The Positive Externalities of CEO Delta∗

Hongrui Feng† Yuecheng Jia‡

January 8, 2018

Abstract

The increases in Delta incentives are dramatic for a small group of firms (leader firms)

but negligible for the majority. We show that leader firms are the ones with larger market

capitalization, higher irreversibility, and in industries with negative shocks. When leader

firms experience substantial growth in Delta incentives, industry peers experience positive

abnormal returns and abnormal improvement in fundamentals despite having no significant

change in Delta. Further, we provide evidence that the abnormal returns are induced by

peer CEOs’ extra effort in response to the increasing competitive pressure caused by leader

firms. To mitigate their competitive pressure and turnover threat, peer CEOs allocate their

extra effort to the firms’ operating efficiency and product differentiation.

JEL Classification: G14, G34

Keywords : CEO Delta incentive, Positive externality, Incentive spillover

∗We would like to thank Ramesh Rao, Brian Boscaljon, Xiangjun Hong, Kuo Zhang (discussant), andseminar participants at European Financial Management 2017 Xiamen Symposium for helpful comments.All errors are our own.†Feng is at the Black School of Business, Pennsylvania State University–Behrend, 4701 College Dr, Erie,

PA 16563, USA; email: hxf51@psu.edu.‡Jia is at the Chinese Academy of Finance and Development, Central University of Finance and Eco-

nomics, Beijing, China; email: jiayuecheng@cufe.edu.cn.

The Positive Externalities of CEO Delta

Abstract

The increases in Delta incentives are dramatic for a small group of firms (leader firms)

but negligible for the majority. We show that leader firms are the ones with larger market

capitalization, higher irreversibility, and in industries with negative shocks. When leader

firms experience substantial growth in Delta incentives, industry peers experience positive

abnormal returns and abnormal improvement in fundamentals despite having no significant

change in Delta. Further, we provide evidence that the abnormal returns are induced by

peer CEOs’ extra effort in response to the increasing competitive pressure caused by leader

firms. To mitigate their competitive pressure and turnover threat, peer CEOs allocate their

extra effort to the firms’ operating efficiency and product differentiation.

JEL Classification: G14, G34

Keywords : CEO Delta incentive, Positive externality, Incentive spillover

1 Introduction

Recent literature documents that the increases in Delta compensation stimulate CEOs to

exert extra efforts in the firms’ operation and enhance corporate stock market performance

(e.g., Cooper et al. (2014); Kale et al. (2009); Lilienfeld-Toal and Ruenzi (2014)). However,

the change in Delta (∆Delta) has much richer dynamics. In reality, ∆Delta is not evenly

distributed among firms. In contrast, it varies substantially across firms, evidenced by dra-

matic for a small group of firms (leader firms) but negligible for the majority. For instance,

Expedia Inc. Chief Executive Dara Khosrowshahi’s incentive pay grew tenfold in 2015 (Wall

Street Journal, 1 May 2016) (As of late August, Mr. Dara Khosrowshahi is the acting CEO

of Uber.). However, for Priceline, Expedia’s primary competitor, Delta incentive merely

decreased by 2% in the same time period.

This study examines potential determinants for why a substantial increase in the level

of Delta incentive compensation is witnessed within small groups of leader firms (such as

Expedia Inc.). We then explore an important economic consequence of the dramatic Delta

increase. We document that a small group of firms’ substantial growth in Delta incentive

compensation predicts the stock returns of their industry peers1. Furthermore, we find that

the positive externality, from large ∆Delta in one firm to its peer firms’ market performance,

is induced by peer CEOs’ extra effort. Peer CEOs exert extra efforts in their firms’ operation

in response to the increasing competitive pressure caused by leader firms.

As a corporate governance device (e.g., Holmstrom (1999); Acharya and Volpin (2009)),

Delta compensation is designed to align CEOs’ interests with that of shareholders. However,

if CEOs have decreasing marginal utility on incentive pay, interest alignment can hardly

explain the substantial increase in Delta such as the tenfold increase in Delta for Expedia

Inc. Chief Executive Dara Khosrowshahi. Reasons beyond interest alignment should be

1For instance, the dramatic growth in Delta for Expedia Inc. is followed by its main competitor–Priceline’sstock price increase from $1,084.87 (1 February 2016) to $1,445.83 per share (8 September 2016). This is notan isolated case. We observe that when large firms such as Bestbuy, P&G, and IBM significantly increasetheir Delta, the stock prices of their industry competitors jump.

1

applied. Our extensive exploration reveals two commonalities among these incentive leader

firms. First, firms with the substantial increase in Delta (incentive leader firms) concentrate

in industries experiencing industry–wide negative shocks such as import tariff cuts. The

industry negative shocks can deteriorate the industry market condition, driving the firms to

stimulate their CEOs to exert more efforts.

Next, within the industries having negative shocks, we explore the determinants of leader

firms. More specifically, we find that large firms and value firms are the ones who react

first. The two commonalities bring us the full picture of who are incentive leader firms.

Facing negative industry–wide shocks, industry leaders with larger market capitalization,

more tangible assets and a higher degree of irreversibility are the ones exposed to severe

losses. In response to the potential losses in profit and in industry positions, industry leaders

initiate the increases in Delta compensation and stimulate CEOs to exert more efforts2.

As a consequence, the extra effort from leader firm CEOs improves the firms’ operating

efficiency (Lilienfeld-Toal and Ruenzi (2014)), putting competitive pressure on its industry

peer firms (e.g., Dixit (1986)). Prior literature in industry organization concludes that peer

firms experience a product price cut and negative market performance, implying a negative

externality from one firm to its peers.

Prior literature emphasizes the negative influence of product market competition on

industry peers while ignores peer CEOs’ active response to the competition. Peer CEOs’

implicit incentives such as mitigating the turnover threat and keeping the reputation can

mitigate the negative impact of product market competition on peer firms. To mitigate the

turnover threat and maintain their reputation, CEOs in peer firms can increase their efforts

to run the firms more efficiently, suggesting a positive change in future stock price among

these firms. Thus, the competitive pressure generated by the substantial increase in one

firm’s Delta can lead to two opposite effects on its peer firms: (i). negative externality on

peer firms’ market performance because of their product price cut generated by the increased

2We thank the referee’s suggestions to identify who are leaders in the first place.

2

industry competition (e.g., Dixit (1986)); and (ii). positive externality on peer firms’ market

performance because of increasing operating efficiency from peer CEOs’ extra effort. It is

then an empirical question to explore which effect dominates.

In this study, we explore the leader firms’ CEO–incentive spillover to their peer firms’

market performance and the fundamentals. We document a positive CEO incentive exter-

nality by showing that a firm’s stock return can be positively predicted by the substantial

increase in the CEO Delta incentives of other firms within the same industry. Moving one

step further, we document that peer firms’ managerial discretion-related operating efficiency

and product differentiation are related to leader firms’ CEO Delta incentives. Following a

substantial increase in Delta incentive in an industry driven by a small group of incentive

leaders, peer firms (in the same industry) experience a significant improvement in their oper-

ating performance as well as better stock market performance despite having no significant

Delta incentive increases by themselves. Our findings reveal that CEO incentive leaders

have positive externality on peer firms’ operation and market performance. That is, peer

CEOs’ extra effort in firms’ operation dominates the increased industry competition brought

in by CEOs of leader firms.

Externalities of Delta incentives on peer firms’ market performance have shown the po-

tential to demonstrate both positive and negative growth patterns. We explore whether a

small group of firms’ dramatic cut in Delta incentives carries negative externalities on peer

firms’ market performance as executives have less incentive to compete. Our evidence indi-

cates that the spillover effect of Delta is one–sided: a small group of firms’ substantial cut in

the use of Delta has no spillover effect on peer firms’ market performance. This asymmetric

positive spillover effect is not surprising based on our further exploration. First, compared

with the magnitude of Delta increases, that of Delta decreases are highly negligible and

have small dynamics. We can rarely find a decrease in Delta of more than 5%. The neg-

ligible decrease of Delta can hardly variate CEOs’ incentives. Second, no commonalities

exist in negative incentive leaders. Our story on positive incentive leaders does not apply to

3

negative leaders. That is, negative leaders do not have larger market capitalization or more

tangible assets. Lastly, the negligible decrease of Delta can hardly drag down CEOs’ effort

in operating firms under internal and external monitoring.

Specifically for our empirical analysis, we begin by identifying positive Delta incentive

leaders and peers in industries that have experienced substantial aggregate Delta incentive

growth. We define positive Delta incentive leaders as a small group of firms that have

experienced significant growth in Delta. We subsequently classify the remaining firms in the

same industry as peers. For our sample period from 1992 to 2015, we identify 851 firm-year

observations as CEO incentive leaders and 7,576 as peers. The average CEO Delta incentive

growth rate for the leaders is 10 times higher than that for the peers (100% vs 9%), indicating

a large distinction between these two types of firms.

Reflexively, we define negative Delta incentive leaders by first identifying industries with

top-decile aggregate Delta incentive decreases3. We then classify negative incentive leaders

as the firms that have experienced top 10% Delta decrease within the industry. In contrast

to positive Delta incentive leaders, the average Delta incentive decrease rate for negative

incentive leaders is only marginally larger than that for peers. Therefore, positive Delta

incentive leaders have much larger variations than their negative counterparts.

We find that both positive Delta incentive leaders and peers experience positive abnor-

mal returns following their identification. An equal-weighted portfolio of leaders has an

annualized Carhart (1997) four–factor alpha of 12.78%. Similarly, an equal-weighted port-

folio of peers has an annualized abnormal return of 9.80%. The value-weighted annualized

abnormal returns for leaders and peers are 12.25% and 9.53%, respectively. In contrast, even

though the equal–weighted portfolio of negative Delta incentive leaders has marginally neg-

ative abnormal return, the equal–weighted portfolio of negative incentive peers experiences

no abnormal returns. Our findings confirm the asymmetry of this Delta-change externality:

3We do not define Delta–decrease event using -5% cutoff because the industry–wide decrease in Delta israther small.

4

positive Delta jumps in firms have a significant positive spillover effect on their peers mar-

ket performance. In contrast, negative Delta jumps considered small in magnitude are less

informative and do not influence their peer firms’ market performance.

The significant positive abnormal returns for the portfolio of Delta leaders are in line

with Giroud and Mueller (2011) and Lilienfeld-Toal and Ruenzi (2014). More important-

ly, the significant positive abnormal returns for the portfolio of peers indicate a positive

Delta incentive spillover from leaders to peers on stock market performance. To identify

the marginal explanatory power of peers on returns, we also perform the Fama–Macbeth

regressions controlling for firm characteristics and peers’ Delta. We find that the peers

continue to have significantly positive abnormal returns even after controlling for multiple

firm characteristics and peers’ Delta. That is, the positive Delta incentive spillover exists

because incentive peer firms experience positive abnormal market performance even without

a significant increase in CEO pay–performance incentives.

We explore multiple explanations for why CEO incentive peers experience positive ab-

normal returns. First, we investigate whether the peer firms’ positive abnormal market

performance is due to the real effect on peer firms or the investors’ inattention. We find

that both Delta leaders and peers experience positive sales growth, gross profitability, and

return on assets in the two years following identification. The results are robust even after

controlling for different firm characteristics and Delta. In contrast, investor inattention has

no explanatory power on peer firms’ market performance. These findings reveal that peer

and leader firms’ market performance is due to the strong fundamentals and unlikely to be

attributed to a behavioral story.

The strong fundamentals of leader and peer firms indicate real improvement in these

firms. Specifically, the strong fundamentals of leaders come from CEOs’ extra effort in

running their firms more efficiently (Lilienfeld-Toal and Ruenzi (2014)). Hinged on leader

CEOs’ allocation of their extra effort, the real improvement in peers’ fundamentals has

two different explanations. First, if CEOs in leader firms allocate their efforts to R&D

5

investment, the strong fundamentals of peers are then caused by technology spillover. In

contrast, if leader firms’ CEOs allocate effort to improve the firms’ operating efficiency, leader

firms’ incentive spillover gives rise to peers’ strong fundamentals. The next two paragraphs

present the two explanations in detail.

The first potential explanation for peer firms’ strong fundamentals assumes leader CEOs

contribute their efforts in increasing R&D investment. Corporate investment in R&D gen-

erates new ideas and technology, creates extra demand for the entire industry, and thus

improves industry-wide productivity (e.g., Hall (1993); Jiang et al. (2015)). If the Delta

incentive increase in leaders drives their CEOs to exert extra efforts in R&D, the peer firms

can benefit from the creation of extra demand and productivity improvement, thus expe-

riencing the increases in fundamentals and better market performance. This explanation

treats CEOs in peer firms as “free-riders” and does not require peer firms’ CEOs to exert

more efforts in running their firms.

The second explanation assumes leader CEOs allocate their efforts to improve their firms’

operating efficiency. If CEOs in incentive leader firms allocate their extra efforts in running

firms more efficiently, the industry-wide production market competition will increase (e.g.,

Chamberlin et al. (1933); Grossman and Shapiro (1984)). Consequently, peer firms encounter

more competitive pressure and the corresponding threat of CEO turnover in peer firms

increases (e.g., Dasgupta et al. (2014)). Due to the increasing threat of turnover, the peer

firms’ CEOs also increase their efforts to run their firms more efficiently and increase peer

firms’ product differentiation. This explanation treats the Delta incentive increase in the

leader companies as an external corporate governance device for peer companies’ CEOs.

To shed light on these two explanations, we conduct multiple tests in sequence. We test

whether the peer firms’ better market performance and strong fundamentals are caused by

peer CEOs’ effort in firm efficiency or in firms’ investment activities. To measure peer CEOs’

extra effort in firms’ operation, we use managerial scores (Demerjian et al. (2012)). More

importantly, we find that high managerial scores in peer firms are associated with peer firms’

6

positive abnormal returns and strong fundamentals.

Moreover, we find that the healthy market performance of peer firms is not due to peers’

increasing investment. We find neither leader nor peer firms experience a significant increase

in contemporaneous or future R&D investment, capital expenditure, or inventory. Thus, the

positive externality on peers’ stock market performance and fundamentals does not come

from leader firms’ R&D spillovers4. Compared with the results on R&D investment, we can

conclude that the spillover effects of incentive leaders on peers come from stimulating peer

CEOs to run firms more efficiently.

The contribution of our paper is at least twofold. First, we document that the increases

in Delta incentives are largely driven by a small group of firms’ substantial increases in Delta

but not by the industry–wide mere increases from a large cross–sectional of firms. To the best

of our knowledge, this paper is the first one to document that the dramatic increases in Delta

share different information contents than either the level of or the average changes in Delta.

Second and more interestingly, we find that the substantial increases in Delta from the small

group of firms have a positive spillover effect on their peer firms’ market performance. The

positive spillover effect complements prior research on the negative externality of one firm’s

increase in the CEO compensation on other firms’ CEO compensation.

The remainder of this paper is structured as follows. Section 2 surveys prior studies to

position this paper in the literature. Section 3 discusses the origin of incentive spillover and

proposes hypotheses. Section 4 discusses the data and our methodology. Section 5 provides

our main results. Section 6 concludes.

2 Literature Review

Our study relates to four strands of literature. First, our paper relates to the prior studies

on the information content of CEO Delta incentives. Even though prior works such as

4In other words, the CEO Delta incentive spillover has different information contents than the R&Dspillover in Jiang et al. (2015)

7

Core and Guay (1999), Core et al. (2003), and Guay (1999) document the determinants of

CEO Delta increase, they ignore the determinants of substantial increase in Delta. Since

the small group of firms’ tremendous Delta increases are the main component driving up

the industry–wide level of Delta, why corporate initiates dramatic Delta increase requires

plausible explanations. Our work is among the first to reveal that the initiation of the large

Delta increases is due to negative industry–wide shocks, firms’ industry status, and the

degree of irreversibility.

Second, in contrast to the prior literature, to the best of our knowledge, this paper is

the first to document a positive externality of CEO pay incentive on peer firms’ market

performance and fundamentals. Per Acharya and Volpin (2009), “firms with weaker gover-

nance offer managers more generous incentive compensation, which induces firms with good

governance to also overpay their management.” As a result, bad governance characteristics

prevail. In contrast, our results show that the increase in CEO incentive payment, although

possibly caused by weak governance Acharya and Volpin (2009), can mitigate or even re-

verse the negative impact of CEOs’ overpayment in peer firms by stimulating them to work

hard. The overpayment of peer firms to their CEOs can eventually be compensated by

their managerial hard work, better operating efficiency, and positive abnormal stock market

performance.

Moreover, our paper also relates to the recent literature on the intersection of corporate

governance, CEO incentive, and asset pricing. Such literature demonstrates that good cor-

porate governance (e.g., Giroud and Mueller (2010); Giroud and Mueller (2011)) and high

CEO incentive (e.g., Cooper et al. (2014); Lilienfeld-Toal and Ruenzi (2014)) benefit the

firms’ stock market performance. Our research elaborates on these works by documenting

that substantially positive Delta jumps have additional information beyond that of Delta

increase. A substantial increase in Delta for a small group of firms can affect not only their

own firms’ but also peer firms’ market performance.

Our baseline results also provide new insight into the theory of industrial organization.

8

The industrial organization literature predicts that an improvement in target firms’ operating

efficiency or productivity creates competitive pressure on peer firms. In response, peer firms

cut their product prices to deal with the increased competitive pressure (e.g., Dixit (1986);

Chamberlin et al. (1933); Tirole (2010)). However, the literature of industrial organization

ignores the impact of managerial discretion on firm decisions. Our results reveal that CEOs

in peer firms respond to competitive pressure from incentive leader firms by inputting extra

effort and running peer firms more efficiently.

3 Hypothesis Development

In this section, we spell out the hypotheses in order to answer the two questions listed in the

introduction: (i). who are incentive leaders? (ii). what is the impact of incentive leaders on

peers’ market performance and fundamentals?

First, we step back to the origin by exploring what are the propensities for firms to

initiate substantial Delta growth? The reason can be traced from leader firms’ usage of the

Delta and V ega combination. We find that the incentive leader firms tend to dramatically

increase Delta incentives but fix the risk–taking motivation (V ega) of CEOs. Therefore,

CEOs, motivated by increasing Delta, enhance the firms’ stock market performance through

the channel of increasing operating efficiency (e.g., Lilienfeld-Toal and Ruenzi (2014)) but

not through that of risk–taking. The “who are incentive leaders” question is equivalent to

the question that what type of firms requires “safe growth” using existing assets–in–place

instead of risk–taking investment in growth options.

The potential answers to the above questions are twofold because Delta incentive leaders

are classified through a two–pass procedure: (i). the firms in Delta increasing industries;

and (ii). the firms themselves in the top–decile of Delta growth rate within the industry.

Correspondingly, we need to explore sequentially what industries and the firms within these

specific industries tend to experience substantial Delta increase?

9

First, at the industry level, exogenous industry–wide negative shocks can result in in-

dustry bust. Under industry bust, risk–averse firms tend to delay risky projects but try

to enhance operating performance (e.g., Hoberg et al. (2014); Hoberg and Phillips (2016);

Hoberg and Phillips (2010)). We argue that industries with significantly increasing Delta

overlap with the ones having exogenous industry–wide negative shocks.

A natural proxy for the exogenous industry–wide shock is the unexpected variations of

the industry–level import tariffs (e.g., Feenstra, Romalis, and Schott (2002); Feenstra and

Romalis (2014); and Fresard (2010))5. Specifically, the import tariff cut triggers a significant

intensification of competitive pressure from foreign rivals and is a good proxy for the negative

shock at the industry level. To explore whether incentive leader firms tend to concentrate

in industries with severe negative industry–wide negative shocks, we spell out the following

hypothesis:

• H1a (Negative Industry Shock Hypothesis): Delta incentive leaders tend to be in

industries with negative industry shocks such as import tariff cut.

After exploring industry features of incentive leaders, we switch to the firm–level com-

monalities of these leader firms. Within industries experiencing negative shocks, we explore

the type of firms that tend to substantially increase the Delta incentives to enhance their

market performance. To form our hypothesis, we first manually search the name, market

capitalization, history, and other firm characteristics of the incentive leaders6. Our manual

search shows that incentive leaders tend to be larger firms within the industry. For instance,

in the year 1997 when computer industry (Fama and French 48 industry code: 35) experi-

enced a 56.8% tariff cut, IBM as a incentive leader more than doubled its Delta from 745

to 1686. Motivated by our manual search, we hypothesize that incentive leaders are the

industry leaders who are guiding the direction of the whole industry.

5We thank the referee for his suggestion to use import tariff cut as the proxy of industry negative shock.

6We explore a sequence of firm characteristics as follows: market capitalization, the book–to–marketratio, CEO overconfidence, CEO entrenchment, firm over/under–investment.

10

On the other hand, we hypothesize that leader firms are value firms because value firms

are the ones deriving their values more from “safe growth” using existing assets–in–place in-

stead of risk–taking investment in growth options. Value firms encounter more risk when neg-

ative industry shocks arrive since their assets–in–place lead to higher irreversibility (Zhang

(2005)). Thus, facing higher risk from negative shocks, leader firms tend to stimulate their

CEOs to improve firm operating efficiency by increasing Delta7.

Furthermore, previous studies document a positive relation between firm idiosyncratic

risk and CEO Delta incentives because higher idiosyncratic volatility requires executives to

exert more efforts and power delegation. Firms with higher uncertainty must pay higher

incentives to their executives (e.g., Albuquerque et al. (2014); Miller et al. (2002); Sarkar

et al. (2014)). We hypothesize that large increases in idiosyncratic volatility also motivate

firms to increase Delta incentive dramatically. To explore whether firms with large market

capitalization, a high degree of irreversibility, and high idiosyncratic volatility tend to be

leader firms, we spell out the following hypothesis:

• H1b (Firm Characteristics Hypothesis): Firms with large market capitalization, higher

degree of irreversibility, and high idiosyncratic volatility tend to be leader firms.

Reflexively, we also examine whether negative incentive leaders with a substantial de-

crease in Delta are the ones with small market capitalization and fewer tangible assets. The

hypothesis is as follows:

• H1c (Negative Incentive Leaders Hypothesis): Negative incentive leader firms tend to

be small and growth firms with few tangible assets.

The substantial Delta increases in leader firms intensify the industry competition and

influence their peer firms. It is well documented in the prior literature regarding CEOs’

concern on the threat of turnover (Murphy (2013), Kaplan and Minton (2012), Jenter and

7In our unreported empirical exploration, we find that the incentive spillover effect is driven by large butnot small incentive leaders and by value but not growth leaders.

11

Kanaan (2015), Dikolli et al. (2013), Dasgupta et al. (2014)). The turnover threat can come

from multiple sources, one of which is the firms’ bad market performance (e.g., Dasgupta

et al. (2014)). That is, firms’ bad market performance leads to a high turnover threat

for CEOs. In this study, peer firms’ bad market performance is generated by increasing

industry competition initiated by CEO Delta incentive leaders. To mitigate the threat of

turnover, peer CEOs must strive to improve their firms’ market performance. Peer CEOs’

effort in mitigating turnover threat is the foundation of CEO Delta incentive spillover. If

the effect of peer CEOs’ effort dominates the impact of product price cut from industry

competition, Delta increases in the incentive leaders give rise to a positive externality on

incentive peers’ market performance. Otherwise, a negative externality arises. To first

explore the sign of CEO Delta spillover on peer firms’ market performance, we spell out the

following hypothesis:

• H2a (Positive Externality Hypothesis): Delta incentive peers’ stock returns can be

positively predicted by the CEO Delta incentive leaders within the same industry.

We test the above hypothesis using multiple approaches in Section 5.2. However, we realize

that peer CEOs’ extra effort is not the only explanation for the potential positive externality

on peer firms’ market performance. The positive abnormal return of incentive peers can

also be attributed to investors’ category learning in incentive event industries (e.g., Mondria

(2010); Peng and Xiong (2006)), those industries with significant industry–average increases

in CEO Delta incentive. With limited attention, investors tend to process the information

on the industry-wide Delta increase rather than the information on the firm-specific Delta

increase. Thus, incentive peers are considered by investors the same as incentive leaders (in

the same industry) and also experience future abnormal returns as incentive leaders. If the

category learning hypothesis holds, the positive market performance of incentive peers comes

from investor attention but not CEOs’ effort in the firms’ operation. In contrast, a strong

increase in peer firms’ fundamentals refutes the category learning hypothesis and supports

12

the CEOs’ extra effort hypothesis. We formally spell out these two contrasting hypotheses

as follows:

• H3a (Category Learning Hypothesis): The positive abnormal returns of incentive peers

are due to investors’ category learning.

• H3b (Strong Fundamentals Hypothesis): The positive abnormal returns of incentive

peers are due to peer firms’ strong fundamentals.

This pair of hypotheses cannot be directly justified through abnormal returns. We test

hypothesis H3a by exploring peer firms’ investor attention and test H3b by looking at firm

fundamentals. Our Hypothesis H2a explores the relation between Delta and stock market

performance in the annual horizon. Behavioral explanations such as category learning can

hardly explain the long–horizon stock market abnormal returns. Therefore, our economic

intuition votes for the Hypothesis H3b.

Strong fundamentals can also come from different resources. At least three methods exist

for peer CEOs to generate strong fundamentals and thus create positive abnormal returns for

their firms. The first explanation is based on the technology spillover. If CEOs in incentive

leaders allocate extra efforts (Lilienfeld-Toal and Ruenzi (2014)) to their firms’ R&D invest-

ment (but not to the operating efficiency), the R&D investment can create new demand for

the industry and thus benefit other firms in the same industry (Jiang et al. (2015)). Second,

if leader firms’ CEOs allocate their efforts in improving their firms’ operating efficiency, the

industry competition increases and gives pressure to peer CEOs. Due to the industry com-

petitive pressure, incentive peer firms’ CEOs increase their efforts to match that of incentive

leaders’ CEOs and thus improve their firms’ market performance. Finally, the strong firm

fundamentals can also come from peer CEOs’ earnings management. We go one step fur-

ther to explore the exact channels for peer CEOs to allocate their efforts and increase firm

fundamentals.

The severe industry competitive pressure from incentive leaders can increase the CEOs’

13

propensity to conduct earnings management (e.g., Giroud and Mueller (2010), Giroud and

Mueller (2011), Lin et al. (2015)). However, the litigation risk from earnings management

dampens CEO career prospect in the labor market, suggesting that the more realistic tool

for CEOs to improve market performance is through exerting extra efforts on either firms’

operation or investment activities. We explore peer CEOs’ effort allocation by providing two

contrasting hypotheses:

• H4a (R&D Spillover Hypothesis): The positive abnormal returns and the strong fun-

damentals of incentive peers come from incentive leader and peer CEOs’ effort in

increasing R&D investment.

• H4b (Incentive Spillover Hypothesis): The positive abnormal returns and the strong

fundamentals of incentive peers come from peer CEOs’ effort in operating their firms

more efficiently and enhancing product differentiation.

The Hypotheses H4a and H4b state that peer CEOs respond positively to the negative

industry shock and the dramatic increase in leader CEOs’ Delta. A natural question from

the conventional wisdom such as Lin et al. (2015) is whether industry competition also drives

peer CEOs to conduct more earnings management. Since earnings management can hardly

boost stock market performance in relatively long horizon such as one year, the implication

from the conventional wisdom is that peer firms with higher propensity to conduct earnings

management would not experience positive abnormal return. We test the implication from

the conventional wisdom by examining the following hypothesis.

• H5: When corporate governance is strong, peer firm CEOs are more likely to put in

more effort and yield better performance.

Since higher earnings management is negatively associated with corporate governance. To

test this hypothesis, we use the Entrenchment Index (E-index) as the proxy for the corporate

governance (e.g., Bebchuk et al. (2008)). Higher magnitude of the E–index indicates weaker

14

corporate governance and higher propensity of earnings management. The empirical tests

for the above hypotheses are in Section 5.4. We will now turn to the empirical tests of our

hypotheses.

4 Data, Sample, and Summary Statistics

In this section, we first define Delta incentive leaders and peers. To better understand

the information contents of leaders and peers, we report the corresponding firm and CEO

characteristics. Based on these characteristics, we discuss who are incentive leaders and why

leaders choose to use more Delta incentives.

CEO compensation data for the sample period of 1992–2015 are from the Standard&Poor’s

Execucomp database for firms in the S&P 500, S&P Midcap 400, and S&P Smallcap 600.

The stock return and accounting data are from CRSP and COMPUSTAT, respectively. We

exclude financial (SIC 6000-6999) and utility (SIC 4900-4999) stocks.

4.1 CEO Incentives and Incentive Leaders & Peers

Our main independent variables are CEO Delta incentive leader and peer. Incentive leader is

the dummy variable indicating a firm has a significant increase in its CEO Delta incentive.

Incentive peer is the dummy variable indicating a firm has a strong economic link with

incentive leaders but no significant increase in CEO pay incentive. Before defining the

incentive leader and peer, we first follow exactly the same approach of Coles et al. (2006) to

construct the CEO Delta.

• Ln(Delta): Logarithm of one plus the dollar change in CEO wealth associated with a

1% change in the firm’s stock price (in $000s).

We then identify incentive “Leader” and “Peer” for each year during the period 1992–

2015. We classify firms into 50 industries following Hoberg and Phillips (2016) classifications.

15

An industry must have at least ten firms in a year to be included in the sample. To identify

Delta leaders and peers, we first define Delta increase events for industries. An industry has

a Delta increase event if its aggregate Delta growth rate is greater than 10% in a given year.

This accounts for approximately 35% of industry-year observations in the whole sample.

Then, within an identified industry year, we define Delta leaders as those firms that have

Delta growth rates ranked among the top 10% (or top 5, whichever has the larger number

of firms) in the industry. We define the remaining firms in the identified industry, excluding

those classified as Delta leaders during the previous three years, as peers. For the industry

years without a significant average increase in CEO Delta incentive, all firms are classified

as non-event firms.

• Industries with Delta increase events: Any industries experience aggregate Delta

growth rate more than 10% in a given year.

• Leader: A dummy equals to 1 if the firm has a Delta growth rates ranking among the

top 10% in an industry with Delta increase event.

• Peer: A dummy equals to 1 if the firm is in the industry with Delta increase events

and not leader firms. We exclude those classified as leaders during the previous three

years.

Thus, our sample contains three types of firms: incentive leaders, peers, and non-event

firms. To isolate the marginal impact of CEO incentive spillover, we also control for other

firm characteristics. These variables include firm’s own market capitalization (Ln(Size)), the

book-to-market ratio (B/M), R&D expenditure, sales growth, capital expenditure (Capex),

and the cumulative monthly stock returns during the previous 11 months from month t-1 to

month t-11. We define all firm characteristics in the Appendix.

Table 1 reports the summary statistics of sample firms from 1992 to 2015. The firm-year

sample contains 21,056 observations, including 851 observations for Delta incentive leaders

and 7,576 as incentive peers. The incentive leaders and peers account for almost 35% of the

16

full sample. Table 1 also presents the CEO and firm characteristics for all three types of firms:

incentive leaders, peers, and non-event firms. More importantly, we test the significance of

the differences in the mean characteristics of three types of firms.

First, the comparison in Table 1 reveals that consistent with our definition, incentive

leaders have a significantly higher Delta growth than that of the other two types of firms.

Specifically, the increase in the level of Delta (∆Delta) for leaders is more than 10 times

larger than that for peers (716.135 versus 68.948). Consequently, the Delta incentive in

leaders becomes 89.57% higher than peers (1437.397 versus 758.091) and 106.46% higher than

non-event firms (1437.397 versus 686.239). In contrast, the Delta growth difference between

peers and non-event firms is only marginally significant (60.753-(-98.318)=159.071). As a

consequence, the Delta incentive difference between peers and non-event firms is insignificant

(758.091 versus 686.239). Thus, the Delta increase events in industries are actually driven

by the small group of incentive leaders. Table 1 also indicates Delta incentive event firms

(leaders and peers) are firms with larger market capitalization than non-event firms.

Moreover, the comparison of firm characteristics across types of firms gives the prelim-

inary tests on the two explanations of the positive incentive spillover. Based on the first

explanation, if the positive incentive spillover is due to peer firms’ additional investment in

R&D, we expect a contemporaneous R&D increase in peers. However, Table 1 indicates that

there is no significant difference in R&D changes (∆R&D) across three types of firms (0.002

for leaders versus 0.002 for peers and 0.001 for non-event firms), casting doubt on the first

explanation.8

In contrast to firm investment, the managerial–driven operating efficiency and sales

growth for leaders are significantly higher than non-event firms. The high operating efficiency

of Delta incentive leaders is consistent with McConnell and Servaes (1990), Lilienfeld-Toal

and Ruenzi (2014), and McConnell et al. (2008): Delta incentive drives CEOs to focus on

running their firms more efficiently rather than investing more in R&D expenditure. More

8The levels of R&D and capital expenditure also have no significant difference across three types of firms.

17

importantly, Table 1 indicates that the CEOs in peers run their firms more efficiently than

CEOs in non-event firms even without significantly higher Delta incentive than CEOs in

non-event firms. We explore the driving force for CEOs in peer firms to run their firms more

efficiently. As argued in our second explanation, it is due to peer CEOs’ extra effort as a

response to the competitive pressure from leaders. To better test our second explanation,

we construct proxies for CEOs’ characteristics in Section 4.2.

4.2 CEOs’ Characteristics

The validation of the second explanation requires peer CEOs’ response to leaders’ com-

petitive pressure. Peer CEOs have heterogeneity which can generate different responses to

incentive leaders’ competitive pressure. To explore the impact of this heterogeneity on in-

centive spillover effect, we first construct four CEOs’ characteristics (managerial score, CEO

overconfidence, CEO age, and CEO tenure) and then interact them with the CEO peer

dummy.

Managerial score9 (Demerjian et al. (2012)), as a measure of CEOs’ extra effort, can

affect the degree of incentive spillover because it captures the peer CEOs’ effort to respond

to the competitive pressure from incentive leaders. CEOs’ personal characteristics such

as overconfidence and age, relates to their risk–taking behavior (e.g., Coles et al. (2006);

Graham et al. (2013); Holmstrom (1999); Serfling (2014)) and thus impacts CEOs’ action

to incentive leaders.

Table 1 presents the managerial score (MGscore) across different types of firms. Even

though leader and peer firms have no significant difference in managerial score, firms in

event industries (leader and peer) have significantly higher managerial scores than firms in

non-event industries, indicating that higher Delta incentive contracts are allocated to CEOs

with a higher level of effort (0.025 and 0.024 versus 0.009). More importantly, the average

9The construction of managerial score strictly follows Demerjian et al. (2012). The procedure is describedwith detail in the Appendix.

18

MGscore for peers is almost three times as large as that for non-event firms (0.024 versus

0.009). In other words, peer CEOs have more effort in response to the competitive pressures

than CEOs from non-event firms.

Based on our second explanation, CEOs with higher effort respond more efficiently to

the competitive pressure brought in by incentive leaders. If the second explanation holds,

we expect to see that the spillover effect concentrates in the group of peers with higher CEO

managerial score. We test this hypothesis in Section 3.

The theory of industrial organization predicts that under the competitive pressure from

leaders, peer CEOs allocate their extra effort in increasing their product differentiation.

We employ the industry similarity measure10 (Hoberg and Phillips (2016)) as the proxy for

product differentiation. Higher industry similarity means low product differentiation. A

negative relation between incentive peer dummy and future industry similarity supports the

prediction of industrial organization.

5 Empirical Results

5.1 Who are Incentive Leaders?

In this section, we explore who are incentive leaders and who are not. Through answering

who are not incentive leaders, we mitigate the concern that our findings are driven by good

industry prospects. Through testing the hypotheses H1a and H1b, we provide a plausible

explanation on who are incentive leaders.

We first identify the industries with substantial increases in Delta incentives within each

sample year. We find that the industries with Delta increase events vary a lot throughout

sample years and are not restricted to hot industries or R&D intensive industries such as

computer or business services. In contrast, cold industries such as Construction, Machinery,

and Chemicals frequently experience industry-wide increases in Delta incentives and thus

10The dataset is from Hoberg Data Library.

19

occupy a large fraction of the sample. In other words, the pattern of the Delta–increasing

event industries indicates that the selection of Delta increasing industries is independent of

industry prospects. Consequently, our results are immune to the concern that our findings

are driven by good industry prospects causing both higher CEO Delta and higher future

stock performance.

We find that the industries with Delta increase event spread out from hot to cold in-

dustries11. What is the commonality for these industries? We test the Hypothesis H1a by

exploring whether the incentive leader firms tend to be in industries experiencing negative

industry–level shocks. A natural proxy for industry–wide shock is the unexpected variations

of industry–level import tariffs (e.g., Feenstra, Romalis, and Schott (2002); Feenstra and

Romalis (2014); and Fresard (2010))12. Specifically, the import tariff cut triggers the signif-

icant intensification of competitive pressures from foreign rivals and is a good proxy for the

negative shock at the industry level.

To measure the magnitude of tariff cut at the industry level, we use the Census Bureau

imports data to get the tariff rate for each Fama–French 48 industry–year. According to

Fresard (2010), we define a dummy variable dramatic tariff cut (CUTt) for the year t which

equals to one if a negative change in the tariff rate is 3 times larger than its median change

and equals to 0 elsewhere. We find a large overlap of 26% between the tariff cut industries

and the Delta increase industries. The finding confirms the Hypothesis H1a and indicates

that firms within industries experiencing negative shocks tend to increase Delta.

A step further, for a certain industry under severe competitive pressure, firms within the

industry can either increase the CEO Delta in the first place or free ride on other firms’

Delta increase. What is the motivation for a small group of firms to increase Delta incentive

in the first place without waiting for their peers to react first?

To identify the commonalities among industry leaders and test the hypothesis H1b, we

11The list of industries with Delta–increasing event is reported in the Online Appendix.

12We thank the referee for his suggestion to use import tariff cut as the proxy of industry negative shock.

20

run a LOGIT regression of incentive leader dummy on a sequence of firm characteristics. We

restrict our sample to Delta increasing industries only. This sample helps to explore the dif-

ferences between incentive leaders and peers within Delta–event industries. The regression

results in the left panel of Table 2 indicate that within Delta increasing industries, firms

with larger market capitalization and more tangible assets tend to be the ones initiating

substantial Delta increase in the first place. Moreover, firm idiosyncratic volatility is posi-

tively associated with the leader–firm dummy variable. These findings are consistent with

the hypothesis H1b that firm size, irreversibility, and firm risk determine the propensity of

firms to initiate substantial Delta increases.

We then test the hypothesis H1c by exploring whether the same set of variables can also

determine the negative incentive leaders. The right panel of Table 2 reveals that the negative

incentive leaders have quite different information content since none of the firm characteristics

are significant determinants. The rejection of the hypothesis H1c is not surprising since the

magnitude of Delta cut is small comparing to the Delta increase. Specifically, the top decile

Delta decrease within Delta–decreasing industry is around 5%, hardly to be the consequence

of variation in firm characteristics.

To sum up, we find that the firms with dramatic increases in Delta concentrate in in-

dustries with negative industry–wide shocks and tend to be the ones with larger market

capitalization, a higher degree of irreversibility, and firm risk. In contrast, CEOs rarely

experience significantly negative incentive cuts. The minor cut in Delta can hardly have

economic meaningful consequence. However, one limitation of the exploration is that too

many underlying mechanisms drive certain firms to increase their managers’ incentive pay

drastically. The true overwhelming reason for dramatic Delta increase remains unclear with-

out a dimension reduction of various channels. A formal exploration of this issue remains

an interesting research question for the future.

21

5.2 Stock Market Performance of Delta Leaders and Peers

In this section, we test the hypothesis H2 by exploring the stock market performance of

Delta incentive leaders and that of peers in the year following their identification. We

employ both portfolio approach and regression approach.

First, we use the portfolio approach to explore the market performance of incentive

leaders and peers. In each calendar year, we identify incentive leaders and peers when

their Delta incentive is released in a specific month t. To ensure the incentive information is

publicly accessible, we skip one quarter after the identification and then form equal-weighted

and value-weighted portfolios for incentive leaders and peers, respectively. Specifically, after

identified at month t, the portfolios of leaders and peers remain the same for the subsequent

12 months from month t + 4 to month t + 15. For the equal–weighted (value–weighted)

portfolio, the average monthly portfolio returns for leaders and peers are 1.11% (1.02%) and

0.89% (0.81%), respectively. The monthly returns are all significantly different from zero at

the 1% level.

We then explore their abnormal market performance by computing Carhart (1997) four-

factor alpha based on the monthly time-series portfolio returns. Table 3 reports the abnormal

returns and factor loadings for incentive leader and peer portfolios. The Panel A of Table 3

reveals that the abnormal monthly return for equal-weighted incentive leader portfolio is

1.065% (i.e., an annualized abnormal return of 12.78%). The abnormal monthly return for

value–weighted incentive leader portfolio is 1.021% (i.e., an annualized abnormal return of

12.252%). The abnormal returns for the incentive leader portfolio confirm previous studies

such as Cooper et al. (2014) and Lilienfeld-Toal and Ruenzi (2014).

More importantly, the incentive peers also experience abnormal positive market perfor-

mance. Specifically, the abnormal monthly return for equal-weighted incentive peer portfolio

is 0.817% (i.e., an annualized abnormal return of 9.804%). Similarly, the abnormal return

for value-weighted portfolio is 0.794% (i.e., an annualized abnormal return of 9.528%). The

factor loadings for the two portfolios are not significantly different from each other. The

22

portfolio alpha for peer portfolio has an even higher statistical significance than that for

leader portfolio (t=3.68 for leaders versus t=5.59 for peers). Although without a significant

increase in the Delta incentive, incentive peer firms experience comparable positive stock

market performance as does the performance of incentive leaders.

To make sure that the results in Panel A of Table 3 are not driven by specific industry

classification, in our unreported results, we use Fama French 48 industry classification as

a robustness check. We find that the pattern in Table 3 still exists using different indus-

try classification. That is, both incentive leaders and peers experience significant monthly

abnormal returns.

One major concern is that the positive market performance of peers is not driven by

the Delta incentive spillover but by peers’ own Delta or other characteristics. To address

this concern, we employ regression analysis to examine the incentive spillover effect while

controlling for other variables. Table 4 presents the results of Fama-Macbeth regressions.

The dependent variable is the stock return during the 12-month holding period (from month

t + 4 to month t + 15) and the main independent variables are incentive leader and peer

dummy variables. To isolate the impact of leaders and peers, we include non-event firms in

the regression. We include firms’ own Delta incentive to mitigate the concern that incentive

peers’ positive abnormal return comes from their own Delta. If peer firms’ abnormal return

comes from incentive spillover, the explanatory power of peer firms’ Delta cannot absorb

that of the incentive peer dummy. We also control for the managerial score because one may

also concern that peer CEOs’ self–initiated effort gives rise to their firms’ positive market

performance. Moreover, we also include CEO V ega incentive to investigate whether the

incentive spillover effect is dominated by peer CEOs’ risk–taking incentives.

Following Cooper et al. (2014) and Lilienfeld-Toal and Ruenzi (2014), we also include in

the regression additional variables such as market capitalization, book-to-market ratio, and

momentum. We measure market capitalization as of June in year t. Book-to-market ratio

(B/M) is the ratio of book equity to the market equity at the end of fiscal year t − 1. The

23

momentum (MOM) is measured by the past 11-month cumulative returns.

Table 4 reports the times-series average coefficients and the corresponding t statistics

using Fama-MacBeth regression of returns on incentive leader and peer dummy variables and

other controls. Column (1) of Table 4 includes only leader and peer dummies as independent

variables. Column (2) includes firms’ own Delta incentive, V ega incentive, and other firm

characteristics. The coefficients of incentive peer dummy are all significant at 1% level under

both specifications, indicating that the positive abnormal returns of incentive peers cannot

be explained by their own Delta or any other firm characteristics.

One may also be concerned that if peer CEOs have a high level of work ethics, their

firms can still experience good market performance without spillovers from leader firms. To

mitigate this concern, we include the managerial score as the proxy for CEOs’ effort in our

regression. Column (3) reports the time-series average coefficients including the managerial

score. Although the managerial score has a significant explanatory power on stock returns,

the managerial score and incentive peer dummy neither drive out nor dominate each other.

The peer dummy even has a greater statistical significance when including the managerial

score. Column (4) shows that the explanatory power of peer dummy is not affected by

including industry similarity as an independent variable.

Across all specifications, the coefficients of incentive leader dummy are significant. How-

ever, when we include leader firms’ Delta and other control variables, the coefficients of

the leader dummy become marginally significant, suggesting that the leader firms’ Delta

and other firm characteristics absorb most of the explanatory power of the incentive leader

dummy. The incentive leader dummy, as an indicator of leaders’ sharp increase in Delta,

contains no additional information other than leader firms’ Delta. Moreover, CEO V ega

incentive is not significant in all the regression models, indicating that the abnormal returns

are not coming from CEOs’ risk–taking behavior.

In unreported results, we also include in the regressions sales growth, R&D growth,

industry momentum, and the change in Delta as control variables. The results are highly

24

comparable to the ones in Table 4. Because the explanatory power of these additional

variables is trivial in our sample, we ignore these additional results for brevity.

5.2.1 Mitigating The Endogeneity Concern

The biggest concern on our baseline results is not on adding additional control variables but

on the causality of leader firms’ impact on peer firms. To further strengthen the causality, we

use three approaches to mitigate the concern of endogeneity. First, we perform a propensity

score matching sample between leader and peer firms based on each industry–year pair. For

leader firms, we find their 1–to–1 matching sample based on the firms’ market capitalization,

profitability, the book–to–market ratio, and idiosyncratic risk. We redo the Fama–MacBeth

regressions of returns on the leader and peer dummies using propensity score matching

sample. The corresponding results are reported in Table 5. The results using propensity

score matching sample indicate that the explanatory power of incentive peer dummy is still

highly robust in determining stock returns but that of leader dummy is only marginally

significant.

Besides propensity score matching, we also use two–stage least square (2SLS) approach

to mitigate the endogeneity concern. In the first stage regressions, we regress CEO Delta

on a set of instrumental variables and get the predicted Delta. We then replace Delta by its

predicted value to rank leaders and peers within the industries with Delta increase event.

Specifically, our instrument variables include CEO gender and CEO MBA degree.13 We also

use another set of instrument variables for CEO compensation such as CEO age and CEO

tenure (e.g., Liu and Mauer (2011); Palia (2001)). To check whether the instruments are

weak, we compute the F–statistics of the reduced form equations in the first stage regressions

for Delta. Following standard convention such as Stock and Watson (2003), the computed F

statistics of 15.79 exceeds 10. We can conclude that our instrument variables are not weak.

13CEO gender is a dummy variable, which is equal to one for male CEO and zero for a female one. Ifa CEO achieves an MBA degree, then CEO MBA is equal to one and zero otherwise. Because we getinstrument variables from Boardex, the 2SLS sample period is from 1999 to 2015.

25

Table 5 presents the regression results of stock returns on the leader and peer dummies

identified by instrument variable predicted value. In all econometric specifications, the peer

dummy is significant at 1% level.

A more persuasive method to tackle the endogeneity concern is the natural experiment.

We use the adoption of SFAS 123(R) in 2004 as a natural experiment in the spirit of Chava

and Purnanandam (2010). This rule makes the option-based compensation less attractive

by requiring firms to expense the option compensation in the earnings statements. By

adopting the new expensing rule, firms switch option–based compensation to restricted stock-

based compensation for their CEOs. Thus, the structure (Delta and V ega) of the CEOs’

compensation experienced significant changes. In contrast, the adoption of the new option–

based compensation expensing rule is exogenous to the firms stock market return. We then

define the leading industry as the one with industry–wide aggregate Delta growth rate greater

than 20% from year 200114 to 2005. We define Delta leaders as those firms that have Delta

growth rates ranked among top 10% in industries with Delta–increasing events. Table A.1

reports the regression coefficients of stock returns on incentive leaders, peers, and non-event

firms during the SFAS 123(R) period. We find that the incentive peer dummy maintains its

strong explanatory power on future stock returns.

5.2.2 Alternative Leader Measure

As we documented, size effect matters for the leader firm classification. A further robustness

check on our main findings is to incorporate size effect in leader firm classification. Our

story is confirmed if the size–adjusted leader measure still leads to positive spillover on

peers’ market performance.

Specifically, for each industry with Delta increase event, we sort the firms into deciles

based on the product of the change in Delta and market capitalization (∆CEODelta×Size).

We then assign the top decile of the firms as leaders and the rest as peers. The regression

14 The discussion of the new regulation started from years 2002 to 2003. Thus we use the year 2001 hereto make a clear–cut.

26

results of stock returns on size–adjusted leader and peer dummies are reported in Table 6.

The results indicate that size–adjusted incentive leaders have a significant spillover effect on

peers.

To sum up, the results in Table 5 and Table 6 indicate that our results are immune from

endogeneity concern and robust to alternative measures.

5.3 Category Learning vs Strong Fundamentals

This section inspects the mechanism for incentive peer firms’ abnormal returns by testing

hypotheses H2a and H2b. To identify the exact channels, we explore the real effects of

incentive spillover on peers’ fundamentals. We use three measures of firm fundamentals by

going from the top to the bottom of the income statement. Sales are on the top of the

income statement. We use sales growth to capture the firms’ market expansion. Gross

profitability (Novy-Marx (2013)) and return on assets (ROA) incorporate the cost of goods

sold (COGS) and we use them to gauge firms’ operating performance. We use the Fama-

MacBeth regression approach. For each regression, the dependent variable is one of the three

fundamental measures in the subsequent year. The main independent variables are the Delta

incentive leader and peer dummies. The rest of the control variables are the corresponding

lagged fundamental measure, sales (Ln(Sales)), R&D, and capital expenditure (Capex).

Table 7 reports corresponding coefficients for the Fama-MacBeth regressions. For all of

the three fundamental measures, the coefficients on incentive peer dummy are significant even

controlling for peer firms’ own Delta and different firm characteristics. The results reveal

that the positive abnormal returns of incentive peers come from their strong fundamentals.

In contrast to that of incentive peer dummy, the coefficients of the incentive leader dummy

lose its explanatory power on fundamentals when we include incentive leaders’ own Delta.

This is consistent with portfolio sorts results for incentive leaders and explains the less

robust positive abnormal returns of incentive leaders. Moreover, we find that CEO V ega

incentive has only marginal explanatory power and even switches signs in explaining the firm

27

fundamentals. This is not surprising because the spillover effect on peer firms is through the

operating performance but not through corporate risk–taking.

We then test whether the positive abnormal returns of incentive peers come from in-

vestors’ category learning. To capture the degree of category learning, we use proxies of

investor attention. Investor attention is inversely related to category learning. With more

attention allocated to one specific firm, more firm–specific information can be incorporated

in the stock price. Following the previous literature (Chichernea et al. (2015); Hou and

Moskowitz (2005); Jiang et al. (2015); Lehavy and Sloan (2005)), we use analyst coverage,

institutional ownership, and advertising expense as proxies for investor attention. Stocks

with high analyst coverage, high institutional ownership, and high advertising expense are

the ones with high investor attention and thus a low degree of category learning.

We divide the incentive peers into two groups: peers with a high degree (above the me-

dian) of investors’ category learning and ones with a low degree (below the median) during

the identification year. If the category learning hypothesis holds, the high category learning

peers would experience significantly higher abnormal returns than the low category learning

peers. We find positive abnormal returns exist in both the high and low category learning

groups. Specifically, for the equal-weighted incentive peer portfolios, the high category learn-

ing peer portfolio experiences a positive monthly abnormal return of 0.83%, while the low

category learning peer portfolio has a monthly abnormal return of 0.79%. The difference in

their abnormal returns is not significant. There is also no significant difference in the future

fundamentals of the high and low category learning incentive peer groups. Moreover, we run

Fama-MacBeth regressions of future investor attention on leaders and peers. The results

reveal that there is no significant increase in incentive peers’ future investor inattention.

These results indicate that the positive abnormal returns of incentive peers is unlikely

due to investors’ category learning but comes from the incentive peers’ strong fundamentals.

The natural next question is: what leads to the strong fundamentals of incentive peers?

28

5.4 Incentive Spillover vs Technology Spillover

Two alternative explanations exist for the peer firms’ strong fundamentals: R&D spillover

and incentive spillover. To test the R&D spillover hypothesis, we divide the incentive peer

firms into two groups in the identification year: the high R&D incentive peers with higher

than median R&D expenditure in the peer identification year and the low R&D incentive

peers. If CEOs’ extra effort relates to higher R&D expenditure, the abnormal returns,

and fundamentals of high and low R&D incentive peers will differ. However, we find no

specific differences in abnormal returns and fundamentals of high and low R&D incentive

peer portfolios. Moreover, we report in the Table 8 the average coefficients of R&D and

investment measures on the leader, peer dummies, and other control variables. We find that

neither incentive leaders nor peers experience any significant increase in R&D investment in

subsequent years, consistent with the results in Table 1. That evidence contradicts the R&D

spillover hypothesis.

We then turn to the tests for incentive spillover hypothesis. CEOs in incentive peers

can respond to the competitive pressure from incentive leaders through two channels: (i).

Exert extra efforts in running their firms more efficiently; (ii). Increase their firms’ product

differentiation. We use managerial scores (MGscore) to capture peer CEOs’ effort in improv-

ing their firms’ operating efficiency because MGscore gauges the impact of CEOs on firms’

sales/cost ratio maximization. If the incentive spillover hypothesis holds, peer CEOs with

higher MGscore can respond more efficiently to the competitive pressure than those with low

MGscores, consequently experiencing significantly better fundamentals and higher abnormal

returns. To measure product differentiation, we use firm-level industry similarity (IndSim)

following Hoberg and Phillips (2016). Higher industry similarity indicates lower product

differentiation. The incentive spillover hypothesis implies that CEOs in peer firms decrease

firms’ product similarity to alleviate competitive pressure. Consequently, low industry simi-

larity peers can experience better fundamentals and higher abnormal returns than can high

industry similarity peers.

29

One common advantage of using MGscore and IndSim to test the incentive spillover

hypothesis is that both measures incorporate managerial discretion. MGscore is driven

by managerial discretion because it filters out the impact of firm characteristics on the

operating efficiency and leaves only the component of operating efficiency contributing to

CEOs’ effort. IndSim is determined by managerial discretion by considering the information

in the Management Discussion and Analysis part (MD&A) which contains managers’ own

opinions regarding product differentiation. Therefore, both measures can gauge CEOs’ effort-

related improvement in firms’ operating efficiency. We test the two hypotheses by using both

portfolio approach and regression approach.

First, we use the portfolio approach to detect the impact of MGscore on peer firms’

abnormal returns. Similar to the methods in Table 3, in each calendar year, we identify

incentive peers when their Delta incentives are released in a specific month t. We then

divide incentive peers into two groups: high MGscore and low MGscore. To ensure the

incentive information is publicly accessible, we skip one quarter after the identification and

then form equal-weighted and value-weighted portfolios for high and low MGscore incentive

peer groups, respectively. After the identification at month t, we carry the portfolios of high

and low MGscore for 12 months from month t+ 4 to month t+ 15 and calculate the return

difference between the two portfolios.

Panel A and Panel B from Table 9 present the corresponding results. For the equal-

weighted high MGscore peer portfolio, the monthly average abnormal return shown in Panel

A is 1.187% ((i.e. an annualized abnormal return of 14.244%). In contrast, the monthly

average alpha for equal-weighted low MGscore peer portfolio is only 0.586% (i.e. annualized

abnormal return of 7.032%). The average monthly return for the high MGscore portfolio

is more than twice as much as that for the low MGscore portfolio. The monthly average

abnormal return difference between the high and low MGscore peer portfolios is 0.601% (i.e.

an annualized return of 7.212%) with the t statistics of 4.92. Panel B indicates that the

value-weighted high and low MGscore peer portfolios yield similar results. Specifically, the

30

magnitude of value-weighted monthly average abnormal return difference is 0.612%, similar

to that of equal-weighted difference.

To examine whether the explanatory power of MGscore on peer portfolio returns is inde-

pendent of the explanatory power of other firm and CEO characteristics, we use regression

analysis by adding additional control variables. Following Fama and MacBeth (1973), we

perform yearly cross-sectional regressions and then compute the time-series average of the

coefficients. The dependent variable is the monthly stock returns from month t + 4 to

month t+15. The main independent variables are HighMGscorePeer and LowMGscorePeer.

HighMGscorePeer is a dummy variable for peers with an above-median managerial score.

LowMGscorePeer is a dummy variable for peers with a below-median managerial score.

Based on incentive spillover hypothesis H2b and the baseline results in Table 9, we expect

HighMGscorePeer has a much stronger explanatory power than that of LowMGscorePeer.

Table 10 presents the average coefficients of Fama and MacBeth regressions with the cor-

responding t statistics. The control variables are defined the same as those in Table 4. Con-

sistent with the portfolio sort results and incentive spillover hypothesis, HighMGscorePeer

has a strong positive explanatory power on peer firms’ future stock returns. In contrast,

LowMGscorePeer has an insignificant influence on peer firms’ future returns when control-

ling for additional variables. These results augment the findings in Tables 4 and 7 by re-

vealing that peer firms’ abnormal returns come from peer CEOs’ extra effort. In other

words, peer firms with CEOs exerting extra efforts to firms perform better to competitive

pressure brought in by incentive leaders. To reinforce this argument, we also test whether

HighMGscorePeer and LowMGscorePeer have real effects on firm fundamentals.

Panel A of Table 11 reports the results of Fama and MacBeth regressions of firm fun-

damentals on HighMGscorePeer and LowMGscorePeer. Again, we still use sales growth,

gross profitability, and return on assets to capture various aspects of fundamentals. Across

all specifications, HighMGscorePeer has a strong positive relationship with a firm’s future

fundamentals even controlling for lagged one period’s firm fundamentals (AutoLag). Thus,

31

peer CEOs’ high effort in firms’ operation is the origin of peer firms’ strong fundamentals

and positive abnormal returns.

Although we set the release date for CEO incentive to be three months ahead of the date

we construct MGscore, one may still concern that the time length is not sufficient for CEOs

to exert their extra efforts in firms’ operation. To mitigate this concern, we explore peer

CEOs’ future effort, that is, their MGscore in the next year. We perform the Fama-MacBeth

regression of CEOs’ future MGscore on Delta incentive leader, peer dummies and also other

control variables. Table 12 reveals that peer CEOs exert significantly more efforts in the

future even after controlling for various firm characteristics. The pattern in Table 12 is also

consistent with Gorton et al. (2014) and Lilienfeld-Toal and Ruenzi (2014), implying market

prices cannot fully reflect the future effort of CEOs.

Since the proxy of managerial effort is constructed as a maximization of the sales and

cost ratio, a natural and interesting question is whether peer CEOs exert more efforts in

future sales promotion or in cost minimization15. A “sales–oriented” CEO may seek to

promote the sales by differentiating product to command a price premium, while a “cost-

minimization” CEO focuses on reducing expenditures on unnecessary or inefficient processes.

To test whether CEOs exert extra efforts in sales maximization or cost minimization, we split

our sample into the high and low sales growth (cost growth) samples. Separately for the high

and low sales growth (cost growth) samples, we run regressions of MGscore on the leader,

peer dummy, and other control variables. The corresponding results in Table A.2 indicate

that the causality from peer dummy to MGscore is more pronounced in high sales–growth

and low cost–growth samples. The findings in Table A.2 reveal that CEOs exert efforts in

both aspects of the operating efficiency which are sales maximization and cost minimization.

Even more importantly, documenting the exact channels for CEOs’ effort also mitigate the

concern that peer firm reaction to leader firm Delta increase is via cutting product price

to remain competitive. The reason is that firms with CEOs exerting efforts in sales and

15We are very grateful to the referee for suggesting the test and the testing hypotheses.

32

cost efficiency are unlikely to be the ones using price cut to maintain competitiveness in the

industry.

We then explore the impact of product differentiation on peer firms’ abnormal returns

by using the exact same approach as that for MGscore. Again, we first discuss the portfolio

approach. In each calendar year after we identify incentive peers, we divide peers into two

groups: peers with industry similarity (IndSim) above- and below-median. We skip one

quarter after the identification of peers and then form equal-weighted and value-weighted

portfolios for high and low IndSim incentive peer groups, respectively. After the identification

at month t, we carry the portfolios of high and low MGscore for 12 months from month t+ 4

to month t+ 15 and calculate the return difference between the two portfolios.

Panel C and Panel D of Table 9 report the portfolio sort results for product differentiation.

For equal-weighted high IndSim peer portfolio, the monthly average abnormal return is

0.676%. In contrast, the monthly average alpha for equal-weighted low IndSim peer portfolio

is 1.157%. That is, the equal-weighted portfolio of low IndSim (high product differentiation)

outperforms that of high IndSim (low product differentiation) of 0.481% with the t statistics

of 3.56. The results are consistent with our incentive spillover hypothesis that incentive peer

CEOs improve their firms’ market performance by increasing product differentiation.

Panel B of Table 10 confirms the results in Table 9 by using Fama-MacBeth regressions

of returns on high and low IndSim peer dummies. It is shown in Table 10 that the positive

externality on peers’ market performance exists only in the low industry similarity incentive

peer group. Table 10 strengthens our hypothesis by showing that only the low IndSim

incentive peer group has a real effect on firm fundamentals.

Table 12 explores the relationship between the incentive peer dummy and future industry

similarity by performing the Fama and MacBeth regression of future industry similarity on

incentive leader and peer dummies. We find peer firms’ future industry similarity significantly

decreases, implying CEOs in peer firms increase their product differentiation to survive in

industry competition.

33

Finally, we test the Hypothesis H5 by examining the relationship between E–Index and

the spillover effect. Specifically, we split the full sample for each year into two sub–samples

with low and high E–Index. We run regressions of future abnormal returns on leader dum-

my, peer dummy, and control variables separately for high and low E–Index samples. Our

untabulated results confirm the Hypothesis H5 by showing that the peer firms’ positive ab-

normal return is more pronounced in the low E–Index (strong corporate governance) sample.

Similarly, we find that CEOs in firms with strong corporate governance tend to exert more

efforts in firms’ operation.

To sum up, the results in this section reject the technology spillover hypothesis H3a and

support the incentive spillover hypothesis H3b. Peer firms’ CEOs respond to competitive

pressure from incentive leaders by exerting extra efforts to improve their firms’ operating

efficiency and product differentiation.

5.5 Turnover Threat, CEO Characteristics, and Incentive Spillover

The previous section proves that peer CEOs’ extra effort is the foundation for the positive

Delta incentive spillover. Going one step further, the motivation hidden below their extra

effort is to mitigate the turnover risk coming from increased competitive pressure generated

by incentive leaders. Larger turnover threat stimulates peer CEOs to exert more efforts in

running their firms. This section explores the hidden motivation of peer CEOs’ extra effort

by testing the relationship between the change in CEO turnover risk and the magnitude of

positive incentive spillover.

Since the ex-ante change in CEO turnover risk cannot be directly estimated, we use three

CEO characteristics to indirectly capture it in different aspects: CEOs’ age, CEOs’ tenure,

and CEOs’ overconfidence. Campbell et al. (2010) among others demonstrate that younger

CEOs and CEOs with shorter tenure have relatively less entrenchment, thus experiencing

higher turnover threat when market performance falls. We expect a negative relation be-

tween CEO turnover risk and CEO age and tenure. That is, to protect their position, peer

34

CEOs who are younger and with shorter tenure will exert more efforts to increase their

firms’ operation and thus to improve their firms’ market performance. Moreover, Dikolli

et al. (2013) shows that overconfidence can lead CEOs to choose the first best level of invest-

ment, thus giving a better response to the industry competitive pressure. We then expect

a positive relation between CEO overconfidence and peer firms’ fundamentals (and market

performance).

We use regression approach for our analysis. Table A.3 examines the effect of CEO

characteristics on the firm fundamentals of incentive peers. The regression specification is

similar to that used in Table 11. The dependent variable is the firm fundamentals in the

subsequent year after the identification of incentive leaders and peers. As in Table 11, we

include control variables such as the lagged one period firm fundamentals, size, the book-to-

market ratio (B/M), past returns (MOM), sales, incentive leader dummy, and CEO Delta.

Unlike Table 11, we include two dummy variables – LowPeer and HighPeer – to classify CEO

characteristics for peers. LowPeer represents peer firms with CEOs of above-median age,

above-median length of tenure, and below-median confidence level. HighPeer represents peer

firms with CEOs of below-median age, above-median length of tenure, and below-median

confidence level.

The results in Table A.3 show that, across all specifications, the coefficients for HighPeer

are significantly larger than those for LowPeer. For instance, when we measure HighPeer as

peers with younger CEOs, the coefficients on HighPeer are 0.005, 0.011, 0.041 respectively.

The corresponding t statistics are all significant at 1% level. In contrast, when we measure

LowPeer as peers with older CEOs, the coefficients are 0.004, 0.001, and 0.006 respectively

with insignificant t statistics. The differences in coefficients (0.001, 0.010, 0.035) are all

significant at 1% level. Therefore, the peers with younger CEOs, shorter tenures, and higher

confidence level outperform those with older CEOs, longer tenure, and lower confidence level

in operating performance. These results are consistent with our hypothesis that peer firms

with CEOs facing higher turnover threat experience more positive externality from incentive

35

leaders.

5.6 Robustness Checks

In this section, we perform multiple robustness checks by considering two alternative methods

to define Delta incentive leaders and peers.

First, when we define the Delta incentive increase event industry, we use 15% (instead of

10%) aggregate Delta growth rate in a given year for industries as the absolute cutoff. We

then define incentive leaders and peers as before. Second, instead of using absolute cutoff

for Delta increase industry, we define an industry to be a Delta increase industry if, in a

given year, its aggregate Delta increase is ranked among the top six in the 50 industries.

The incentive leaders and peers are defined as before.

We perform portfolio sorts, stock return regressions, and operating performance regres-

sions with the sample created using these two alternative methods. The results are even

stronger than those presented in Tables 3 to A.3. We only describe the tests and results

here without presenting the tabulated results for brevity. The tabulated results are available

upon request.

6 Conclusions

The impact of Delta incentive increase in one firm goes beyond the firm itself. This study

documents that a small group of firms‘ increases in their CEO Delta incentive results in

their peer firms’ improvement in the market performance. Even though prior literature

argues that one firm’s Delta incentive increase has a negative externality on its peer firms’

corporate governance, our findings suggest the increase has a positive externality on peer

firms’ market performance.

We find the positive externality comes from CEOs’ effort spillover. Specifically, the

incentive increase in leader firms stimulates their CEOs to exert more efforts to run their

36

firms more efficiently and increase the industry competition. The increased competitive

pressure consequently raises up the turnover threat for peer firms’ CEOs. To mitigate the

threat of turnover, peer CEOs also exert more efforts to improve the operating efficiency

and product differentiation. Peer CEOs’ effort results in peer firms’ strong fundamentals

and good market performance. Further, we find evidence that peer CEOs’ effort level is

related to the degree of their turnover threat. More severe turnover threat stimulates more

effort from CEOs and thus stronger firm fundamentals.

37

References

Acharya, Viral V, and Paolo F Volpin, 2009, Corporate governance externalities, Review of

Finance 14, 1–33.

Albuquerque, Ana M, George Papadakis, and Peter D Wysocki, 2014, The impact of risk on

ceo equity incentives: evidence from customer concentration .

Bebchuk, Lucian, Alma Cohen, and Allen Ferrell, 2008, What matters in corporate gover-

nance?, The Review of financial studies 22, 783–827.

Campbell, Timothy Colin, Michael F Gallmeyer, Shane A Johnson, Jessica Rutherford, and

Brooke Stanley, 2010, CEO optimism and forced turnover, Journal of Financial Economics

101, 695–712.

Carhart, Mark M, 1997, On persistence in mutual fund performance, The Journal of finance

52, 57–82.

Chamberlin, Edward Hastings, et al., 1933, The theory of monopolistic competition, volume 6

(JSTOR).

Chava, Sudheer, and Amiyatosh Purnanandam, 2010, Ceos versus cfos: Incentives and cor-

porate policies, Journal of Financial Economics 97, 263–278.

Chichernea, Doina C, Michael F Ferguson, and Haimanot Kassa, 2015, Idiosyncratic risk,

investor base, and returns, Financial Management 44, 267–293.

Coles, Jeff, Naveen Daniel, and Lalitha Naveen, 2006, Managerial incentives and risk taking,

Journal of Financial Economics 79, 431–468.

Cooper, Michael J., Huseyin Gulen, and Raghavendra P. Rau, 2014, Performance for pay?

The relation between CEO incentive compensation and future stock price performance,

University of Utah working paper.

38

Core, John, and Wayne Guay, 1999, The use of equity grants to manage optimal equity

incentive levels , Journal of Accounting and Economics 28, 151–184.

Core, John E, Wayne R Guay, and David F Larcker, 2003, Executive equity compensation

and incentives: A survey, Social Science Electronic Publishing 9, 27–50.

Dasgupta, Sudipto, Xi Li, and Yan Albert Wang, 2014, Product market competition shocks,

firm performance, and CEO turnover, Firm Performance, and CEO Turnover (November

30, 2014) .

Demerjian, Peter R, Baruch Lev, Melissa F Lewis, and Sarah E McVay, 2012, Managerial

ability and earnings quality, The Accounting Review 88, 463–498.

Dikolli, Shane S, William J Mayew, and Dhananjay Nanda, 2013, CEO tenure and the

performance-turnover relation, Review of Accounting Studies 19, 281–327.

Dixit, Avinash, 1986, Comparative statics for oligopoly, International Economic Review

107–122.

Fama, Eugene F, and James D MacBeth, 1973, Risk, return, and equilibrium: Empirical

tests, Journal of political economy 81, 607–636.

Feenstra, Robert C, and John Romalis, 2014, International prices and endogenous quality,

The Quarterly Journal of Economics 129, 477–527.

Feenstra, Robert C, John Romalis, and Peter K Schott, 2002, US imports, exports, and tariff

data, 1989-2001, Technical report, National bureau of economic research.

Fresard, Laurent, 2010, Financial strength and product market behavior: The real effects of

corporate cash holdings, The Journal of finance 65, 1097–1122.

Giroud, Xavier, and Holger M Mueller, 2010, Does corporate governance matter in compet-

itive industries?, Journal of Financial Economics 95, 312–331.

39

Giroud, Xavier, and Holger M Mueller, 2011, Corporate governance, product market com-

petition, and equity prices, The Journal of Finance 66, 563–600.

Gorton, Gary, Ping He, and Lixin Huang, 2014, Agency-based asset pricing, Journal of

Economic Theory 149, 311–349.

Graham, John R., Campbell R. Harvey, and Manju Puri, 2013, Managerial attitudes and

corporate actions, Journal of Financial Economics 109, 103–121.

Grossman, Gene M, and Carl Shapiro, 1984, Informative advertising with differentiated

products, The Review of Economic Studies 51, 63–81.

Guay, Wayne R., 1999, The sensitivity of CEO wealth to equity risk: An analysis of the

magnitude and determinants, Journal of Financial Economics 51, 43–71.

Hall, Bronwyn H, 1993, The stock market’s valuation of r&d investment during the 1980’s,

The American Economic Review 83, 259–264.

Hoberg, Gerard, and Gordon Phillips, 2010, Real and financial industry booms and busts,

Journal of Finance 65, 4586.

Hoberg, Gerard, and Gordon Phillips, 2016, Text-based network industries and endogenous

product differentiation, Journal of Political Economy 124, 1423–1465.

Hoberg, Gerard, Gordon Phillips, and Nagpurnanand Prabhala, 2014, Product market

threats, payouts, and financial flexibility, The Journal of Finance 69, 293–324.

Holmstrom, Bengt, 1999, Managerial incentive problems: A dynamic perspective, The Re-

view of Economic Studies 66, 169–182.

Hou, Kewei, and Tobias J Moskowitz, 2005, Market frictions, price delay, and the cross-

section of expected returns, Review of Financial Studies 18, 981–1020.

40

Jenter, Dirk, and Fadi Kanaan, 2015, CEO turnover and relative performance evaluation,

Journal of Finance 70, 2155–2184.

Jiang, Yi, Yiming Qian, and Tong Yao, 2015, R&d spillover and predictable returns, Review

of Finance 20, 1769–1797.

Kale, Jayant R, Ebru Reis, and Anand Venkateswaran, 2009, Rank-order tournaments and

incentive alignment: The effect on firm performance, The Journal of Finance 64, 1479–

1512.

Kaplan, Steven N, and Bernadette A Minton, 2012, How has CEO turnover changed, Inter-

national Review of Finance 12, 57–87.

Lehavy, Reuven, and Richard G Sloan, 2005, Investor recognition and stock returns, Review

of Accounting Studies 13, 327–361.

Lilienfeld-Toal, Ulf Von, and Stefan Ruenzi, 2014, CEO ownership, stock market perfor-

mance, and managerial discretion, The Journal of Finance 69, 1013–1050.

Lin, Chen, Micah S Officer, and Xintong Zhan, 2015, Does competition affect earnings

management? evidence from a natural experiment .

Liu, Yixin, and David C. Mauer, 2011, Corporate cash holdings and CEO compensation

incentives, Journal of Financial Economics 102, 183–198.

McConnell, John J., and Henri Servaes, 1990, Additional evidence on equity ownership and

corporate value, Journal of Financial Economics 27, 595–612.

McConnell, John J., Henri Servaes, and Karl V. Lins, 2008, Changes in insider ownership

and changes in the market value of the firm, Journal of Corporate Finance 14, 92–106.

Miller, Janice S, Robert M Wiseman, and Luis R Gomez-Mejia, 2002, The fit between ceo

compensation design and firm risk, Academy of Management Journal 45, 745–756.

41

Mondria, Jordi, 2010, Portfolio choice, attention allocation, and price comovement, Journal

of Economic Theory 145, 1837–1864.

Murphy, Kevin J, 2013, Executive compensation: Where we are, and how we got there,

Handbook of The Economics of Finance 2, 211–356.

Newey, Whitney K., and Kenneth D. West, 1987, Hypothesis testing with efficient method

of moments estimation, International Economic Review 777–787.

Novy-Marx, Robert, 2013, The other side of value: The gross profitability premium, Journal

of Financial Economics 108, 1–28.

Palia, Darius, 2001, The endogeneity of managerial compensation in firm valuation: A solu-

tion, Review of Financial Studies 14, 735–764.

Peng, Lin, and Wei Xiong, 2006, Investor attention, overconfidence and category learning,

Journal of Financial Economics 80, 563–602.

Richardson, Scott, 2006, Over-investment of free cash flow, Review of accounting studies 11,

159–189.

Sarkar, Salil., Humnath. Panta, and Sanjiv Sabherwal., 2014, The effects of idiosyncratic

risk on ceo incentives, Southwest Finance Association Annual Meeting Working Paper .

Serfling, Matthew A., 2014, CEO age and the riskiness of corporate policies, Journal of

Corporate Finance 25, 251–273.

Stock, James H, and Mark W Watson, 2003, Has the business cycle changed? evidence and

explanations, 9–56 (Federal Reserve Bank of Kansas City Kansas City).

Tirole, Jean, 2010, The Theory of Corporate Finance (Princeton University Press).

Zhang, Lu, 2005, The value premium, The Journal of Finance 60, 67–103.

42

Table 1: Summary Statistics

This table presents the summary statistics of sample firms for the period from 1992 to 2015.Panel A reports the summary statistics of CEO characteristics. Delta is the CEO Deltaincentive in year t. ∆Delta is the growth in Delta from year t− 1 to year t. MGscore is themanagerial score capturing CEOs’ extra effort. IndSim is the industry similarity as definedin Hoberg and Phillips (2016). Confidence is the dummy variable which equals to one if theCEO is overconfident and equals to zero elsewhere. Panel B reports the summary statisticsof firm characteristics. GP (t + 1) is the firm’s gross profitability in year t + 1. ROA(t + 1)is the return on asset in year t + 1. B/M is the book–to–market equity. Size is the marketcapitalization. ∆Sales is the sales growth from year t− 1 to year t. Ln(Sales) is the naturallogarithm of sales in year t. R&D is the research and development expenditure divided bytotal assets. Columns (1) through (3) report the summary statistics for leader, peer, andnon-event firms, respectively. Columns (4) through (6) report the differences in mean forthe three types of firms in the sample. The t statistics for the differences in mean are inparentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

(1) Leader (2) Peer (3) Non-Event (1)–(2) (1)–(3) (2)–(3)

Panel A: CEO Characteristics

Delta 1437.397 758.091 686.239 679.306*** 751.158*** 71.852∆Delta 719.574 60.753 -98.318 658.821*** 817.892*** 159.071***MGscore 0.025 0.024 0.009 0.001 0.016*** 0.015***IndSim 6.658 5.542 5.855 1.116*** 0.803*** -0.313

Confidence 0.549 0.472 0.284 0.077 0.265 0.188***

Panel B: Firm Characteristics

GP(t+1) 0.3 0.259 0.283 0.041 0.017 -0.024ROA(t+1) 0.06 0.045 0.042 0.015 0.018 0.003

B/M 0.473 0.489 0.506 -0.016 -0.033 -0.017Size 9192.219 7563.083 7400.689 1629.136*** 1791.53*** 162.394***

∆Sales 0.173 0.123 0.067 0.05*** 0.106*** 0.056***Ln(Sales) 7.878 7.619 7.767 0.259 0.111 -0.148

R&D 0.017 0.014 0.013 0.003 0.004 0.001

43

Table 2: Logit Regressions of Positive and Negative leaders

This table reports the coefficients and corresponding Wald Chi − Square statistics of logitregressions that examine positive and negative leaders to incentive leaders and peers. Thedependent variable is positive and negative leader dummies. Positive Leader is the dummyvariable for CEO Delta incentive leaders. Negative Leader is the dummy variable for firmswith top decile negative CEO Delta movements in industries ranked top decile CEO Deltadrop. Ln(Delta) is the natural logarithm of one plus CEO Delta incentive. Ln(Vega) is thenatural logarithm of one plus CEO V ega incentive. Ln(Tenure) is the natural logarithm of1 plus CEO tenure. Ln(Age) is the natural logarithm of 1 plus CEO Age. Ln(Size) is thenatural logarithm of market capitalization. MtB is the market to book ratio. Over Inv isthe over-investment measure used by Richardson (2006). R&D is research and developmentexpenditure scaled by total assets. Hard Assets is the property, plant and equipment scaledby total assets. Capex is capital expenditure scaled by total assets. IVOL is a firm’sidiosyncratic risk during past 60 months. ***, **, and * denote significance at the 1%, 5%,and 10% levels, respectively.

Positive Leader Negative Leader

Intercept -1.648 -0.427 3.664 6.669(0.54) (0.03) (2.05) (5.73)***

Ln(Delta) -0.339 -0.366 0.453 0.44(40.98)*** (43.87)*** (70.93)*** (54.49)***

Ln(Vega) -0.009 -0.081 -0.04 -0.079(0.04) (2.99) (0.77) (2.47)

Ln(Tenure) 0.214 0.227 -0.633 -0.635(6.24)*** (6.33)*** (46.86)*** (40.19)***

Ln(Age) 0.527 0.641 -2.064 -2.283(4.88)*** (5.23)*** (10.27)*** (10.8)***

Ln(Size) 0.175 0.27 -0.064 -0.098(12.87)*** (25.53)*** (0.98) (1.09)

MtB 0.132 0.115 -0.102 -0.163(1.42) (1.16) (1.11) (1.5)

Over Inv -0.545 -0.58 -0.134 -0.173(6.09)*** (6.47)*** (0.11) (0.15)

R&D -0.05 -0.038 -0.199 -0.22(5.09)*** (2.95) (0.71) (0.79)

Hard Assets 0.289 0.279 -0.488 -0.745(6.28)*** (5.78)*** (1.22) (1.83)

Capex 0.623 0.467 -0.075 -0.534(6.34)*** (6.01)*** (0.02) (0.9)

IVOL 0.876 0.758 0.479 0.663(6.04)*** (6.01)*** (1.73) (0.94)

44

Table 3: Performance of the Portfolios of Leaders and Peers

This table reports the abnormal returns for Hoberg–Philip 50 Industries equal- and value-weighed portfolios of incentive leaders and peers. To form portfolios for incentive leaders andpeers, we skip one quarter after their identification. Specifically, after identified at montht, the portfolios of leaders and peers remain the same for the subsequent 12 months frommonth t + 4 to month t + 15. The abnormal returns are estimated using Carhart (1997)four-factor model. MKTRF, SMB, HML, and UMD are the market excess return, the size,book-to-market, and momentum factors from Kenneth French’s website. The Newey andWest (1987) adjusted t statistics with six lags are reported in the parentheses. ***, **, and* denote significance at the 1%, 5%, and 10% levels, respectively.

Equal–Weighted Value–WeightedLeader Peer Leader Peer

Alpha 1.065 0.817 1.021 0.794(3.68)*** (5.59)*** (3.54)*** (5.51)***

MKTRF 1.087 0.988 1.088 0.983(16.55)*** (29.36)*** (16.63)*** (29.65)***

SMB 0.414 0.344 0.382 0.307(4.97)*** (8.01)*** (4.61)*** (7.26)***

HML 0.794 0.641 0.769 0.625(8.98)*** (14.06)*** (8.72)*** (13.91)***

UMD -0.167 -0.131 -0.165 -0.125(-3.18)*** (-4.81)*** (-3.15)*** (-4.69)***

45

Table 4: Fama-MacBeth Regressions of Stock Returns

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine stock returns to incentive leaders, peers and non-event firms. Thedependent variable is the annual stock returns from month t+ 4 to month t+ 15. The mainindependent variables are leader and peer. Leader is the dummy variable for CEO Deltaincentive leaders. Peer is the dummy variable for incentive peers. Ln(Delta) is the naturallogarithm of one plus CEO Delta incentive. Ln(Vega) is the natural logarithm of one plusCEO V ega incentive. Ln(Size) is the natural logarithm of market capitalization. B/M isthe book to market ratio. MOM is the cumulative monthly return of past 11 months up toone month ago. MGscore is the managerial score. IndSim is the industry similarity. Thereported coefficients are the time-series average of cross-sectional regression coefficients. Thecorresponding Newey and West (1987) adjusted t statistics with six lags are reported in theparentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

(1) (2) (3) (4)

Intercept 1.309 1.941 1.182 1.714(3.68)*** (1.49) (0.84) (1.24)

Leader 0.413 0.455 0.491 0.469(1.76)* (1.81)* (1.99)** (1.87)*

Peer 0.295 0.254 0.241 0.296(3.32)*** (2.84)*** (2.79)*** (3.44)***

Ln(Delta) . 0.131 0.123 0.123. (4.32)*** (4.72)*** (4.65)***

Ln(Vega) . -0.002 -0.003 -0.003. (-0.92) (-1.11) (-1.32)

Ln(Size) . -0.16 -0.174 -0.188. (-1.64) (-1.57) (-1.64)

BTM . 0.71 0.737 0.748. (4.2)*** (5.47)*** (5.35)***

MOM . -0.669 -0.718 -0.759. (-0.35) (-0.6) (-0.7)

MGscore . . 2.5 .. . (7.53)*** .

IndSim . . . 0.084. . . (1.55)

R2 0.006 0.072 0.075 0.071

46

Table 5: Cross–Sectional Regressions of Stock Returns with Endogeneity–Mitigation Treat-ment

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine stock returns to incentive leaders, peers and non-event firms eitherwithin propensity score matching sample or using instrument variable for Delta Leaderand Peer Dummies. The dependent variable is the annual stock returns from month t +4 to month t + 15. The main independent variables are leader and peer. Leader is thedummy variable for CEO Delta incentive leaders. Peer is the dummy variable for propensityscore matching incentive peers. Ln(Delta) is the natural logarithm of one plus CEO Deltaincentive. Ln(Vega) is the natural logarithm of one plus CEO V ega incentive. Panel Areports the results corresponding to the propensity score matching sample. Panel B presentsthe results using instrument variable approach. In both panels, Model (1) stands for theregression results of returns only on leader and peer dummies. Model (2) add in five controlvariables in the regression, Ln(Delta), Ln(Vega), Ln(Size), B/M, and MOM. Ln(Size) isthe natural logarithm of market capitalization. B/M is the book to market ratio. MOMis the cumulative monthly return of past 11 months up to one month ago. Model (3)augments Model (2) by adding MGscore (managerial score) as the additional control. Model(4) augments Model (2) by adding InSim (industry similarity) instead of MGscore. Thereported coefficients are the time-series average of cross-sectional regression coefficients. Thecorresponding Newey and West (1987) adjusted t statistics with six lags are reported in theparentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

Panel A: Propensity Score Matching Sample

(1) (2) (3) (4)

Leader 0.391 0.37 0.581 0.462(1.95)* (1.71)* (1.99)** (1.92)*

Peer 0.311 0.35 0.381 0.37(3.45)*** (3.66)*** (3.81)*** (3.71)***

Ln(Delta) . 0.151 0.171 0.152. (4.41)*** (4.87)*** (4.11)***

Ln(Vega) . -0.034 -0.049 -0.051. (-1.07) (-1.93)* (-2.33)**

Control No Yes Yes+MGscore Yes+IndSimR2 0.006 0.071 0.076 0.072

Panel B: Instrument Variable Sample

(1) (2) (3) (4)

Leader 0.468 0.461 0.47 0.475(2.62)** (2.29)** (2.74)*** (2.06)**

Peer 1.197 1.192 1.207 1.21(3.53)*** (4.59)*** (4.08)*** (4.45)***

Ln(Delta) . 0.153 0.162 0.167. (4.42)*** (4.88)*** (4.11)***

Ln(Vega) . -0.031 -0.047 -0.049. (-1.06) (-1.92)* (-2.32)**

Control No No Yes+MGscore Yes+IndSimR2 0.006 0.069 0.071 0.063

47

Table 6: Fama-MacBeth Regressions of Stock Returns with Size-adjusted Leaders

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine stock returns to size-adjusted incentive leaders, peers and non-eventfirms. The dependent variable is the annual stock returns from month t+4 to month t+15.The main independent variables are leader and peer. Leader is the dummy variable for∆CEODelta× Size. Peer is the dummy variable for ∆CEODelta× Size peers. Ln(Delta)is the natural logarithm of one plus CEO Delta incentive. Ln(Vega) is the natural logarithmof one plus CEO V ega incentive. Ln(Size) is the natural logarithm of market capitalization.B/M is the book to market ratio. MOM is the cumulative monthly return of past 11 monthsup to one month ago. MGscore is the managerial score. IndSim is the industry similarity.The reported coefficients are the time-series average of cross-sectional regression coefficients.The corresponding Newey and West (1987) adjusted t statistics with six lags are reportedin the parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels,respectively.

(1) (2) (3) (4)

Intercept 1.289 1.94 1.115 1.684(3.8)*** (1.62) (0.86) (1.23)

Leader 0.38 0.419 0.463 0.339(1.77)* (1.37) (2.17)** (1.57)

Peer 0.398 0.475 0.451 0.446(4.02)*** (4.38)*** (3.95)*** (3.81)***

Ln(Delta) . 0.128 0.125 0.138. (3.55)*** (4.77)*** (3.67)***

Ln(Vega) . -0.002 -0.003 -0.003. (-1.01) (-1.93)* (-1.34)

Ln(Size) . -0.103 -0.146 -0.1. (-1.44) (-1.57) (-1.3)

B/M . 0.662 0.79 0.756. (4.71)*** (5.49)*** (5.81)***

MOM . -0.731 -0.778 -0.822. (-0.45) (-0.61) (-0.72)

MGscore . . 2.426 .. . (7.43)*** .

IndSim . . . 0.086. . . (1.34)

R2 0.008 0.071 0.081 0.069

48

Tab

le7:

Fam

a-M

acB

eth

Reg

ress

ions

ofF

undam

enta

ls

This

table

rep

orts

the

aver

age

coeffi

cien

tsan

dco

rres

pon

din

gt

stat

isti

csof

Fam

a-M

acB

eth

regr

essi

ons

that

exam

ine

firm

fundam

enta

lsto

ince

nti

vele

ader

s,p

eers

and

non

-eve

nt

firm

s.T

he

dep

enden

tva

riab

les

are

the

firm

fundam

enta

lsm

easu

res

afte

rth

eid

enti

fica

tion

ofC

EO

ince

nti

vele

ader

san

dp

eers

.W

euse

thre

efirm

fundam

enta

lsm

easu

res

toca

ptu

rediff

eren

tas

pec

tsof

firm

s’op

erat

ion:

sale

sgr

owth

(Sal

esG

row

th),

gros

spro

fita

bilit

y(G

P),

and

retu

rnon

asse

ts(R

OA

).T

he

mai

nin

dep

enden

tva

riab

les

are

lead

eran

dp

eer.

Lea

der

isth

edum

my

vari

able

for

CE

ODelta

ince

nti

vele

ader

s.P

eer

isth

edum

my

vari

able

for

ince

nti

vep

eers

.A

uto

Lag

isth

eco

rres

pon

din

gla

gged

one

per

iod

fundam

enta

ls.

Ln(D

elta

)is

the

nat

ura

llo

gari

thm

ofon

eplu

sC

EODelta

ince

nti

ve.

Ln(V

ega)

isth

enat

ura

llo

gari

thm

ofon

eplu

sC

EOVega

ince

nti

ve.

Ln(S

ales

)is

the

nat

ura

llo

gari

thm

ofsa

les.

R&

Dis

rese

arch

and

dev

elop

men

tex

pen

dit

ure

scal

edby

tota

las

sets

.C

apex

isca

pit

alex

pen

dit

ure

scal

edby

tota

las

sets

.T

he

rep

orte

dco

effici

ents

are

the

tim

e-se

ries

aver

age

ofcr

oss-

sect

ional

regr

essi

onco

effici

ents

.T

he

corr

esp

ondin

gN

ewey

and

Wes

t(1

987)

adju

sted

tst

atis

tics

wit

hsi

xla

gsar

ere

por

ted

inth

epar

enth

eses

.**

*,**

,an

d*

den

ote

sign

ifica

nce

atth

e1%

,5%

,an

d10

%le

vels

,re

spec

tive

ly.

GP

RO

ASal

esG

row

th(1

)(2

)(1

)(2

)(1

)(2

)

Auto

Lag

0.66

80.

670.

211

0.24

30.

004

0.01

4(2

6.17

)***

(26.

47)*

**(4

.59)

***

(5.8

8)**

*(0

.1)

(0.3

3)L

eader

0.00

30.

004

0.01

50.

015

0.04

70.

059

(0.3

5)(0

.53)

(2.4

6)**

(2.6

8)**

*(3

.04)

***

(3.9

1)**

*P

eer

0.00

40.

004

0.00

90.

009

0.01

20.

014

(2.7

2)**

*(2

.81)

***

(3.8

3)**

*(3

.81)

***

(2.6

8)**

*(2

.97)

***

Ln(D

elta

)0.

004

.0.

013

.0.

025

.(0

.39)

.(6

.34)

***

.(6

.18)

***

.L

n(V

ega)

-0.0

02-0

.002

0.00

10.

002

-0.0

01-0

.009

(-1.

73)*

(-1.

69)*

(0.8

5)(1

.92)

*(-

0.31

)(-

1.04

)L

n(S

ales

)-0

.013

-0.0

12-0

.009

-0.0

09-0

.239

-0.2

33(-

3.52

)***

(-3.

4)**

*(-

2.08

)**

(-2.

29)*

*(-

12.2

4)**

*(-

12.0

7)**

*R

&D

0.15

40.

151

0.30

20.

412

-0.0

33-0

.052

(1.6

3)(1

.61)

(2.2

3)**

(3.0

1)**

*(-

0.13

)(-

0.2)

Cap

ex-0

.146

-0.1

430.

116

0.09

5-0

.233

-0.1

57(-

2.83

)***

(-2.

82)*

**(2

.07)

**(1

.86)

*(-

1.68

)*(-

1.17

)R

20.

467

0.46

60.

819

0.81

70.

373

0.36

9

49

Table 8: Fama-MacBeth Regressions of Investment Measures on Leader and Peer

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine three future Investment Measures (R&D, Capex and Total Inv) toincentive leaders, peers and non-event firms. The dependent variable is Investment Measures(R&D, Capex and Total Inv) in the subsequent year after the identification of CEO incentiveleaders and peers. R&D is research and development expenditure scaled by total assets.Capex is capital expenditure scaled by total assets. Total Inv is the sum of research anddevelopment expenditure, capital expenditure and acquisition expenditure subtracted thesales of property and scaled by total assets. The main independent variables are incentiveleader and peer. Leader is the dummy variable for CEO Delta incentive leaders. Peer isthe dummy variable for incentive peers. Ln(Delta) is the natural logarithm of one plusCEO Delta incentive. Ln(V ega) is the natural logarithm of one plus CEO V ega incentive.Ln(Sales) is the natural logarithm of sales. The reported coefficients are the time-seriesaverage of cross-sectional regression coefficients. The corresponding Newey and West (1987)adjusted t statistics with six lags are reported in the parentheses. ***, **, and * denotesignificance at the 1%, 5%, and 10% levels, respectively.

R&D Capex Total Inv

Auto Lag 0.102 0.204 0.016(2.03)** (9.87)*** (2.29)**

Leader 0.001 0.005 0.004(0.14) (0.95) (0.89)

Peer 0.001 0.001 0.001(0.59) (0.26) (1.11)

Ln(Delta) -0.001 0.004 0.007(-0.5) (6.91)*** (7.52)***

Ln(Vega) 0.07 -0.001 -0.002(2.31)** (-3.17)*** (-2.51)**

Ln(Sales) -0.003 -0.002 -0.001(-0.45) (-0.92) (-0.4)

R2 0.421 0.562 0.517

50

Table 9: High and Low MGscore Peer Portfolio Performance

This table reports the abnormal returns for equal- and value-weighed portfolios of incentivepeers with high and low managerial score and industry similarity. To form portfolios for highand low managerial score (MGscore) incentive peers, we skip one quarter after their identifi-cation. Specifically, after the identification of incentive peers at month t, we divide incentivepeers into two groups: the peers with lower than median managerial scores (LowMGPeer)and the ones with higher than median managerial scores (HighMGPeer) in the identificationyear. The portfolios of LowMGPeer and HighMGPeer remain the same for the subsequent12 months from month t+ 4 to month t+ 15. To form portfolios for high and low industrysimilarity (IndSim), we divide incentive peers into high (HighIndSimPeer) and low IndSim(LowIndSimPeer) groups based whether their IndSim are higher than median IndSim in theidentification year. The portfolios of high and low industry similarity peers remain the samefor the subsequent 12 months from month t + 4 to month t + 15. The abnormal returnsare estimated using Carhart (1997) four-factor model. MKTRF, SMB, HML, and UMD arethe market excess return, the size, book-to-market, and momentum factors from KennethFrench’s website. The Newey and West (1987) adjusted t statistics with six lags are report-ed in the parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels,respectively.

Panel A: EW High and Low MGscore Portfolio Performance

LowMGPeer HighMGPeer H-LAlpha 0.586 1.187 0.602

(3.85)*** (6.1)*** (4.92)***

Panel B: VW High and Low MGscore Portfolio Performance

LowMGPeer HighMGPeer H-LAlpha 0.562 1.173 0.612

(3.77)*** (6.1)*** (4.81)***

Panel C: EW High and Low IndSim Portfolio Performance

HighIndSimPeer LowIndSimPeer L-HAlpha 0.676 1.157 0.481

(4.29)*** (5.93)*** (3.56)***

Panel D: VW High and Low IndSim Portfolio Performance

HighIndSimPeer LowIndSimPeer L-HAlpha 0.654 1.129 0.475

(4.21)*** (5.92)*** (3.79)***

51

Table 10: Fama-MacBeth Regressions of Returns on High and Low Managerial Score Peers andHigh and Low industry Similarity Peers

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine stock returns to incentive leaders, peers and non-event firms. Thedependent variable is the annual stock returns from month t + 4 to month t + 15. Themain independent variables are leader and peer. Leader is the dummy variable for CEODelta incentive leaders. HighMGscorePeer is the dummy variable indicating incentive peerswith higher than median managerial score in the identification year. LowMGscorePeer isthe dummy variable indicating incentive peers with lower than median industry similarityin the identification year. HighIndSimPeer is the dummy variable indicating incentive peerswith higher than median industry similarity in the identification year. LowIndSimPeer isthe dummy variable indicating incentive peers with lower than median industry similarity inthe identification year. Ln(Delta) is the natural logarithm of one plus CEO Delta incentive.Ln(Vega) is the natural logarithm of one plus CEO V ega incentive. Ln(Size) is the naturallogarithm of market capitalization. B/M is the book to market ratio. MOM is the cumulativemonthly return of past 11 months up to one month ago. The reported coefficients are thetime-series average of cross-sectional regression coefficients. The corresponding Newey andWest (1987) adjusted t statistics with six lags are reported in the parentheses. ***, **, and* denote significance at the 1%, 5%, and 10% levels, respectively.

y=Rett+1 High & Low MGscorePeer High& Low IndSimPeer

Intercept 1.82 1.92(1.43) (1.52)

Leader 0.401 0.411(1.78)* (1.71)*

HighMGscorePeer 0.44 .(3.47)*** .

LowMGscorePeer 0.151 .(1.75)* .

LowIndSimPeer . 0.371. (2.96)***

HighIndSimPeer . 0.21. (2.16)**

Ln(Delta) 0.141 0.151(4.18)*** (4.34)***

Ln(Vega) -0.03 -0.03(-0.95) (-0.96)

Ln(Size) -0.129 -0.13(-1.56) (-1.65)

B/M 0.68 0.641(4.41)*** (4.11)***

MOM -1.15 -1.009(-0.4) (-0.36)

R2 0.076 0.079

52

Table 11: Fama-MacBeth Regressions of Fundamentals on Managerial Scores and IndustrySimilarity

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine firm fundamentals to incentive leaders, peers and non-event firms.The dependent variables are the firm fundamentals measures in the subsequent year after theidentification of CEO incentive leaders and peers. We use three firm fundamentals measuresto capture different aspects of firms’ operation: sales growth (SalesGrowth), gross profitabil-ity (GP), and return on assets (ROA). The main independent variables are dummy variablesindicating leaders, peers with above- and below-median managerial score (HighMGscorePeerand LowMGscorePeer) or with above- and below-median industry similarity (LowIndSim-Peer and HighIndSimPeer). Auto Lag is the corresponding lagged one period fundamentals.Ln(Delta) is the natural logarithm of one plus CEO Delta incentive. Ln(Vega) is the nat-ural logarithm of one plus CEO V ega incentive. Even though not reported in the table,we include a sequence of variables including Ln(Sales), R&D, and Capex. Ln(Sales) is thenatural logarithm of sales. R&D is research and development expenditure scaled by totalasset. Capex is capital expenditure scaled by total asset. The reported coefficients are thetime-series average of cross-sectional regression coefficients. The corresponding Newey andWest (1987) adjusted t statistics with six lags are reported in the parentheses. ***, **, and* denote significance at the 1%, 5%, and 10% levels, respectively.

Panel A: Regressions of Fundamentals on MGscore

GP ROA SalesGrowth

Auto Lag 0.667 0.206 0.003(26.09)*** (4.51)*** (0.07)

Leader 0.013 0.015 0.047(2.35)** (2.41)** (3.02)***

HighMGscorePeer 0.009 0.013 0.031(2.87)*** (3.91)*** (3.79)***

LowMGscorePeer 0.002 -0.009 0.009(0.83) (-2.33)** (0.99)

Ln(Delta) 0.001 0.007 0.025(0.83) (6.32)*** (6.12)***

Ln(Vega) -0.002 0.001 -0.001(-1.7)* (0.87) (-0.25)

Panel B: Regressions of Fundamentals on IndSim

GP ROA SalesGrowth

Auto Lag 0.668 0.211 0.004(26.17)*** (4.59)*** (0.1)

Leader 0.014 0.015 0.047(2.36)** (2.46)** (3.03)***

LowIndSimPeer 0.008 0.006 0.035(2.89)*** (2.71)*** (3.06)***

HighIndSimPeer 0.001 0.002 0.01(0.59) (0.69) (1.49)

Ln(Delta) 0.001 0.003 0.025(0.87) (2.34)** (6.17)***

Ln(Vega) -0.002 0.001 -0.001(-1.71)* (0.86) (-0.3)

53

Table 12: Fama-MacBeth Regressions of MGscore and IndSim on Leader and Peer

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine future managerial score and industry similarity to incentive leaders,peers and non-event firms. The dependent variables are managerial score and industrysimilarity in the subsequent year after the identification of CEO incentive leaders and peers.The main independent variables are incentive leader and peer. Leader is the dummy variablefor CEO Delta incentive leaders. Peer is the dummy variable for incentive peers. Ln(Delta)is the natural logarithm of one plus CEO Delta incentive. Ln(V ega) is the natural logarithmof one plus CEO V ega incentive. Ln(Sales) is the natural logarithm of sales. R&D is researchand development expenditure scaled by total assets. Capex is capital expenditure scaled bytotal assets. The reported coefficients are the time-series average of cross-sectional regressioncoefficients. The corresponding Newey and West (1987) adjusted t statistics with six lagsare reported in the parentheses. ***, **, and * denote significance at the 1%, 5%, and 10%levels, respectively.

Managerial Score Industry Similarity(1) (2) (1) (2)

Leader 0.011 0.009 0.204 0.166(3.28)*** (2.54)** (1.49) (1.71)*

Peer 0.009 0.013 -0.74 -0.748(5.17)*** (7.25)*** (-6.8)*** (-7.14)***

Ln(Delta) 0.008 . 0.028 .(15.22)*** . (2.3)** .

Ln(Vega) -0.006 -0.013 -0.142 -0.122(-3.21)*** (-5.13)*** (-2.37)** (-2.24)**

Ln(Sales) 0.006 0.008 0.244 0.289(2.23)** (3.12)** (3.48)** (3.13)***

R&D -0.135 -0.137 9.376 9.333(-2.71)*** (-2.75)*** (4.1)*** (4.15)***

Capex 0.061 0.053 0.716 0.74(2.83)*** (2.27)** (1.27) (1.23)

R2 0.623 0.617 0.871 0.865

54

Appendix

A.1 Firm Characteristics

The stock return and accounting data are from the CRSP and COMPUSTAT. We excludefinancial (SIC 6000-6999) and utility (SIC 4900-4999) stocks. We winsorize all continuousvariables at 1% and 99% levels to ensure that our results are not driven by outliers. We applythe log transformation to most variables to make them more symmetrically distributed.

• Capex: Capital expenditure, the ratio of capital expenditures (CAPX) to total assets.

• R&D: Research and development expenditure (XRD, replaced by 0 when missing)divided by total assets.

• Ln(Sales): The logarithm of Sales (e.g., Coles et al. (2006)).

• Ret: The next-month stock returns.

• Ln(Size): The logarithm of market capitalization.

• B/M : The ratio of the book value of equity to the market value of equity.

• MOM : The 11-month cumulative return up to one month ago.

A.2 Interaction Terms with Peers

• Low Similarity Peer: A dummy variable that takes the value of 1 if a firm is a Peerwith low (below-median) industrial similarity.

• High Similarity Peer: A dummy variable that takes the value of 1 if a firm is a Peerwith high (above-median) industrial similarity.

• Older Peer: A dummy variable that takes the value of 1 if a firm is a Peer with oldermanager.

• Y ounger Peer: A dummy variable that takes the value of 1 if a firm is a Peer withyounger manager.

• Shorter Tenured Peer: A dummy variable that takes the value of 1 if a firm is a Peerwith shorter tenured manager.

• Longer Tenured Peer: A dummy variable that takes the value of 1 if a firm is a Peerwith longer tenured manager.

• Less Confident Peer: A dummy variable that takes the value of 1 if a firm is a Peerwith less confident manager.

• More Confident Peer: A dummy variable that takes the value of 1 if a firm is a Peerwith more confident manager.

55

A.3 Construction of Managerial Score

To estimate the managerial score, we first estimate the firm efficiency for each of the Hoberg–Philip 50 industry groups. We follow Demerjian et al. (2012) to estimate firm efficiencywithin industries, comparing the sales generated by each firm, conditional on the followinginputs used by the firm: Cost of Goods Sold (COGS, Selling and Administrative Expenses(SG&A), Net PP&E (PPE), Net Operating Leases (OL), Net Research and Development(R&D), Purchased Goodwill (GW), and Other Intangible Assets:

maxvΘ =Sales

(v1COGS + v2SG&A + v3PPE + v4OL + v5R&D + v6GW ). (1)

The firm efficiency measure, Θ, takes a value between 0 and 1, reflecting constraints in theoptimization program. We then run the following regression by industry:

Θi = α0 + α1TAi + α2MSi + α3FCFi + α4Agei + α5Concti + α6FCi + Y eari + vi, (2)

where Θ is the firm efficiency measure; TA is the total assets; MS stands for the market share;FCF is the free cash flow; Age is the firm age; Conct is the business segment concentration;FC is the foreign currency indicator. The residual of the above regression in Demerjian et al.(2012) is largely attributable to the manager and is defined as managerial effort.

56

Table A.1: Cross-sectional Regressions of Stock Returns During the SFAS 123(R) EventPeriod

This table reports the coefficients and corresponding t statistics of cross-sectional regressionsthat examine stock returns to incentive leaders, peers and non-event firms during the SFAS123(R) event period. The dependent variable is the annual stock returns from month t + 4to month t + 15 for fiscal year 2006. The main independent variables are leader and peer.Leader is a dummy variable equal to 1 if the firm belongs to an industry that experiencesan aggregate 20% Delta increase event ((Delta2005 − Delta2001)/Delta2001) and the firmDelta growth rate ranks among the top 10% in the industry. Peer is the dummy variablefor incentive peers. Ln(Delta) is the natural logarithm of one plus CEO Delta incentive.Ln(Vega) is the natural logarithm of one plus CEO V ega incentive. Ln(Size) is the naturallogarithm of market capitalization. B/M is the book to market ratio. MOM is the cumulativemonthly return of past 11 months up to one month ago. MGscore is the managerial score.IndSim is the industry similarity. ***, **, and * denote significance at the 1%, 5%, and 10%levels, respectively.

(1) (2) (3) (4)

Intercept 1.128 1.756 1.418 1.596(1.36) (0.63) (0.53) (0.6)

Leader 2.78 2.386 2.348 2.409(3.72)*** (3.32)*** (2.96)*** (3.36)***

Peer 1.302 1.191 1.116 1.124(3.5)*** (3.36)*** (2.94)*** (3.12)***

Ln(Delta) . 0.053 0.052 0.056. (0.29) (0.36) (0.36)

Ln(Vega) . -0.012 -0.015 -0.013. (-0.23) (-0.21) (-0.2)

Ln(Size) . -0.094 -0.051 -0.082. (-0.67) (-0.46) (-0.62)

BTM . 0.818 0.797 0.819. (4.57)*** (3.75)*** (4.32)***

MOM . 0.481 0.343 0.301. (0.76) (0.34) (0.48)

MGscore . . 1.846 .. . (2.16)** .

IndSim . . . -0.007. . . (-0.92)

R2 0.007 0.071 0.076 0.077

57

Table A.2: Fama-MacBeth Regressions of MGscore on Leader and Peer in Sub-samples

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine future managerial score to incentive leaders, peers and non-eventfirms in four sub-samples. The dependent variable is managerial score in the subsequent yearafter the identification of CEO incentive leaders and peers. The main independent variablesare incentive leader and peer. Leader is the dummy variable for CEO Delta incentiveleaders. Peer is the dummy variable for incentive peers. Ln(Delta) is the natural logarithmof one plus CEO Delta incentive. Ln(V ega) is the natural logarithm of one plus CEO V egaincentive. Ln(Sales) is the natural logarithm of sales. R&D is research and developmentexpenditure scaled by total assets. Capex is capital expenditure scaled by total assets. Thereported coefficients are the time-series average of cross-sectional regression coefficients. Thecorresponding Newey and West (1987) adjusted t statistics with six lags are reported in theparentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

Sales Growth Cost Growth(High) (Low) (High) (Low)

Leader 0.019 0.007 0.017 0.019(3.64)*** (2.37)** (2.46)** (2.41)**

Peer 0.019 0.004 0.006 0.018(4.62)*** (1.69)* (2.25)** (4.21)***

Ln(Delta) 0.018 0.01 0.01 0.016(17.22)*** (13.11)*** (13.57)*** (16.69)***

Ln(Vega) -0.007 -0.005 -0.005 -0.007(-3.32)*** (-2.93)*** (-3.01)*** (-3.24)***

Ln(Sales) 0.005 0.006 0.006 0.005(2.38)** (2.24)** (2.25)** (2.33)**

R&D -0.116 -0.171 -0.244 -0.031(-2.11)** (-2.76)*** (-3.34)*** (-0.19)

Capex 0.075 0.043 0.069 0.057(3.14)*** (2.17)** (2.88)*** (2.04)**

R2 0.712 0.541 0.571 0.685

58

Tab

leA

.3:

CE

OC

har

acte

rist

ics

and

Ince

nti

veSpillo

ver

This

table

rep

orts

the

aver

age

coeffi

cien

tsan

dco

rres

pon

din

gt

stat

isti

csof

Fam

a-M

acB

eth

regr

essi

ons

that

exam

ine

firm

fundam

enta

lsto

ince

nti

vele

ader

s,p

eers

and

non

-eve

nt

firm

s.T

he

dep

enden

tva

riab

les

are

the

firm

fundam

enta

lsm

easu

res

inth

esu

bse

quen

tye

araf

ter

the

iden

tifica

tion

ofC

EO

ince

nti

vele

ader

san

dp

eers

.W

euse

thre

efirm

fundam

enta

lsm

easu

res

toca

ptu

rediff

eren

tas

pec

tsof

firm

s’op

erat

ion:

sale

sgr

owth

(Sal

esG

row

th),

gros

spro

fita

bilit

y(G

P),

and

retu

rnon

asse

ts(R

OA

).T

he

mai

nin

dep

enden

tva

riab

les

are

lead

ers

and

dum

my

vari

able

sin

dic

atin

gp

eers

wit

hab

ove-

and

bel

ow-m

edia

nC

EO

char

acte

rist

ics

inth

eid

enti

fica

tion

year

.W

eex

plo

reth

ree

CE

Och

arac

teri

stic

s:C

EO

age,

CE

Ote

nure

,an

dC

EO

over

confiden

cele

vel.

Pee

rsw

ith

bel

ow-m

edia

nC

EO

age,

bel

ow-m

edia

nte

nure

,an

dab

ove-

med

ian

over

confiden

cele

vel

are

Hig

hP

eer.

Lea

der

isth

edum

my

vari

able

for

CE

ODelta

ince

nti

vele

ader

s.A

uto

Lag

isth

eco

rres

pon

din

gla

gged

one

per

iod

fundam

enta

ls.

Ln(D

elta

)is

the

nat

ura

llo

gari

thm

ofon

eplu

sC

EODelta

ince

nti

ve.

Ln(V

ega)

isth

enat

ura

llo

gari

thm

ofon

eplu

sC

EOVega

ince

nti

ve.

Ln(S

ales

)is

the

nat

ura

llo

gari

thm

ofsa

les.

R&

Dis

rese

arch

and

dev

elop

men

tex

pen

dit

ure

scal

edby

tota

las

sets

.C

apex

isth

eca

pti

alex

pen

dit

ure

scal

edby

tota

las

sets

.T

he

rep

orte

dco

effici

ents

are

the

tim

e-se

ries

aver

age

ofcr

oss-

sect

ional

regr

essi

onco

effici

ents

.T

he

corr

esp

ondin

gN

ewey

and

Wes

t(1

987)

adju

sted

tst

atis

tics

wit

hsi

xla

gsar

ere

por

ted

inth

epar

enth

eses

.**

*,**

,an

d*

den

ote

sign

ifica

nce

atth

e1%

,5%

,an

d10

%le

vels

,re

spec

tive

ly.

Pan

elA

:G

PP

anel

B:

RO

AP

anel

C:

Sal

esG

row

thA

ge

Ten

ure

Con

fid

ence

Age

Ten

ure

Con

fid

ence

Age

Ten

ure

Con

fid

ence

Au

toL

ag0.6

69

0.6

69

0.6

680.

212

0.21

20.

208

0.00

40.

005

0.00

4(2

6.1

7)**

*(2

6.1

4)*

**(2

6.1

8)**

*(4

.6)*

**(4

.62)

***

(4.5

1)**

*(0

.11)

(0.1

1)(0

.08)

Lea

der

0.01

40.

013

0.0

140.

015

0.01

50.

016

0.04

60.

047

0.04

7(2

.44)*

*(2

.26)

**

(2.3

7)**

(2.4

)**

(2.3

)**

(2.4

5)**

(2.9

9)**

*(3

.06)

***

(3.0

7)**

*L

owP

eer

0.0

040.

002

0.0

020.

001

0.00

30.

001

0.00

60.

006

0.00

6(1

.17)

(0.7

1)(1

.12)

(0.3

)(0

.85)

(0.4

4)(0

.59)

(0.7

3)(0

.77)

Hig

hP

eer

0.0

050.0

060.

077

0.01

10.

010.

012

0.04

10.

044

0.04

8(2

.75)

***

(3.2

6)*

**(3

.72)

***

(3.3

9)**

*(3

.21)

***

(3.0

9)**

*(3

.8)*

**(4

.09)

***

(4.5

3)**

*L

n(D

elta

)0.0

010.

001

0.0

010.

009

0.00

90.

009

0.02

50.

025

0.02

5(0

.87)

(0.8

6)(0

.76)

(6.3

3)**

*(6

.17)

***

(6.1

8)**

*(6

.12)

***

(6.0

1)**

*(5

.94)

***

Ln

(Veg

a)

-0.0

02

-0.0

02-0

.002

0.00

10.

001

0.00

1-0

.001

-0.0

01-0

.001

(-1.

72)*

(-1.

75)*

(-1.

58)

(0.8

6)(0

.76)

(1.0

9)(-

0.3)

(-0.

33)

(-0.

31)

Ln

(Sal

es)

-0.0

13-0

.013

-0.0

13-0

.009

-0.0

09-0

.01

-0.2

39-0

.239

-0.2

39(-

3.52

)***

(-3.

53)*

**(-

3.5

5)**

*(-

2.07

)**

(-2.

06)*

*(-

2.16

)**

(-12

.25)

***

(-12

.27)

***

(-12

.27)

***

R&

D0.1

530.

154

0.1

540.

403

0.43

0.42

6-0

.033

-0.0

3-0

.034

(1.6

3)

(1.6

3)

(1.6

3)(3

.23)

***

(4.2

5)**

*(4

.22)

***

(-0.

13)

(-0.

11)

(-0.

13)

Cap

ex-0

.145

-0.1

46

-0.1

460.

115

0.11

40.

113

-0.2

33-0

.233

-0.2

37(-

2.82

)***

(-2.

85)*

**(-

2.8

5)**

*(2

.06)

**(2

.04)

**(2

.05)

**(-

1.68

)*(-

1.68

)*(-

1.71

)*R

20.4

72

0.4

70.

475

0.81

70.

822

0.81

90.

375

0.37

10.

369

59

Table A.4: Fama-MacBeth Regressions of MGscore on Leader and Peer in Sub-samples

This table reports the average coefficients and corresponding t statistics of Fama-MacBethregressions that examine future managerial score to incentive leaders, peers and non-eventfirms in four sub-samples. The dependent variable is managerial score in the subsequent yearafter the identification of CEO incentive leaders and peers. The main independent variablesare incentive leader and peer. Leader is the dummy variable for CEO Delta incentiveleaders. Peer is the dummy variable for incentive peers. Ln(Delta) is the natural logarithmof one plus CEO Delta incentive. Ln(V ega) is the natural logarithm of one plus CEO V egaincentive. Ln(Sales) is the natural logarithm of sales. R&D is research and developmentexpenditure scaled by total assets. Capex is capital expenditure scaled by total assets. Thereported coefficients are the time-series average of cross-sectional regression coefficients. Thecorresponding Newey and West (1987) adjusted t statistics with six lags are reported in theparentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

Sales Growth Cost Growth(High) (Low) (High) (Low)

Leader 0.019 0.007 0.017 0.019(3.64)*** (2.37)** (2.46)** (2.41)**

Peer 0.019 0.004 0.006 0.018(4.62)*** (1.69)* (2.25)** (4.21)***

Ln(Delta) 0.018 0.01 0.01 0.016(17.22)*** (13.11)*** (13.57)*** (16.69)***

Ln(Vega) -0.007 -0.005 -0.005 -0.007(-3.32)*** (-2.93)*** (-3.01)*** (-3.24)***

Ln(Sales) 0.005 0.006 0.006 0.005(2.38)** (2.24)** (2.25)** (2.33)**

R&D -0.116 -0.171 -0.244 -0.031(-2.11)** (-2.76)*** (-3.34)*** (-0.19)

Capex 0.075 0.043 0.069 0.057(3.14)*** (2.17)** (2.88)*** (2.04)**

R2 0.712 0.541 0.571 0.685

60