Managerial Ownership and Firm Performance 2

Managerial Ownership and Firm Performance:
Evidence From the 2003 Tax Cut ∗
Xing Li Stephen Teng Sun Constantine Yannelis

Peking University GSM † CityUHK ‡ Chicago Booth §
September 2018
Abstract
Causal evidence on the effect of managerial ownership on firm performance is elusive due to a lack
of within-firm variation and credible empirical designs. We identify this causal effect by exploiting the
2003 Tax Cut as a natural experiment, which increased net-of-tax effective managerial ownership, and
examine changes in Tobin’s Q as well as alternative measures of operating performance. Consistent with
theoretical predictions, our difference-in-difference empirical design uncovers a hump-shaped improve-
ment in firm performance concerning the level of managerial ownership. The increase in performance
is more pronounced for firms with more severe agency problems or under weaker governance, further
demonstrating managerial ownership as the underlying channel.
Keywords: Firm valuation, Director and officer ownership, Corporate governance, Institutional owner-
ship concentration
JEL Classification: M48, G34, G32, D86
∗
We greatly appreciate the comments and suggestions of Kelly Shue (our AFA discussant), Ilona Babenko, Shai
Bernstein, John Beshears, Nick Bloom, Tim Bresnahan, Francisco Perez-Gonzalez, Charles M. C. Lee, Josh Rauh,
Florian Scheuer, Luke Stein, Yiqing Xing, Ding Yuan, Miao (Ben) Zhang and seminar participants at the AFA Meet-
ings and Stanford. We are grateful to Katerina Nikalexi for superb research assistance. Stephen Teng Sun gratefully
acknowledges summer financial support from the Olin Program in Law and Economics and the Rock Center for Cor-
porate Governance at Stanford University. Constantine Yannelis wishes to thank the Alexander S. Onassis Foundation,
the Kapnick Foundation and SIEPR.
†
Guanghua School of Management, Peking University, 5 Yi He Yuan Road, Haidian District, Beijing 100871,
China, xingli@gsm.pku.edu.cn, (+86 10) 6276-7407.
‡
Department of Accountancy, City University of Hong Kong, 83 Tat Chee Ave., Kowloon, Hong Kong. teng-
sun@cityu.edu.hk, (+852) 9068-0127.
§
Department of Finance, University of Chicago Booth School of Business, 5807 S Woodlawn Ave, Chicago, IL
60637, constantine.yannelis@chicagobooth.edu, (217) 721-0587.
Electronic copy available at: https://ssrn.com/abstract=2285638

1 Introduction
Does managerial ownership mitigate the conflict of interest between shareholders and managers in modern
corporations? This classic agency problem has been studied in theoretical work dating to Jensen and Meck-
ling (1976), who predict that a higher level of managerial ownership will improve firm performance due to
a closer alignment of interests between managers and shareholders. Despite theoretical and practical impor-
tance, there is a lack of credible evidence on whether managerial ownership matters for firm performance.
This paper presents novel evidence on the importance of managerial ownership in determining firm perfor-
mance. We exploit the 2003 Tax Cut, which serves as a natural experiment that varies effective managerial
ownership by pre-determined levels across different firms, to identify causal effects (Chetty and Saez, 2005,
2010; Cheng, Hong, and Shue, 2013; Masulis and Reza, 2015; Ma, 2017).
Empirical evidence on managerial ownership and firm performance has been elusive due to two well-
known identification challenges: endogenous managerial ownership and a lack of within-firm variation in
ownership levels.1 For example, the firm-specific contracting environment could endogenously affect both
ownership and performance (Demsetz and Lehn, 1985). Fixed-effect estimates using within-firm variation
traditionally have found null effects (Himmelberg, Hubbard, and Palia, 1999), but this approach has raised
concerns due to a lack of within-firm variation in managerial ownership. Small annual changes in owner-
ship are unlikely to cause substantial shifts in managerial incentives, and thus relying only on within-firm
variation could lack enough statistical power to identify any effect, even if it exists (Zhou, 2001). More
recent work on this topic has used structural models (Coles, Lemmon, and Meschke, 2012), examined CEO
ownership instead of managerial ownership (Kim and Lu, 2011), and studied a particular industry such as
banks (Calomiris and Carlson, 2014).
We overcome these empirical challenges by exploiting the 2003 Tax Cut in the US that created an abrupt
change in the net-of-tax effective managerial ownership as a natural experiment design (Cheng, Hong, and
Shue, 2013; Masulis and Reza, 2015). The 2003 Tax Cut, formally known as the Jobs and Growth Tax Relief
Reconciliation Act of 2003, reduced the highest statutory dividend tax rate from 38.6% to 15% (Chetty and
Saez, 2005, 2010). Since dividend income accounts for over 20% of total compensation for an average S&P
1
See Demsetz and Lehn (1985) for a discussion of endogeneity problems in identifying the effect of managerial ownership on
firm performance. Controversy remains on this topic. Some papers such as Morck, Shleifer, and Vishny (1988) show significant
non-linear relations between insider ownership and firm performance in the cross-section. Other works, such as Demsetz and Lehn
(1985), Himmelberg, Hubbard, and Palia (1999), and others, question the interpretation of existing cross-sectional results as causal
evidence for the effect of managerial ownership on firm performance. Other related work includes Cho (1998) and Calomiris and
Carlson (2014).

500 company CEO (Ma, 2017), the 2003 Tax Cut had a substantial effect on managerial incentives.
The 2003 Tax Cut provides an ideal setting to examine the causal relationship between managerial
ownership incentive and firm performance. One can think of the tax cut as a natural experiment, as it appears
to have been largely unanticipated by the market (Auerbach and Hassett, 2006). More importantly, the
2003 Tax Cut creates a differential increase in managers’ effective ownership across firms, which depends
crucially on the ex-ante level of managerial ownership. Intuitively, when managers own close to zero share
of the firm, their incentives will not be changed much by the tax cut, therefore we should not expect a large
change in performance for these firms. Firms with very high managerial ownership already have managers’
interests well-aligned with shareholders’ and are not likely to display a large change in performance either.
At the same time, firms with intermediate levels of managerial ownership could respond strongly to the
tax cut, which greatly increased the marginal benefit of firm-value enhancing behaviors. These features
enable us to examine the effect of increased managerial incentive on firm value by employing a difference-
in-difference empirical strategy.
Our results show that by boosting managerial ownership incentive, the 2003 Tax Cut significantly in-
creased firm performance as measured by Tobin’s Q.2 Results from both non-parametric and parametric
specifications detect a hump-shaped improvement in firm performance: it is larger for firms with intermedi-
ate managerial ownership while smaller for firms with very small or large managerial ownership. In addition
to the impact on Tobin’s Q, we find that managerial ownership affects other measures of firm performance,
such as the Operating Return on Assets (OROA), the Return on Assets (ROA) and the Return on Equity
(ROE), and also observe the hump-shaped pattern predicted by standard agency theory. While Tobin’s Q
is the focus of the debate on managerial ownership and firm performance, these alternative measures add
support to our main findings. Besides, we find a similar effect on performance metrics not directly related to
a firm’s market value or asset, such as Net Profit Margin (NPM), Total Factor Productivity (TFP), and value
added. These results present direct evidence that managerial ownership impacts firm outcomes, and that the
results are not driven by stock price manipulation or other channels. We also find an effect of the 2003 Tax
Cut on R&D spending and firm leverage, suggesting that when effective ownership increases, managers’
incentives are better aligned with the long run interests of the firm.
To provide evidence that agency problems lie behind our empirical findings, we show that informational
2
To gauge the average effect of the 2003 Tax Cut in the economy, we calculate that the asset-weighted average of managerial
ownership is 6%. Combined with our model estimates, it implies the rise in managerial ownership incentive leads to an around 5%
increase in Tobin’s Q.

asymmetries caused by agency problems impact changes in firm performance. Firms facing more serious
agency problems leave managers more discretionary power and are harder to monitor. At the same time,
these firms have more room for improvement when the interests between managers and shareholders are
better aligned.3 Dividing the sample into subsamples using standard proxies for agency problems, we find a
statistically and economically more pronounced effect of the Tax Cut on firm performance for firms subject
to more severe agency problems.
We provide further evidence that manager incentives impact firm performance by examining how man-
agerial ownership, as a governance mechanism, interacts with other channels of corporate governance. In
particular, we divide firms into strong and weak governance sub-samples based on the pre-determined level
in 2002 and repeat our empirical exercise as before in both sub-samples. We examine anti-takeover in-
dex such as the G-index or E-index as well as institutional ownership.4 Our results show that managerial
ownership as a governance mechanism substitutes with other channels of corporate governance; only firms
with weak internal corporate governance have a significant and hump-shaped effect from the 2003 Tax Cut.
We also uncover a more pronounced effect due to the Tax Cut for firms with lower leverage. These re-
sults strongly support the underlying channel for the increase in firm performance being through managerial
incentives.
We finally implement a battery of robustness checks. We first demonstrate that our results are not driven
by pre-existing differential trends in firm performance before the 2003 Tax Cut by conducting pre-trend tests
as well as placebo tests in which we choose years before 2003 and simulate a tax cut. These results strongly
support our identifying assumption that firms with different levels of managerial ownership would not have
evolved differentially without the 2003 Tax Cut, and therefore the validity of our empirical strategy. We
also use Canadian firms in Compustat to do a placebo test, and find no significant effects in this sample
3
As agency problems are likely to arise from moral hazard or informational asymmetries within firms, we use the following
metrics to proxy agency problems: R&D intensity for opaqueness of the firm (Sufi, 2007), cash holding for free cash-flow (Chava
and Roberts, 2008; Bharath, Jayaraman, and Nagar, 2013) and idiosyncratic volatility of stock returns for informational asymmetries
(Moeller, Schlingemann, and Stulz, 2007).
4
The first governance mechanism we examine is a firm’s anti-takeover strength proxied by the G-index (Gompers, Ishii, and
Metrick, 2003) or E-index (Bebchuk, Cohen, and Ferrell, 2009). Many studies have uncovered a persistent and negative association
between these governance indices and Tobin’s Q or operating performance and interpret these anti-takeover provisions as entrench-
ment protection to managers. The second governance mechanism is concentrated institutional ownership, which is measured by
shares of top-5 largest institutional holders. Previous studies such as Shleifer and Vishny (1986), Edmans and Manso (2011) and
Bharath, Jayaraman, and Nagar (2013) demonstrate blockholders and institutional investors have non-negligible influence on firm
policies and performance. Calomiris and Carlson (2014) also find a negative relationship between managerial ownership and cor-
porate governance for banks. The third governance mechanism we consider is firm leverage, as theoretical work argues a higher
leverage ratio could help reduce the agency problem induced by free cash-flow and discourage managers to enjoy the quiet life
(Jensen, 1986).

around the 2003 event year. This placebo test is valid since few executives of Canadian firms reside in the
US and pay US dividend taxes. Our results are also confirmed in a battery of complementary robustness
tests that rule out (i) potential multiplicative effects, (ii) dividend issuance, (iii) changes in the incentives of
institutional investors, (iv) serial correlation of error terms, (v) industry-specific trends, (vi) changes in cost
of capital, and (vii) the 2004 Foreign Tax Holiday.
Our paper contributes to an important literature in corporate governance where credible causal evidence
is lacking. Coles, Lemmon, and Meschke (2012) summarize findings from a considerable body of work
on this topic as "The unfortunate conclusion is that, at least in the ownership-performance context, proxy
variables, fixed effects, and instrumental variables do not generally provide reliable solutions to simultaneity
bias”. Several recent studies such as Fahlenbrach and Stulz (2009) have turned to examining evidence
from relatively sharp changes in managerial ownership to sidestep the lack of within-firm variation issue
raised by Zhou (2001). Our paper joins this literature, using a natural experiment to examine the effect of
managerial ownership on firm performance. This circumvents problems in studies of managerial incentives,
such as simultaneity and other biases, by exploiting policy-induced variation in managerial ownership, that
is unrelated to other determinants of firm performance. Using our empirical framework, the paper makes
three key contributions.
The primary contribution of our paper is to provide new evidence that managerial ownership does affect
firm performance exploiting the 2003 Tax Cut as a natural experiment. This empirical design has a number
of advantages. First, the 2003 Tax Cut is applied to virtually all public firms in the US, while other notable
within-firm variations in managerial ownership are only available in specific contexts and remain elusive
in large samples. Second, compared to existing studies that look at the actual change in firm managerial
ownership spanning a long period, we focus on a relatively short window, which helps alleviate the concern
about changes in unobservables at the firm level. Third, our empirical design avoids many complications
involved in direct changes in managerial ownership.5 Fourth, we can examine pre-determined trends and do
falsification tests to deal with potential endogeneity concerns.
Our paper contributes to an important literature in corporate governance where credible causal evidence
is lacking. Coles, Lemmon, and Meschke (2012) summarize findings from a considerable body of work
5
Changes in managerial ownership have been found to be associated with large changes in ownership structure at the same
time (Denis and Sarin, 1999). For example, a retiring CEO with a substantial ownership stake could sell shares upon retirement.
Fahlenbrach and Stulz (2009) find firms experiencing a large drop in managerial ownership are more likely to have a concurrent
change in CEO or the board chairman.

on this topic as "The unfortunate conclusion is that, at least in the ownership-performance context, proxy
variables, fixed effects, and instrumental variables do not generally provide reliable solutions to simultaneity
bias”.6 Several recent studies such as Core and Larcker (2002) and Fahlenbrach and Stulz (2009) have
turned to examining evidence from relatively sharp changes in managerial ownership to sidestep the lack of
within-firm variation issue raised by Zhou (2001). Our paper joins this literature, using a natural experiment
to examine the effect of managerial ownership on firm performance. This circumvents problems in studies
of managerial incentives, such as simultaneity and other biases, by exploiting policy-induced variation in
managerial ownership, that is unrelated to other determinants of firm performance. Using our empirical
framework, the paper makes three key contributions.
Lastly, our paper provides a direct test to the agency interpretation of the effects of 2003 Tax Cut on firm
behaviors (Chetty and Saez, 2010), which differs from other contemporary theoretical research motivated
by evidence from the 2003 Tax Cut, such as Korinek and Stiglitz (2009). For example, Korinek and Stiglitz
(2009) assume that retained earnings are allocated efficiently by the manager. They obtain a neutrality result
that permanent dividend tax policy changes have no effects on economic efficiency. Our results show that
there are substantial economic efficiency implications as a result of tax policy changes, providing strong
support for the agency view in Chetty and Saez (2010).
The rest of this paper is organized as follows. Section 2 describes the data, the background for the
2003 Tax Cut and our empirical design. Section 3 presents our main results on managerial ownership and
firm performance, our diagnostic tests for a pretrend, as well as the heterogeneity results on firm agency
problems. Section 4 demonstrates the interactive effects of managerial ownership and other governance
channels. Section 5 presents placebo and robustness checks. Section 6 concludes and discusses avenues for
further research.
6
For examples of recent work, see Holderness, Kroszner, and Sheehan (1999), Core and Guay (2002), Anderson and Reeb
(2003), Adams and Santos (2006), McConnell, Servaes, and Lins (2008), Kim and Lu (2011) and Bova et al. (2014).

2 Data and empirical design
2.1 Data
Our main regressions use a panel of firms from 2000 to 2005, combining information of managerial owner-
ship data in Compact Disclosure,7 accounting data in Compustat and stock return data from CRSP. We use
the aggregate percentage ownership of equity securities owned by all officers and directors of a company as
a measure of managerial ownership.8 Following Chetty and Saez (2005) and Cheng, Hong, and Shue (2013),
we do not include option holdings in managerial ownership for a number of reasons. First, as many options
are not dividend protected, there is no clear effect from the dividend tax cut. Second, options instead of
shares were granted to managers for reasons beyond aligning incentives, including favorable tax treatment,
workarounds for firm financial constraints, as well as accounting manipulation (Lie, 2005; Cicero, 2009).9
We obtain a final sample of 15,846 firm-year observations for 3,656 different firms. We exclude utilities and
financial firms. We require that our sample firms have no missing sales or total assets and they end their
fiscal year in December. We also exclude dual class firms.
We use Tobin’s Q as our main measure of firm performance, which has been the primary outcome studied
in the literature on firm performance and thus allows us to compare our estimates to other findings.10 Tobin’s
Q is proxied by the ratio of market value of assets divided by the book value of assets, where the market value
is calculated as the market value of common equity plus the estimated market value of preferred stock plus
the book value of total liabilities. Following Almeida and Campello (2007), we eliminate firm-years with
Q exceeding 10.0, as their Q is likely to be grossly mismeasured. To guard against outliers, we winsorize
7
Compact Disclosure collects information on firms that have assets in excess of $5 million from Securities and Exchange
Commission (SEC) filings. Data is not provided after 2005.
8
This is a standard measure of managerial ownership, used by many recent studies such as Himmelberg, Hubbard, and Palia
(1999), Holderness, Kroszner, and Sheehan (1999), Zhou (2001) and Fahlenbrach and Stulz (2009). Earlier studies used other
measures. For instance, Morck, Shleifer, and Vishny (1988) use the company’s directors, and Demsetz and Lehn (1985) examine
ownership by the five or twenty largest shareholders of a corporation. The literature on managerial ownership and firm performance
focused on Jensen and Murphy type measures, which is the dollar change in wealth for a dollar change in firm value (Jensen and
Murphy, 1990)
9
Specifically for our study, a large number of firms have been found to engage in option backdating during 1996 to 2005, with
the majority occurring before the implementation of the Sarbanes-Oxley Act (SOX) in August of 2002 (Heron and Lie, 2009). That
is, firms alter the date a stock option was granted to usually an earlier date when the underlying stock price was lower. Accordingly,
option backdating has come to be seen as one of the largest corporate scandals of recent times (Bizjak, Lemmon, and Whitby, 2009).
The climax of the scandal occurred in 2006-2007, when an investigation by the Securities and Exchange Commission resulted in
the resignations of more than 50 senior executives and CEOs at firms across the industry spectrum from restaurants and recruiters to
home builders and healthcare. Therefore, large option grants in these periods may suggest serious agency issues at the firm, instead
of a better alignment of interest between insiders and shareholders, as insider ownership would (Cheng, Hong, and Shue, 2013).
10
We also examine several alternative measures of firm performance, such as return on assets and equity, net profit margin, value
added, TFP, as well as firm behaviors such as R&D spending and the leverage ratio in Section 3.6.

both variables at the 1st and 99th percentiles of their empirical distribution. We obtain similar results if
winsorizing at the 5th and 95th percentiles or at the 10th and 90th percentiles, or using quantile regressions
instead.11
We report summary statistics of all variables used in our study in Table 2. In our sample, the mean and
median of managerial ownership is 23.8% and 16.3%.12 We also calculate the asset-weighted average of
managerial ownership to be 6%. Consistent with existing literature that shows a skewed distribution of in-
sider ownership, the managerial ownership rate is highly skewed. Within our sample of 15,846 observations,
8,788 (55.56%) display managerial ownership of no more than 20%. Among 7,058 (44.54%) observations
that show ownership greater than 20%, 3,570 (22.52%) are greater than 40%; and 1,108 (9.07%) are greater
than 60%. The maximum managerial ownership in our sample is 100%.
2.2 The 2003 Tax Cut
The Jobs and Growth Tax Relief Reconciliation Act of 2003 was signed into law by President George W.
Bush on May 28, 2003. The tax reform was proposed in early January and was applied retroactively from
January 1st, 2003. The main provision of the act is to reduce the highest statutory dividend tax rate from
38.6% to 15%, the most sizeable decrease in dividend tax in recent decades (Yagan, 2015). The Act also
affected three other relevant provisions. It reduced the top capital gains tax rate (which also affected share-
holders) from 20% to 15%.13 It expanded temporary accelerated depreciation for equipment and light struc-
tures investment through 2004, and it accelerated the already-legislated phase-in of reductions in individual
ordinary income tax rates.14 The Act did not change the corporate income tax schedule.
The 2003 Tax Cut offers several advantages in examining the causal relationship between higher man-
agerial ownership incentives and firm performance. First, the tax cut was largely unanticipated by the market
(Auerbach and Hassett, 2006). Auerbach and Hassett (2006) also show that expectations were in line with
the 2003 Tax Cut being permanent. Since dividend income accounts for over 20% of total compensation for
an average S&P 500 company CEO (Ma, 2017), the 2003 Tax Cut has a substantial effect on managerial
11
The results of these perturbations are shown in Appendix Table A.1
12
Summary statistics of our data are similar to previous studies such as Himmelberg, Hubbard, and Palia (1999) and Fahlenbrach
and Stulz (2009).
13
The decrease in capital gains tax also affected managerial incentive in the same direction as the dividend tax decrease. Thus,
we can identify the effects of higher managerial incentive on firm performance from all firms regardless of whether they issue
dividends or not.
14
For example, it immediately reduced the top rate from 38.6 percent to 35 percent rather than waiting for it to fall to 37.6 percent
in 2004 and 35 percent in 2006.

incentives. A reduction in the tax rate works to increase managers’ effective ownership, as a manager’s net
of tax cash-flow increases following the Tax Cut. Other factors such as private benefits and control rights are
unchanged by the tax provision and we discuss other potential changes from the Tax Cut in sections 4 and
5. Second, as shown in the previous paragraph, the Tax Cut was narrow in focus, which is important for our
identification purpose. Other important tax reforms like the Tax Reform Act of 1986, which implemented a
widespread overhaul of the tax system that affected much more than shareholder taxes, would not serve as
a good natural experiment for our purpose. Third and most importantly, the 2003 Tax Cut induces a differ-
ential increase in managers’ effective ownership across firms, which depends crucially on the ex-ante level
of managerial ownership. Intuitively, when managers own close to zero share of the firm, their incentives
will not be changed much by the tax cut. Therefore we should not expect a large change in performance for
these firms. Firms with very high managerial ownership already have managers’ interests well-aligned with
shareholders’ and are not likely to display a large change in performance either. At the same time, firms with
intermediate levels of managerial ownership could respond strongly to the tax cut, which greatly increased
the marginal benefit of firm-value enhancing behaviors. These features enable us to examine the effect of
increased managerial incentive on firm value by employing a difference-in-difference empirical strategy. We
present a more formal illustrative model in the Appendix for interested readers.
The 2003 Tax Cut affected managers’ incentives due to an increase in effective ownership. Given man-
agers’ higher effective ownership, following the Tax Cut they have higher incentives to choose productive
investments and increase future dividend payouts. Chetty and Saez (2005) find that the Tax Cut led to a
sharp increase in corporate dividend payout. Chetty and Saez (2010) interpret this stylized finding in a
simple principal-agent model where the manager trades off private benefits from control against net-of-tax
benefits from cash-flow rights as a shareholder. Our paper can be viewed as a specific empirical application
of this idea to study the important and classic question of managerial ownership and firm performance.
2.3 Empirical Model and Identification Strategy
In this Section, we present our empirical model and discuss our identification strategy. We will first derive
our empirical model from standard model specifications in the managerial ownership and performance lit-
erature, then we will show this specification can also be motivated as a difference-in-difference empirical
design.
We assume that firm performance, as measured by Tobin’s Q, follows a "hump-shaped" or quadratic

relationship with respect to managerial ownership.15 This specification follows classic work such as Stulz
(1988), Himmelberg, Hubbard, and Palia (1999) and Kim and Lu (2011). In comparison to the earlier
literature, we specify the relationship between Tobin’s Q and effective managerial ownership (instead of
nominal managerial ownership) as the following:
yit = γ1 EffectiveShareit + γ2 EffectiveShare2it + Xit β + c̃i + dt + ηit , (1)
where yit is a measure of firm performance, c̃i represents firm fixed-effect, dt is time fixed effect, and
parameters γ1 and γ2 capture respectively the linear and quadratic effects of effective managerial ownership
on firm performance. EffectiveShareit is the net-of-tax ownership rate and is given by (1−τt )Shareit , where
τt is the effective tax rate.16 In our empirical context, the effective tax rate τpre before 2003 does not equal
the effective tax rate τpost during and after 2003 due to the 2003 Tax Cut. We can rewrite
EffectiveShareit = (1 − τpre + ∆τ · P ost2003t ) · Shareit , (2)
where ∆τ = τpre − τpost , and P ost2003t is a dummy that equals one for years including and after 2003.
Plugging (2) into (1), and using pre-determined managerial ownership in 2002, Share2002i , to proxy
Shareit , we derive estimation equation as the following17
yit = α1 Share2002i × P ost2003t + α2 Share20022i × P ost2003t + Xit β + ci + dt + εit . (3)
Our main measure of firm performance yit is Tobin’s Q for firm i at the end of calendar year t, P ost2003t
equals one for years including and after 2003. We include year fixed effects dt to absorb any economy-wide
temporal shocks and firm fixed effects ci to control for the firm-specific contracting environment, following
Himmelberg, Hubbard, and Palia (1999).18 We include a set of firm-level controls Xit to prevent omitted
15
The quadratic specification nests a linear effect of managerial ownership on firm performance. A simple test for the effect
of managerial ownership on firm performance following a quadratic specification is that the estimated coefficient on the quadratic
term is equal to zero. In appendix table A.2, we also present results using a linear specification. These confirm the relationship
between managerial ownership and firm performance.
16
The earlier literature implicitly assumes that the effective tax rate is time-invariant. However, Zhou (2001) argued the nominal
ownership rate lacks variation over time, therefore fixed effects approaches will not be able to identify the effect of managerial
ownership on performance, even if it exists.
17
Simple calculation reveals that α1 = ∆τ · γ1 , α2 = ∆τ 2 + 2∆τ (1 − τpre ) · γ2 , and other time-invariant changes are

reflected in the change of firm fixed effect from c̃i to ci .

18
Our results are robust when we fully saturate the model with 3-digit industry by year fixed effects to guard against concerns
from confounding macro shocks, for example, firms in different industries could have differential growth opportunities.

variable bias.19 The full list of controls is explained in Table 1 and presented in Table 4 along with our main
regression. We cluster standard errors by firm to allow for within-firm serial correlation of error terms.
Alternatively, our main empirical specification (3) can be motivated from a standard difference-in-
difference (DID) design which exploits the 2003 Tax Cut. The 2003 Tax Cut generated variation in ef-
fective (net-of-tax) managerial ownership within firms, and our identification strategy uses a DID approach
exploiting this abrupt change in effective ownership to estimate the effect of managerial ownership on firm
performance. The logic can be understood intuitively. When managers own close to zero share of the firm,
their incentives will not be changed much by the tax cut. Therefore we should not expect a large change
in performance for these firms. Firms with very high managerial ownership already have managers’ inter-
ests well-aligned with shareholders’ and are not likely to display a large change in performance either. At
the same time, firms with intermediate levels of managerial ownership could respond strongly to the tax
cut, which greatly increased the marginal benefit of firm-value enhancing behaviors. More specifically, in
our DID design, the first difference is before and after 2003, while the second difference is among firms
with different levels of ex-ante managerial ownership in 2002. Our treatment dummy is P ost2003t , and the
treatment intensity depends on Share2002i . To account for the non-linear relationship, we interact the treat-
ment dummy with both linear and quadratic terms of the pre-determined managerial ownership ratio in 2002,
Share2002i and Share20022i . In the appendix, we develop a model drawing from Chetty and Saez (2010)
and Cheng, Hong, and Shue (2013) to illustrate that firms with different levels of managerial ownership
are affected differentially, displaying a hump-shaped relationship between change in firm performance and
managerial ownership.20 Therefore, our empirical model is identical to a difference-in-difference approach
by comparing performance of firms with intermediate levels of managerial ownership to firms with very
low or very high managerial ownership before and after the 2003 Tax Cut, which faced different incentive
structures following the reform.

19
We include all controls from Himmelberg, Hubbard, and Palia (1999) and other miscellaneous controls. Construction for all
variables is in Table 1. Three sets of variables are generally regarded to be related to both Tobin’s Q and managerial ownership: firm
size, moral hazard and firm idiosyncratic risk. Firm size controls are log of sales, log(Sale) and the squared term (log(Sale))2 .
Controls for moral hazard include capital-to-sales, the ratio of property, plant, and equipment (PPE) to sales, as well as its square;
R&D-to-capital, the ratio of R&D expenditures to PPE, and ad-to-capital, the ratio of advertising expenditures to PPE. Since not all
firm-years report R&D-to-capital and ad-to-capital, we set missing values of these two variables to 0 and use two indicator variables
for missing observation of R&D-to-capital and ad-to-capital to maintain sample size. Capex-to-capital, capital expenditures divided
by PPE, and OI-to-sales, operating income normalized by sales are also used by Himmelberg, Hubbard, and Palia (1999) to proxy
for the link between high growth and discretionary investment opportunities. Firm idiosyncratic risk, Sigma, is calculated as the
standard error of the residuals from a CAPM model estimated using daily return for the period covered by the annual sample. We
apply the same procedure as in R&D-to-capital and ad-to-capital to deal with missing values of Sigma to maintain sample size.
20
Cheng, Hong, and Shue (2013) first showed that the effect of the tax cut with respect to ownership levels should be hump-
shaped in the context of corporate social responsibility.
10

Using the 2003 Tax Cut allows us to overcome the fact that managerial ownership is not randomly
assigned. Our identification strategy relies on the assumption that, had the 2003 Tax Cut not taken place,
the performance of firms with different levels of managerial ownership would have evolved similarly. Even
though we can not prove this identifying assumption, there is sufficient supporting evidence both from prior
work and our results. Many studies document that there is very little within-firm variation in managerial
ownership (Himmelberg, Hubbard, and Palia, 1999; Zhou, 2001). Consistent with this work, we find little
evidence of firms being on differential trends in ownership prior to the 2003 Tax Cut. We provide standard
tests for pre-trend in Section 3.4 and find that before the 2003 Tax Cut, there is no clear association between
managerial ownership and firm performance. In addition, when we regard year 1999 or 2000 as two separate
placebo years of "Tax Cut" to conduct falsification tests in Section 5.1.1, we cannot replicate the results in
our main regression.
It is important that we make the distinction between endogenous contracting environment and efficient
contracting. While the level of managerial ownership could be endogenously affected by characteristics of
the firm, it is not necessarily efficient. In fact, the efficient contracting view is questioned by a strand of
studies that show incentive contracts are often captured by the management team, in particular the CEO. If
the shareholders can always set the efficient contract, then a change in managerial ownership should have any
positive effect on firm performance (inconsistent with Fahlenbrach and Stulz (2009)) and the 2003 Tax Cut
will have no effects in agency problems. An underlying assumption for the strand of literature that studies
the 2003 Tax Cut in agency perspective from Chetty and Saez (2005) is that while the level of managerial
ownership is endogenous, there are enough contracting frictions that prevent efficient contracting.
Figure 1 illustrates the intuition behind our empirical strategy through a stylized example. The figure
shows the effective ownership for two firms with different levels of nominal managerial ownership. In
both firms, managers pay the highest statutory dividend tax rate. The black long dashed line depicts a
firm with 8% nominal managerial ownership, while the gray shorter dashed lines depict a firm with 2%
of nominal managerial ownership. In 2003, there is an abrupt change in dividend taxes which leads to a
sharp discontinuity in effective managerial ownership through cash-flow rights, and it affects the two firms
differentially. We exploit this change in effective managerial ownership, and compare the differential change
in firm performance before and after the tax reform.
11

3 Results
In this section we present the main results of the paper, and show that managerial ownership affects firm
performance using the 2003 Tax Cut. We begin by illustrating the challenges of using traditional fixed effects
estimators. We then present non-parametric evidence of the effect of pre-determined managerial ownership
on changes in firm performance, as measured by Tobin’s Q. After that, we present our main difference-in-
difference estimates, followed by supporting evidence of our identifying parallel trends assumption. We
then look at heterogeneity by monitoring costs, which strengthens the interpretation that the effect is driven
by an agency channel. Finally, we show robustness in our findings with many other performance measures.
3.1 Fixed Effects Estimates
Table 3 presents a regression model similar to that in Himmelberg, Hubbard, and Palia (1999), with pooled
cross-sectional OLS and firm fixed effects. This table illustrates the challenges associated with obtaining
within-firm variation in managerial ownership. The analysis in Table 3 use a larger and more recent sample
of firms in a shorter horizon while maintaining the same empirical specification and variable construction.
The results are quite similar to those in Himmelberg, Hubbard, and Palia (1999). The pooled cross sectional
OLS shows a significant negative association between ownership and performance, which is the opposite of
what theory would predict. OLS estimates are potentially biased due to endogenous ownership assignment
within a firm’s contracting environment. With firm fixed effects to control for the firm-specific contract-
ing environment, there does not appear to be a significant association between managerial ownership and
performance. However, the estimated coefficients are very imprecise and the specification lacks enough
statistical power to detect even large effects of managerial ownership on firm performance. This is likely
due to a lack of within-firm variation in managerial ownership, see Zhou (2001) and Coles, Lemmon, and
Meschke (2012). We will present results from a difference-in-difference estimation, which deals with both
the endogenous determination of the contracting environment and the lack of within-firm variation.
3.2 Nonparametric Evidence
Before estimating our parametric specification developed in Section 2.3, we present nonparametric evidence
on the relationship between the change in Tobin’s Q and the pre-determined level of managerial ownership in
2002. This non-parametric evidence, along with theory, informs our choice of the parametric specification.
12

A kernel regression avoids imposing a specific functional form on the relationship between the two variables
under study. We use an Epanechnikov kernel with optimal bandwidth based. Figure 2 plots the change in
Tobin’s Q from 2002 to 2004 on the top panel and conditional changes in Tobin’s Q from 2002 to 2004 in
the bottom panel,21 against the managerial share in 2002. The results from both graphs suggest a non-linear
pattern of the relationship between the change in Tobin’s Q and insider ownership: with a positive and
almost linear relationship from zero to around 50%, followed by a decreasing effect until zero as managerial
ownership becomes higher. Because only approximately 14% of firms have managerial ownership above
50%, the kernel regression for that region displays a large confidence band.
3.3 Main Results
Table 4 presents the main results of our difference-in-difference empirical model in (3). The results in Table
4 indicate the increase in effective managerial ownership brought about by the Tax Cut on firm performance
was significant. Moreover, the results indicate that the effect is non-linear, displaying a hump shaped pattern.
More specifically, Table 4 shows results from a specification in which the dependent variable is Tobin’s
Q, regressed on the linear and quadratic terms of the managerial ownership ratio in 2002, both interacted with
treatment dummy P ost2003t . In column (1), we include no controls and see a significant and hump-shaped
treatment effect across firms. When we add controls sequentially from column (2) to column (4), the results
continue to display a significant hump-shaped effect. In column (4), which includes all controls, we see the
coefficient for linear term is 1.92 (s.e.=0.30) and the coefficient for the quadratic term is -1.95 (s.e.=0.40).22
Both estimates are significant at the 1% level. The estimated asset-weighted mean of managerial ownership
in our sample is 6.01%. Informed by these coefficients, this translates into an increase of 0.11 in Tobin’s Q,
or a 5.41% improvement, given an average of Tobin’s Q of 2.03 in the sample.23
3.4 Results by Year and Pre-trend Tests
We provide further evidence of our identifying parallel trends assumption by analyzing the effect of the
reform in each year. The key assumption underlying our analysis is that, in the absence of any reform, firms
21
We regress Tobin’s Q on controls as described in equation (3) and take the difference of residuals between 2004 and 2002 for
the conditional change in Tobin’s Q.
22
The term s.e. stands for standard error.
23
We interpret our results using the asset-weighted mean of managerial ownership instead of the unweighted mean as firm size is
negatively correlated with managerial ownership. The asset-weighted mean of managerial ownership provides a better illustration
of the average effects of the 2003 Tax Cut in the economy. Alternatively, we calculate every firm’s improvement in Tobin’s Q
implied by our regression estimates and find the asset-weighted mean of the improvement in Tobin’s Q to be 3.45%.
13

with differing levels of managerial ownership would have trended similarly. Thus prior to the reform, we
should see no differential effect of 2002 levels of managerial ownership on firm performance.
Results reported in Table 5 confirm the hypothesis. We estimate a model similar to the main specification
above, except that we replace the treatment dummy P ost2003t with a dummy for each year in our sample,
excluding the base year 2002. These results provide support for our identifying assumption, and thus the
validity of our difference-in-difference empirical strategy. It is also worth noting that the treatment effect of
2003 Tax Cut, when examined in each separate year, is quite persistent. The treatment effects in 2003 for
linear and quadratic ownership rates are respectively 1.53 (s.e.=0.31) and -1.91 (s.e.=0.43) , both significant
at the 1% level. These treatment effects increase in magnitude in 2004, with estimated effects for linear and
quadratic ownership rates becoming 2.09 (s.e.=0.34) and -2.56 (s.e.=0.47), both significant at the 1% level.
Year 2005 displays similar results as in previous years.
We extend our sample from 1998 to 2006 and run a similar regression as in Table 5. Results are presented
in Figure 3, giving further evidence that there are no confounding pre-trends. In Figure 3, we plot the
estimated coefficients as well as 95% confidence intervals for the year dummies interacted with treatment
intensity. We see no pre-trend for years before 2003, and a significant effect following the tax reform.
3.5 Heterogeneous Effects by Proxies of Agency Problems
Our results in previous subsections demonstrate that increased managerial incentives due to the 2003 Tax
Cut affected firm performance for firms with differential managerial ownership. If such a change in firm
performance is indeed driven by a boost in managerial incentives, then the effects of our tax cut should move
systematically with the severity of agency problems, indicated by informational asymmetry and/or higher
monitoring costs between managers and outside shareholders (Fahlenbrach and Stulz, 2009). We show
that the 2003 Tax Cut induced a larger effect on firms with more severe agency problems or informational
asymmetries.
To test the above hypothesis, we divide the sample according to a set of proxies for these agency prob-
lems and see if there is any difference in the treatment effects. Our first proxy is the ratio of R&D spending
to capital. This split is based on the fact that these discretionary spending items are hard to observe and diffi-
cult to monitor (Aboody and Lev, 2000; Sufi, 2007). Our second proxy is the fraction of assets held in cash,
also used in Chava and Roberts (2008) and Bharath, Jayaraman, and Nagar (2013). Motivated by Jensen
(1986), uncommitted or free cash-flow could provide managers with an incentive to over-invest. Admati and
14

Pfleiderer (2009) suggest that cash-rich firms are more likely to display agency problems in the context of
shareholder activism. Thus higher free cash-flow could proxy for more severe agency problem. Lastly, we
follow Moeller, Schlingemann, and Stulz (2007) and use idiosyncratic volatility of stock returns as our third
proxy for informational asymmetry. For each of the three specifications, we use the pre-determined level of
the proxy in 2002 to determine whether it is above or below the sample median in 2002.
We find a larger effect in firms subject to more severe agency problems, which is consistent with our
predictions. The results of our analysis are presented in Table 6. Columns (1) and (2) separately estimate
the treatment effects of managerial ownership on firm performance for firms which have R&D spending
to capital ratios below and above the sample median. While both subsamples are significantly affected by
the Tax Cut, we find the economic magnitude of the improvement in firm performance for high opaqueness
firms proxied by a higher R&D spending ratio is over twice the size of low opaqueness firms. We pool
these two samples together in column (3) to test the statistical significance of the difference, which produces
a p-value of 0.03, rejecting the null that these effects are not different. The economic magnitude of the
difference between the two subsamples is also quite large. Columns (4) and (5) examine the heterogeneous
effects on firms with different free cash-flow, as proxied by the cash holding ratio of assets. Similarly, we
find firms with more cash holding improve their performance significantly more than firms with less cash
holding. Again we jointly test whether the significance of the coefficients in two subsamples differ from
each other and obtain a p-value below 0.01. Results on tests for firm asymmetric information proxied by
idiosyncratic volatility of stock returns reveal a similar statistically significant pattern.
3.6 Alternative Performance Measures and Other Firm Outcomes
In previous sections, we followed much of the preexisting literature (Morck, Shleifer, and Vishny, 1988;
Himmelberg, Hubbard, and Palia, 1999; Kim and Lu, 2011) and use Tobin’s Q to measure firm perfor-
mance, which has been at the center of the debate on managerial ownership and firm performance. Tobin’s
Q has a number of advantages, for example it reflects investors’ expectations regarding future firm prospects,
and is less subject to manipulation by managers as opposed to many measures of operating performance.
However, we also examine alternative measures of firms’ operating performance and additional firm be-
havior outcomes to provide further evidence. Effects on these firm behaviors such as R&D spending and
firm leverage provide further insights in regards to the channels through which managerial incentives affect
firm performance and short-term behavior. The evidence also shows the robustness of our results to other
15

measures of firm performance, such as profits or return on equity, and also rules out concerns that the results
could be driven by direct effects of the reform on market valuation.
First, we examine alternative measures of firm performance that strengthen our main findings, and pro-
vide evidence against other explanations for the channel of the observed effect. We use four standard ac-
counting measures of firm operating performance: operating return on asset (OROA), return on asset (ROA),
net profit margin (NPM) and return on equity (ROE).24 While Tobin’s Q as a measure of firm performance
contains investors’ expectation about future firm prospects, these measures for operating performance cap-
ture the concurrent performance of the firm. We use these measures separately as the dependent variable in
a regression similar to our earlier difference-in-difference specification. The results are presented in Table 7.
In column (1), the dependent variable is OROA. Consistent with results in the main regression, we find a
hump-shaped effect due to the Tax Cut, with the coefficient on the linear term being 0.11 (s.e.=0.03) and the
quadratic term being -0.12 (s.e.=0.04), both statistically significant at 1% level. Results on other measures
of operating performance display similar patterns from columns (2) to (4).
The next four columns of Table 7 present additional outcomes: Value Added (VA), Total Factor Pro-
ductivity (TFP), R&D spending, and the leverage ratio. Columns (5)-(6) show the difference-in-difference
estimates in specifications for which the dependent variable is logarithm of VA and TFP. In both specifi-
cations we see the now familiar hump shaped pattern, and highly significant effects. This provides further
evidence of a direct effect of managerial incentives on real performance outcomes, and that the observed
effect is not due to manipulation or other changes in firm market value that could have been induced by the
reform.
Column (7) indicates managerial ownership affects research & development (R&D) spending.25 The
result indicates that firms with higher levels of managerial ownership spend more on R&D, which can
benefit firms in the long run. The coefficient on the linear term is significant at the 10% level, while the
coefficient on the quadratic term is insignificant. The observed effects are consistent with managers’ long-
term incentives being better aligned with those of shareholders. Several studies have found that firms and
managers underinvest in R&D due to earnings targets and short-term incentive structures (Budish, Roin,
24
These variables are defined according to Pérez-González (2006) respectively as: operating income over book value of assets,
net income over book value of assets, net income over sales and net income over book value of common stock.
25
The existing literature finds dark side of higher equity incentive of managers, for example, they are more likely to manage earn-
ings (Cheng and Warfield, 2005), hide bad news (Kim et al., 2011), reduce accounting conservatism (Lafond and Roychowdhury,
2008), result in asset-substitution problem and increase risk (Bizjak et al., 1993; Begley and Feltham, 1999; Rajgopal and Shevlin,
2002) or reduce long-term investment such as R&D (Bushee, 1998).
16

and Williams, 2015). Our result is consistent with managers investing more in long term projects rather than
focusing on short term targets. This suggests that changes in managerial incentives, and the alignment of
interests between managers and shareholders, led to a long term increase in firm value, rather than a focus
on short-term profitability.
Lastly, column (8) indicates that managerial ownership affects firm leverage ratios. This finding is
consistent with managerial ownership altering incentives affecting the short-term behavior of managers.
Firm leverage can amplify shocks, and impair firm performance (Myers, 1977). We again find a hump-
shaped pattern, and firms with higher levels of effective managerial ownership after the Tax Cut see lower
levels of firm debt. Overall, the results from columns (7)-(8) indicate that managerial ownership impacts
R&D and firm leverage, which can have long run performance implications.
4 Interactive effects of managerial ownership and other governance chan-
nels
In this section we investigate how the change in managerial ownership incentives brought about by the Tax
Cut interacts with other channels of corporate governance in the literature by examining the heterogeneous
treatment effects across firms. The section provides causal evidence of a substitution relationship between
managerial ownership and several alternative channels of governance mechanisms. The governance mech-
anisms that we focus on are proxied by an anti-takeover index, institutional blockholding, and leverage.
Kim and Lu (2011) find a hump-shaped relationship between CEO ownership and firm valuation only in
firms with high external governance as measured by strong product market competition or high institutional
ownership. While these works suggest that there may be substitution between different governance channels
in mitigating agency problems, other studies show complementary relationships as well. For example Cre-
mers and Nair (2005) find firms with strong corporate governance, as proxied by a low anti-takeover index,
out-perform poorly governed firms only when blockholder ownership is high. Our work thus contributes to
research that examines how different governance mechanisms interact with each other.
Looking at governance channels also strengthens our interpretation that the 2003 Tax Cut causally affects
firm performance through the channel of providing higher incentive to managers. If the increase in perfor-
mance is due to better alignment of interests between managers and shareholders, then we would expect
this treatment effect to be particularly significant for firms with relatively poor corporate governance, which
17

leaves substantial room for agency problems. In this section, we provide a set of novel and comprehensive
evidence on the substitution between managerial shareholding, a form of governance, and other co-existing
channels of corporate governance. We focus on a set of channels of corporate governance including (1)
an anti-takeover index such as the G-index and E-index, (2) the strength of concentrated institutional own-
ership, and (3) firm leverage. We use firms’ pre-determined governance measures in 2002 to avoid direct
effects from the tax reform to these governance measures.
4.1 Interactive Effects of Managerial Ownership and Antitakeover Provisions
We first examine how corporate governance interacts with managerial ownership. We find a larger effect of
the reform on firms with weaker governance, and the results strengthen our interpretation that the effect of
the reform was through a managerial ownership channel. Our first set of results use the G-index measure
of corporate governance (Gompers, Ishii, and Metrick, 2003), which consists of 24 anti-takeover and share-
holder rights provisions. Firms with strong anti-takeover protections appear to underperform as they protect
entrenched managers against correcting forces from the capital market.26 We focus on a subsample of firms
with an available G-index, and this reduces our sample observations substantially to 4,381 firm-years.
We estimate our main regression separately for firms with their 2002 G-index above and below the
sample median. The results are reported in columns (1)-(3) of Table 8. Column (1) is for the low G-
index (good corporate governance) sub-sample and column (2) is for the high G-index (weak corporate
governance) sub-sample. Clearly for those firms with relatively good corporate governance, there is no
statistically significant treatment effect from the 2003 Tax Cut. In column (1) with the full set of controls,
the magnitude of treatment effects for the linear term is -0.12 (s.e.=0.86), and for the quadratic term is 0.60
(s.e.=1.41). In this sample of firms with strong governance, both estimates are close to zero, and neither is
significantly different from zero at conventional levels. In comparison, for firms with a high G-index score,
the coefficients for linear and quadratic terms are 1.86 (s.e.=0.80) and -2.10 (s.e.=1.62), with the linear term
significant at the 5% level. We also examine if the treatment effects for the two sub-samples are indeed
statistically different in a regression pooling the sample together as in column (3) in Table 8. The p-value
for the null hypothesis that these coefficients are the same is below 0.01, strongly rejecting the null.
We obtain similar results if we use another common measure of corporate governance, the E-index
26
Bebchuk, Cohen, and Wang (2013) find the correlation of stock returns and these governance indices only occurred during
1991-1999, but did not persist from 2000-2008. They also provide evidence this is due to market participants’ learning. However,
they still find a negative association between these indices and Tobin’s Q or operating performance.
18

(Bebchuk, Cohen, and Ferrell, 2009), which consists of 6 of the 24 provisions listed in the G-index. The
results are presented in columns (4)-(6) of Table 8. Again we find only firms with weak corporate governance
have significantly improved their performance, and the treatment effect displays a hump-shaped pattern,
consistent with the main results. We also test if the effects are statistically the same across the two sub-
samples and obtain a p-value below 0.01.
The results above suggest that managerial ownership seems to substitute for internal corporate gover-
nance as proxied by the strength of anti-takeover protection and shareholder rights. Providing increased
managerial ownership incentives could potentially help align the interests of managers with sharehold-
ers, particularly when managers enjoy entrenchment from anti-takeover provisions that weaken shareholder
rights. At the same time, the results from examining the interaction of corporate governance and managerial
ownership greatly strengthen our argument for the specific channel of improvement in firm performance.
Poorly governed firms are those that are most likely to have agency problems and room for improvement.
Several earlier studies have found that better governed firms have a higher Tobin’s Q (Gompers, Ishii,
and Metrick, 2003; Bebchuk, Cohen, and Wang, 2013). These results also strengthen our findings. If
the increase in Tobin’s Q came from channels other than increased managerial ownership incentives, for
example, an increased interest in the market for corporate control for some firms, then those firms with a
low G-index or E-index should have been more positively affected. Results from our empirical analysis
clearly refute this story, as the observed effect is only present in weakly governed firms, through channels
such as managerial incentives.
4.2 Interactive Effects of Managerial Ownership and Institutional Ownership
The corporate governance literature demonstrates blockholders and institutional investors with a substantial
ownership stake could affect firm performance in important ways.27 We therefore examine how the increase
in managerial ownership incentives interact with external governance from concentrated institutional own-
ership. Compared with firm internal governance using the G-index or E-index, we are able to better utilize
our sample size as most firms have information on their institutional holdings. The results again provide
evidence that the channel through which the tax reform impacted firm performance was via managerial
incentives, and suggest that managers’ insider governance substitutes for investors’ external governance.
27
See works such as Shleifer and Vishny (1986), Hartzell and Starks (2003), Cremers and Nair (2005), Edmans (2009), and
Edmans and Manso (2011).
19

We present results similar to our main analysis regarding groups of firms with above-median and below-
median top-5 institutional investors’ share ownership in 2002, which is characterized as the total share of
a firm held by the largest institutional owners.28 We split the sample by whether it is above or below the
median of the share of top-5 institutional ownership. Results are shown in columns (7) to (9) in Table 8. For
firms with below-median institutional ownership, i.e., those firms with relatively weaker external governance
from institutional shareholders, we find treatment effects for the linear term of 2.26 (s.e.=0.44) and for the
quadratic term of -2.64 (s.e.=0.53), both significant at the 1% level in column (7). In column (8) we do not
find any significant effects for firms with above-median institutional ownership, i.e., firms subject to strong
institutional governance. We also pool these two groups of firms together in a joint regression to see if the
treatment effect for firms with below-median institutional ownership is indeed statistically different from the
other group of firms. Results in column (9) of Table 8 confirm this is indeed the case. The p-value for testing
if the coefficients for the two sub-samples are the same is 0.00. Consistent with our agency interpretation,
those firms with relatively weak external governance from below-median institutional ownership are those
that have left managers a lot of leeway in firms’ operation and have indeed improved their performance more
as a result of increased managerial ownership incentives.
One could be concerned that since the 2003 Tax Cut affected not only the incentive of share-holding
managers but also tax-paying institutional investors with substantial stakes, those institutional investors
could have pushed for an increase in firm performance as they themselves also have a higher incentive.
We want to point out a number of reasons why this is not likely to be a major concern. First, managerial
ownership is mechanically a substitute for institutional ownership. In our sample, there is a -0.3 negative
correlation between these two measures. If the first-order underlying channel is that institutional investors
push for better performance in firms they own, the effect should be particularly strong in firms with low
managerial ownership, which goes against our main results.29 Second, our results in this section are clearly
contrary to this story, as we find a statistically significant stronger effect in firm performance for those firms
with below-median institutional ownership, ceteris paribus. Third, we show in section 5 that our results
remain even if we only consider firms with non-affected institutional investors. In addition to ruling out
this important confound, our results suggest insider governance from managerial ownership substitutes for
28
The use of top-5 institutional ownership as a proxy for institutional governance or blockholder governance follows Hartzell
and Starks (2003) and Kim and Lu (2011).
29
We also try including the top-5 institutional ownership in 2002 interacted with the treatment dummy of P ost2003t as an extra
control and find our results are robust to this specification as well.
20

external governance from institutional investors.
4.3 Interactive Effects of Managerial Ownership and Firm Leverage
Theoretical work demonstrates that firm leverage could be used as a means for corporate governance. A
higher leverage ratio could help reduce agency problems induced by free cash-flow and discourage managers
from enjoying “the quiet life” (Jensen, 1986). A recent growing literature argues that creditors also play
an important role in corporate governance well before a firm nears bankruptcy, notably through the use
of debt covenants as a tripwire (Chava and Roberts, 2008; Nini, Smith, and Sufi, 2012). We therefore
examine leverage as an alternative governance mechanism as in Atanassov (2013). The results confirm that
managerial ownership substitutes for firm leverage, reducing agency problems.
We divide our sample into firms with above and below sample-median leverage ratios and examine how
the treatment effects differ across these two groups. As reported in columns (10)-(12) of Table 8, we find
both groups of firms have been significantly affected by the increase in managerial incentive. We see the
effect for lower leverage firms is 2.52 (s.e.=0.49) for the linear term and -2.52 (s.e.=0.66) for the quadratic
term. The effect for higher leverage firms is dampened with the linear effect being 1.51 (s.e.=0.34) and
the quadratic effect being -1.70 (s.e.=0.46). While both groups of firms have been affected significantly
in a hump-shaped manner, lower leverage firms on average achieve a 50% higher improvement in firm
performance than an otherwise similar higher leverage firm. When estimating the difference between these
two groups in a joint regression in column (12), we find the p-value that these coefficients are the same across
two groups is 0.04. The result that managerial ownership also substitutes for firm leverage in corporate
governance is consistent with other evidence, for example, Atanassov (2013) finds leverage alleviates the
negative effect of state anti-takeover laws on corporate innovation.
5 Placebo and Robustness Tests
In this section, we present further evidence that our identifying assumptions are valid, and that our results
are robust to various threats. We first present placebo tests using simulated reforms and Canadian firms to
further strengthen our identifying assumption. We then conduct a number of robustness checks beyond our
main results and our tests for pre-determined trends presented in our main results. The checks in this section
provide strong evidence in support of our identifying assumption and the robustness of our results.
21

5.1 Placebo Tests
As mentioned above, a key identifying assumption for our difference-in-difference empirical design is that
in the absence of a tax cut there would be no differential change in performance for firms with different
levels of managerial ownership. We provide more support for this assumption by repeating our previous
empirical exercises in cases without any changes in the tax rate. We do this in two ways. First, we simulate
placebo reforms in years in which there was no reform. Second, we repeat our analysis on Canadian firms,
which faced similar market conditions to American firms but were unaffected by the tax reform. In both
cases we find no significant effect, providing further evidence of the parallel trends assumption.
5.1.1 Placebo Tests Using Non-Event Year
We examine results from two placebo difference-in-difference empirical designs in 1999 and 2000 respec-
tively, using sample periods 3 years before and after the placebo tax cut year. To determine treatment
intensity, we use firms’ managerial ownership one year immediately before our placebo tax cut year. That
is, in the placebo test of 1999, we use the managerial ownership in year 1998 as our treatment intensity. The
empirical model follows the main specification given in section 2.3. We present the results of these exercises
in columns (1) to (4) of Table 9. The estimates for the treatment effect are not significantly different from 0
in either cases. They are also much smaller in magnitude compared with our 2003 Tax Cut results. In the
1999 placebo test, the linear and quadratic treatment coefficients are -0.17 (s.e.=0.29) and 0.19 (s.e.=0.37)
(Column 2) while we obtain 0.11 (s.e.=0.33) and 0.22 (s.e.=0.44) (Column 4) in the 2000 placebo test.
Overall, the findings of our placebo tests add credence to our identifying assumption that firm performance
could have evolved similarly with regard to managerial ownership without the 2003 Tax Cut.
Moreover, these results can rule out certain endogeneity concerns about changes in managerial owner-
ship and firms’ future performance. If managers possess private information about a firm’s future perfor-
mance, which is unobservable to econometricians, and change their ownership accordingly, then we could
detect a spurious relationship rather than a causal effect of higher managerial incentive to firm performance.
This concern does not hold given evidence from our placebo tests. First, the lack of within-firm variation in
the ownership ratio (Zhou, 2001) could imply that this concern is not critical, otherwise we would expect
to see managerial ownership changes more frequently and dramatically. This is not the case as is shown in
Table 2. Second, this story would predict that firms with generally higher managerial ownership should real-
22

ize a higher performance improvement in the following years, whether we examine those firms in treatment
years 1999, 2000 or 2003. This clearly does not reconcile with results from our placebo tests.
5.1.2 A Placebo Test Using Canadian Firms
Another potential concern to our identification strategy using the 2003 Tax Cut is that there might exist
other concurrent macro-economic shocks which affect firms with certain characteristics and these firm char-
acteristics are associated with the level of managerial ownership. We follow Chetty and Saez (2005) and use
Canadian firms in Compustat to do another placebo test. Canadian firms are likely to share similar confound-
ing macro shocks as US firms. This placebo test is valid since few executives of Canadian firms reside in the
US and pay US dividend taxes. Managerial ownership of these Canadian firms is unobserved in Compact
Disclosure, so we predict the Canadian firms’ managerial ownership based on known firm characteristics.30
Specifically, we regress actual managerial ownership on all controls in our main regression for US firms, and
take the regression coefficients as well as the value of control variables of Canadian firms to predict their
managerial ownership. In columns (5) and (6) of Table 9, we do not see any statistically or economically
significant treatment effects for the sample of Canadian firms. To add further credence to our empirical
results, we show in columns (7) and (8) of Table 9 that a subsample of US firms matched by characteristics
of these Canadian firms display significant treatment effects. The matching is done using nearest-neighbor
estimator and based on log sales, capital over sales, income over sales, investment over capital, R&D over
sales, firm age as well as industry. Overall, the placebo tests using Canadian firms demonstrate our results
are not due to confounding macro shocks that coincide with the tax cut.
5.2 Robustness Checks
In this section we conduct a battery of complementary robustness tests beyond our tests for pre-determined
trends and placebo tests in preceding sections. We rule out the following concerns: (i) potential multi-
plicative effects, (ii) dividend issuance, (iii) changes in the incentives of institutional investors, (iv) serial
correlation of error terms, (v) industry-specific trends, (vi) changes in cost of capital, and (vii) the 2004
Foreign Tax Holiday.

30
If there were confounding macro or other shocks that affect firms with different characteristics and these characteristics were
also correlated with managerial ownership in a way to generate our main results, then we should expect to find significant treatment
effects even using predicted managerial ownership, irrespective of whether the predicted managerial ownership is a good measure
of the true managerial ownership.
23

5.2.1 Logarithmic Tobin’s Q
A potential concern is that the dividend tax cut could have a multiplicative effect on firm valuation. To
address concerns regarding multiplicative effects, we repeat our main regression using logarithmic Tobin’s
Q. As shown in columns (1) to (2) in Table 10, our results are robust to this modification in specification.
5.2.2 Dividend Issuance
As shown in Chetty and Saez (2005), the 2003 Tax Cut induced a sudden increase in dividend payout, par-
ticularly in firms with high managerial ownership. Recently Hartzmark and Solomon (2013) find a dividend
month premium, i.e., abnormal return in the month of expected dividend issuance, as some investors chase
dividends and drive up stock return. In our case, one could conjecture that investors foresee the large payout
in firms with higher managerial ownership, prefer to buy and drive up the price of these stocks. Since the
change in Tobin’s Q is a measure closely related to stock returns, we could have identified a spurious effect
unrelated to the agency problems of interest. We deal with this concern by controlling for dividend issuance,
as dividend issuance is not our object of interest. This specification also deals with the concern that dividend
payouts could have directly affected firm value. In columns (3) and (4) of Table 10, we include two variables
to control for the effect of dividend issuance: the amount of dividend payout (measured as dividend yield)
as well as an indicator of whether the firm issues dividend. Even though these two control variables are
correlated with Tobin’s Q and both are significant at the 1% level, our results are robust to this considera-
tion. The results are still robust when we include higher order series of these two controls, as well as when
they are fully interacted with treatment dummies. We also note that in Section 3.6, we examine a host of
alternative firm performance measures like ROA, NPM, Value Added and TFP, and find consistent results as
for Tobin’s Q. These measures capture firms’ operating performance and are not directly related with firms’
stock market valuation, and should not be affected by firm dividend issuance. Therefore, we conclude that
dividend payout is not a major confounding driving force for the change in firm performance.
5.2.3 Firms with Non-affected Institutional Investors
Since the 2003 Tax Cut affected incentives of all equity-holders, another potential concern is that institu-
tional investors could play a larger role in corporate governance and affect firm performance, given their
identified effects in the literature. To show that this is not the underlying driver of our results, we exploit
24

the fact that some institutional investors have tax-favored accounts and were thus not affected by the reform.
For example, pension funds were not subject to taxation for their dividend income. If our results are largely
driven by an increased incentive for institutional investors that push for better firm performance, then we
would not expect to see any response from firms where such nonaffected entities could play an important
role. This is similar to the strategy used in Chetty and Saez (2005). We isolate a sub-sample of firms
whose largest institutional owner is not affected by the 2003 Tax Cut—they are pension funds, insurance
companies, nonprofit organizations, nonfinancial corporations and government agencies—in the Thomson
financial institutional ownership database. Specifically, nonaffected entities are those classified as insurance
companies (type 2) and "others" (type 5) whose names indicate whether they are a pension fund, nonprofit
organization, government agency, or nonfinancial corporation.31
Isolating these non-affected firms gives us a much smaller sample of 4,030 firm-year. We follow our
main specifications for this sub-sample in columns (5) and (6) of Table 10. We again obtain very similar
results for firms with non-affected institutional investors, showing a non-linear treatment effect, with coeffi-
cients on linear and quadratic terms being 1.72 (s.e.=0.72) and -2.19 (s.e.=0.91), both significant at 5% level.
This finding strongly demonstrates that our results are not driven by influence from institutional investors.
In section 4 we also examine the interaction of institutional governance and managerial ownership and find
firms with weak institutional governance have much larger improvement in firm performance. These results
provide further evidence that effects from institutional investors do not confound our results.
5.2.4 Potential Serial Correlation
In our main regressions, we cluster standard errors at the firm level, to allow for potential serial correlation
within-firm. As an alternative, we generate bootstrapped standard errors for 2000 iterations. As shown in
column (7) of Table 10, our results are robust to considerations about serial correlation. We also allow for
serial correlation at both firm and year level. Results in column (8) of Table 10 again confirm that this does
not affect our conclusion.

31
After 1998, the Thomson financial database misclassified new institutions which actually should belong to type 1-4 categories
as type 5 (other). Chetty and Saez (2005) hand-classify type 5 institutions throughout to address this data problem. We use their
reclassification to correct for the errors in 13F.
25

5.2.5 Industry Specific Trends
One may be concerned that the pattern we uncover could be driven by differential industry trends during
the 2002 to 2007 expansion periods. For example, we know less capital intensive firms tend to have higher
managerial ownership and these firms have indeed performed well for a long time for reasons well beyond
the agency issues. We deal with this concern in two approaches. First, we extend our observation of the
performance of firms under study until 2008 before the Great Recession, and plot the year-specific treatment
effects (the coefficients for linear and quadratic terms of treatment variable) in Figure A.1. We find that
treatment effects peak at around year 2004, and then gradually decrease to almost zero in year 2008. This
pattern is not consistent with the overall trends of persistent over-performance of certain industries. Second,
we further control for industry specific trends in our empirical specifications. In column 1 of Table A.3 we
include year by 3-digit SIC industry fixed effects in our baseline specification. We still uncover a hump-
shaped effect of managerial ownership on Tobin’s Q. Although the coefficients have smaller magnitudes,
they are still statistically indistinguishable from our baseline estimates. We further interact a dummy of
whether a firm is above industry median in terms of capital intensity with year fixed effects in column 2
or with year and 3-digit SIC industry fixed effects in column 3 and find our results remain. In the last
two columns, we further include each firm’s capital intensity ratio in 2002 interacted with the post 2003
treatment dummy to see if indeed firms with differential capital intensities are affected differentially in
treatment years. There is no evidence that this is the case, and controlling for this channel has no material
effects on our estimates.
5.3 Cost of Capital
A potential concern for our empirical design is that since the 2003 Dividend Tax Cut is first and foremost
a shock to dividend taxes, it directly changes cost of capital as personal tax on capital gains and dividends
affect the value of stocks to investors (Lin and Flannery, 2013). Firms with a lower cost of capital are
able to put newly raised capital into better use, and may raise their performance. However, this channel
is unlikely to drive our main results in our context. First, a recent study finds that the 2003 Tax Cut did
not induce any change in corporate investment in the cost of capital channel (Yagan, 2015). Second, even
if cost of capital was affected by the Tax Cut, it is unlikely that the effect of cost of capital is correlated
with pre-determined managerial ownership. In fact, in our earlier sub-sample analysis in 5.2.3, we examine
26

a sample of firms whose largest institutional investor is tax-exempt, thus not subject to the cost of capital
effect, and find results to be consistent with main results. In the appendix we provide another test to address
this concern in our empirical tests following (Lin and Flannery, 2013). In the first two columns of Appendix
Table A.4, we calculate the total holdings directly owned by individual investors (excluding managers and
officers), who are likely to be affected by the cost of capital channel. We then examine the sample of firms
with low individual shareholding, and find indeed results are consistent with our main results. Overall, both
the conceptual analysis and empirical tests confirm that changes in the cost of capital do not pose a threat to
our identification strategy.
5.4 Foreign Tax Holiday in 2004
Our empirical design employs the 2003 Dividend Tax Cut, which was followed by the American Jobs
Creation Act (AJCA) of 2004, signed into law by President George W. Bush on October 22, 2004. The AJCA
creates a one-time tax holiday for multinational corporations to repatriate undistributed foreign earnings at
an unusually low tax cost, resulting in a significant tax rate reduction from a maximum of 35% to 5.25%. One
might be concerned that our main results could be driven by the Foreign Tax Holiday due to the immediacy
of the two acts. However, Dharmapala, Foley, and Forbes (2011) study this event carefully and find that
firms paid out one dollar for each dollar of repatriated earnings due to the tax holiday event, without any
other firm-level changes in firm performance. A priori, it is also unclear why the event should deliver the
nonlinear results on firm performance that we documented. We further rule out this concern in our empirical
test by studying a sub-sample of non-multinational corporations,32 who are unlikely to be affected by the
Tax Holiday. As shown in the last two columns of Appendix Table A.4, the results are still consistent to our
main findings, suggesting the 2004 Foreign Tax Holiday is not a main concern for our analysis.
6 Conclusion
This paper uses the 2003 Tax Cut to assess how managerial ownership affects firm performance. The
2003 Tax Cut created an abrupt change in the managerial net-of-tax effective ownership, providing quasi-
experimental variation to study the effect of managerial ownership on firm performance. We use this frame-
work to shed light on the controversy regarding whether managerial ownership affects firm performance,
32
Their foreign income taxes are lower than or equal to $1 million USD in our sample years, following Harford, Wang, and
Zhang (2017).
27

and our empirical framework alleviates two long-standing concerns: the concern over endogeneity and the
concern over lack of within-firm variation. Using the tax cut, we find strong evidence that managerial
ownership affects firm performance, and that the nature of this effect is consistent with theory.
We present three main findings in this paper. First, we find robust evidence that managerial ownership
affects firms performance. We see this effect in a number of outcomes, including Tobin’s Q, and our re-
sults are robust to examination of a pre-determined trend or placebo tests as well as other checks. Second,
consistent with standard agency theory, we show that an increase in the effective managerial ownership
significantly leads to an increase in firm performance in a non-linear fashion. Firms with an intermediate
level of managerial ownership show large improvements in firm performance, while the effect is smaller for
firms with very low or very high managerial ownership. Third, the increase in performance is greatest for
firms with more severe agency problems as well as firms with weaker governance. Interacting managerial
ownership with alternative channels of governance mechanisms, we find that only firms with weak internal
corporate governance, as proxied by the G-Index or E-Index, or weak institutional governance, are signifi-
cantly affected in a hump-shaped manner by the tax reform and subsequent change in effective managerial
ownership. Furthermore, firms with low leverage increase their performance significantly more than firms
with high leverage.
This paper demonstrates that firm performance is affected by managerial ownership and manager’s in-
centives, and that this effect is quantitatively important. Moreover we provide clear evidence consistent
with an agency model of the firm. However, many important policy questions remain unanswered, and
future work should focus on how improved governance structures and compensation policies can improve
firm performance by better aligning incentives between managers and shareholders.
28

References
Aboody, D. and B. Lev (2000). Information asymmetry, R&D, and insider gains. The Journal of Fi-
nance 55(6), 2747–2766.
Adams, R. B. and J. A. Santos (2006). Identifying the effect of managerial control on firm performance.
Journal of Accounting and Economics 41(1), 55–85.
Admati, A. R. and P. Pfleiderer (2009). The “wall street walk” and shareholder activism: Exit as a form of
voice. Review of Financial Studies 22(7), 2645–2685.
Almeida, H. and M. Campello (2007). Financial constraints, asset tangibility, and corporate investment.
Review of Financial Studies 20(5), 1429–1460.
Anderson, R. C. and D. M. Reeb (2003). Founding-family ownership and firm performance: Evidence from
the S&P 500. The Journal of Finance 58(3), 1301–1327.
Atanassov, J. (2013). Do hostile takeovers stifle innovation? Evidence from antitakeover legislation and
corporate patenting. The Journal of Finance 68(3), 1097–1131.
Auerbach, A. J. and K. A. Hassett (2006). Dividend taxes and firm valuation: New evidence. The American
Economic Review 96(2), 119–123.
Bebchuk, L., A. Cohen, and A. Ferrell (2009). What matters in corporate governance? Review of Financial
Studies 22(2), 783–827.
Bebchuk, L. A., A. Cohen, and C. C. Wang (2013). Learning and the disappearing association between
governance and returns. Journal of Financial Economics 108(2), 323–348.
Begley, J. and G. A. Feltham (1999). An empirical examination of the relation between debt contracts and
management incentives. Journal of Accounting and Economics 27(2), 229–259.
Bharath, S. T., S. Jayaraman, and V. Nagar (2013). Exit as governance: An empirical analysis. The Journal
of Finance 68(6), 2515–2547.
Bizjak, J., M. Lemmon, and R. Whitby (2009). Option backdating and board interlocks. Review of Financial
Studies 22(11), 4821–4847.
Bizjak, J. M., J. A. Brickley, and J. L. Coles (1993). Stock-based incentive compensation and investment
behavior. Journal of Accounting and Economics 16(1-3), 349–372.
Bova, F., K. Kolev, J. K. Thomas, and X. F. Zhang (2014). Non-executive employee ownership and corporate
risk. The Accounting Review 90(1), 115–145.
Budish, E., B. N. Roin, and H. Williams (2015). Do firms underinvest in longterm research? Evidence from
cancer clinical trials. American Economic Review 105, 2044–2085.
Bushee, B. J. (1998). The influence of institutional investors on myopic r&d investment behavior. The
Accounting Review, 305–333.
Calomiris, C. W. and M. Carlson (2014). Corporate governance and risk management at unprotected banks:
National banks in the 1890s. Journal of Financial Economics 119(3), 512–532.
29

Chava, S. and M. R. Roberts (2008). How does financing impact investment? The role of debt covenants.
The Journal of Finance 63(5), 2085–2121.
Cheng, I.-H., H. Hong, and K. Shue (2013). Do managers do good with other peoples’ money? Chicago
Booth Research Paper (12-47).
Cheng, Q. and T. D. Warfield (2005). Equity incentives and earnings management. The Accounting Re-
view 80(2), 441–476.
Chetty, R. and E. Saez (2005). Dividend taxes and corporate behavior: Evidence from the 2003 dividend
tax cut. The Quarterly Journal of Economics 120(3), 791–833.
Chetty, R. and E. Saez (2010). Dividend and corporate taxation in an agency model of the firm. American
Economic Journal: Economic Policy 2(3), 1–31.
Cho, M.-H. (1998). Ownership structure, investment, and the corporate value: An empirical analysis.
Journal of Financial Economics 47(1), 103–121.
Cicero, D. C. (2009). The manipulation of executive stock option exercise strategies: Information timing
and backdating. The Journal of Finance 64(6), 2627–2663.
Coles, J. L., M. L. Lemmon, and F. J. Meschke (2012). Structural models and endogeneity in corporate
finance: The link between managerial ownership and corporate performance. Journal of Financial Eco-
nomics 103(1), 149–168.
Core, J. and W. Guay (2002). Estimating the value of employee stock option portfolios and their sensitivities
to price and volatility. Journal of Accounting Research 40(3), 613–630.
Core, J. E. and D. F. Larcker (2002). Performance consequences of mandatory increases in executive stock
ownership. Journal of Financial Economics 64(3), 317–340.
Cremers, K. and V. B. Nair (2005). Governance mechanisms and equity prices. The Journal of Fi-
nance 60(6), 2859–2894.
Demsetz, H. and K. Lehn (1985). The structure of corporate ownership: Causes and consequences. The
Journal of Political Economy 93(6), 1155–1177.
Denis, D. J. and A. Sarin (1999). Ownership and board structures in publicly traded corporations. Journal
of Financial Economics 52(2), 187–223.
Dharmapala, D., C. F. Foley, and K. J. Forbes (2011). Watch what I do, not what I say: The unintended
consequences of the homeland investment act. The Journal of Finance 66(3), 753–787.
Edmans, A. (2009). Blockholder trading, market efficiency, and managerial myopia. The Journal of Fi-
nance 64(6), 2481–2513.
Edmans, A. and G. Manso (2011). Governance through trading and intervention: A theory of multiple
blockholders. Review of Financial Studies 24(7), 2395–2428.
Fahlenbrach, R. and R. M. Stulz (2009). Managerial ownership dynamics and firm value. Journal of
Financial Economics 92(3), 342–361.
Gompers, P., J. Ishii, and A. Metrick (2003). Corporate governance and equity prices. The Quarterly Journal
of Economics 118(1), 107–156.
30

Harford, J., C. Wang, and K. Zhang (2017). Foreign cash: Taxes, internal capital markets, and agency
problems. The Review of Financial Studies 30(5), 1490–1538.
Hartzell, J. C. and L. T. Starks (2003). Institutional investors and executive compensation. The Journal of
Finance 58(6), 2351–2374.
Hartzmark, S. M. and D. H. Solomon (2013). The dividend month premium. Journal of Financial Eco-
nomics 109(3), 640–660.
Heron, R. A. and E. Lie (2009). What fraction of stock option grants to top executives have been backdated
or manipulated? Management Science 55(4), 513–525.
Himmelberg, C. P., R. G. Hubbard, and D. Palia (1999). Understanding the determinants of managerial
ownership and the link between ownership and performance. Journal of Financial Economics 53(3),
353–384.
Holderness, C. G., R. S. Kroszner, and D. P. Sheehan (1999). Were the good old days that good? Changes
in managerial stock ownership since the great depression. The Journal of Finance 54(2), 435–469.
Imrohoroglu, A. and S. Tüzel (2014). Firm-level productivity, risk, and return. Management Science 60(8),
2073–2090.
Jensen, M. C. (1986). Agency costs of free cash flow, corporate finance, and takeovers. American Economic
Review 76(2), 323–329.
Jensen, M. C. and W. H. Meckling (1976). Theory of the firm: Managerial behavior, agency costs and
ownership structure. Journal of Financial Economics 3(4), 305–360.
Jensen, M. C. and K. J. Murphy (1990). Performance pay and top-management incentives. The Journal of
Political Economy 98(2), 225–264.
Kim, E. H. and Y. Lu (2011). Ceo ownership, external governance, and risk-taking. Journal of Financial
Economics 102(2), 272–292.
Kim, J.-B., Y. Li, and L. Zhang (2011). Cfos versus ceos: Equity incentives and crashes. Journal of
Financial Economics 101(3), 713–730.
Korinek, A. and J. E. Stiglitz (2009). Dividend taxation and intertemporal tax arbitrage. Journal of Public
Economics 93(1), 142–159.
Lafond, R. and S. Roychowdhury (2008). Managerial ownership and accounting conservatism. Journal of
Accounting Research 46(1), 101–135.
Lie, E. (2005). On the timing of CEO stock option awards. Management Science 51(5), 802–812.
Lin, L. and M. J. Flannery (2013). Do personal taxes affect capital structure? Evidence from the 2003 tax
cut. Journal of Financial Economics 109(2), 549–565.
Ma, S. (2017). Agency, incentives, and motivating innovation. Unpublished Mimeo.
Masulis, R. W. and S. W. Reza (2015). Agency problems of corporate philanthropy. Review of Financial
Studies 28(2), 592–636.
McConnell, J. J., H. Servaes, and K. V. Lins (2008). Changes in insider ownership and changes in the market
value of the firm. Journal of Corporate Finance 14(2), 92–106.
31

Moeller, S. B., F. P. Schlingemann, and R. M. Stulz (2007). How do diversity of opinion and information
asymmetry affect acquirer returns? Review of Financial Studies 20(6), 2047–2078.
Morck, R., A. Shleifer, and R. W. Vishny (1988). Management ownership and market valuation: An empir-
ical analysis. Journal of Financial Economics 20, 293–315.
Myers, S. C. (1977). Determinants of corporate borrowing. Journal of Financial Economics (5), 147–75.
Nini, G., D. C. Smith, and A. Sufi (2012). Creditor control rights, corporate governance, and firm value.
Review of Financial Studies 25(6), 1713–1761.
Pérez-González, F. (2006). Inherited control and firm performance. American Economic Review 96(5),
1559–1588.
Rajgopal, S. and T. Shevlin (2002). Empirical evidence on the relation between stock option compensation
and risk taking. Journal of Accounting and Economics 33(2), 145–171.
Shleifer, A. and R. W. Vishny (1986). Large shareholders and corporate control. The Journal of Political
Economy, 461–488.
Stulz, R. (1988). Managerial control of voting rights: Financing policies and the market for corporate
control. Journal of Financial Economics 20, 25–54.
Sufi, A. (2007). Information asymmetry and financing arrangements: Evidence from syndicated loans. The
Journal of Finance 62(2), 629–668.
Yagan, D. (2015). Capital tax reform and the real economy. American Economic Review 63(1), 3531–63.
Zhou, X. (2001). Understanding the determinants of managerial ownership and the link between ownership
and performance: comment. Journal of Financial Economics 62(3), 559–571.
32

Figure 1: The Effect of the Tax Reform
This figure shows a stylized example of two firms affected by the tax reform. In both firms, managers pay the highest statutory
dividend tax rate. The two firms have different nominal managerial ownership that is unchanged before and after the Tax Cut. The
black long dashed line depicts the effective ownership for Firm 1 with a nominal ownership of 8%, while the gray shorter dashed
line depicts the effective ownership for Firm 2 with a nominal ownership of 2%. The thin vertical dashed line depicts the tax reform.
Effective Ownership
EffectiveFirm1
Difference
Post-Reform
Difference
Pre-Reform
EffectiveFirm2
1998 2000 2002 2004 2006 2008
33

Figure 2: Non-parametric Evidence of Nonlinear Changes
Kernel regression estimates of the change in firm performance from 2002 to 2004 plotted against the pre-determined managerial
ownership in 2002. In Panel A, the vertical-axis is the change in Tobin’s Q from 2002 to 2004. In Panel B, the vertical-axis is the
change in conditional Tobin’s Q from 2002 to 2004, where conditional Tobin’s Q is defined as the residual of regressing Tobin’s
Q on the full set of controls used In the main regression. The grey area shows error bounds, which are 95% confidence intervals
(+/-1.96 standard deviations).
Panel A: Difference in Tobin’s Q between 2002 and 2004

.6
.4
.2
0
-.2
0 .2 .4 .6 .8 1
Share02
Panel B: Conditional Difference in Tobin’s Q between 2002 and 2004

.2
0
-.2
-.4
-.6
0 .2 .4 .6 .8 1
Share02
34

Figure 3: Pre-trend Tests
This figure plots the coefficients in front of the year-specific linear and quadratic terms of the treatment variable
from 1998 to 2006, where 2002 is omitted as the base year. The dependent variable is Tobin’s Q, defined as
the ratio of the market value of assets to the book value of assets, where the market value of assets is calculated
as the market value of common equity plus book value of total liability plus book value of preferred stock.
For each year in our sample, we create year-specific treatment effects (for example, Share2002*Y ear2000
and Share20022 *Y ear2000) by interacting the year dummy with linear and quadratic terms of the treatment
intensity. Robust standard errors clustered at the firm-level. The error bounds are 95% confidence intervals
3 2
Linear Term
1 0
-1
-4 -3 -2 -1 +1 +2 +3 +4
Year
0
Quadratic Term
-3 -2-4 -1
-4 -3 -2 -1 +1 +2 +3 +4
Year
35

Table 1: Variable Definition
The table presents the main analysis variables. The first column presents the variable name used in the paper,
while the second column presents a description of the variable.
Performance and ownership variables
Tobin’s Q The market value of common equity plus the book value of total liability
plus the book value of preferred stock, divided by the book value of total
assets
Share The sum of the fractions of shares held by all officers and directors, from
Compact Disclosure
Share2002 The sum of the fractions of shares held by all officers and directors in 2002,
from Compact Disclosure
Event variable
Post2003 A dummy variable equal to one if the observation is in or after year 2003
Firm size controls
Log(sales) The natural logarithm of sales
Log(assets) The natural logarithm of assets
Moral hazard controls
Capital-to-sales The ratio of property, plant and equipment to sales
Missing Capital-to-sales A dummy variable equal to one if the data for the variable Capital-to-sales
is missing
OI-to-sales The ratio of operating income before depreciation to sales
Missing OI-to-sales A dummy variable equal to one if the data for the variable OI-to-sales is
missing
R&D-to-capital The ratio of R&D expenditures to property, plant and equipment
Missing R&D-to-capital A dummy variable equal to one if the data for the variable R&D-to-capital
is missing
Ad-to-capital The ratio of ads expenditures to property, plant and equipment
Missing Ad-to-capital A dummy variable equal to one if the data for the variable Ad-to-capital is
missing
Capex-to-capital The ratio of capital expenditure to property, plant and equipment
Missing Capex-to-capital A dummy variable equal to one if the data for the variable Capex-to-capital
is missing
Firm risk controls
Sigma The idiosyncratic stock return risk, calculated as the standard error of the
residuals from a CAPM model estimated using daily stock return data for
the period covered by the annual sample, from CRSP
Missing Sigma A dummy variable equal to one if the data for the variable Sigma is missing
Miscellaneous controls
Leverage Long-term debt divided by the book value of total assets
Missing Leverage A dummy variable equal to one if the data for the variable Leverage is miss-
ing
Cash-to-assets The ratio of cash and short-term investment to assets
Age The listing age of a firm as measured by the number of years since its IPO
as reported in CRSP
Missing Age A dummy variable equal to one if the data for the variable Age is missing
Alternative measures of performance and outcomes
36

OROA Operating return on asset defined as the ratio of operating income over book
value of assets
ROA Return on asset defined as the ratio of net income over book value of assets
NPM Net profit margin defined as the ratio of net income over sales
ROE Return on equity defined as the ratio of net income over book value of com-
mon stock
Log(value added) The natural logarithm of value added, where value added is defined as the
gross output net of expenditures on materials as well as the other expensed
items such as advertisement, R&D expenditures, and rental expenses, fol-
lowing Imrohoroglu and Tüzel (2014).
TFP Firm-level total factor productivity, estimated using semi-parametric meth-
ods used in Imrohoroglu and Tüzel (2014).
Log(R&D) The natural logarithm of R&D
Log(leverage) The natural logarithm of Leverage
Corporate governance
G-Index2002 G-Index in 2002 from Gompers, Ishii, and Metrick (2003)
E-Index2002 E-Index in 2002 from Bebchuk, Cohen, and Ferrell (2009)
Inst. holding2002 Institutional ownership concentration proxied by the total share of the top
five institution holdings in 2002
Other variables
Dividend Dividend yield calculated as dividend payout over stock price
Positive dividend A dummy variable equal to one if dividend payout is positive
37

Table 2: Summary Statistics for All Variables Used in the Paper
S.D. is the sample standard deviation of the corresponding variable while p10(/50/90) is the 10(/50/90)th percentile respectively of
the variable of interest. The number of observations for each variable varies based on data availability. Definitions of all variables
are in Table 1.
Obs. Mean S.D. p10 p50 p90

Valuation and ownership variables
Tobin’s Q 15,846 2.027 1.514 0.845 1.507 3.957
Share 13,346 0.238 0.219 0.024 0.163 0.574
Share2002 15,846 0.238 0.221 0.022 0.161 0.576
Firm size controls
Log(sales) 15,846 4.928 2.462 1.862 4.981 8.012
Log(assets) 15,846 5.129 2.246 2.210 5.115 8.015
Moral hazard controls
Capital-to-sales 15,829 1.366 44.996 0.034 0.183 1.140
Missing Capital-to-sales 15,846 0.001 0.033 0.000 0.000 0.000
OI-to-sales 15,802 -2.680 68.596 -0.613 0.083 0.269
Missing OI-to-sales 15,846 0.003 0.053 0.000 0.000 0.000
R&D-to-capital 10,213 2.617 39.442 0.000 0.315 4.109
Missing R&D-to-capital 15,846 0.355 0.479 0.000 0.000 1.000
Ad-to-capital 5,609 0.489 2.935 0.010 0.098 0.838
Missing Ad-to-capital 15,846 0.646 0.478 0.000 1.000 1.000
Capex-to-capital 15,600 0.330 4.702 0.060 0.212 0.596
Missing Capex-to-capital 15,846 0.016 0.124 0.000 0.000 0.000
Firm risk controls
Sigma 13,825 0.041 0.027 0.016 0.034 0.076
Missing Sigma 15,846 0.128 0.334 0.000 0.000 1.000
Miscellaneous controls
Leverage 15,841 0.172 0.251 0.000 0.085 0.444
Missing Leverage 15,846 0.000 0.018 0.000 0.000 0.000
Cash-to-assets 15,846 0.209 0.234 0.009 0.108 0.586
Age 9,293 7.480 5.282 2.000 7.000 14.000
Missing Age 15,846 0.414 0.492 0.000 0.000 1.000
Alternative measures of performance and outcomes
OROA 15802 0.054 0.148 -0.259 0.097 0.224
ROA 15846 0.004 0.081 -0.107 0.021 0.121
NPM 15846 0.001 0.089 -0.122 0.017 0.137
ROE 15845 0.053 0.150 -0.139 0.063 0.313
Log(value added) 15455 10.201 2.281 7.392 10.122 13.235
TFP 11970 -0.409 0.650 -1.016 -0.352 0.198
Log(R&D) 8360 -0.734 1.978 -3.321 -0.511 1.584
Log(leverage) 12916 -1.981 1.682 -4.327 -1.453 -0.538
Corporate governance
G-Index2002 4,381 9.193 2.815 6.000 10.000 13.000
E-Index2002 4,904 2.472 1.263 1.000 2.000 4.000
Inst. holding2002 13,365 0.206 0.139 0.010 0.211 0.381
Other variables
Dividend 4,145 0.055 0.312 0.000 0.014 0.045
Positive dividend 15,846 0.262 0.440 0.000 0.000 1.000
38

Table 3: Fixed Effects Specifications
The table replicates the specifications of OLS and FE in Himmelberg, Hubbard, and Palia (1999). Definitions of all variables are
provided in Table 1. Share ownership, Share, is the sum of the fractions of shares held by all directors and officers. The data are
from Compact Disclosure and range from 2000 to 2005. Regressions in all columns include year-fixed effects and in columns (3)
and (4) include firm-fixed effects. See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are
reported in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively.
Dependent Variable: Tobin’s Q

OLS FE
(1) (2) (3) (4)
Share -0.657** -1.333*** 0.112 -0.009
(0.273) (0.272) (0.366) (0.363)
(Share)2 0.235 1.048*** 0.189 0.263
(0.349) (0.337) (0.429) (0.425)
Firm size controls No Yes No Yes
Moral hazard controls No Yes No Yes
Firm risk controls No Yes No Yes
Miscellaneous controls No Yes No Yes
Year fixed effects Yes Yes Yes Yes
Firm fixed effects No No Yes Yes
Observations 13,346 13,346 13,346 13,346
R2 0.005 0.163 0.001 0.022
39

Table 4: Main Results
This table reports results from our difference-in-difference (DID) estimations for the effect of managerial ownership on firm per-
formance. The dependent variable is Tobin’s Q, defined as the ratio of the market value of assets to the book value of assets, where
the market value of assets is calculated as the market value of common equity plus book value of total liability plus book value of
preferred stock. The key independent variables Share2002*Post2003 and (Share2002)2 *Post2003 identify our treatment effects,
which are the post dummy interacted with linear and quadratic terms of treatment intensity. Post dummy, Post2003, takes the value
of one for the year 2003 and onward and zero otherwise. Treatment intensity, Share2002, is the sum of the fractions of shares held
by all directors and officers in 2002 from Compact Disclosure, while (Share2002)2 is the squared of Share2002. Definitions of
all other control variables are provided in Table 1. Robust standard errors clustered at firm-level are reported in parentheses. Our
sample ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients marked
with *, **, and *** are significant at 10%, 5%, and 1%, respectively.
(1) (2) (3) (4)
Share2002 * Post2003 1.943*** 1.938*** 1.921*** 1.923***
(0.295) (0.296) (0.296) (0.295)
(Share2002)2 * Post2003 -1.973*** -1.982*** -1.963*** -1.945***
(0.401) (0.402) (0.401) (0.397)
Log(sales) -0.127*** -0.118** -0.063
(0.047) (0.054) (0.055)
(Log(sales))2 -0.001 -0.003 -0.000
(0.005) (0.006) (0.006)
Capital-to-sales -0.001 -0.001
(0.001) (0.001)
(Capital-to-sales)2 0.000 0.000
(0.000) (0.000)
Missing Capital-to-sales -0.139 -0.174
(0.260) (0.256)
OI-to-sales -0.000 -0.000
(0.000) (0.000)
Missing OI-to-sales 0.005 0.050
(0.273) (0.289)
R&D-to-capital -0.001** -0.001**
(0.000) (0.000)
Missing R&D-to-capital 0.064 0.075
(0.092) (0.091)
Ad-to-capital -0.002 -0.004
(0.008) (0.008)
Missing Ad-to-capital -0.001 0.005
(0.061) (0.061)
Capex-to-capital -0.001 -0.001*
(0.001) (0.001)
Missing Capex-to-capital -0.081 -0.058
(0.134) (0.135)
Sigma -1.959***
(0.594)
Missing Sigma 0.271***
(0.074)
Leverage 0.201*
40

(0.118)
Missing Leverage -0.168
(0.438)
Cash-to-assets 0.959***
(0.162)
Age -0.018
(0.014)
Missing Age 1.473***
(0.041)
Firm fixed effects Yes Yes Yes Yes
Observations 15,846 15,846 15,846 15,846
R2 0.008 0.012 0.014 0.029
41

Table 5: Pre-trend Tests
This table examines pre-trend and heterogeneous treatment effects across years. The dependent variable is Tobin’s Q, defined as
the ratio of the market value of assets to the book value of assets, where the market value of assets is calculated as the market value
of common equity plus book value of total liability plus book value of preferred stock. For each year in our sample, we create
year-specific treatment effects (for example, Share02*2000 and ShareSq02*2000) by interacting the year dummy with linear and
quadratic terms of treatment intensity. Treatment intensity, Share02, is the sum of the fractions of shares held by all directors and
officers in 2002 from Compact Disclosure while ShareSq02, is the squared of Share02. We omit year of 2002 as our base year. See
Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are reported in parentheses. Our sample
ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients marked with *,
**, and *** are significant at 10%, 5%, and 1%, respectively.

(1) (2) (3) (4)
Share2002 * Year2000 -0.220 -0.210 -0.226 -0.300
(0.381) (0.380) (0.379) (0.377)
(Share2002)2 * Year2000 -0.502 -0.510 -0.492 -0.441
(0.488) (0.486) (0.485) (0.481)
Share2002 * Year2001 -0.113 -0.112 -0.137 -0.159
(0.280) (0.279) (0.279) (0.278)
(Share2002)2 * Year2001 -0.352 -0.352 -0.326 -0.324
(0.364) (0.363) (0.363) (0.361)
Share2002 * Year2003 1.562*** 1.548*** 1.526*** 1.525***
(0.311) (0.311) (0.311) (0.309)
(Share2002)2 * Year2003 -1.940*** -1.932*** -1.908*** -1.912***
(0.432) (0.433) (0.432) (0.428)
Share2002 * Year2004 2.140*** 2.144*** 2.115*** 2.089***
(0.338) (0.339) (0.339) (0.337)
(Share2002)2 * Year2004 -2.596*** -2.617*** -2.584*** -2.560***
(0.470) (0.471) (0.471) (0.467)
Share2002 * Year2005 1.869*** 1.889*** 1.847*** 1.786***
(0.374) (0.376) (0.376) (0.371)
(Share2002)2 * Year2005 -2.298*** -2.336*** -2.290*** -2.196***
(0.506) (0.508) (0.508) (0.502)
Firm size controls No Yes Yes Yes
Moral hazard controls No No Yes Yes
Firm risk controls No No No Yes
Miscellaneous controls No No No Yes
Observations 15,846 15,846 15,846 15,846
R2 0.010 0.014 0.016 0.032
42

Table 6: Heterogeneous Effects on Agency Problem
This table examines the heterogeneous effects of managerial ownership on firm performance for firms with different levels of agency problems. We use three proxies of the
agency problem between managers and shareholders: columns (1)-(3) uses R&D spending ratio following Aboody and Lev (2000), columns (4)-(6) uses the cash holding ratio
following Chava and Roberts (2008) and columns (7)-(9) uses idiosyncratic stock volatility following Moeller, Schlingemann, and Stulz (2007). We divide firms into Low
(High) groups based on whether their 2002 proxy value is below (above) the sample median of the proxy in 2002. The dependent variable is Tobin’s Q, defined as the ratio of
the market value of assets to the book value of assets, where the market value of assets is calculated as the market value of common equity plus book value of total liability
plus book value of preferred stock. The key independent variables Share02*Post2003 and ShareSq02*Post2003 identify our treatment effects, which are the treatment dummy,
Post2003, interacted with linear and quadratic terms of treatment intensity. Treatment intensity, Share02, is the sum of the fractions of shares held by all directors and officers
in 2002 from Compact Disclosure while ShareSq02, is the squared of Share02. See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are
reported in parentheses. In columns (1), (4) and (7), we restrict our sample to firms with below-sample-median agency proxies. In columns (2), (5) and (8) we restrict our sample
to firms with above-sample-median agency proxies. In columns (3), (6) and (9), we pool all samples and allow treatment to vary with subsamples with below(above) median
agency proxies. The number of observations varies with the availability of different agency proxies. We report the F statistics and p-value for a joint test of whether the linear
and quadratic terms are different for the below- and above-median subsamples (Share02*Post*Above Med and ShareSq02*Post*Above Med). Our sample ranges from 2000 to
2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively.

R&D / Capital Cash / Asset Idiosyncratic Volatility
43
LOW HIGH ALL LOW HIGH ALL LOW HIGH ALL

(1) (2) (3) (4) (5) (6) (7) (8) (9)
Share2002 * Post2003 1.283*** 3.064*** 1.621*** 1.273*** 2.696*** 1.555*** 0.801*** 1.417** 1.231***
(0.421) (0.734) (0.437) (0.270) (0.560) (0.278) (0.302) (0.638) (0.319)
(Share2002)2 * Post2003 -1.067** -2.850*** -1.362** -1.448*** -2.579*** -1.746*** -0.619 -1.491** -0.916**
(0.543) (0.974) (0.579) (0.382) (0.736) (0.385) (0.416) (0.756) (0.443)
Share2002*Post2003*High 1.211** 0.626* 1.453***
(0.525) (0.343) (0.339)
(Share2002)2 *Post2003*High -1.232 -0.213 -1.803***
(0.850) (0.543) (0.522)
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 5,107 5,106 10,213 7,923 7,923 15,846 6,628 6,628 13,256
R2 0.058 0.027 0.034 0.039 0.026 0.031 0.057 0.020 0.033
F stat 4.274 9.129 10.898
p-value 0.014 0.000 0.000

Table 7: Alternative Measures of Performance and Outcomes
This table reports results from our difference-in-difference (DID) estimation of the effect of managerial ownership on firm performance, using alternative measures of firm perfor-
mance. The dependent variables in columns (1) - (8) are OROA, ROA, NPM, ROE, Log(value added), TFP, Log(R&D), Log(leverage). Their definitions are listed in Table 1. The
key independent variables Share02*Post2003 and ShareSq02*Post2003 identify our treatment effects, which are the treatment dummy, Post2003, interacted with linear and quadratic
terms of treatment intensity. Treatment intensity, Share02, is the sum of the fractions of shares held by all directors and officers in 2002 from Compact Disclosure while ShareSq02,
is the squared of Share02. We control for log(Sale), log(Sale)2 , age, dummy for non-missing age in columns (1)-(2), log(Asset), log(Asset)2 , age, dummy for non-missing age in
columns (3)-(4), and a full set of controls in Table 4 in columns (5)-(8). Our sample ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed
effects. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively.
Dependent variable
OROA ROA NPM ROE VA TFP R&D Leverage
44
(1) (2) (3) (4) (5) (6) (7) (8)

Share2002 * Post2003 0.111*** 0.125*** 0.268* 0.309** 0.281*** 0.161** 0.433* -0.559**
(0.031) (0.045) (0.145) (0.124) (0.093) (0.074) (0.241) (0.252)
(Share2002)2 * Post2003 -0.115*** -0.123* -0.355* -0.302* -0.284** -0.116 -0.426 0.602*
(0.044) (0.065) (0.214) (0.175) (0.129) (0.096) (0.343) (0.332)
Controls Yes Yes Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 15,802 15,846 15,846 15,845 15,455 11,970 8,360 12,916
R2 0.124 0.069 0.025 0.014 0.398 0.451 0.077 0.220

Table 8: Interaction with Alternative Governance Channels
This table examines how the effect of managerial ownership on firm performance interacts with alternative channels of corporate governance. columns (1)-(3) uses the G-index following
Gompers, Ishii, and Metrick (2003); columns (4)-(6) uses E-Index following Bebchuk, Cohen, and Ferrell (2009); columns (7)-(9) uses strength of concentrated institutional ownership,
measured as the sum of the fractions of shares held by top-five institutions following Hartzell and Starks (2003); columns (10)-(12) uses leverage ratio. We divide firms into Low (High)
group based on whether their 2002 governance measure is below (above) the sample median in 2002. The dependent variable is Tobin’s Q, defined as the ratio of the market value of
assets to the book value of assets, where the market value of assets is calculated as the market value of common equity plus book value of total liability plus book value of preferred stock.
The key independent variables Share02*Post2003 and ShareSq02*Post2003 identify our treatment effects, which are the treatment dummy, Post2003, interacted with linear and quadratic
is the squared of Share02. See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are reported in parentheses. In columns (1), (4), (7) and (10),
we restrict our sample to firms with below-sample-median corporate governance measures. In columns (2), (5), (8) and (11), we restrict our sample to firms with above-sample-median
corporate governance measures. In columns (3), (6), (9) and (12), we pool all samples with non-missing corporate governance measures and allow treatment to vary with subsamples
with below/above median corporate governance measures. The number of variables varies with the availability of different corporate governance measures. We report the F statistics and
p-value for a joint test of whether the linear and quadratic terms are different for the below- and above-median subsamples (Share02*Post*Above Med and ShareSq02*Post*Above Med).
Our sample ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients marked with *, **, and *** are significant at 10%, 5%,
and 1%, respectively.

45
G-Index E-Index Institutional Holding Leverage

LOW HIGH ALL LOW HIGH ALL LOW HIGH ALL LOW HIGH ALL
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Share2002 * Post2003 -0.12 1.86** 0.05 0.70 2.34*** 0.29 2.26*** 0.75 2.75*** 2.52*** 1.51*** 2.52***
(0.86) (0.80) (0.70) (0.83) (0.84) (0.67) (0.44) (0.49) (0.35) (0.49) (0.34) (0.49)
(Share2002)2 * Post2003 0.60 -2.10 0.39 -0.50 -3.39** -0.05 -2.64*** -0.01 -3.13*** -2.52*** -1.70*** -2.52***
(1.41) (1.62) (1.28) (1.46) (1.58) (1.30) (0.53) (0.86) (0.48) (0.66) (0.46) (0.66)
Share2002*Post2003*High 2.07** 2.84*** -2.72*** -1.04*
(0.83) (0.79) (0.46) (0.60)
(Share2002)2 *Post2003*High -2.89 -4.60*** 4.04*** 0.86
(1.78) (1.71) (0.85) (0.81)
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 2,167 2,214 4,381 2,455 2,449 4,904 6,434 6,931 13,365 7,242 6,994 14,236
R2 0.04 0.09 0.05 0.05 0.09 0.06 0.04 0.05 0.04 0.03 0.05 0.04
Fstat 5.00 7.69 19.33 3.19
P-value 0.01 0.00 0.00 0.04

Table 9: Placebo Tests
This table reports two types of placebo tests. In the first set of placebos, we use 1999 and 2000 as two separate year of placebo ‘Tax Cut’. We replicate our difference-in-difference
empirical strategy, setting a placebo Tax Cut year as 1999, which is shown in columns (1) and (2) and to year 2000 shown in columns (3) and (4). In the second placebo test, we
use Canadian firms as the placebo group. Columns (5)-(6) restrict to Canadian firms in Compustat where the managerial ownership ratio is predicted using all controls in a linear
regression. Columns (7)-(8) restrict the sample to US firms that match Canadian firms based on log sales, capital over sales, income over sales, investment over capital, R&D over
sales, firm age as well as industry. The dependent variable is Tobin’s Q, defined as the ratio of the market value of assets to the book value of assets, where the market value of assets
is calculated as the market value of common equity plus book value of total liability plus book value of preferred stock. For each placebo test, we create treatment effects, ShareYR(-
1)*PostYR and ShareSqYR(-1)*PostYR by interacting the Post-treatment year dummy, PostYR with linear and quadratic terms of treatment intensity. Treatment intensity, ShareYR(-1),
is the sum of the fractions of shares held by all directors and officers one year before the placebo treatment year (1999, 2000 and 2003, respectively) from Compact Disclosure while
ShareSqYR(-1), is the square of ShareYR(-1). See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are reported in parentheses. Columns (1)
to (4) use samples three years before and after the respective placebo treatment year, and columns (5) to (8) use years between 2000 and 2005. Regressions in all columns include
year-fixed effects and firm-fixed effects. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively.

46
Canadian Firms as Placebo Group (YR=2003)

YR = 1999 YR = 2000
Canadian Firms Matched U.S. Firms
(1) (2) (3) (4) (5) (6) (7) (8)
ShareYR(-1) * PostYR -0.449 -0.166 -0.203 0.114 -0.346 0.554 2.337*** 2.257***
(0.288) (0.285) (0.360) (0.333) (0.844) (0.818) (0.632) (0.634)
(ShareYR(-1))2 * PostYR 0.589 0.188 0.595 0.218 -0.937 -2.234 -2.656*** -2.576***
(0.370) (0.369) (0.480) (0.443) (1.801) (1.720) (0.880) (0.875)
Controls No Yes No Yes No Yes No Yes
Observations 18,822 18,822 18,739 18,739 4,572 4,572 3,482 3,482
R2 0.000 0.079 0.001 0.076 0.005 0.037 0.010 0.028

Table 10: Other Robustness Checks
This table reports results from our difference-in-difference (DID) estimates of the effect of managerial ownership on firm performance with different robustness checks. Column
(1)-(2) use the log of Tobin’s Q instead of Tobin’s Q itself. Columns (3)-(4) include dividend payout variables as additional controls. Columns (5)-(6) restrict to firms whose largest
institutional holder is not affected by the 2003 Tax Cut. Columns (7)-(8) use different methods to calculate standard errors, with columns (7) using bootstrap for 2000 times and
column (8) allowing two-way clustering with both time and firm. The dependent variable in columns (3) to (8) is Tobin’s Q, defined as the ratio of the market value of assets to the
book value of assets, where the market value of assets is calculated as the market value of common equity plus book value of total liability plus book value of preferred stock. The
key independent variables Share02*Post2003 and ShareSq02*Post2003 identify our treatment effects, which are the treatment dummy, Post2003, interacted with linear and quadratic
is the square of Share02. See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are reported in parentheses in columns (1)-(6). Our sample
ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%,
respectively.

Dependent Variable: Two-way
Controlling Dividend Non-Tax Inst. Holders Bootstrap
47
Log Tobin’s Q Cluster

(1) (2) (3) (4) (5) (6) (7) (8)
Share2002 * Post2003 0.877*** 0.868*** 1.876*** 1.834*** 1.677** 1.716** 1.923*** 1.923***
(0.114) (0.114) (0.294) (0.293) (0.721) (0.719) (0.239) (0.367)
(Share2002)2 * Post2003 -0.884*** -0.869*** -1.937*** -1.888*** -2.122** -2.189** -1.945*** -1.945
(0.157) (0.155) (0.400) (0.396) (0.927) (0.908) (0.322) (0.413)
Dividend -0.210*** -0.229***
(0.054) (0.061)
Positive dividend 0.364*** 0.384***
(0.044) (0.044)
Controls No Yes No Yes No Yes Yes Yes
Observations 15,846 15,846 15,846 15,846 4,030 4,030 15,846 15,846
R2 0.012 0.037 0.016 0.039 0.003 0.044 0.059 0.698

7 A PPENDIX
7.1 A Stylized Model and Empirical Implications

Here we develop a model drawing from Chetty and Saez (2010) and Cheng, Hong, and Shue (2013) to
illustrate the mechanism of the 2003 Tax Cut and a hump-shaped relationship between managerial ownership
and changes in firm performance. For any firm i in period 0, the representative manager allocates a firm’s
cash Γi among three alternatives uses. The manager either invests Ki in a productive project, spends Ji in
less productive investment or pays out a dividend Di at period 0.33 The return from all investment in period
1 will be f (Ki ) + Ki + Ji and the manager obtains a private benefit from the less productive investment
g(Ji ). We assume that f (·) and g(·) are positive, strictly increasing and concave functions, with g(0) = 0.
The time discount rate is r.
We denote the fraction of managerial ownership as αi and take it as given in the model. Effective
managerial ownership stake ωi is the residual share of payouts from equity holdings of the firm accruing to
the manager, i.e., ωi = αi (1 − τ ), where τ is the tax rate. The manager trades off the private benefit due
to control rights against the stake in the dividend payout and the investment return due to cash-flow rights,
under the budget constraint. The manager solves,

f (Ki ) + Ki + Ji g(Ji )
max αi (1 − τ ) Di + +
Ki ,Ji ,Di 1+r 1+r
subject to Ki + Ji + Di ≤ Γi . Denote the pre-2003 tax rate as τipre and the after-2003 tax rate as τipost ,
the 2003 Tax Cut leads to a fall in the tax rate in our simple model τ post < τ pre , or equivalently an
increase in effective ownership stake ωipost > ωipre . Denote firm value under the pre-2003 tax rate as
Vipre = f (Kipre ) + Kipre and that under the after-2003 tax rate as Vipost = f (Kipost ) + Kipost . As the
manager’s marginal benefit of productive projects increases, we expect a (weak) increase in the productive
investment subject to the budget constraint, and subsequently a (weak) increase in firm value, ceteris paribus.
Under reasonable assumptions, we derive the following predictions regarding the effect of the 2003 Tax Cut
on firm value.
P REDICTION 1: A lower value of tax rate τ , or equivalently, a higher value of effective net-of-tax
ownership stake ωi , leads to a (weak) increase in firm value V : Vipost − Vipre ≥ 0, ceteris paribus.
P REDICTION 2: The increase in firm value, i.e., Vipost − Vipre is hump-shaped with respect to the
pre-determined managerial ownership ratio αi , ceteris paribus.
We make the following assumptions to prove our predictions. For simplicity, we omit the firm-specific
subscript i in the derivations to follow. We note that the model does not assume that the pre-2003 levels
of managerial ownership are randomly distributed across firms and we do not require the optimal level of
ownership to be the same across all firms.
Assumption 1. Both f (·) and g(·) are increasing and concave in their arguments. In addition,
1. 0 < f 0 (Γ) < r < f 0 (0), i.e., ∃K ∗ ∈ (0, Γ), such that f 0 (K ∗ ) = r
2. 0 ≤ g 0 (Γ) < g 0 (Γ − K ∗ ) < r < g 0 (0)
Let w = α(1 − τ ). Under Assumption 1, we can characterize the manager’s behavior as follows:
33
This model ignores the possibility that firms can repurchase shares instead of issue dividends. We follow Chetty and Saez
(2010) and Cheng, Hong, and Shue (2013) in assuming that firms issue dividends due to frictions. Firms that do not issue dividends
could also be affected as the market expects the tax cut to be permanent and they anticipate paying dividends in the future.
48

Proposition 1. Under Assumption 1, there exists two thresholds
g 0 (Γ)
ω=
f 0 (0)
g 0 (Γ − K ∗ )
ω̄ =
r
such that
1. When ω ≤ ω, D = 0, K = 0, J = Γ.
2. When ω < ω ≤ ω̄, D = 0, 0 < K ≤ K ∗ , J > 0, with
ωf 0 (K) = g 0 (Γ − K)
J =Γ−K
3. When ω > ω̄, D > 0, K = K ∗ ,J > 0, with
g 0 (J) = ωr
D = Γ − K∗ − J
The intuition of the proposition above is as follows. When effective ownership is very low, gains from
private investment outweigh gains from non-private investment or dividends, so the manager will invest
zero in productive investments. As effective ownership grows above threshold ω̄1 , managers start to invest
positive amounts in productive investments, until the productive investment level attains optimal level K ∗ .
After that, productive investment stays constant while the manager chooses between non-private instant
dividend D and less productive private investment J.
Proof. We first simplify the optimization, and then use Kuhn-Tucker conditions to prove the proposition.
1. The budget constraint is always binding because f 0 > 0. So the original problem is equivalent to
max ω (Γ + rD + f (K)) + g (J)

D,K,J
subject to
D+K +J =Γ
The new program is to maximize the manager’s payoff in the second period, which equals the share
from firm value including interest from dividends, as well as the return from the project. The La-
grangian is
L = ω (rD + f (K)) + g (J) − λ (D + K + J)
with Kuhn-Tucker conditions
∂L
= ωr − λ ≤ 0 with equality if D > 0 (4)
∂D
∂L
= ωf 0 (K) − λ ≤ 0 with equality if K > 0 (5)
∂K
∂L
= g 0 (J) − λ ≤ 0 with equality if J > 0 (6)
∂J
2. When ω ≤ ω, D = 0, K = 0, J = Γ.
49

(a) If K > 0, λ = ωf 0 (K) < ωf 0 (0) ≤ ωf 0 (0) = g 0 (Γ) < g 0 (J), so ∂L
∂J > 0, contradiction.
(b) If D > 0, λ = ωr, so ∂L
∂K = ω · (f 0 (0) − r) > 0, contradiction.
3. K ≤ K ∗ . If not, K > K ∗ , λ = ωf 0 (K) < ωr, so ∂L

∂D > 0, contradiction.
4. When ω < ω ≤ ω̄, D = 0, 0 < K ≤ K ∗ , J > 0.

∂L
(a) If D > 0, λ = ωr, ∂K = ω (f 0 (K) − r) ≤ 0 implies K ≥ K ∗ , so ∂K
∂L
= 0, and K = K ∗ .
∂L 0 ∗ 0 ∗
Then ∂J = g (Γ − K − D) − ωr > g (Γ − K ) − ωr = r (ω̄ − ω) ≥ 0, contradiction.
(b) If K = 0, J = Γ, λ = g 0 (Γ), ∂L
∂K = ωf 0 (0) − g 0 (Γ) > ωf 0 (0) − g 0 (Γ) = 0, contradiction.
(c) If J = 0, K = Γ ≤ K ∗ , contradiction.
5. When ω > ω̄, D > 0, K = K ∗ ,J > 0.
(a) Note that from super-modularity, K is increasing in ω. And we know when ω = ω̄, K = K ∗ .
So when ω > ω̄, K = K ∗ , and λ = ωr.
(b) If D = 0, J = Γ − K ∗ . ∂L
∂J = g 0 (Γ − K ∗ ) − ωr = r (ω̄ − ω) < 0 implies J = 0, contradiction.
(c) If J = 0, ∂L 0 0
∂J = g (0) − ωr > g (0) − r > 0, with the last inequality resulting from the
assumption. Contradiction.
Since V (K) is increasing in K and we show in the above that K is increasing in ω, thus Prediction
1 follows. Now we proceed to prove Prediction 2 in the paper. We assume both f (·) and g (·) are power
functions for simplicity, and the following two Lemmas provide conditions to derive the hump-shaped rela-
tionship between the change in firm value and managerial ownership. Note that since f 0 (0) g 0 (Γ), we
have ω = 0.
Lemma 1. Suppose f (K) = AK γ , and g (J) = BJ γ with γ < 1, define σ = 1

1−γ , δ = γr , and
(δA)σ + (δB)σ < Γ
then assumptions for Proposition 1 are satisfied, with
ω=0
K ∗ = (δA)σ
1
(δB)σ

σ
ω̄ = σ
Γ − (δA)
Proof. We follow several steps to prove this Lemma.
1. Since f 0 (0) = +∞, ω = 0.

1
γA
= (δA)σ
1−γ
2. f 0 (K) = γAK γ−1 = r, so K ∗ = r
1
3. To show 0 < f 0 (Γ) < r < f 0 (0), it suffices to show f 0 (Γ) < r, which is equivalent to γAΓ− σ < r,
so (δA)σ < Γ
50

4. To show 0 ≤ g 0 (Γ) < g 0 (Γ − K ∗ ) < r < g 0 (0), it suffices to show g 0 (Γ − K ∗ ) < r, which is
1
equivalent to γB (Γ − K ∗ )− σ < r, which means (δA)σ + (δB)σ < Γ
1 1
g 0 (Γ−K ∗ ) γB(Γ−K ∗ )− σ (δB)σ

σ
5. Lastly, ω̄ = r = r = Γ−(δA)σ
We define firm value as V = f (K) + K, the following condition characterizes one property of this
function.
Lemma 2. Suppose conditions for Lemma 1 are satisfied, and in addition,

1+γ
(δA)σ < Γ
γ
when ω ≤ ω̄,
∂2V
<0
∂α∂τ
Proof. From first order condition of ωf 0 (K) = g 0 (Γ − K), we have
1
K ωA 1−γ
= =C
Γ−K B
C
and K = 1+C Γ. So
∂C C
=
∂ω ω(1 − γ)
∂K K
=
∂ω ω(1 − γ)(1 + C)
and
∂V ∂K
= f 0 (K) + 1
∂ω ∂ω
γf (K) + K
=
ω (1 − γ) (1 + C)
∂2V ∂V γf 0 (K) + 1 ∂K

1 ∂C/∂ω
= · − −
∂ω 2 ∂ω γf (K) + K ∂ω ω 1+C
2
∂V 1 γ f (K) + K 1
= · −C −
∂ω (1 − γ)(1 + C)ω γf (K) + K ω
and
∂2V ∂2V ∂V
= − ·ω−
∂α∂τ ∂ω 2 ∂ω 2
∂V 1 γ f (K) + K
= − · −C
∂ω (1 − γ)(1 + C) γf (K) + K
∂V 1 1
= − · · ((1 − C)K + γ (γ − C) f (K))
∂ω (1 − γ)(1 + C) γf (K) + K
51

Let φ = (1 − C)K + γ (γ − C) f (K), what remains to show is φ > 0 for ω ∈ (0, ω̄). Since C (ω) is
increasing in ω, it suffices to show
γ − C (ω̄) > 0
which is equivalent to σ σ
(δB)σ (δA)σ

ω̄A A
γ> = · =
B Γ − (δA)σ B Γ − (δA)σ
so
(δA)σ
Γ − (δA)σ >
γ
or
1+γ
Γ> (δA)σ
γ
In the next proposition, we build our implication with respect to the change in firm value due to the tax
cut. Let τpre = 0.35 and τpost = 0.15 be the dividend tax rate before and after tax, and define two threshold
values of ownership ᾱ1 and ᾱ2 as ᾱ1 (1 − τpost ) = ω̄ and ᾱ2 (1 − τpre ) = ω̄. Let the change in firm value
be ∆V (α) = V (α (1 − τpost )) − V (α (1 − τpre )). The following proposition provides implications of
∆V with respect to ownership α.
Proposition 2. Suppose conditions for Lemma 1 and Lemma 2 hold, then ∆V (α) is hump-shaped with
respect to α. More specifically, ∆V is increasing in α for α < ᾱ1 , and ∆V is decreasing in α for α > ᾱ1 .
Proof. We show the proposition for α belonging to three different ranges.
1. When α ≤ ᾱ1 , V (·) is a smooth function, define ∆τ = τpost − τpre < 0,
∂V (α (1 − τ̃ ))
∆V (α) = ∆τ
∂τ
so
∂∆ (α) ∂ 2 V (α (1 − τ̃ ))
= ∆τ > 0
∂α ∂τ ∂α
∂2V
for ∂α∂τ < 0 shown from Lemma 2.
2. When ᾱ1 < α ≤ ᾱ2 , Kpost = K ∗ , V (α (1 − τpost )) = f (K ∗ ) + K ∗ = V̄post , so ∆ (α) =

V̄post − f (Kpre ) − Kpre . Since Kpre is increasing in α, so ∆V is decreasing in α.
3. When α > ᾱ2 , Kpost = Kpre = K ∗ , so ∆V (α) = 0
52

Figure A.1: Pre-trend Tests for Longer Periods
This figure plots the coefficients in front of the year-specific linear and quadratic terms of the treatment variable
from 1998 to 2008, where 2002 is omitted as the base year. The dependent variable is Tobin’s Q, defined as
the ratio of the market value of assets to the book value of assets, where the market value of assets is calculated
as the market value of common equity plus book value of total liability plus book value of preferred stock.
For each year in our sample, we create year-specific treatment effects (for example, Share2002*Y ear2000
and Share20022 *Y ear2000) by interacting the year dummy with linear and quadratic terms of the treatment
intensity. Robust standard errors clustered at the firm-level. The error bounds are 95% confidence intervals
2
Linear Term
0 -1
-2 1
1998 1999 2000 2001 2003 2004 2005 2006 2007 2008
Year
1 0
Quadratic Term
-2 -1
-3
1998 1999 2000 2001 2003 2004 2005 2006 2007 2008
Year
53

Table A.1: Additional Robustness Results
This table reports results from our difference-in-difference (DID) estimates of the effect of managerial ownership on firm performance, with robustness checks. Columns (1)-(2)
winsorize the dependent variable at 5%, columns (3)-(4) winsorize the dependent variable at 10%, columns (5)-(6) use quantile regression. The dependent variable is Tobin’s Q,
defined as the ratio of the market value of assets to the book value of assets, where the market value of assets is calculated as the market value of common equity plus book value
of total liability plus book value of preferred stock. The key independent variable Share2002*Post2003 identifies our treatment effects, which is the post dummy interacted with the
treatment intensity. Post dummy, Post2003, takes the value of one for the year 2003 and onward and zero otherwise. Treatment intensity, Share2002, is the sum of the fractions of
shares held by all directors and officers in 2002 from Compact Disclosure. See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are reported
in parentheses. Our sample ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients marked with *, **, and *** are
significant at 10%, 5%, and 1%, respectively.

Winsorize at 5% Winsorize at 10% Quantile Regression
(1) (2) (3) (4) (5) (6)
Share2002 * Post2003 1.688*** 1.664*** 1.383*** 1.364*** 1.494*** 0.929***
(0.244) (0.243) (0.192) (0.192) (0.166) (0.258)
(Share2002)2 * Post2003 -1.718*** -1.681*** -1.424*** -1.394*** -1.364*** -0.949***
54
(0.332) (0.330) (0.265) (0.263) (0.228) (0.181)

Controls No Yes No Yes No Yes
Year fixed effects No No No No Yes Yes
Firm fixed effects No No No No Yes Yes
Observations 15,846 15,846 15,846 15,846 15,846 15,846
R2 0.702 0.709 0.719 0.727

Table A.2: Linear Specification
This table reports the linear specification for our difference-in-difference (DID) estimates of the effect of managerial ownership
on firm performance. The dependent variable is Tobin’s Q, defined as the ratio of the market value of assets to the book value of
assets, where the market value of assets is calculated as the market value of common equity plus book value of total liability plus
book value of preferred stock. The key independent variable Share2002*Post2003 identifies our treatment effects, which is the post
dummy interacted with the treatment intensity. Post dummy, Post2003, takes the value of one for the year 2003 and onward and
zero otherwise. Treatment intensity, Share2002, is the sum of the fractions of shares held by all directors and officers in 2002 from
Compact Disclosure. See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are reported in
parentheses. Our sample ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects.
Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively.

(1) (2) (3) (4)
Share2002 * Post2003 0.566*** 0.554*** 0.549*** 0.562***
(0.097) (0.097) (0.097) (0.095)
Firm size controls No Yes Yes Yes
Moral hazard controls No No Yes Yes
Firm risk controls No No No Yes
Miscellaneous controls No No No Yes
Observations 15,846 15,846 15,846 15,846
R2 0.674 0.675 0.676 0.681
55

Table A.3: Capital Intensity
This table reports results from our difference-in-difference (DID) estimates of the effect of managerial ownership on firm performance, robust to concerns on capital intensity. The
dependent variable is Tobin’s Q, defined as the ratio of the market value of assets to the book value of assets, where the market value of assets is calculated as the market value of
common equity plus book value of total liability plus book value of preferred stock. The key independent variables Share02*Post2003 and ShareSq02*Post2003 identify our treatment
effects, which are the treatment dummy, Post2003, interacted with linear and quadratic terms of treatment intensity. Treatment intensity, Share02, is the sum of the fractions of shares
held by all directors and officers in 2002 from Compact Disclosure while ShareSq02, is the squared of Share02. See Table 4 for a complete list of controls. Robust standard errors
clustered at the firm-level are reported in parentheses. Our sample ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients
marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively.

(1) (2) (3) (4) (5)
Share2002 * Post2003 1.670*** 1.597*** 1.693*** 1.676*** 1.696***
(0.322) (0.320) (0.347) (0.323) (0.347)
(Share2002)2 * Post2003 -1.619*** -1.518*** -1.591*** -1.625*** -1.594***
(0.428) (0.424) (0.463) (0.429) (0.463)
Capital-to-sales * Post2003 -0.001 -0.005
56
(0.009) (0.008)
(Capital-to-sales)2 * Post2003 0.000 0.000
(0.000) (0.000)
Controls Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes
Year*SIC3 fixed effects Yes Yes Yes Yes Yes
Year*Capital-to-sales-above-median fixed effects No Yes Yes No Yes
Year*SIC3*Capital-to-sales-above-median fixed effects No No Yes No Yes
Observations 15,846 15,829 15,829 15,829 15,829
R2 0.707 0.709 0.719 0.707 0.719

Table A.4: Other confounding channels
This table reports results from our difference-in-difference (DID) estimates of the effect of managerial ownership on firm per-
formance, robust to other confounding channels including cost of capital and tax holidays. Columns (1) and (2) restricts to the
subsample of firms whose total holdings by individual investors (excluding managers and officers) are below median in 2002.
Columns (3) and (4) restricts to the subsample of firms with foreign tax below or equal to one million USD for all years in our data
period. The key independent variables Share02*Post2003 and ShareSq02*Post2003 identify our treatment effects, which are the
treatment dummy, Post2003, interacted with linear and quadratic terms of treatment intensity. Treatment intensity, Share02, is the
sum of the fractions of shares held by all directors and officers in 2002 from Compact Disclosure while ShareSq02, is the squared of
Share02. See Table 4 for a complete list of controls. Robust standard errors clustered at the firm-level are reported in parentheses.
Our sample ranges from 2000 to 2005. Regressions in all columns include year-fixed effects and firm-fixed effects. Coefficients
marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively.

Low Ind. Ownership Non Multinational
(1) (2) (3) (4)
Share2002 * Post2003 0.748** 1.124*** 1.900*** 1.812***
(0.344) (0.343) (0.450) (0.447)
(Share2002)2 * Post2003 -0.188 -0.635 -2.278*** -2.176***
(0.434) (0.429) (0.588) (0.582)
Controls No Yes No Yes
Observations 7361 7361 9603 9603
R2 0.697 0.710 0.668 0.677
57

Managerial Ownership and Firm Performance 2

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Managerial Ownership and Firm Performance 2

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Managerial Ownership and Firm Performance 2

Uploaded by

Copyright:

Available Formats

Managerial Ownership and Firm Performance:

Evidence From the 2003 Tax Cut ∗

Xing Li Stephen Teng Sun Constantine Yannelis

JEL Classification: M48, G34, G32, D86

Electronic copy available at: https://ssrn.com/abstract=2285638

banks (Calomiris and Carlson, 2014).

Electronic copy available at: https://ssrn.com/abstract=2285638

in-difference empirical strategy.

Electronic copy available at: https://ssrn.com/abstract=2285638

to more severe agency problems.

Electronic copy available at: https://ssrn.com/abstract=2285638

of capital, and (vii) the 2004 Foreign Tax Holiday.

three key contributions.

falsification tests to deal with potential endogeneity concerns.

Electronic copy available at: https://ssrn.com/abstract=2285638

framework, the paper makes three key contributions.

support for the agency view in Chetty and Saez (2010).

Electronic copy available at: https://ssrn.com/abstract=2285638

fiscal year in December. We also exclude dual class firms.

Electronic copy available at: https://ssrn.com/abstract=2285638

than 60%. The maximum managerial ownership in our sample is 100%.

2.2 The 2003 Tax Cut

Electronic copy available at: https://ssrn.com/abstract=2285638

increased managerial incentive on firm value by employing a difference-in-difference empirical strategy. We

2.3 Empirical Model and Identification Strategy

We assume that firm performance, as measured by Tobin’s Q, follows a "hump-shaped" or quadratic

Electronic copy available at: https://ssrn.com/abstract=2285638

nominal managerial ownership) as the following:

yit = γ1 EffectiveShareit + γ2 EffectiveShare2it + Xit β + c̃i + dt + ηit , (1)

EffectiveShareit = (1 − τpre + ∆τ · P ost2003t ) · Shareit , (2)

Shareit , we derive estimation equation as the following17

yit = α1 Share2002i × P ost2003t + α2 Share20022i × P ost2003t + Xit β + ci + dt + εit . (3)

reflected in the change of firm fixed effect from c̃i to ci .

Electronic copy available at: https://ssrn.com/abstract=2285638

managerial ownership.20 Therefore, our empirical model is identical to a difference-in-difference approach

structures following the reform.

Electronic copy available at: https://ssrn.com/abstract=2285638

our main regression.

in firm performance before and after the tax reform.

Electronic copy available at: https://ssrn.com/abstract=2285638

3.1 Fixed Effects Estimates

3.2 Nonparametric Evidence

Electronic copy available at: https://ssrn.com/abstract=2285638

3.3 Main Results

or a 5.41% improvement, given an average of Tobin’s Q of 2.03 in the sample.23

3.4 Results by Year and Pre-trend Tests

Electronic copy available at: https://ssrn.com/abstract=2285638

Year 2005 displays similar results as in previous years.

3.5 Heterogeneous Effects by Proxies of Agency Problems

Electronic copy available at: https://ssrn.com/abstract=2285638

idiosyncratic volatility of stock returns reveal a similar statistically significant pattern.

3.6 Alternative Performance Measures and Other Firm Outcomes

Electronic copy available at: https://ssrn.com/abstract=2285638

could be driven by direct effects of the reform on market valuation.

of operating performance display similar patterns from columns (2) to (4).

Electronic copy available at: https://ssrn.com/abstract=2285638

4 Interactive effects of managerial ownership and other governance chan-