do better connected executives have longer incentive horizons? · do better connected executives...
TRANSCRIPT
Do Better Connected Executives Have Longer Incentive Horizons?
Minjie Huang
Department of Finance
University of Louisville
Louisville, KY 40292
June 2016
I am deeply grateful to George Bittlingmayer and Felix Meschke for their guidance and suggestions. I thank
Varouj Aivazian, Anup Agrawal, Christopher W. Anderson, George Cashman, Elif Sisli Ciamarra, David
Cicero, Lauren Cohen, Douglas O. Cook, Robin Cowan, Bob DeYoung, Ahmed Elnahas, Stuart Gillian,
Jennifer Itzkowitz, Hui Liang James, Chao Jiang, Darren Kisgen, Paul Koch, Tunde Kovacs, Tom Kubick,
Lei Li, Yun Liu, William McCumber, Darius Miller, Debarshi K Nandy, M. Fabricio Perez, Lynnette Purda,
Selim Topaloglu, Kevin Tseng, Shane Underwood, Albert Wang, Jide Wintoki, Hong Zhao, and seminar
and conference participants at the University of Kansas, the Midwest Finance Association 2015 Annual
Meeting, the Southwestern Finance Association 2015 Annual Meeting, the Eastern Finance Association
2015 Annual Meeting, the Southern Finance Association 2015 Annual Meeting, and the University of
Louisville for helpful comments. All errors remain my own.
Do Better Connected Executives Have Longer Incentive Horizons?
Abstract
I examine the role of executive connections in mitigating managerial short-termism. I hypothesize that
better connected executives have less short-termism because larger networks give them better access to
information and protect them against adverse job outcomes. Using the duration of managerial compensation
to measure incentive horizon, I find that better connected executives have longer incentive horizons.
Executives with larger networks also engage in less earnings management, and executive connections affect
earnings management through the channel of incentive horizon. These findings suggest that social networks
mitigate managerial short-termism and facilitate corporate long-term behaviors.
Keywords: social networks, managerial short-termism, incentive horizons, earnings management
JEL classification: G30, G34, J33, L14
1
1 Introduction
The recent financial crisis has reignited researchers and practitioners’ interests in the issue of
managerial short-termism. Researchers have argued that managers prefer short-term projects due to career
concerns (Narayanan 1985), near-term stock price pressure (Stein 1989), and herding behaviors (Zwiebel
1995). In a survey of financial managers, the majority of executives would pass on a long-term project with
positive net present value (NPV) to avoid missing earnings forecasts of the current quarter (Graham,
Harvey, and Rajgopal 2005). Long-term projects usually have highly uncertain payoffs and substantial
probability of failure (Kothari, Laguerre, and Leone 2002). Top executives are also hesitant to engage in
long-term behaviors due to career concerns. Therefore, risk-averse managers are reluctant to choose long-
term projects and exhibit short-termism. In this paper, I examine the role of executive connections in
mitigating managerial short-termism through access to information and labor market insurance.
Prior literature has established that executives with more connections have better access to
information via their networks (e.g., Fracassi 2014; Engelberg, Gao, and Parsons 2012; Faleye, Kovacs,
and Venkateswaran 2014). Burt (2004) and McDonald, Khanna, and Westphal (2008) also show that
network connections expose the manager to alternative perspectives that improve the quality of strategic
decisions in settings with great uncertainties. Well-connected executives are able to access network
information that helps them identify and implement long-term projects with positive NPV. The access-to-
information effect can also reduce managerial risk aversion to long-term focus by diminishing the
uncertainties of long-term projects.
Executives with more connections also have better labor market insurance against losses of
employment. Better connected managers are more likely to find new and better jobs after forced turnovers,
and have more outside employment options and higher compensation (e.g., Nguyen 2012; Liu 2014;
Engelberg, Gao, and Parsons 2013). A network of acquaintances, former colleagues, and alumni provides
a safety net that protects the manager against adverse job outcomes by increasing the probability of re-
employment after a job turnover. The labor-market-insurance effect reduces the manager’s exposure to the
downside risk of long-term projects without affecting her exposure to the upside gains.
2
An emerging literature has provided theoretical frameworks and empirical evidence that
compensation schemes with short incentive horizons reflect managerial short-termism (e.g., Bolton,
Scheinkman, and Xiong, 2006; Manso, 2011; Peng and Roell, 2014; Edmans, Gabaix, Sadzik, and
Sannikov, 2012). Gopalan, Milbourn, Song, and Thakor (2014) use pay duration - the weighted average of
vesting periods of different pay components - to measure incentive horizon, and find that shorter pay
duration is associated with more earnings-increasing accruals, a sign of short-termism. Baranchuk,
Kieschnick, and Moussawi (2014) show that, in a sample of newly public firms, corporations become more
innovative after they award the CEO equities with longer vesting periods. This implies that incentive
horizon not only captures managerial short-termism but also acts as a channel through which long-term
corporate behaviors are induced.
To test whether better connected executives mitigate managerial short-termism through longer
incentive horizons, I construct measures of incentive horizons and executive connections in a sample of
more than 13,000 executives in S&P 1500 firms from 2000 to 2013. Consistent with recent papers (e.g.,
Gopalan, Milbourn, Song, and Thakor 2014; Engelberg, Gao, and Parsons 2013), I estimate the duration of
executive compensation to measure incentive horizons, and I use the number of first-degree links an
executive has to other individuals in elite corporations to measure the executive’s network centrality. I find
that better connected executives have significantly longer incentive horizons in the full sample and
subsamples. I explore variations in employment history and education history to construct instruments for
executive connections. Results from instrumental variable regressions show that network connectedness of
an executive significantly and positively affects her incentive horizon. This positive relationship remains
significant after a battery of robustness tests. Furthermore, the effect of the executive’s network centrality
on her incentive horizon is stronger in firms with more growth options and weaker in firms with greater
risk. Finally, executives with larger networks engage in less earnings management, and executive
connectedness reduces earnings management through the channel of incentive horizons. Therefore, the
empirical evidence strongly supports that better connected executives have longer incentive horizons and
thus exhibit less managerial short-termism.
3
These findings contribute to two strands of literatures. First, I contribute to the literature on
managerial short-termism. Prior studies have examined whether and how managerial short-termism can be
alleviated by enhancing monitoring (e.g., Bushee 1998; Farber 2005; Cadman and Sunder 2014). I extend
this literature by showing that managers’ access to network information and career concerns also play
significant roles in affecting managerial short-termism. Second, I contribute to the literature on social
networks in finance. Fracassi (2014) show that external networks help the firm obtain information
advantages to improve its investment decisions, and firms that are more central in the networks exhibit
better performance. Engelberg, Gao, and Parsons (2012) find that when banks and firms have managers
interconnected through professional or education networks interest rates are significantly reduced. My
results suggest that executive connections may have a positive effect by inducing long-term corporate
behaviors. Specifically, executive connections may help the manager choose desirable yet risky long-term
projects by providing valuable network information and labor market insurance. In addition, my results
suggest that the manager’s network centrality may facilitate corporate long-term behaviors via the channel
of her incentive horizon.
2 Background and Hypothesis Development
Prior studies indicate that executive pay with long horizon improves long-term firm performance
and thus alleviates managerial short-termism (e.g., Bebchuk and Fried, 2010; Bhagat and Romano, 2010).
Subsequently, an emerging literature has built theoretical frameworks for the optimal incentive horizon,
and the general view is that equity awards with long vesting periods alleviate managerial short-termism
(e.g., Bolton, Scheinkman, and Xiong, 2006; Manso, 2011; Peng and Roell, 2014; Edmans, Gabaix,
Sadzik, and Sannikov, 2012). Gopalan, Milbourn, Song, and Thakor (2014) use pay duration - the weighted
average of vesting periods of different pay components - to measure incentive horizon, and find that firms
with longer pay duration for their CEOs have lower earnings-increasing accruals. Baranchuk, Kieschnick,
and Moussawi (2014) show that, in a sample of newly public firms, corporations become more innovative
after they award the CEO equities with longer vesting periods. They conclude that this finding is consistent
4
with the theoretical prediction that equity awards with longer vesting periods provide executives with better
incentives to pursue long-term objectives. While the growing literature has focused on the economic
outcomes and firm characteristics that are related to managerial incentive horizon, few papers attempt to
investigate the effect of individual attributes on pay duration.1
One crucial individual attribute for top executives is their network connectedness. Top executives
in Corporate America are linked to high-ranking managers or directors at other firms through professional,
educational, and social networks, and these networks have been shown to have significant impacts on
economic outcomes, both at the firm level and the individual level. At both levels, one fundamental
advantage of networks is that information flows across network nodes and creates spillovers (e.g., Glaeser
et al. 1992; Jaffe, Trajtenberg, and Henderson 1993), meaning that knowledge generated in one node of the
network becomes accessible to other nodes.
The executive’s network may accrue value to the firm. For example, Fracassi (2014) show that
external networks help the firm obtain information advantages to improve its investment decisions, and
firms that are more central in the networks exhibit better performance. Engelberg, Gao, and Parsons (2012)
find that when banks and firms have managers interconnected through professional or education networks
interest rates are significantly reduced. They show that firms with connected deals subsequently improve
performance in future credit ratings or stock returns, implying that networks lead to knowledge spillovers.
Faleye, Kovacs, and Venkateswaran (2014) document that firms with more-connected CEOs have higher
R&D investments, and these firms obtain more and better patents. They conclude that external networks
confer the firm better access to information and alleviate career concerns of executives in risk investments.
Cohen, Frazzini, and Malloy (2008) use college connections between mutual fund managers and corporate
directors to infer information flows in stock markets. They find that portfolio managers invest more in
connected firms and their connected holdings perform significantly better than the non-connected holdings.
1 For recent studies that related individual attributes to investment decisions, productivity, and compensation level,
see Bertrand and Schoar (2003); Schoar (2007); Malmendier and Tate (2009); Kaplan, Klebanov, and Sorensen
(2012).
5
Overall, these papers suggest that network connections may be an important channel for information
spillovers among organizations and appear to be valuable to the firm.
Networks may also benefit individuals with connections. For instance, Nguyen (2012) find that a
CEO with social connections to directors is less likely to be fired for underperformance and more likely to
find new and better employers after a forced turnover. Liu (2014) studies the network effect on the labor
market conditions for CEOs. She concludes that the CEO’s connectedness expands outside employment
options and reduces job search frictions. Engelberg, Gao, and Parsons (2013) document that CEOs with
larger networks have higher compensation. They show that the wage premium for connections is consistent
with network literature on information spillovers, whereby CEOs are compensated for their valuable and
portable connections that facilitate knowledge flows into the firm. Cohen, Frazzini, and Malloy (2010) find
that stock analysts outperform significantly on stock recommendations for the firm where they have
educational links to the senior executives. They conclude that analysts obtain superior information through
their social networks, and the information advantage benefits them on stock recommendations. Berkman,
Koch, and Westerholm (2014) find that corporate directors gain abnormal returns when they purchase
stocks from firms to which they are connected via interlocking board seats and when they have high network
centrality. To sum up, prior studies support that executives with large networks obtain comparative
advantage that alleviate their career concerns and facilitate better personal performance, and hence
networks appear to be valuable to executives.
While recent papers on networks study economic outcomes at the individual level and the firm
level, no paper, to the best of my knowledge, has focused on the network effect on the executive’s incentive
horizon. Network connectedness may affect pay horizon from the perspective of both the firm and the
executive. On one hand, the board may prefer awarding long-term contracts to retain the executive for
continuous access to valuable information through her network. In particular, the board may use long
vesting periods to reward managers for their long-term success and at the same time make early voluntary
departures costly for executives (e.g., Gopalan, Milbourn, Song, and Thakor 2014). On the other hand,
networks expand outside options for executives and thus alleviate their dismissal risk. Previous papers
6
document that CEOs bear significant costs in forced turnovers. For example, Peters and Wagner (2013)
show that turnover risk is a determinant in managerial compensation. They document a significantly
positive relationship between turnover risk and CEO pay. High turnover risk increases the probability for
the executive to leave the firm early, and thus makes the compensation contract with long vesting periods
less desirable to the executive. Laux (2012) also show that dismissal risk is a major friction that makes long
vesting periods costly to managers.
Therefore, in equilibrium, the executive’s connectedness may be an effective contracting factor that
leads to long incentive horizon. Better connected executives may benefit the firm by providing access to
valuable information, so the board awards compensation with longer vesting periods to retain the executive.
Larger networks reduce the executive’s dismissal risk by providing outside options and hence labor market
insurance, so compensation with long vesting periods becomes less costly to better connected executives.
Prior studies (Bolton, Scheinkman, and Xiong 2006) have suggested that managerial compensation tends
to have longer vesting schedules if doing so is more valuable to the firm and less costly to the manager.
However, compared to the efficient contracting hypothesis that better connected executives have
longer incentive horizon, it is also possible that greater network centrality also coincides with more
powerful executives. Prior studies have shown that the executive’s connections, especially internal
connections within the company, may be detrimental to firm value. For example, Fracassi and Tate (2012)
document that firms with more CEO-director connections exhibit lower value and are involved in in more
value-destroying acquisitions. Coles, Daniel, and Naveen (2014) find that in firms with directors appointed
by the CEO board monitoring decreases. Managers are typically risk-averse and prefer compensation with
short incentive horizon. If the executive’s network has little long-term benefit to the firm, a powerful
executive with high network centrality may be able to have a contract with short incentive horizon.
Therefore, the managerial power hypothesis would predict that better connected executives have shorter
incentive horizon.
7
3 Data and Specifications
3.1 Data Construction
To measure executive connections I obtain network connection data from BoardEx. BoardEx is a
business intelligence database that provides detailed profiles of over 400,000 of the world's business leaders
in over 14,800 public and large private companies in North America and Europe. BoardEx uniquely
provides in-depth biographical information about directors and executives, including complete employment
record, education background, professional qualifications, and non-business-related activities. Based on
these detailed biographical information, network connectedness can be measured by constructing networks
among executives and directors via their shared common histories in employment and education. Following
prior studies (e.g., Engelberg, Gao, and Parsons 2013; Faleye, Kovacs, and Venkateswaran 2014), I measure
executive network connectedness by the total number of individuals with whom the executive shares an
overlapping employment history in a public firm or overlapping educational history in a university, which
excludes those individuals who are employed by the executive’s current firm.
A recent paper by Gopalan, Milbourn, Song, and Thakor (2014) proposes to measure incentive
horizon by the weighted average duration of components of compensation (salary, bonus, restricted stock,
and stock option). Following their specification in Equation (1) of Gopalan, Milbourn, Song, and Thakor
(2014), I measure incentive horizon by Incentive duration in the equation below:
Incentive duration = (𝑆𝑎𝑙𝑎𝑟𝑦+𝐵𝑜𝑛𝑢𝑠)×0 + ∑ 𝑅𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑 𝑠𝑡𝑜𝑐𝑘𝑖× 𝑡𝑖+ ∑ 𝑂𝑝𝑡𝑖𝑜𝑛𝑗× 𝑡𝑗
𝑛𝑜𝑗=1
𝑛𝑠𝑖=1
𝑆𝑎𝑙𝑎𝑟𝑦+𝐵𝑜𝑛𝑢𝑠 + ∑ 𝑅𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑 𝑠𝑡𝑜𝑐𝑘𝑖+ ∑ 𝑂𝑝𝑡𝑖𝑜𝑛𝑗𝑛𝑜𝑗=1
𝑛𝑠𝑖=1
, (1)
where Salary and Bonus are the dollar values of yearly salary and bonus, Restricted stocki is the dollar value
of the ith equity compensation of restricted stocks with vesting period ti (in months), and Optionj is the dollar
value of the jth equity compensation of stock options with vesting period tj (in months). The dollar value of
equity awards are estimated at the end of a fiscal year. In the year t, an executive may be awarded multiple
equity grants with different vesting periods, and ns and no are the total number of such grants in stock and
options, respectively. I also construct alternative measures of incentive horizon. Incentive durationGD is
measured in Eq. (1) except that the dollar value of equity awards are estimated at the date when these awards
8
are granted. Incentive durationFV is measured in Eq. (1) except that the dollar value of equity awards are the
grant date fair value reported by the firm.
I obtain detailed compensation data to construct Incentive duration from Incentive Lab. Incentive
Lab is a comprehensive compensation database that contains in-depth information from corporate reports
and proxy statements about compensation of executive officers and directors (e.g., Bettis, Bizjak, Coles,
and Kalpathy 2013). In particular, it provides grant-by-grant information of equity awards such as vesting
schedules, vesting periods, and fair values, which is usually not available in standard compensation
databases. The in-depth compensation data from Incentive Lab allow me to construct Incentive duration to
measure incentive horizon.
Finally, I intersect the executive network data from BoardEx with the incentive horizon data from
Incentive Lab and complement firm and executive characteristics using data from Compustat, CRSP, and
Execucomp. The sample period is from 2000 through 2013. The boundary of the sample is set by BoardEx
and Incentive Lab, which mostly covers S&P 1500 firms. The final sample contains executive level data of
network connections and incentive horizon with 51,452 observations, which covers 13,485 executive
officers from 1,242 firms during the time period of 2000 – 2013.
I am facing several data limitations when drawing inferences from the final sample. First, the social
network of an individual is dynamically changing and difficult to fully capture. To the extent that the
network data I use from BoardEx covers mainly S&P 1500 firms in the U.S., the network measures in this
paper may underestimate the actual network connectedness for executives. Be that as it may, it is perhaps
relevant to focus on network connectedness among executives and directors who have worked in an S&P
1500 firm, through which the information access and potential job opportunities may be more valuable to
both firms and executives. Second, in this version of the paper, I focus on the incentive duration of new
equity awards granted to executives every year. While the new equity awards reflect the most current
incentive horizon, grants that were awarded before but are yet to fully vest remain important to the overall
incentive horizon. I plan to address this issue by including in the next version of the paper alternative
measures of incentive horizon that take into account the current and past equity grants.
9
3.2 Summary Statistics
The summary statistics, reported in Table 2, show that the average Incentive duration for an
executive is 18.39 months, while the median is 18.80 months. An increase of 7.21 months moves the 25th
percentile of Incentive duration to the median, and another increase of 5.85 months moves the median of
Incentive duration to the 75th percentile. Taken together, the distribution statistics of Incentive duration
seem to suggest that its distribution in the final sample is quite symmetric. Alternative measures of incentive
horizon, Incentive durationGD and Incentive durationFV, also show similar distributions. In untabulated tests,
I compare average incentives horizon for executives across the Fama-French 49 industries. The
considerable amount of variation across industries implies that we should control for industry heterogeneity.
Similar to Gopalan, Milbourn, Song, and Thakor (2014), I also find that executive incentive horizon is
longer in industries that usually have long project horizon such as Aircraft, Medical Equipment,
Shipbuilding, Pharmaceuticals and that have significant intangible assets such as Computers and Computer
Software. I also compare in untabulated tests the average incentive horizon for executives along the time
period of 2000 – 2013. While Incentive duration is generally stable during 2000 – 2005 and 2006 – 2013,
the apparent jump in 2006 warrants further investigation. One potential explanation may be that after the
adoption of FAS 123R in 2005, large public firms in the U.S. are required to expense their stock options in
income statements using fair value on the grant date, which leads to more compensation expenses and thus
results in the decline of option awards to executives.2 Another consequence of the adoption of FAS 123R
may be the increase of vesting periods, which may help reduce the compensation expenses in the income
statement and may also accidentally increase Incentive duration. I further investigate this concern by
splitting the sample by the time period of 2000-2005 and the time period of 2006-2013 and conduct
robustness tests in these subsamples.
Executive network connectedness is mainly measured by Total connections, which is the total
number of individuals with whom the executive shares an overlapping employment history in a public firm
2 For example, see Hayes, Lemmon, and Qiu (2012) for detailed discussion.
10
or overlapping educational history in a university, which excludes those individuals who are employed by
the executive’s current firm. On average, the typical executive in the final sample is connected via
employment or educational network to 203 individuals who have worked in an S&P 1500 firm. 126 of the
connections come from overlapping employment history, and the remaining 77 connections come from
alumni network based on overlapping university education. Alternatively, BoardEx also provides the
number of first-degree connections to all other individuals in the BoardEx universe as the variable Network
size (Akbas, Meschke, and Wintoki 2015). Since Network size captures internal and external networks for
an individual through multiple organizations, it is larger than Total connections in both mean and median.
Compared to the mean, the median value of Total connections is 121 and thus indicates that the distribution
is skewed to the right. I therefore use the natural logarithm of measures of network connectedness in
regressions to reduce the influence of outliers. Similarly, the final sample mainly covers S&P 1500 firms
and thus has distributions of assets, sales, and cash compensation to be also skewed to the right, and I use
the natural logarithm of these variables in regressions to reduce the influence of outliers.
In particular, I use two instruments for network connections in instrumental variable regressions to
address the endogeneity concern that network connectedness may also capture unobservables such as
ability. The motivation and validity of these instruments are discussed in later sections. Number of prior
firms is the number of public firms the executive has previously worked for, and Graduate degree measures
whether the executive has earned a graduate degree. On average, an executive in the sample has worked for
2.59 firms, and 61% of the executives in the final sample have earned a graduate degree. I use the natural
logarithm of Number of prior firms in regressions to reduce the influence of outliers. To measure earnings
management, I use the performance-adjusted cross-sectional variation of the modified Jones model
(Dechow et al. 1995; Kothari et al. 2005). I estimate the residual term in the model as the measure of
discretionary accruals.3 Table 2 shows that the variable Accruals has a mean value of 0.002, which is similar
3 I use the signed value of discretionary accruals to test for earnings management. See Hribar and Nichols (2007) for
discussion of the implications of using signed value and absolute value of discretionary accruals.
11
to the mean value of 0.00 reported by Kothari et al. (2005). To reduce the influence of extreme values, I
winsorize distributions of all continuous variables at the 1st and 99th percentiles.
3.3 Econometric Specifications
I use the following empirical model to examine the effects of executive connections on incentive
horizon,
𝑦𝑖𝑗𝑘𝑡 = 𝛼 + 𝛽′𝑛𝑒𝑡𝑤𝑜𝑟𝑘𝑖𝑗𝑘𝑡 + 𝛾′𝑥𝑖𝑗𝑘𝑡 + 𝛿′𝜇𝑘 + 𝜑′𝜈𝑡 + 휀𝑖𝑗𝑘𝑡, (2)
where i indicate individuals, j indicates firms, k indicates industries, and t indicates years. The outcome
variable, 𝑦𝑖𝑗𝑘𝑡, represents the incentive horizon, Incentive duration, for an executive at a firm in a year. The
covariate 𝑛𝑒𝑡𝑤𝑜𝑟𝑘𝑖𝑗𝑘𝑡 is the natural logarithm of network connections, ln(Total connections). The vector
𝑥𝑖𝑗𝑘𝑡 controls for firm and executive characteristics, 𝜇𝑘 represents two-digit SIC industry fixed effects, and
𝜈𝑡 represents year fixed effects. I assume that the executive-firm-year specific error term, 휀𝑖𝑗𝑘𝑡, is
heteroskedastic and correlated within individuals and follow Petersen (2009) in reporting robust standard
errors clustered by executives in all regressions.
Individuals with larger network size often tend to be those with high ability, and the better network
connectedness of these individuals may merely capture their unobserved ability. Furthermore, firms may
also use compensation contracts with longer incentive horizon to retain executives with high aptitude. To
the extent that unobservables such as ability drives both network connections and incentive horizon, any
findings from OLS regressions should be interpreted with caution against causal inferences. Coles,
Lemmon, and Meschke (2012) document that finance researchers face many challenges in addressing
endogeneity, and I do not claim to fully address all conceivable concerns. To mitigate endogeneity issues
as outlined above, I apply two-stage instrumental variable regressions in the following specification,
𝑦𝑖𝑗𝑘𝑡 = 𝛼 + 𝛽𝑛𝑒𝑡𝑤𝑜𝑟𝑘̅̅ ̅̅ ̅̅ ̅̅ ̅̅ ̅̅𝑖𝑗𝑘𝑡 + 𝛾′𝑥𝑖𝑗𝑘𝑡 + 𝛿′𝜇𝑘 + 𝜑′𝜈𝑡 + 휀𝑖𝑗𝑘𝑡, (3)
where 𝑛𝑒𝑡𝑤𝑜𝑟𝑘̅̅ ̅̅ ̅̅ ̅̅ ̅̅ ̅̅𝑖𝑗𝑘𝑡 is estimated from firs-stage instrumental variable regressions using control variables
in Equation (2) and two excluded instruments: the natural logarithm of Number of prior firms and Graduate
12
degree. Similar instruments are also used in Faleye, Kovacs, and Venkateswaran (2014). The rationale to
use these instruments is that on the one hand the past working experience and graduate school education
may be mechanically associated with network size, on the other hand it is not clear how the number of prior
firms to work for and the graduate degree could be indicative of an individual’s ability that is relevant in
business world. For example, a high-tech company may be more likely to hire a long-term corporate senior
officer who spends most of her career in a large IT company than some job hoppers who frequently switch
among companies. Moreover, one may also argue that executives who never find the need or time to earn
a graduate degree may be individuals of high aptitude. In later sections, I further discuss the relevancy and
exclusion conditions for these instruments.
4 Results
4.1 Univariate Analysis
Table 3 reports Pearson correlation coefficients between variables of the full sample. Consistent
with the hypothesis, Panel A shows that incentive horizon and executive network connections are positively
correlated, no matter whether the connectedness is measured by total connections, connections via
overlapping employment history, connections via overlapping education history in the same university, or
network size provided by BoardEx. Number of prior firms and Graduate degree are also positively
correlated with measures of incentive horizon. More interestingly, prior work history and graduate
education are indeed positively correlated with network connections with correlation coefficients as high
as 0.56, which seems to support the relevancy condition for instrumental variable regression. Intuitively,
prior work history is more correlated with employment connections than with educational connections,
while graduate education seems opposite. It also suggests that prior work history and graduate education
are much more relevant for network connections than for incentive horizon, which may alleviate the concern
that these excluded instruments directly affect incentive horizon. Panel B provides Pearson correlations for
variables in main regression models. Similar to prior studies (e.g., Gopalan, Milbourn, Song, and Thakor
2014), I also find that incentive horizon is negatively correlated with measures of firm risk (Debt ratio,
13
Volatility, and Spread) and corporate governance measures (Entrenchment index and G-index). Firms with
larger size, more long-term assets and R&D investments, more growth options (Market-to-book), better
performance (Excess return), and higher institutional ownership have longer incentive horizon. Executives
with older age, larger cash compensation, and low optimism (Holder30) have shorter incentive horizon.4
Overall, univariate results from correlation matrix are consistent with my hypotheses and prior studies,
which motivate further investigation in multivariate regression frameworks.
4.2 Baseline
Table 4 reports multivariate results for incentive horizon in the full sample. Column (1) include
ln(Total connections), firm characteristics, industry fixed effect, and year fixed effect as independent
variables. Column (2) through (4) further controls for executive characteristics such as position, age and
compensation. Column (5) includes both firm characteristics and executive characteristics and hence
provides the baseline regression. The coefficients on ln(Total connections) are positive for all five
specifications and are statistically significant at the 1% level. Consistent with the hypothesis, better
connected executives have longer incentive horizon. In terms of economic significance, a one-standard-
deviation increase in ln(Total connections) is associated with 6.28% of one standard deviation shock to
Incentive duration.5
Most of the control variables in Table 4 have signs consistent with prior literature. Incentive horizon
for executives is longer in firms that have larger sales, more long-term assets and R&D investments, more
growth opportunities, better stock performance, and higher shares turnover. Firms with higher debt ratio
and more volatile stock performance have shorter incentive horizon for their executives. CEOs have longer
4 For discussion about measures of executives’ overconfidence and low-optimism, see Malmendier and Tate (2005)
and Campbel et al. (2011) 5 1.17 × 0.566 ÷ 10.55 = 0.0628, where 1.17 is the sample standard deviation of ln(Total connections, 0.566 is the
coefficient on ln(Total connections) in Column (5), Table 4, and 10.55 is the sample standard deviation of Incentive
duration.
14
incentive horizon, while executives that are older and have higher cash compensation have shorter incentive
duration.
Consistent with the univariate results, the multivariate results in Table 4 also find that executives
with larger networks have longer incentive horizon. Controlling for common determinants suggested by
prior studies and executive characteristics, I show that network connections are significantly and positively
associated with incentive horizon.
4.3 Subsample Analysis
In Table 5 I examine whether previous results hold in subsample analysis. To account for a potential
exogenous shock to incentive horizon around 2006 due to changes in accounting practices, I divide the full
sample into before-2006 subsample and post-2006 subsample to investigate whether the effect of network
connection on incentive horizon is concentrated in a certain subsample. Columns (1) and (2) of Table 5
show that the coefficients on ln(Total connections) are both positive and significant at the 1% level. It
suggests that despite a potential shock from changes in accounting practices, the positive association
between network connections and incentive horizon is significant before and after such shock. Prior studies
find that firm size is an important determinant of incentive horizon for executives. I then divide the full
sample into small firm subsample and large firm subsample to investigate whether the effect of network
connection on incentive horizon is concentrated in a certain size of firms. The small firm subsample
includes firms with sales smaller than the sample median, while the large firm subsample includes firms
with sales larger than the sample median. Columns (3) and (4) of Table 5 show that the coefficients on
ln(Total connections) are both positive and significant at the 1% level. It seems that the marginal effect of
network connections on incentive horizon is more pronounced in small firm, which may imply that
executives with larger networks are particularly valued by small firms. Lastly, I split the sample into CEO
sample and non-CEO sample. While in both samples the relationship between connections and incentive
duration is positive and significant at the 1% level, the CEO sample exhibits strong economic significance.
In terms of economic significance, a one-standard-deviation increase in ln(Total connections) for a CEO is
15
associated with 7.01% of one standard deviation shock to Incentive duration.6 Overall, results of Table 5
continue to support the hypothesis that executive network connectedness is an effective and important
determinant of incentive horizon.
4.4 Instrumental Variable Regressions
So far I have shown a positive and significant association between an executive’s network
connectedness and her incentive horizon in the full sample and subsamples. Yet it is difficult to estimate
the causal effect of network connectedness on incentive horizon as they may be jointly determined. First,
unobservables such as individual ability are likely to affect both network connections and incentive
horizons. Individuals of high ability may be central in social networks and thus have high connectedness,
while firms may provide long term contracts in order to retain top talents. As a result, such unobservables
are hidden in the error term in Equation (2), and they may cause biases in estimation as they are correlated
with both incentive horizon and network connections. Second, managerial contract is the equilibrium
outcome of a matching mechanism between executives and the board. It is possible that the board has
private information about the growth potential of the firm, and it seeks to hire or retain executives that may
help fulfil the firm’s growth opportunities. For example, when a pharmaceutical firm is planning to start a
new pipeline, the board may hire a well-connected executive officer and offer her long term contracts. The
matching between network connectedness and horizon of executive incentives also challenge the estimation
of causal effect.
To address the endogeneity concern that unobservables such as individual aptitude jointly
determine network connections and incentive horizon, I use two-stage instrumental variable regressions in
Table 6. Consistent with prior studies (Faleye, Kovacs, and Venkateswaran 2014), I explore variations in
the prior employment history and education background and use the number of firms an executive has
6 1.07 × 0.751 ÷ 11.47 = 0.0701, where 1.07 is the CEO sample standard deviation of ln(Total connections, 0.751 is
the coefficient on ln(Total connections) in Column (5), Table 5, and 11.47 is the CEO sample standard deviation of
Incentive duration.
16
previously worked for and whether she has graduate degree as excluded instruments. I examine whether
these instruments are correlated with network connections and are uncorrelated with the error term in
Equation (2).
Panel A of Table 6 reports regressions results in the full sample. Columns (1) through (3) show
reduced form results of instrumental variable regressions, where instruments (ln(Number of prior firms)
and Graduate degree) are directly included in regressions of incentive horizon. The positive and significant
coefficients on the instruments with respect to incentive horizon support the relevancy condition. To
examine the exclusion condition, one needs to argue that the instruments (ln(Number of prior firms) and
Graduate degree) affect incentive horizon only through executive connections. Column (4) reports that
after control for ln(Total connections), coefficients of ln(Number of prior firms) and Graduate degree on
incentive horizon are no longer significant, which is consistent with the exclusion condition. Lastly, Column
(5) reports second-stage results of instrumental variable regressions. The coefficient on ln(Total
connections), instrumented by ln(Number of prior firms), Graduate degree, and second-stage control
variables, is positive and significant at the 1% level. The first stage F statistics are significant at the 1%
level. In unreported tests, weak instrument test are significant at the 1% level and well above the critical
values (around 20) of Stock-Yogo test for weak instruments. Consistent with the reduced form results, these
test statistics reject null hypotheses that the instruments are weak. Hence I find strong evidence to support
the relevancy condition for instruments. Moreover, the Hansen J statistics are not significant at all, not
rejecting the null hypothesis of overidentification. That is, the exclusion condition cannot seem to be
rejected. Therefore, the reduced form results and the econometric tests suggested by prior literature all show
that the instruments used in Table 6 seem to satisfy both the relevancy condition and the exclusion
condition, and hence they are valid instruments. Panel B of Table 6 reports regression results in the CEO
sample and find similar and strong results. In particular, instrumental variable regression in the CEO sample
suggests a strong economic effect of connections on incentive horizon. In terms of economic significance,
a one-standard-deviation increase in ln(Total connections) for a CEO is associated with 12.97% of one
17
standard deviation shock to Incentive duration.7 Taken together, the instrumental variable regressions in
Table 6 confirm what I find in previous multivariate analysis: Network connectedness of an executive
positively affects her incentive horizon.
4.5 Robustness Checks
In Table 7, I investigate the effect of connections on incentive horizon using alternative measures
of incentive horizon, alternative measures of network connectedness, and additional control variables. Panel
A reports regression results with alternative measures of incentive horizon. Instead of measuring incentive
horizon with dollar value of equity grants estimated at the end of a fiscal year, I estimate the dollar value
of equity grants at the grant date and use the grant date fair value of equity grants reported by the company,
respectively. Columns (1) through (2) show that the positive effect of executive connections is robust to
alternative measures of incentive horizon.
Panel B of Table 7 reports regression results with alternative measures of executive connections.
Instead of ln(Total connections), I use ln(Employment connections), ln(Educational connections), and
ln(Network size) to measure network connectedness. Columns (1) and (2) of Table 6 show that connections
via employment history and education history are positively associated with the horizon of executive
compensation, respectively. When put together in Column (3), these two subcategories of network
connectedness also jointly determine the horizon of executive incentives. Moreover, Column (3) of Table
6 finds that the coefficient on ln(Employment connections) is significantly larger than that on
ln(Educational connections) at the 1% level. It seems to imply that the marginal effect of employment-
induced connections on incentive horizon is stronger than that of education-induced connections. Columns
(4) and (5) of Table 6 include ln(Network size) in regressions and show that network connectedness
measured by the size of the total network of an executive is also positively and significantly associated with
7 1.01 × 1.466 ÷ 11.42 = 0.1297, where 1.01 is the CEO sample standard deviation of ln(Total connections, 1.466 is
the coefficient on ln(Total connections) in Column (5) of Panel B, Table 6, and 11.42 is the CEO sample standard
deviation of Incentive duration.
18
incentive horizon. In sum, while I find that incentive horizon is more affected by network connections based
on past employment than past education, alternative measures of network connectedness point to the same
finding: better connected executives have longer horizon of incentives.
Panel C of Table 7 reports regression results with additional control variables. While previous
regressions include control variables suggested by prior studies (Gopalan, Milbourn, Song, and Thakor
2014), it is possible that I haven’t exhausted the potential set of determinants of incentive horizon. Columns
(1) through (7) show that firms with better corporate governance (Entrenchment index, G-index, and Board
independence) have longer incentive horizon for their executives and CEOs that are optimistic have longer
incentive horizon (Holder67, Holder30). In untabulated tests, I also find that firms with institutional
ownership higher than its sample median and spread lower than its sample median have longer incentive
horizon. This is consistent with hypotheses that long-term contracts are desirable for firms with more long-
term perspective. Overall, the positive effect of executive connections on incentive horizon is robust to
alternative measures of incentive horizon, alternative measures of network connectedness, and additional
control variables.
4.6 Heterogeneity in the Effect of Executive Connections on Incentive Horizons
Previous results show that executive connections are positively and significantly associated with
incentive horizon, I then examine through what channel executive connections affect incentive horizon in
Table 8 by investigating interaction terms between connections and relevant control variables.
The effect of network connection may be dampened for firms with greater risk. On one hand the
executive may demand short-term contracts due to greater uncertainty of the firm. On the other hand the
board may find it less feasible to implement long-term strategy due to current risk imposed to the firm.
Columns (1) and (2) of Table 8 show results that support this conjecture. The effect of network
connectedness on incentive horizon decreases in debt ratio and stock volatility.
Network connections may be more valuable to firms with more growth options, as the potential
access to information through network may help explore these growth options in the future. Hence firms
19
are more likely to design long-term contracts to match their growth options with the executive’s network
capital. Column (3) of Table 8 reports a positive coefficient on the interaction term of network
connectedness and market-to-book ratio, although not statistically significant. Untabulated tests show that
in the CEO sample the effect of CEO connectedness on incentive horizon significantly increases in market-
to-book ratio at the 5% level.
Column (4) reports that the effect of executive connectedness on incentive horizon significantly
decreases in high institutional ownership at the 1% level. Prior studies find that institutional ownership is
positively related to the quality of corporate governance (Chung and Zhang 2011). Corporate governance
and executive connectedness may be substitutes to affect incentive horizon as the firm may not need long-
term contract to restrict the executive if the quality of governance is good enough to foster long-term bond
between the firm and the executive. In that sense, results in Column (4) support the substitution effect of
network connectedness and corporate governance.
Lastly, Column (5) of Table 8 shows that the effect of executive connectedness on incentive horizon
significantly decreases in firms with high spread at the 1% level. Consistent with Bolton, Scheinkman and
Xiong (2006), it becomes more costly to implement long-term contracts between the firm and the executive
as short-term mispricing increases. Hence short-term mispricing dampens the effect of executive
connectedness on incentive horizon. However, perhaps more evidence is needed to further test the short-
term mispricing channel.
Taken together, I find evidence on the heterogeneity in the effect of executive connections on
incentive horizon by testing interaction terms between connections and relevant control variables. I show
that the effect of executive connections on incentive horizon is stronger in firms with more growth options
and weaker in firms with greater risk and more short-term mispricing. I also find that for firms with good
governance, long incentive horizon is not used to retain executives with large networks. These results seem
consistent with the hypotheses that executive’s network connectedness affects incentive horizon through
the long-term perspective channel.
20
4.7 Additional Endogeneity Concerns
While the effect of executive connections on incentive horizons is strong in instrumental variable
regressions and remains significant after a battery of robustness tests, I may face additional challenges in
addressing endogeneity as follows. First, the exclusion restriction in instrumental variable regressions may
be violated if instruments directly affect outcome variables. For instance, the propensity for a manager to
have a graduate degree may proxy her aptitude or risk aversion and thus directly impact her compensation.
In untabulated tests, I use only one instrument, ln(Number of prior firms), to instrument for executive
connections and find robust results. It suggests that even if the instrument Graduate degree may not satisfy
the exclusion restriction, I still find consistent results using only ln(Number of prior firms) as the instrument.
Second, I measure a manager’s connections by counting the number of her personal associations with top
executives or directors through employment and education networks. Previous studies have shown that it
is not suitable to include manager fixed effects or firm fixed effects because this type of “elite” executive
connections display very little time-series variation over the manager’s tenure (Engelberg, Gao, and Parsons
2013). I could alternatively rely on the cross-sectional variations to perform additional tests. For example,
school fixed effects could be included so that the effect of executive connections on incentive horizons is
estimated within alumni from the same school. I could also examine connections that were built at the early
stage of a manager’s career and hence are less likely to be predetermined by the current career objective of
the manager. Third, it is also possible that executive connections capture managerial power. While
managerial power is unobserved, powerful executives are often accompanied by weak corporate boards. I
control for various measures of corporate governance in Table 7 and show that the positive effect of
executive connections on incentive horizons is robust and thus not the byproduct of managerial power.
Lastly, while I hypothesize that better connected executives have less short-termism because larger
networks give them better access to information and protect them against adverse job outcomes, these
channels haven’t been directly examined in this version of the paper. I plan to test the implications of these
channels in the next version of the paper. For instance, the access-to-information channel would predict
stronger effects in connections that contain more valuable information. The labor-market-insurance channel
21
would predict stronger effects in connections that are more likely to alleviate career concerns. The detailed
biographical information I have on managers would allow me to further explore cross-sectional variations
in executive connections and examine the channels through which social networks affect incentive
horizons.
4.8 The Effect of Executive Connections on Earnings Management
I examine the effect of executive connections on earnings management by analyzing discretionary
accruals. Using the performance-adjusted cross-sectional variation of the modified Jones model (Dechow
et al. 1995; Kothari et al. 2005), I estimate the residual term in the model as the measure of discretionary
accruals and also split discretionary accruals into positive and negative accruals to examine whether
executive connectedness is associated with earnings-enhancing accruals.
The results are reported in Table 9. Panel A shows that for total accruals, discretionary accruals,
and positive accruals, the coefficients on ln(Total connections) are negative and statistically significant at
the 1% level. In contrast, the relation between executive connectedness and negative accruals is only
significant at the 10% level. This indicates that executives with larger networks engage less in earnings-
enhancing accruals, and it is consistent with the hypothesis that network connections reduce managerial
myopia. Panel B tests whether executive connectedness affects earnings management through incentive
horizon. I first regress Incentive duration on ln(Total connections) and control variables for accruals and
then use the predicted Incentive duration in the accruals regression models. If executive connectedness
affects earnings management through incentive horizon, the coefficient of Predicted (Incentive duration)
should be negative, as variations in the fitted value come from ln(Total connections) after the first-stage
control variables are also included in the second-stage regression. Results in Panel B show that the
coefficient of Predicted (Incentive duration) are negative and significant at the 1% level for total accruals,
discretionary accruals, and positive accruals.
To test whether the negative relationship between executive connections and accruals may be more
than correlation, I use instrumental variable regressions, where ln(Total connections) is estimated by
22
ln(Number of prior firms), Graduate degree, and second-stage control variables. Panel C of Table 9 shows
that the instrumented ln(Total connections) has negative and significant coefficients at the 5% level on total
accruals and discretionary accruals, although the effect is only significant at the 10% level on positive
accruals and negative accruals. The first stage F statistics are significant at the 1% level, and the Hansen J
statistics are insignificant. This seems to support the relevancy condition and exclusion condition of
instrumental variable regressions. In sum, I show strong evidence that executive with large networks engage
in less earnings management, especially in earnings-enhancing accruals.
5 Conclusions
The excessive focus of corporate managers on short-term results has attracted heightened attention
from researchers and practitioners since the recent financial crisis. While prior literature examine the role
of firm characteristics such as monitoring in managerial short-termism, few papers have looked into
managerial attributes that may alleviate short-termism. In this paper, I study the role of executive
connections in mitigating managerial short-termism through access to information and labor market
insurance. I hypothesize that better connected executives have less short-termism because larger networks
give them better access to information and provide them with more labor market insurances against career
concerns, which lead to more willingness to choose long-term projects.
I construct measures of incentive horizons and executive connections in a sample of more than
13,000 executives in S&P 1500 firms from 2000 to 2013. I use the duration of managerial compensation to
measure incentive horizons and the total number of executive connections to other individuals in elite
corporations to measure the executive’s network centrality. I find strong evidence that better connected
executives have significantly longer horizons in compensation incentives in the full sample and subsamples.
To mitigate the endogeneity concern due to omitted variables, I explore variations in past employment and
education history to construct instruments for executive connections. In instrumental variable regressions,
I find that network connectedness of an executive significantly and positively affects her incentive horizon.
This positive relationship is robust to alternative measures of incentive horizons, alternative measures of
23
network connections, and additional control variables. Furthermore, I show that the effect of executive
connections on incentive horizons is stronger in firms with more growth options and weaker in firms with
greater risk. Lastly, I show strong evidence that executives with larger networks engage in less earnings
management, and executive connectedness affects earnings management through the channel of incentive
horizons. Taken together, the empirical evidence strongly supports that better connected executives have
longer incentive horizons and thus exhibit less managerial myopia.
I contribute to two strands of literatures. First, I contribute to the literature on managerial short-
termism. Existing studies have investigated whether and how managerial short-termism can be mitigated
by enhancing monitoring (e.g., Bushee 1998; Farber 2005; Cadman and Sunder 2014). I extend this
literature by documenting that managers’ access to network information and career concerns also play
significant roles in affecting managerial short-termism. Second, I contribute to the literature on social
networks in finance. Fracassi (2014) find that external networks of managers help the firm obtain
information advantages to improve its investment decisions and to achieve better performance. Engelberg,
Gao, and Parsons (2012) find that firms with managers connected to bankers have lower interest rates. My
results imply that executive connections may have a positive effect in inducing long-term corporate
behaviors. In particular, executive connections increase the manager’s willingness to choose desirable yet
risky long-term projects by providing valuable network information and a safety net against adverse job
outcomes. In addition, my results suggest that the manager’s network centrality may facilitate corporate
long-term behaviors via the channel of her incentive horizon.
24
Appendix
Table A1 - First-stage of IV Regressions
The table reports first-stage results of instrumental variable regressions that examine the effect of executive
connections on incentive horizons in the full sample and the CEO sample. The sample period is from 2000
through 2013. Dependent variables are ln(Total connections). The excluded instruments are ln(Number of
prior firms) and Graduate degree. See Table 6 for corresponding second-stage results. Variable definitions
are provided in Table 1. 2-digit SIC industry and year fixed effects are included. Robust standard errors are
adjusted for clustering by executive and presented in the parenthesis. ***, **, and * indicate significance
at the 1%, 5%, and 10% levels, respectively.
All executives CEO
(1) (2)
ln(Number of prior firms) 1.211*** 1.032***
(0.0292) (0.0637)
Graduate degree 0.345*** 0.352***
(0.0201) (0.0444)
ln(Sales) 0.253*** 0.289***
(0.00898) (0.0186)
Long-term assets -0.0316 -0.00479
(0.0589) (0.125)
R&D 3.009*** 2.886***
(0.248) (0.544)
Debt ratio 0.0671 0.0711
(0.0541) (0.114)
Market-to-book 0.0243*** 0.0341**
(0.00818) (0.0169)
Excess return -0.785*** -0.864***
(0.149) (0.287)
Volatility 2.958*** 2.170
(0.751) (1.572)
Share turnover 0.000531 -0.00373
(0.00129) (0.00293)
CEO 0.350*** -
(0.0253)
ln(Age) -0.319*** -0.720***
(0.0763) (0.175)
ln(Cash pay) 0.144*** 0.0645**
(0.0166) (0.0328)
Constant 0.518 3.473***
(0.350) (0.800)
First stage F statistic 1070.11*** 175.76***
Industry and Year FE Yes Yes
Observations 44,538 9,040
Adjusted R2 0.356 0.381
25
References
Akbas, Ferhat and Meschke, Felix and Wintoki, M. Babajide. 2016. Director Networks and Informed
Traders (April 5, 2016). Journal of Accounting & Economics, Forthcoming. Available at SSRN
2378655.
Baranchuk, Nina, Robert Kieschnick, and Rabih Moussawi. 2014. Motivating innovation in newly public
firms. Journal of Financial Economics 111.3: 578-588.
Bebchuk, Lucian A., and Jesse M. Fried. 2010. Paying for long-term performance. University of
Pennsylvania Law Review (2010): 1915-1959.
Berkman, Henk, Paul D. Koch, and P. Joakim Westerholm. 2014. Inside the director network: When
insiders trade outside stocks (June 2014). Available at SSRN 2424527.
Bertrand, Marianne, and Antoinette Schoar. 2003. Managing with style: The effect of managers on firm
policies. The Quarterly Journal of Economics (2003): 1169-1208.
Bettis, J. Carr, John M. Bizjak, Jeffrey L. Coles, and Swaminathan L. Kalpathy. 2013. Performance-vesting
provisions in executive compensation (December 18, 2013). Available at SSRN 2289566.
Bhagat, Sanjai, and Roberta Romano. 2010. Reforming executive compensation: Simplicity, transparency
and committing to the long-term. European Company and Financial Law Review 7.2: 273-296.
Bolton, Patrick, Jose Scheinkman, and Wei Xiong. 2006. Executive compensation and short-termist
behaviour in speculative markets. The Review of Economic Studies 73.3: 577-610.
Burt, Ronald S. 2004. Structural holes and good ideas. American Journal of Sociology 110.2: 349-399.
Bushee, Brian J. 1998. The influence of institutional investors on myopic R&D investment behavior. The
Accounting Review 73.3: 305-333.
Cadman, Brian, and Jayanthi Sunder. 2014. Investor horizon and CEO horizon incentives. The Accounting
Review 89.4: 1299-1328.
Campbell, T. Colin, Michael Gallmeyer, Shane A. Johnson, Jessica Rutherford, and Brooke W. Stanley.
2011. CEO optimism and forced turnover. Journal of Financial Economics 101.3: 695-712.
Cheng, Qiang, Young Jun Cho, and Jae B. Kim. 2014. Managers’ pay duration and voluntary disclosures
(April 9, 2014). 2014 Canadian Academic Accounting Association (CAAA) Annual Conference.
Chung, Kee H, John Elder, Jang-Chul Kim. 2010. Corporate governance and liquidity. Journal of Financial
and Quantitative Analysis 45.2: 265-291.
Chung, Kee H, and Hao Zhang. 2011. Corporate governance and institutional ownership. Journal of
Financial and Quantitative Analysis 46.1: 247-273.
Cohen, Lauren, Andrea Frazzini, and Christopher Malloy. 2008. The small world of investing: board
connections and mutual fund returns. Journal of Political Economy 116.5: 951-979.
26
Cohen, Lauren, Andrea Frazzini, and Christopher Malloy. 2010. Sell‐side school ties. The Journal of
Finance 65: 1409-1437.
Coles, Jeffrey L., Naveen D. Daniel, and Lalitha Naveen. 2014. Co-opted boards. Review of Financial
Studies 27.6: 1751-1796.
Coles, Jeffrey L., Michael L. Lemmon, and J. Felix Meschke. 2012. Structural models and endogeneity in
corporate finance: The link between managerial ownership and corporate performance. Journal of
Financial Economics 103.1: 149-168.
Dechow, Patricia M., Richard G. Sloan, and Amy P. Sweeney. 1995. Detecting earnings management. The
Accounting Review (1995): 193-225.
Duchin, Ran, and Denis Sosyura. 2013. Divisional managers and internal capital markets. The Journal of
Finance 68: 387-429.
Edmans, Alex, Xavier Gabaix, Tomasz Sadzik, and Yuliy Sannikov. 2012. Dynamic CEO compensation.
The Journal of Finance 67.5: 1603-1647.
Edmans, Alex, Vivian W. Fang, and Emanuel Zur. 2013. The effect of liquidity on governance. Review of
Financial Studies 26.6: 1443-1482.
Engelberg, Joseph, Pengjie Gao, and Christopher A. Parsons. 2012. Friends with money. Journal of
Financial Economics 103.1: 169-188.
Engelberg, Joseph, Pengjie Gao, and Christopher A. Parsons. 2013. The price of a CEO's rolodex. Review
of Financial Studies 26.1: 79-114.
Faleye, Olubunmi, Tunde Kovacs, and Anand Venkateswaran. 2014. Do better-connected CEOs innovate
more? Journal of Financial and Quantitative Analysis 49.5-6: 1201-1225.
Farber, David B. 2005. Restoring trust after fraud: Does corporate governance matter? The Accounting
Review 80.2: 539-561.
Fracassi, Cesare, 2014. Corporate Finance Policies and Social Networks (April 17, 2014). Available at
SSRN 1121503.
Fracassi, Cesare, and Geoffrey Tate. 2012. External networking and internal firm governance. The Journal
of Finance 67.1: 153-194.
Glaeser, Edward L., Hedi D. Kallal, Jose A. Scheinkman, and Andrei Shleifer. 1992. Growth in cities.
Journal of Political Economy 100: 1126-1152.
Graham, John R., Campbell R. Harvey, and Shiva Rajgopal. 2005. The economic implications of corporate
financial reporting. Journal of Accounting and Economics 40.1: 3-73.
Gopalan, Radhakrishnan, Todd Milbourn, Fenghua Song, and Anjan V. Thakor. 2014. Duration of
executive compensation. The Journal of Finance 69.6: 2777-2817.
Hayes, Rachel M., Michael Lemmon, and Mingming Qiu. 2012. Stock options and managerial incentives
for risk taking: Evidence from FAS 123R. Journal of Financial Economics 105.1: 174-190.
27
Hribar, Paul, and D. Craig Nichols. 2007. The use of unsigned earnings quality measures in tests of earnings
management. Journal of Accounting Research 45.5: 1017-1053.
Jaffe, Adam B., Manuel Trajtenberg, and Rebecca Henderson. 1993. Geographic localization of knowledge
spillovers as evidenced by patent citations. Quarterly Journal of Economics 108: 577-598.
Kaplan, Steven N., Mark M. Klebanov, and Morten Sorensen. 2012. Which CEO characteristics and
abilities matter? The Journal of Finance 67.3: 973-1007.
Kothari, S. P., Ted E. Laguerre, and Andrew J. Leone. 2002. Capitalization versus expensing: Evidence on
the uncertainty of future earnings from capital expenditures versus R&D outlays. Review of
Accounting Studies 7.4: 355-382.
Kothari, Sagar P., Andrew J. Leone, and Charles E. Wasley. 2005. Performance matched discretionary
accrual measures. Journal of Accounting and Economics 39.1: 163-197.
Laux, Volker. 2012. Stock option vesting conditions, CEO turnover, and myopic investment. Journal of
Financial Economics 106.3: 513-526.
Liu, Yun. 2014. Outside options and CEO turnover: The network effect. Journal of Corporate Finance 28:
201-217.
McDonald, Michael L., Poonam Khanna, and James D. Westphal. 2008. Getting them to think outside the
circle: Corporate governance, CEOs' external advice networks, and firm performance. Academy of
Management Journal 51.3: 453-475.
Malmendier, Ulrike, and Geoffrey Tate. 2005. CEO overconfidence and corporate investment. The Journal
of Finance 60.6: 2661-2700.
Malmendier, Ulrike, and Geoffrey Tate. 2009. Superstar CEOs. The Quarterly Journal of Economics 124.4:
1593-1638.
Manso, Gustavo. 2011. Motivating innovation. The Journal of Finance 66.5: 1823-1860.
Narayanan, M. P. 1985. Managerial incentives for short‐term results. The Journal of Finance 40.5: 1469-
1484.
Nguyen, Bang Dang. 2012. Does the Rolodex matter? Corporate elite's small world and the effectiveness
of boards of directors. Management Science 58.2: 236-252.
Peng, Lin, and Ailsa Röell. 2014. Managerial incentives and stock price manipulation. The Journal of
Finance 69.2: 487-526.
Petersen, Mitchell A. 2009. Estimating standard errors in finance panel data sets: Comparing approaches.
Review of Financial Studies 22.1: 435–480.
Schoar, Antoinette. 2007. CEO careers and style. NBER Working Paper (2007).
Stein, Jeremy C. 1989. Efficient capital markets, inefficient firms: A model of myopic corporate behavior.
The Quarterly Journal of Economics 655-669.
28
Zwiebel, Jeffrey. 1995. Corporate conservatism and relative compensation. Journal of Political Economy
1-25.
29
Table 1 - Variable Definitions
Variable Definition
Incentive horizon
Incentive duration The weighted average duration of components of compensation (salary,
bonus, restricted stock, and stock options) as in Equation (1) of Gopalan,
Milbourn, Song, and Thakor (2014):
(𝑆𝑎𝑙𝑎𝑟𝑦+𝐵𝑜𝑛𝑢𝑠)×0 + ∑ 𝑅𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑 𝑠𝑡𝑜𝑐𝑘𝑖× 𝑡𝑖+ ∑ 𝑂𝑝𝑡𝑖𝑜𝑛𝑗× 𝑡𝑗
𝑛𝑜𝑗=1
𝑛𝑠𝑖=1
𝑆𝑎𝑙𝑎𝑟𝑦+𝐵𝑜𝑛𝑢𝑠 + ∑ 𝑅𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑 𝑠𝑡𝑜𝑐𝑘𝑖+ ∑ 𝑂𝑝𝑡𝑖𝑜𝑛𝑗𝑛𝑜𝑗=1
𝑛𝑠𝑖=1
,
where Salary and Bonus are the dollar values of yearly salary and bonus,
Restricted stocki is the dollar value of the ith equity of restricted stocks
with vesting period ti (in months), and Optionj is the dollar value of the jth
equity compensation of stock options with vesting period tj (in months).
In the year t, an executive may be awarded multiple equity grants with
various vesting periods, and ns and no are the total number of such grants
in stock and options, respectively. Dollar value of equity awards are
estimated at the end of a fiscal year.
Incentive durationGD The weighted average duration of components of compensation (salary,
bonus, restricted stock, and stock options) as in Equation (1) of Gopalan,
Milbourn, Song, and Thakor (2014). Compared to Incentive duration,
Incentive durationGD estimates dollar value of equity awards at the date
when the awards are granted.
Incentive durationFV The weighted average duration of components of compensation (salary,
bonus, restricted stock, and stock options) as in Equation (1) of Gopalan,
Milbourn, Song, and Thakor (2014). Compared to Incentive duration,
Incentive durationFV uses the grant date fair value reported by the firm to
measure the dollar value of equity awards.
Executive network
Total connections The number of individuals with whom the executive shares an overlapping
employment history in a public firm or an overlapping educational history
in a university, which excludes those individuals who are employed by the
executive’s current firm.
Employment connections The number of individuals with whom the executive shares an overlapping
employment history in a public firm, which excludes those individuals
who are employed by the executive’s current firm.
Educational connections The number of individuals with whom the executive shares an overlapping
educational history in a university, which excludes those individuals who
are employed by the executive’s current firm.
Network size The size of the executive’s network, provided by BoardEx.
Number of prior firms The number of public firms the executive has previously worked for.
30
Table 1 (Continued)
Graduate degree Equal to one if the executive has earned a graduate degree and zero
otherwise.
Firm and executive characteristics
Accruals Signed discretionary accruals, the residual term estimated from the
performance-adjusted cross-sectional variation of the modified model
(Dechow et al. 1995; Kothari et al. 2005).
Age The executive’s age in the data year.
Assets Firm’s total assets.
Board independence The number of independent directors/board size.
Cash pay The sum of the executive’s annual salary and bonus
Cashflows Cash flows from operating activities/lagged total assets.
CEO Equal to one if the executive is a CEO and zero otherwise.
Debt ratio (Long-term debt + short-term debt)/total assets.
Entrenchment index The Bebchuk, Cohen, and Ferrell (2009) entrenchment index.
Excess return The average monthly stock returns in excess of market returns.
G-index The Gompers, Ishii, and Metrick (2001) governance index.
High inst. ownership Equal to one if the stock ownership by institutional investors is greater
than the sample median and zero otherwise.
High spread Equal to one if the average daily stock bid-ask spread in a year is greater
than the sample median and zero otherwise.
Holder30 Equal to one if starting from the first year when the person has a pattern
that average per-exercised-option value < 30% of average strike price
for at least two years during the career in a firm and zero otherwise. See
Malmendier and Tate (2005) and Campbel et al. (2011).
Holder67 Equal to one if starting from the first year when the person has a pattern
that average per-unexercised-option value > 67% of average strike price
for at least two years during the career in a firm and zero otherwise. See
Malmendier and Tate (2005) and Campbel et al. (2011).
Institutional ownership The stock ownership by institutional investors.
Long-term assets (PP&E + goodwill) / non-cash assets.
Market-to-book Market value of assets / book value of assets.
Negative accruals Negative discretionary accruals, which is the negative value of the the
residual term estimated from the performance-adjusted cross-sectional
variation of the modified Jones model (Dechow et al. 1995; Kothari et al.
2005).
Positive accruals Positive discretionary accruals, which is the positive value of the the
residual term estimated from the performance-adjusted cross-sectional
31
Table 1 (Continued)
variation of the modified Jones model (Dechow et al. 1995; Kothari et al.
2005).
R&D R&D expenses/total assets, where missing values in R&D are coded as
zero.
Sales Firm’s sales.
Sales growth The firm’s annual growth rate of sales.
S.D. (Cashflows) The standard deviation of the ratio of cash flows to lagged total assets over
previous five years.
S.D. (Sales growth) The standard deviation of yearly sales growth rate of the firms during the
previous five years.
Share turnover The average of (daily share trading volume/shares outstanding) during the
data year.
Spread The average daily stock bid-ask spread in a year.
Total accruals (Income before extraordinary items – operating cash flows)/lagged total
assets.
Volatility Standard deviation of daily stock returns in a year.
32
Table 2 - Summary Statistics
The table provides the summary statistics of the full sample. The sample period is from 2000 to 2013.
Variable definitions are provided in Table 1.
Variable N Mean S.D. 25th % Median 75th %
Incentive horizon
Incentive duration (months) 51,452 18.39 10.55 11.59 18.80 24.65
Incentive durationGD (months) 51,089 18.28 10.60 11.43 18.76 24.59
Incentive durationFV (months) 37,959 19.29 9.40 12.79 19.02 24.67
Executive network
Total connections 51,452 203.03 233.36 46 121 267
Employment connections 51,452 125.75 184.49 26 55 134
Educational connections 51,452 76.57 107.16 1 27 114
Network size 51,452 539.28 538.88 142 357 751
Number of prior firms 51,438 2.59 0.89 2 2 3
Graduate degree 44,548 0.61 0.48 0 1 1
Firm & executive characteristics
Assets ($ millions) 51,452 18,118 46,041 1,754 4,278 13,659
Sales ($ millions) 51,452 8,537 15,866 1,211 2,990 8,227
Long-term assets 51,452 0.42 0.24 0.24 0.43 0.60
R&D 51,452 0.026 0.050 0 0 0.030
Debt ratio 51,452 0.24 0.19 0.098 0.22 0.35
Market-to-book 51,452 1.96 1.23 1.16 1.53 2.25
Excess return 51,452 0.0087 0.031 -0.0084 0.0059 0.023
Volatility 51,452 0.026 0.014 0.016 0.022 0.031
Share turnover 51,452 10.83 8.26 5.27 8.45 13.68
CEO 51,452 0.18 0.39 0 0 0
Age 51,452 51.87 6.74 47 52 57
Cash pay 51,452 849,698 830,004 405,381 587,090 955,001
Entrenchment index 19,784 2.56 1.28 2 3 4
G-index 28,403 9.46 2.57 8 9 11
Board independence 51,419 0.73 0.16 0.67 0.77 0.86
High Inst. ownership 50,780 0.50 0.50 0 0 1
Institutional ownership 50,780 0.78 0.40 0.65 0.79 0.91
Holder67 43,883 0.43 0.49 0 0 1
Holder30 37,070 0.072 0.26 0 0 0
High spread 51,452 0.50 0.50 0 1 1
Spread 51,452 0.033 0.017 0.022 0.029 0.040
Sales growth 51,419 0.10 0.24 -0.010 0.074 0.17
Cashflows 50,431 0.11 0.095 0.055 0.10 0.16
S.D. (Sales growth) 50,127 0.24 0.41 0.071 0.13 0.24
S.D. (Cashflows) 51,354 0.056 0.065 0.020 0.037 0.067
Earnings management
Total accruals 50,367 -0.060 0.071 -0.088 -0.051 -0.022
Accruals 50,367 0.0024 0.062 -0.026 0.0035 0.034
Positive accruals 27,072 0.043 0.040 0.014 0.032 0.059
Negative accruals 23,295 -0.044 0.048 -0.058 -0.029 -0.012
33
Table 3 - Correlation Matrix
The table reports Pearson correlation coefficients for variables in the full sample. The sample period is from
2000 to 2013. Variable definitions are provided in Table 1. Bolded coefficients are significant at the 5%
level.
Panel A: Pearson correlations for incentive horizon and executive network measures
Variable 1 2 3 4 5 6 7 8
1 Incentive duration
2 Incentive duration GD 0.95
3 Incentive duration FV 0.88 0.89
4 ln(Total connections) 0.17 0.17 0.15
5 ln(Employment connections) 0.18 0.19 0.17 0.78
6 ln(Educational connections) 0.08 0.08 0.07 0.71 0.24
7 ln(Network size) 0.15 0.15 0.14 0.77 0.59 0.59
8 ln(Number of prior firms) 0.07 0.07 0.05 0.42 0.56 0.12 0.30
9 Graduate degree 0.04 0.03 0.02 0.21 0.10 0.20 0.17 0.07
34
Table 3 (Continued)
Panel B: Pearson correlations for variables in main regression models
V
aria
ble
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
1
Ince
nti
ve
dura
tion
2
ln(T
ota
l co
nnec
tions)
0.1
7
3
ln(S
ales
) 0.1
4
0.2
6
4
Long
-ter
m a
sset
s 0.0
2
-0.0
1
0.0
8
5
R&
D
0.1
1
0.1
8
-0.3
4
-0.1
9
6
Deb
t ra
tio
-0.0
6
0.0
2
0.0
9
0.2
5
-0.1
7
7
Mar
ket
-to
-book
0.1
6
0.0
3
-0.2
6
-0.0
8
0.3
5
-0.1
8
8
Exce
ss r
eturn
0.0
5
-0.0
7
-0.1
2
-0.0
1
0.0
0
-0.0
2
0.2
4
9
Vola
tili
ty
-0.0
3
-0.0
1
-0.2
9
-0.1
1
0.2
3
0.0
3
0.0
2
0.1
3
10
Shar
e tu
rnover
0.1
0
0.0
5
-0.1
5
-0.0
5
0.2
3
-0.0
3
0.0
9
0.0
1
0.5
2
11
CE
O
0.0
8
0.2
0
0.0
2
-0.0
0
-0.0
0
0.0
1
0.0
1
0.0
1
-0.0
0
-0.0
2
12
ln(A
ge)
-0
.04
0.0
2
0.1
9
0.0
3
-0.1
1
0.0
2
-0.0
9
-0.0
2
-0.1
4
-0.1
0
0.2
1
13
ln(C
ash p
ay)
-0.0
1
0.2
7
0.4
5
-0.0
5
-0.1
6
0.0
6
-0.0
6
0.0
1
-0.1
9
-0.1
5
0.4
3
0.2
2
14
Entr
ench
men
t in
dex
-0.0
9
-0.0
7
-0.0
7
0.0
5
-0.1
2
0.0
7
-0.1
3
0.0
1
-0.1
2
-0.0
8
0.0
1
0.0
4
-0.0
4
15
G-i
ndex
-0
.06
0.0
1
0.1
1
0.0
4
-0.1
1
0.0
5
-0.1
2
0.0
0
-0.1
7
-0.1
8
0.0
1
0.0
9
0.0
4
0.7
1
16
Boar
d i
ndep
enden
ce
0.1
0
0.1
6
0.2
4
0.0
9
-0.0
4
0.0
0
-0.1
0
-0.1
2
-0.2
1
0.0
8
-0.0
1
0.0
8
0.0
1
0.1
3
0.1
9
17
Inst
ituti
onal
ow
ner
ship
0.0
3
0.0
0
-0.0
3
0.0
3
0.0
1
0.0
3
0.0
1
-0.0
3
0.0
0
0.1
7
-0.0
1
-0.0
3
-0.0
2
0.0
2
-0.0
4
0.1
2
18
Hold
er67
0.0
8
-0.0
5
-0.0
8
-0.0
1
0.0
0
-0.0
9
0.2
3
0.1
0
0.0
1
0.0
4
0.1
9
0.1
4
0.1
5
-0.0
4
-0.0
9
-0.0
8
0.0
3
19
Hold
er30
-0.0
1
0.0
4
0.1
0
0.0
2
-0.0
6
0.0
1
-0.0
6
-0.0
3
-0.0
8
-0.0
7
0.0
6
0.1
1
0.0
6
0.0
1
0.0
7
0.0
6
-0.0
5
-0.0
6
20
Spre
ad
-0
.03
-0.0
2
-0.3
1
-0.1
1
0.2
5
0.0
3
0.0
3
0.1
3
0.9
7
0.5
1
-0.0
0
-0.1
5
-0.1
9
-0.1
3
-0.1
7
-0.2
2
-0.0
0
0.0
1
-0.0
8
35
Table 4 - The Effect of Executive Connections on Incentive Horizons
The table examines the effect of executive connections on incentive horizons in the full sample. The sample
period is from 2000 to 2013. Dependent variables are Incentive duration. Variable definitions are provided
in Table 1. 2-digit SIC industry and year fixed effects are included. Robust standard errors are adjusted for
clustering by executive and presented in the parenthesis. ***, **, and * indicate significance at the 1%, 5%,
and 10% levels, respectively.
(1) (2) (3) (4) (5)
ln(Total connections) 0.681*** 0.509*** 0.676*** 0.786*** 0.566***
(0.0667) (0.0670) (0.0666) (0.0677) (0.0669)
ln(Sales) 1.377*** 1.419*** 1.432*** 1.608*** 2.002***
(0.0661) (0.0658) (0.0662) (0.0712) (0.0712)
Long-term assets 2.552*** 2.509*** 2.525*** 2.512*** 2.354***
(0.456) (0.454) (0.454) (0.457) (0.453)
R&D 18.48*** 19.33*** 18.20*** 18.58*** 19.94***
(1.845) (1.834) (1.836) (1.851) (1.825)
Debt ratio -1.943*** -1.927*** -1.938*** -1.889*** -1.794***
(0.438) (0.434) (0.437) (0.438) (0.431)
Market-to-book 1.304*** 1.302*** 1.295*** 1.307*** 1.295***
(0.0717) (0.0720) (0.0711) (0.0716) (0.0711)
Excess return 17.63*** 17.63*** 17.75*** 18.91*** 20.44***
(1.627) (1.625) (1.622) (1.621) (1.605)
Volatility -26.87*** -25.41*** -29.83*** -30.84*** -35.91***
(6.329) (6.304) (6.302) (6.328) (6.259)
Share turnover 0.0962*** 0.0967*** 0.0928*** 0.0975*** 0.0958***
(0.0106) (0.0105) (0.0106) (0.0107) (0.0105)
CEO 2.115*** 4.017***
(0.196) (0.222)
ln(Age) -4.302*** -5.143***
(0.546) (0.552)
ln(Cash pay) -1.068*** -2.221***
(0.134) (0.147)
Constant 1.592* 1.702* 18.36*** 13.45*** 46.52***
(0.894) (0.887) (2.311) (1.728) (2.750)
Industry fixed effect Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes
Observations 51,452 51,452 51,452 51,452 51,452
Adjusted R2 0.147 0.153 0.150 0.150 0.167
36
Table 5 - Subsample Analysis
The table examines the effect of executive connections on incentive horizons in subsamples. The sample
period is from 2000 to 2013. Dependent variables are Incentive duration. Columns 1 and 2 report regression
results in subsamples with time period from 2000 to 2005 and from 2006 to 2012, respectively. Columns 3
and 4 report regression results in subsamples with sales size below the sample median and above the sample
median, respectively. Columns 5 and 6 report regression results in CEO sample and non-CEO sample,
respectively. Variable definitions are provided in Table 1. 2-digit SIC industry and year fixed effects are
included. Robust standard errors are adjusted for clustering by executive and presented in the parenthesis.
***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.
2000-2005 2006-2013 Small Firm Large Firm CEO Non-CEO
(1) (2) (3) (4) (5) (6)
ln(Total connections) 0.586*** 0.519*** 0.612*** 0.374*** 0.751*** 0.528***
(0.0967) (0.0780) (0.0883) (0.0937) (0.188) (0.0700)
ln(Sales) 1.556*** 2.238*** 1.638*** 2.200*** 1.955*** 1.999***
(0.107) (0.0808) (0.138) (0.120) (0.188) (0.0756)
Long-term assets 2.605*** 2.498*** 2.001*** 3.031*** 1.521 2.586***
(0.659) (0.556) (0.600) (0.660) (1.205) (0.475)
R&D 17.86*** 18.79*** 13.15*** 22.80*** 19.23*** 19.97***
(2.745) (2.171) (2.159) (3.680) (4.641) (1.947)
Debt ratio -1.412** -1.015** -0.980** -3.512*** -1.139 -1.957***
(0.665) (0.463) (0.497) (0.775) (1.172) (0.451)
Market-to-book 1.110*** 1.280*** 1.111*** 1.589*** 0.898*** 1.388***
(0.0961) (0.0940) (0.0840) (0.120) (0.175) (0.0759)
Excess return 9.165*** 34.56*** 18.39*** 24.47*** 20.78*** 20.40***
(2.439) (1.975) (2.114) (2.306) (3.922) (1.749)
Volatility 15.47 -116.9*** -36.34*** -48.62*** -33.69** -35.84***
(9.768) (7.768) (8.098) (9.858) (16.37) (6.692)
Share turnover 0.122*** 0.0897*** 0.139*** 0.0117 0.0876*** 0.0966***
(0.0171) (0.0116) (0.0129) (0.0173) (0.0300) (0.0110)
CEO 3.096*** 4.521*** 3.724*** 4.265*** - -
(0.308) (0.269) (0.317) (0.291)
ln(Age) -5.877*** -3.996*** -5.397*** -4.505*** -9.533*** -4.214***
(0.776) (0.658) (0.708) (0.791) (1.579) (0.569)
ln(Cash pay) -1.334*** -2.788*** -2.141*** -2.289*** -2.295*** -2.193***
(0.196) (0.198) (0.213) (0.193) (0.346) (0.159)
Constant 35.84*** 49.15*** 50.23*** 44.13*** 70.08*** 42.34***
(3.764) (3.367) (3.734) (3.845) (7.502) (2.885)
Industry fixed effect Yes Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes Yes
Observations 23,249 28,203 25,724 25,728 9,491 41,961
Adjusted R2 0.128 0.187 0.160 0.184 0.153 0.166
37
Table 6 - Instrumental Variable Regressions
The table examines the effect of executive connections on incentive horizons using instrumental variable
regressions in the full sample and the CEO sample, respectively. The sample period is from 2000 to 2013.
Dependent variables are Incentive duration. Columns (1) through (4) report reduced form results. Column
(5) reports second-stage results of instrumental variable regressions, where ln(Total connections) is
estimated by ln(Number of prior firms), Graduate degree, and second-stage control variables. See
Appendix Table A1 for corresponding first-stage results. Variable definitions are provided in Table 1. 2-
digit SIC industry and year fixed effects are included. Robust standard errors are adjusted for clustering by
executive and presented in the parenthesis. ***, **, and * indicate significance at the 1%, 5%, and 10%
levels, respectively.
Panel A: All executives Reduced form of IV IV
(1) (2) (3) (4) (5)
ln(Total connections) 0.474*** 0.617***
(0.0823) (0.164)
ln(Number of prior firms) 0.696*** 0.665*** 0.0910
(0.216) (0.216) (0.235)
Graduate degree 0.341** 0.316** 0.152
(0.154) (0.154) (0.156)
ln(Sales) 2.156*** 2.151*** 2.149*** 2.029*** 1.995***
(0.0739) (0.0739) (0.0738) (0.0755) (0.0846)
Long-term assets 2.280*** 2.300*** 2.273*** 2.288*** 2.292***
(0.485) (0.485) (0.485) (0.486) (0.486)
R&D 21.52*** 21.75*** 21.34*** 19.91*** 19.50***
(1.878) (1.872) (1.878) (1.882) (1.962)
Debt ratio -2.068*** -2.052*** -2.083*** -2.115*** -2.123***
(0.463) (0.463) (0.464) (0.463) (0.463)
Market-to-book 1.294*** 1.293*** 1.296*** 1.284*** 1.280***
(0.0747) (0.0748) (0.0747) (0.0747) (0.0747)
Excess return 19.01*** 19.00*** 18.99*** 19.36*** 19.48***
(1.714) (1.713) (1.714) (1.713) (1.721)
Volatility -30.32*** -29.77*** -30.42*** -31.83*** -32.29***
(6.644) (6.638) (6.643) (6.628) (6.643)
Share turnover 0.0869*** 0.0882*** 0.0869*** 0.0866*** 0.0864***
(0.0113) (0.0112) (0.0112) (0.0113) (0.0113)
CEO 4.336*** 4.347*** 4.334*** 4.168*** 4.117***
(0.229) (0.229) (0.229) (0.228) (0.239)
ln(Age) -4.907*** -4.897*** -4.922*** -4.771*** -4.723***
(0.593) (0.594) (0.593) (0.590) (0.595)
ln(Cash pay) -2.302*** -2.280*** -2.297*** -2.365*** -2.389***
(0.157) (0.157) (0.157) (0.157) (0.157)
Constant 47.50*** 47.66*** 47.40*** 47.15*** 47.62***
(2.937) (2.933) (2.936) (2.928) (3.280)
First stage F statistic 1070.11***
Hansen J statistic 0.57
Industry and Year FE Yes Yes Yes Yes Yes
Observations 44,538 44,538 44,538 44,538 44,538
Adjusted R2 0.162 0.162 0.162 0.164 0.163
38
Table 6 (Continued)
Panel B: CEO only Reduced form of IV IV
(1) (2) (3) (4) (5)
ln(Total connections) 0.526** 1.466***
(0.219) (0.443)
ln(Number of prior firms) 1.416*** 1.357*** 0.814
(0.518) (0.520) (0.566)
Graduate degree 0.733* 0.686* 0.501
(0.392) (0.392) (0.399)
ln(Sales) 2.162*** 2.161*** 2.149*** 1.997*** 1.727***
(0.185) (0.185) (0.184) (0.193) (0.222)
Long-term assets 1.897 2.003 1.911 1.914 1.905
(1.219) (1.226) (1.222) (1.221) (1.218)
R&D 20.22*** 20.85*** 19.78*** 18.26*** 15.54***
(4.664) (4.646) (4.635) (4.658) (4.758)
Debt ratio -0.778 -0.690 -0.811 -0.848 -0.920
(1.210) (1.205) (1.210) (1.208) (1.206)
Market-to-book 0.950*** 0.959*** 0.956*** 0.938*** 0.904***
(0.178) (0.179) (0.178) (0.177) (0.176)
Excess return 19.53*** 19.40*** 19.39*** 19.84*** 20.69***
(4.011) (4.015) (4.011) (4.014) (4.027)
Volatility -26.16 -23.48 -25.65 -26.79 -29.18*
(16.42) (16.38) (16.38) (16.44) (16.50)
Share turnover 0.0710** 0.0739** 0.0705** 0.0725** 0.0758**
(0.0307) (0.0305) (0.0306) (0.0307) (0.0310)
ln(Age) -8.991*** -9.106*** -9.063*** -8.685*** -7.986***
(1.601) (1.606) (1.595) (1.588) (1.662)
ln(Cash pay) -2.356*** -2.297*** -2.343*** -2.377*** -2.446***
(0.353) (0.353) (0.351) (0.353) (0.357)
Constant 69.33*** 69.82*** 69.20*** 67.37*** 63.88***
(7.670) (7.683) (7.655) (7.600) (8.341)
First stage F statistic 175.76***
Hansen J statistic 0.25
Industry and Year FE Yes Yes Yes Yes Yes
Observations 9,040 9,040 9,040 9,040 9,040
Adjusted R2 0.147 0.147 0.148 0.149 0.145
39
Table 7 - Robustness Checks
The table examines the robustness of the effect of executive connections on incentive horizons using
alternative measures of incentive horizon, alternative measures of connections, and additional control
variables. The sample period is from 2000 to 2013. Dependent variables are Incentive duration. Variable
definitions are provided in Table 1. 2-digit SIC industry and year fixed effects are included. Robust standard
errors are adjusted for clustering by executive and presented in the parenthesis. ***, **, and * indicate
significance at the 1%, 5%, and 10% levels, respectively.
Panel A: Alternative measures of incentive horizons
Incentive durationGD Incentive durationFV
(1) (2)
ln(Total connections) 0.531*** 0.434***
(0.0670) (0.0688)
ln(Sales) 2.032*** 2.005***
(0.0715) (0.0709)
Long-term assets 2.182*** 2.145***
(0.460) (0.461)
R&D 18.77*** 14.15***
(1.868) (1.818)
Debt ratio -1.889*** -2.024***
(0.440) (0.459)
Market-to-book 1.355*** 1.367***
(0.0709) (0.0763)
Excess return -17.50*** -11.14***
(1.627) (1.702)
Volatility -21.86*** -44.19***
(6.305) (6.661)
Share turnover 0.116*** 0.0704***
(0.0104) (0.0108)
CEO 3.888*** 4.962***
(0.223) (0.226)
ln(Age) -4.886*** -3.055***
(0.551) (0.573)
ln(Cash pay) -2.094*** -2.414***
(0.146) (0.169)
Constant 43.93*** 43.40***
(2.748) (2.922)
Industry fixed effect Yes Yes
Year fixed effect Yes Yes
Observations 51,089 37,959
Adjusted R2 0.163 0.197
40
Table 7 (Continued)
Panel B: Alternative measures of executive connections
(1) (2) (3) (4) (5)
ln(Employment connections) 0.555*** 0.502*** 0.392***
(0.0650) (0.0656) (0.0741)
ln(Educational connections) 0.193*** 0.154*** 0.0920**
(0.0319) (0.0321) (0.0373)
ln(Network size) 0.595*** 0.286***
(0.0725) (0.0949)
ln(Sales) 1.969*** 2.130*** 1.954*** 2.005*** 1.928***
(0.0717) (0.0702) (0.0718) (0.0711) (0.0720)
Long-term assets 2.352*** 2.369*** 2.355*** 2.427*** 2.386***
(0.451) (0.453) (0.452) (0.454) (0.452)
R&D 19.82*** 22.00*** 19.37*** 20.33*** 19.03***
(1.836) (1.816) (1.830) (1.822) (1.829)
Debt ratio -1.835*** -1.702*** -1.819*** -1.801*** -1.841***
(0.430) (0.431) (0.430) (0.432) (0.431)
Market-to-book 1.297*** 1.295*** 1.293*** 1.299*** 1.294***
(0.0709) (0.0713) (0.0709) (0.0711) (0.0709)
Excess return 20.44*** 20.13*** 20.49*** 20.38*** 20.55***
(1.605) (1.607) (1.605) (1.603) (1.603)
Volatility -36.23*** -33.59*** -36.16*** -35.26*** -36.43***
(6.268) (6.255) (6.249) (6.255) (6.246)
Share turnover 0.0975*** 0.0960*** 0.0957*** 0.0966*** 0.0958***
(0.0105) (0.0105) (0.0105) (0.0105) (0.0105)
CEO 4.071*** 4.118*** 3.958*** 3.974*** 3.905***
(0.222) (0.222) (0.222) (0.225) (0.224)
ln(Age) -5.498*** -5.254*** -5.327*** -4.689*** -5.015***
(0.552) (0.553) (0.551) (0.561) (0.567)
ln(Cash pay) -2.199*** -2.169*** -2.245*** -2.257*** -2.279***
(0.147) (0.147) (0.147) (0.147) (0.147)
Constant 48.18*** 47.31*** 48.16*** 44.41*** 46.60***
(2.745) (2.757) (2.744) (2.790) (2.835)
Industry fixed effect Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes
Observations 51,452 51,452 51,452 51,452 51,452
Adjusted R2 0.167 0.166 0.168 0.167 0.168
41
Table 7 (Continued)
Panel C: Additional control variables
(1) (2) (3) (4) (5) (6) (7)
ln(Total connections) 0.577*** 0.594*** 0.556*** 0.569*** 0.564*** 0.561*** 0.567***
(0.101) (0.0879) (0.0671) (0.0669) (0.0730) (0.0842) (0.0668)
Entrenchment index -0.330***
(0.0821)
G-index -0.106***
(0.0369)
Board independence 1.083**
(0.485)
Institutional ownership 0.211
(0.228)
Holder67 1.419***
(0.160)
Holder30 -0.900***
(0.310)
Spread -6.973
(13.54)
ln(Sales) 1.602*** 1.874*** 1.983*** 2.003*** 1.984*** 1.943*** 1.999***
(0.115) (0.0948) (0.0712) (0.0713) (0.0766) (0.0872) (0.0718)
Long-term assets 2.970*** 2.144*** 2.362*** 2.450*** 2.173*** 2.153*** 2.353***
(0.682) (0.601) (0.453) (0.455) (0.499) (0.569) (0.453)
R&D 15.55*** 20.09*** 20.06*** 19.99*** 19.33*** 18.16*** 20.01***
(2.897) (2.638) (1.831) (1.832) (2.073) (2.522) (1.838)
Debt ratio -0.599 -1.317** -1.820*** -1.893*** -1.837*** -1.894*** -1.792***
(0.802) (0.625) (0.429) (0.432) (0.505) (0.581) (0.431)
Market-to-book 1.366*** 1.191*** 1.296*** 1.283*** 1.228*** 1.279*** 1.295***
(0.108) (0.0932) (0.0711) (0.0715) (0.0788) (0.0845) (0.0713)
Excess return 10.81*** 22.03*** 20.43*** 20.39*** 20.90*** 18.48*** 20.46***
(2.947) (2.484) (1.604) (1.609) (1.772) (2.017) (1.611)
Volatility 25.61** 11.54 -34.68*** -34.14*** -39.99*** -25.47*** -27.89*
(11.70) (10.08) (6.287) (6.328) (6.864) (8.411) (16.02)
Share turnover 0.170*** 0.121*** 0.0945*** 0.0955*** 0.0836*** 0.0854*** 0.0959***
(0.0194) (0.0160) (0.0105) (0.0107) (0.0117) (0.0139) (0.0105)
CEO 3.384*** 3.662*** 4.011*** 4.030*** 3.826*** 3.921*** 4.017***
(0.327) (0.279) (0.222) (0.223) (0.235) (0.257) (0.222)
ln(Age) -4.853*** -5.548*** -5.153*** -5.137*** -5.124*** -4.359*** -5.147***
(0.826) (0.703) (0.552) (0.553) (0.604) (0.677) (0.552)
ln(Cash pay) -1.579*** -1.872*** -2.212*** -2.229*** -2.186*** -2.060*** -2.222***
(0.215) (0.177) (0.147) (0.147) (0.156) (0.178) (0.147)
Constant 38.30*** 43.72*** 45.79*** 47.34*** 46.20*** 42.32*** 46.58***
(3.881) (3.332) (2.767) (2.691) (3.001) (3.372) (2.751)
Industry fixed effect Yes Yes Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes Yes Yes
Observations 19,784 28,403 51,419 50,780 43,883 37,070 51,452
Adjusted R2 0.154 0.149 0.167 0.167 0.170 0.168 0.167
42
Table 8 – Heterogeneity in the Effect of Executive Connections on Incentive Horizons
The table reports the heterogeneity in the effect of executive connections on incentive horizons. The sample
period is from 2000 to 2013. Dependent variables are Incentive duration. Variable definitions are provided
in Table 1. For conciseness, estimates of control variables are suppressed. 2-digit SIC industry and year
fixed effects are included. Robust standard errors are adjusted for clustering by executive and presented in
the parenthesis. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5)
ln(Total connections) 0.714*** 0.741*** 0.520*** 0.858*** 0.697***
(0.102) (0.117) (0.117) (0.0882) (0.0857)
ln(Total connections) × Debt ratio -0.625**
(0.305)
ln(Total connections) × Volatility -6.596*
(3.410)
ln(Total connections) × Market-to-book 0.0229
(0.0537)
ln(Total connections) × High inst. own. -0.574***
(0.111)
High inst. own. 3.041***
(0.556)
ln(Total connections) × High spread -0.247***
(0.0909)
High spread 0.726
(0.448)
Controls Yes Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes
Observations 51,452 51,452 51,452 50,780 51,452
Adjusted R2 0.167 0.167 0.167 0.168 0.168
43
Table 9 - The Effect of Executive Connections on Earnings Management
The table examines the effect of executive connections on earnings management in the CEO sample. The
sample period is from 2000 to 2013. Dependent variables are Total accruals, Accruals, Positive accruals,
and Negative accruals. Panel A reports the baseline regression results. Panel B uses fitted values from
regressing Incentive duration on ln(Total connections) and control variables from Panel A in fixed effects
OLS regressions to estimate the effect of executive connections on earnings management through incentive
horizons. Panel C reports second-stage results of instrumental variable regressions, where ln(Total
connections) is estimated by ln(Number of prior firms), Graduate degree, and second-stage control
variables. Variable definitions are provided in Table 1. 2-digit SIC industry and year fixed effects are
included. Robust standard errors are adjusted for clustering by executive and presented in the parenthesis.
***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.
Panel A: Baseline
Total accruals Accruals Positive accruals Negative accruals
(1) (2) (3) (4)
ln(Total connections) -0.00443*** -0.00383*** -0.00279*** -0.00165*
(0.00102) (0.000921) (0.000689) (0.000934)
ln(Sales) 0.00732*** 0.00658*** 0.00282*** 0.00421***
(0.000951) (0.000837) (0.000512) (0.000955)
Debt ratio -0.0296*** -0.00802 0.000576 -0.00462
(0.00737) (0.00642) (0.00472) (0.00651)
Market-to-book 0.00983*** 0.00682*** 0.00727*** 0.000929
(0.00157) (0.00134) (0.000963) (0.00125)
Sales growth 0.0187*** 0.00970** 0.0176*** -0.0133***
(0.00576) (0.00478) (0.00358) (0.00506)
Cashflows -0.338*** -0.339*** -0.207*** -0.107***
(0.0229) (0.0191) (0.0166) (0.0190)
S.D. (Sales growth) -0.0121** -0.0111*** -0.00583** -0.00333
(0.00513) (0.00429) (0.00250) (0.00420)
S.D. (Cashflows) -0.123*** -0.103*** 0.0544*** -0.152***
(0.0261) (0.0222) (0.0182) (0.0229)
Constant -0.0559*** -0.00482 0.0191*** -0.0348***
(0.0123) (0.0106) (0.00705) (0.0118)
Industry fixed effect Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes
Observations 8,688 8,688 4,674 4,014
Adjusted R2 0.288 0.247 0.290 0.207
44
Table 9 (Continued)
Panel B: Executive connections on earnings management through incentive horizons
Total accruals Accruals Positive accruals Negative accruals
(1) (2) (3) (4)
Predicted (Incentive duration) -0.00520*** -0.00450*** -0.00328*** -0.00194*
(0.00120) (0.00108) (0.000809) (0.00110)
ln(Sales) 0.0128*** 0.0113*** 0.00627*** 0.00625***
(0.00187) (0.00168) (0.00115) (0.00176)
Debt ratio -0.0348*** -0.0125* -0.00270 -0.00655
(0.00743) (0.00648) (0.00480) (0.00659)
Market-to-book 0.0164*** 0.0125*** 0.0114*** 0.00338*
(0.00222) (0.00195) (0.00148) (0.00191)
Sales growth 0.0246*** 0.0148*** 0.0213*** -0.0111**
(0.00598) (0.00500) (0.00365) (0.00528)
Cashflows -0.353*** -0.351*** -0.216*** -0.113***
(0.0232) (0.0193) (0.0170) (0.0194)
S.D. (Sales growth) -0.0134*** -0.0122*** -0.00661*** -0.00379
(0.00511) (0.00428) (0.00252) (0.00420)
S.D. (Cashflows) -0.103*** -0.0853*** 0.0674*** -0.144***
(0.0263) (0.0224) (0.0188) (0.0232)
Constant -0.00973 0.0351** 0.0483*** -0.0176
(0.0176) (0.0154) (0.0115) (0.0164)
Industry fixed effect Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes
Observations 8,688 8,688 4,674 4,014
Adjusted R2 0.288 0.247 0.290 0.207
45
Table 9 (Continued)
Panel C: Instrumental variable regressions
Total accruals Accruals Positive accruals Negative accruals
(1) (2) (3) (4)
ln(Total connections) -0.00527** -0.00407** -0.00263* -0.00380*
(0.00227) (0.00201) (0.00155) (0.00206)
ln(Sales) 0.00757*** 0.00665*** 0.00277*** 0.00489***
(0.00111) (0.000985) (0.000634) (0.00115)
Debt ratio -0.0296*** -0.00801 0.000539 -0.00480
(0.00735) (0.00640) (0.00468) (0.00645)
Market-to-book 0.00992*** 0.00684*** 0.00725*** 0.00114
(0.00157) (0.00134) (0.000970) (0.00124)
Sales growth 0.0184*** 0.00963** 0.0176*** -0.0137***
(0.00581) (0.00482) (0.00357) (0.00503)
Cashflows -0.339*** -0.339*** -0.207*** -0.110***
(0.0228) (0.0190) (0.0165) (0.0191)
S.D. (Sales growth) -0.0120** -0.0111*** -0.00586** -0.00313
(0.00509) (0.00426) (0.00248) (0.00414)
S.D. (Cashflows) -0.123*** -0.103*** 0.0543*** -0.152***
(0.0261) (0.0221) (0.0181) (0.0226)
Constant -0.0475*** 0.0162 0.0476*** -0.0272*
(0.0163) (0.0140) (0.0100) (0.0154)
First stage F statistic 160.43*** 160.43*** 118.77*** 127.62***
Hansen J statistic 1.93 0.71 0.00 0.03
Industry fixed effect Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes
Observations 8,688 8,688 4,674 4,014
Adjusted R2 0.288 0.247 0.290 0.206