as well as the number of
accounting experts on the board in the preceding year who left the board (DAE Departure). A
new addition or departure of a director is more likely when a board vacancy occurs, which is
closely related to CEO turnover or insiders leaving in the preceding year. CEO Turnover equals
one if the firm experiences a CEO turnover in the preceding year. Insider Departure is the
-
28
number of directors leaving the board in the preceding year.
The results in Panel A of Table 5 confirm our expectation that higher investor optimism is
associated with lower likelihood of appointment of DAEs. All coefficient estimates of High
Indus EPS Growth dummy are negative and significant. The negative effect is most pronounced
among the appointment of audit committee members. Negative and significant coefficients of
Lagged %AE suggest that a firm is more likely to appoint a DAE to the board when the existing
representation of DAE is low.
In sharp contrast, the low investor optimism dummy, which takes a value of one if Indus.
EPS Growth is below the 33rd percentile of forecasts of a given year and zero otherwise, is
positive and significant in all three models, see Panel B. This finding suggests that when
investors are pessimistic about general economic conditions, insiders are more likely to appoint
DAE to the board, consistent with our prediction.
Estimates in Panels C and D indicate that the odds of departure of DAE are unrelated to investor
optimism, which is also consistent with our expectation.20
4.4. Accounting Restatements, Investor Optimism and Board Accounting Expertise -
Propensity Score Matching (PSM)
Now we turn to more detailed tests on our second hypothesis which predict that (a) investor
optimism raises the likelihood of accounting irregularities, and (b) the likelihood of
misstatements (both irregularities and errors) decreases as the proportion of directors with
accounting expertise increases. Our research design centers on estimating variants of the
following basic model:
Prob (Accounting Misstatement = 1) = β0 + β1 Accounting Expertise +β2 Indus. EPS
20 The spline tests (untabulated, for brevity) offer strong support for our revised argument that insiders behave opportunistically by (a) exploiting investor overoptimism to weaken the accounting expertise of the board, and (b) refraining from dampening DAE and independent directors) when investor optimism is at a normal level. However, at lower levels of investor optimism (first and second quintile), the coefficients on Indus EPS Growth forecast are positive but not significant at 5%.
-
29
Growth + β3 Indus. EPS Growth Squared + β4 ROA + β5 Leverage + β6 External Financial
Need + β7 Insider Ownership + β8 Big N Auditors +β9 M&A Expenditure + β10 CAPX + β11
R&D Expenditure + β12 Log Assets + β13 Analyst Coverage + β14 SOX + ε
(2)
Prob (Accounting Misstatement = 1) represents either the likelihood of committing
irregularities or errors. It equals one for firms committing irregularities or errors firms and zero
for control firms with no restatements. Our test variables are Accounting Expertise and Indus.
EPS Growth, both measured prior to the first misstated year. Similar to Equation (1), we use
three accounting expertise (DAE) measures of independent directors, all directors and audit
committee members. To test potential non-linear relation between investor beliefs and the
likelihood of misstatements documented in Wang et al. (2010), we include Indus. EPS Growth
Squared.
The (ex-ante) control variables are measured prior to the year of commission to minimize
spurious relationship with the outcome variable. We use return on assets (ROA) to measure
accounting performance because previous studies find that manipulating firms have strong
financial performance prior to the misstatements (see Beasley, 1996; Crutchley et al., 2007;
Dechow et al.,2010; Kinney and McDaniel, 1989; Summers and Sweeney, 1998; Erickson et
al. 2006).21 We control for Leverage based on the debt covenant hypothesis (DeFond and
Jiambalvo; 1994, Sweeney, 1994; Dichev and Skinner, 2002).22
High external financing needs affect earnings management as well as the commission of
accounting misstatements (Dechow et al., 1996; Teoh et al., 1998; Wang, 2013; Wang and
Winton, 2014). We follow Dechow et al. (1996) and define Ext. Fin. Need as a dummy variable
21 Beasley (1996) finds that companies reporting repeated losses are more likely to engage in financial fraud. Kinney and McDaniel (1989) find that less profitable companies are more likely to misreport. But Summers and Sweeney (1998) find a
positive association between misstatements and profitability, and Erickson, Hanlon and Maydew (2006) find insignificant or
mixed evidence. 22 However, empirical evidence on the impact of leverage is mixed (DeAngelo et al., 1994; Dechow et al. 1996; DeFond and Jiambalvo, 1991; Dechow et al. 2011).
-
30
equal to one if the firm’s free cash flow is less than –0.5 and zero otherwise.23 Free cash flow
is defined as cash from operations minus average capital expenditures during the previous three
years and then divided by prior year current assets. Cash from operations equals earnings minus
accruals. Accruals are changes in current assets minus changes in current liabilities minus
changes in cash/cash equivalents plus changes in debt included in current liabilities, and minus
depreciation and amortization expense.
Warfield et al. (1995) find that greater Insider Ownership is associated with greater earnings
informativeness and better accruals quality. However, Goldman and Slezak (2006) and Denis
et al. (2006) suggest that a positive relation between firm performance and insiders’
compensation can induce misreporting. Previous studies (such as Bhattacharya and Marshall,
2012; Agrawal and Cooper, 2014) indicate mixed results. We construct this proxy using the
percentage of shares owned by insiders.24 We control for M&A expenditure. Kinney et al.
(2004) suggest that acquisitions may increase the probability of misstatement because of new
accounting issues and possible business integration problems. In addition, we add capital
expenditure (CAPX), R&D Expenditure, Log Assets, the SOX dummy, and Analyst Coverage
as controls (Wang (2013) and Wang and Winton (2014)). We include a dummy variable for big
4 or big 5 auditors to control for audit firm quality (Big N Auditors).25 In robustness tests
discussed in Section 4.8, we add controls for additional board monitoring mechanisms, such as
board and audit committee independence, governance quality, board size and executive
compensation, which also could affect the incidence of accounting failures.
Since the accounting profile of directors, like many other board attributes and corporate
arrangements, is not randomly assigned (Adams, Hermalin, and Weisbach, 2010), tests of the
monitoring impact of directors with accounting expertise are vulnerable to sample selection
23 While Wang (2013) and Wang and Winton (2014) focus on accounting fraud, we study accounting restatements. 24 We construct Fama-French 49 industry medians for each fiscal year and replace missing observations with industry medians. 25 Since some firms might restate their accounts only once in our sample period, we cannot control for firm fixed effect with accounting misstatement as our binary dependent variable.
-
31
and endogeneity biases. The propensity score matching methodology models the investors’
choice of directors with accounting expertise as a treatment variable. This matching method
allows us to construct a counterfactual (weighted) control sample with observable firm
characteristics virtually identical to those of the treatment sample except for board accounting
expertise.26 It provides reasonable assurance that our test results on the link between the
likelihood of restatements and board accounting expertise are not attributable to observable
heterogeneity between firms with directors with accounting expertise and those without, but it
does not control for unobserved heterogeneity. To conduct rigorous analyses, we begin with
binary treatment under propensity score matching and supplement it with:, categorical
treatment and continuous treatment (see Appendix T.6). Under binary treatment, firm-years
with at least one director with accounting expertise are included in the treatment sample, while
firms without any director with accounting expertise are assigned to the control sample.27 We
use the predicted probabilities (propensity scores) from the probit regressions in Models (1),
(3) and (5) of Table 4 to match each firm-year observation on a firm with accounting expertise
to that on a firm without accounting expertise such that the absolute value of the difference in
propensity scores between the treated and control groups based on the nearest neighborhood
matching technique with replacement is minimized (Rosenbaum and Rubin, 1983).28
26 Propensity score matching does not allows us to control unobservable characteristics. In later sections we provide a battery of tests using other empirical strategies including bivariate Probit regressions, two-stage bivariate Probit regressions and
instrumental variables to mitigate this concern. 27 As a diagnostic test, we report in Appendix T.5 univariate comparisons of key observable firm characteristics between
those with accounting expertise among independent directors and their matched peers. The covariate balance between the
matched pairs show that most of the differences in observable firm attributes are trivial and are not significantly different
from zero, which suggests that the propensity score matching process has formed a control group of closely-matched firms
with director accounting expertise which are highly similar to firms with director accounting expertise. In particular, the lack
of significance of both the mean and median difference tests shows that the treated (i.e., firms with at least one accounting
expert on the board) and control samples are very similar in other types of board monitoring mechanisms such as board size,
governance mechanisms, board and audit committee independence. This covariate balancing process should strengthen our ability to identify a relation between director accounting expertise, our test variable, on financial misreporting. In
untabulated analyses, we obtain similar results for firms with director accounting expertise among all directors and audit
committee members. 28 We repeat our tests using (1) nearest neighbor 1:1 with replacement, (2) nearest neighbor 1:4 without replacement and (3) a sample including all firms within the region of common support of their propensity scores. Results are quite similar except
that the coefficient estimates are more significant when using nearest neighbor 1:4 and common support due to a larger sample
size. In unreported results, we test average treatment effects of having DAE on board on the outcome of irregularity by using
different matching methods including nearest neighbor 1:1, nearest neighbor 1:4 and Kernel weighting. In all the methods, the
coefficient estimates of DAE are negative and statistically significant with t-statistics greater than 2.
-
32
Table 6 presents results regarding the impact of investor optimism and board accounting
expertise on the likelihood of irregularities (Panel A) and of errors (Panel B). Models (1), (2)
and (3) are estimated using three firm-level binary measures of director accounting expertise.
For example, in Model (1) of Panel A, there are 2,088 firm-year observations with at least one
independent director with accounting expertise (treatment sample), which are matched with
2,088 firm-year observations with no independent director with accounting expertise (control
sample). In all models the estimated coefficients on board accounting expertise are negative
and significant, suggesting the presence of accounting experts on the board is associated with
a reduction in the likelihood of accounting misreporting. In unreported results, the predicted
probability of irregularities in firms with accounting expertise among independent directors is
1.2% while that of firms without accounting expertise is 2.4%, holding other variables at their
means (based on Model (1)). This is a huge difference (marginal effect) because the predicted
probability of firms without accounting expertise is twice as large as the probability of firms
with DAE.29 For unintentional misstatement (errors), the predicted probability of errors in
firms with director accounting expertise is 8% while that of firms without DAE is 10%, holding
other variables at their means (based on Model (1)). The difference is economically significant,
though not quite as large as that in irregularities. Further, we typically observe a positive
relation between investor optimism and the likelihood of accounting misstatements, both
irregularities and errors. Results on control variables are consistent with prior studies. We find
similar results with categorical and continuous treatments in Appendix T.6.
As accounting failures shake investor trust in capital markets and institutions, they are very
costly not only to the firm but also to the economy at large. Our analyses thus far indicate that
29 We note that board accounting expertise in Table 5 is a binary variable. The marginal effect is calculated as the difference in the predicted probability of firms with director accounting expertise and that of firms without director accounting expertise.
The predicted probability of irregularity for firms with accounting expertise among all directors is 1.7% given that all predictors
are set to their mean values, while that of firms without DAE is 2.7%. The corresponding numbers for DAE among audit
committee members are 1.2% and 2.6%, respectively. Note that, it is difficult to compare the marginal effect of Indus. EPS
Growth with that of DAE because the former is a categorical variable while the latter is a binary variable. See our discussion
on Table 8 for the comparison of the effect on investor belief and that of DAE on the likelihood of accounting misstatements.
-
33
while investor optimism is correlated with a decrease in the proportion of accounting experts
on the board and an increase in the incidence of misstatements, DAE is associated negatively
with the probability of misconduct. Is the weakened monitoring intensity of DAE sufficiently
strong to offset the increased likelihood of irregularities and errors due to investor optimism?
To explore this question, we repeat the tests in Panel A of Table 6 by interacting the three binary
measures of DAE with Indus. EPS Growth. The results presented in Panel C suggest that all
the interaction coefficient estimates are negative and two out of three are significant. All the
three net effects (sum of the negative coefficient estimate of DAE dummy and positive estimate
of Indus. EPS Growth) of DAE and investor beliefs are insignificant (with p-values of 0.932,
0.874, and 0.501). Following Ai and Norton 2003; Norton, Wang, and Ai 2004; Cornelißen and
Sonderhof 2009), we estimate the marginal effects of investor optimism on the likelihood of
irregularity. Our tests indicate these effects are insignificant in the presence of board accounting
expertise, holding other control variables at their means. But the marginal effects of investor
optimism are positive and significant in the absence of DAE. These results imply that investors,
boards and regulators could help counteract the increased likelihood of accounting restatements
due to investor optimism by appointing more directors with accounting expertise.
4.5. Bivariate Probit Model
Investors are unaware of financial misstatements at the time when they are actually
committed, and they typically come to learn about the beginning of misconduct at a later time
when the misstatements are disclosed. Therefore, the occurrence of accounting mistakes is only
partially observable to investors. We use a bivariate probit model to address the partial
observability (incomplete detection) concern. There are two latent (indicator) variables,
whether there is a mistake in accounting statements (M) and whether the misstatement is
disclosed or detected (D). We only observe misstatements that have been committed and
restated, for which both M and D equal one. In addition, the probability of observed
-
34
misstatements is different from the probability of committed misstatements if the
detection/disclosure process is not perfect.
Our bivariate probit model includes (a) an equation of misstatement commission, (P(M=1),
and (b) an equation of misstatement disclosure, P(D=1|M=1). Equation P(M=1) has the same
specification as in Equation (2), with the latent dependent variable equal to one if a firm
committed an accounting misstatement, zero otherwise. The ex-ante explanatory variables are
measured prior to the year of commission. In equation P(D=1|M=1), the dependent variable is
a latent variable equal to one if a firm committed an accounting misstatement and then restated
or got caught, zero otherwise. We include a set of ex-ante disclosure/detection factors
(measured prior to the disclosure of reporting failures) as well as a set of ex-post
disclosure/detection factors (measured as of the year of disclosure) as follows:30
Prob (Disclosure of Accounting Misstatement = 1 | Commission of Accounting
Misstatement = 1) = β0 + β1 Abnormal Restatement Risk +β2 Disastrous Stock Return + β3
Abnormal Volatility + β4 Abnormal Turnover + β5 M&A Expenditure + β6 CAPX + β7 R&D
Expenditure +β8 Log Assets +β9 Analyst Coverage + β10 SOX + ε (3)
Our ex-ante determinants of misstatement disclosure include M&A Expenditure, R&D
Expenditure, CAPX, Analyst Coverage, Log Assets and SOX dummy. 31 For the ex-post
determinants of misstatement disclosure, we add a proxy for restatement risk (defined as the
logarithm of the sum of market value of restated firms in an industry) to control for industry
restatement intensity. Abnormal Restatement Risk is the yearly deviation from the industry
average restatement intensity. Regulators and investors pay more attention to a particular
industry which has more misstatements. Firms that experience large negative returns, high
30 For robustness, we also test our model by constructing ex post disclosure/detection factors as of one year after the first misstated period and find similar results. 31 Wang (2013) suggests that R&D investment, capital expenditure and mergers and acquisitions tend to affect the likelihood of detection. Dyck et al. (2010) suggest that analysts are important external monitors of firms. Accordingly, we include Analyst Coverage as the number of stock analysts that follow an industry. Following Wang et al. (2010) and Wang (2013), we include
firm size and the SOX dummy.
-
35
stock turnover and high return volatility are more likely to be watched by investors and
shareholders. Disastrous Stock Return is an indicator variable equal to one if the firm’s stock
return is in the bottom 10% of all the firm-year return observations in the COMPUSTAT-CRSP
merged database.32 Abnormal Return Volatility is the difference between the yearly standard
deviation of the firm’s returns and its time-series average. Similarly, Abnormal Stock Turnover
is the difference between the monthly share turnover of the firm and its time-series average.
Panel A (irregularities) and Panel B (errors) of Table 7 report bivariate probit regression
results. For both irregularities and errors, coefficient estimates of the three DAE measures are
negative and typically significant at 1%, consistent with our previous results using propensity
score matching. In addition, the probability of committing irregularities is significant and
positively related to the level of investor optimism, but it is significant and negatively related
to the squared optimism term. This hump-shaped relation between investor optimism and the
incidence of irregulariti