Employee Stock Ownership Plans: Employee Compensation and Firm Value

E. Han Kim and Paige Ouimet*

Abstract

Changes in wages and firm value following ESOP adoptions are related to the ESOP size. When

it is small (less than 5% of outstanding shares), both mean wages and firm values increase. Since

shareholders and employees are the two main claimants of firm surplus, these changes suggest

small ESOPs increase productivity. Importantly, employees gain more and shareholder gain less

when employee job mobility increases after ESOP initiations, implying the productivity gains are

shared between employees and shareholders according to their bargaining power. Large ESOPs

have neutral effects on wages and shareholder value, indicating productivity gains no greater than

the value of ESOP shares granted. Some large ESOPs seem to be motivated by reasons unrelated

to improving group incentives and co-monitoring: cash conservation by small and young firms,

leading to wage cuts, and worker-management alliance to thwart takeover threats, causing wage

increases unrelated to productivity gains.

April 14, 2011

JEL classification: G32, M52, J54, J33

Keywords: ESOPs, Group Incentives, Worker Mobility, Worker-Management Alliance, and

Cash-constrained firms.

\* E. Han Kim, the corresponding author, is at Ross School of Business, University of Michigan, Ann Arbor, Michigan

48109; email: ehkim@umich.edu. Paige Ouimet is at Kenan-Flagler Business School, University of North Carolina,

Chapel Hill, NC 27599; email: Paige_Ouimet@unc.edu. A previous version of this paper was circulated under the title

―Employee Capitalism or Corporate Socialism: Broad-based Employee Stock Ownership.‖ We are grateful for helpful

comments and suggestions by Sreedhar Bharath, Joseph Blasi, Amy Dittmar, Charles Hadlock, Diana Knyazeva, Doug

Kruse, Francine Lafontaine, Margaret Levenstein, Randall Morck, Daniel Paravisini, Joel Shapiro, Clemens Sialm,

Jagadeesh Sivadasan; seminar participants at INSEAD, North Carolina State Universiy, University of Hawaii,

University of Michigan, University of Oxford, the US Bureau of Census, and Washington University at St. Louis; and

participants at 2010 American Finance Association Annual Meetings, 2009 Conference on Financial Economics and

Accounting, Labour and Finance Conference at University of Oxford, Madrid Conference on Understanding Corporate

Governance, the Census Research Data Center Annual Conference, and the International Conference on Human

Resource Management in Banking Industry. We acknowledge financial support from Mitsui Life Financial Research

Center at the Ross School of Business. The research was conducted while the authors were Special Sworn Status

researchers of the U.S. Census Bureau at the University of Michigan and Triangle Census Research Data Center. We

thank Clint Carter and Bert Grider for their diligent assistance with the data and clearance requests. Any opinions and

conclusions expressed herein are those of the author(s) and do not necessarily represent the views of the U.S. Census

Bureau. All results have been reviewed to ensure that no confidential information is disclosed.

1

Broad-based employee stock ownership plans (ESOPs) are common in the U.S. The National

Center for Employee Ownership (NCEO) estimates that as of 2010 there were over 10,500

private and public U.S. firms with ESOPs. The plans show an increasing long-term trend; the

NCEO estimates the number of participants in ESOPs was one-quarter million in 1975, five

million in 1990, and 13 million in 2007. In our sample, the median ESOP controls 6.7% of the

firm’s shares outstanding, representing an average market value of $18,643 (in 2006 dollars) per

employee.

Earlier studies based on stock price reactions to announcements of ESOP adoptions show

ESOPs are, on average, good news for shareholders, unless they are used to entrench managers.1

However, the literature on ESOPs offers little evidence on how employees – the participants – are

affected by this plan. In this paper, we attempt to shed light on how ESOPs affect employee

welfare by examining their wage changes following ESOP adoptions.

Using micro-data on employee compensation at the establishment (workplace) level from

the U.S. Bureau of Census, we find, on average, cash wages increase following the

implementation of an ESOP. Our definition of cash wages includes all forms of taxable ordinary

income, such as regular pay checks, bonuses, and commissions. The data also reveal significant

cross-sectional variation. Employees benefit the most when participating in a small ESOP,

defined as an ESOP which never controls more than 5% of the firm’s outstanding stock.

Following the adoption of a small ESOP, cash wages increase on average by 6.1%. Cash wages

do not include ESOP shares granted to employees; hence, the average compensation increase is

greater than 6.1%. The wage estimates are obtained by panel regressions using a control group of

non-ESOP firms matched by industry, year, size, average firm wage, and trends in wage changes

prior to the ESOP initiation. Our estimates control for establishment- and year fixed effects and

industry- and state mean wages, among others.

1 See Gordon and Pound (1990), Chang and Mayers (1992), Chaplinsky and Niehaus (1994), Beatty (1995), and

Cramton, Mehran, and Tracy (2008).

2

These gains to employees do not come at the expense of shareholders. Shareholders also

benefit. We find an average increase in industry-adjusted Tobin’s Q of 17% following the

adoption of small ESOPs.2 Since employees and shareholders are the two main claimants to total

firm surplus, from these gains we infer small ESOPs are associated with an increase in employee

productivity.

The evidence of employee and shareholder gains can be interpreted as causal effect of

ESOPs on firm surplus or as evidence of selection biases associated with the timing of the ESOP

implementation. ESOPs may increase productivity by improving group-based incentives and co-

monitoring (Kandel and Lazear, 1992; Jones and Kato, 1995; Blasi, Conte, and Kruse, 1996;

Kruse, Freeman and Blasi, 2010). Alternatively, a generous manager may implement an ESOP

when she anticipates higher future profits and wants to share them with employees.

To differentiate between these interpretations, we explore a prediction unique to the

causal interpretation. The causal story suggests that small ESOPs increase productivity and

employees and shareholders share the gains according to their relative bargaining power. Thus,

exogenous changes to employees’ bargaining power whould change the gains to shareholders and

employees. We proxy for exogenous shocks to employee bargaining power by changes in

employee mobility, as measured by a Herfindahl index of employer concentration within each

industry and geographic location of work place. We find that when employee mobility increases

at ESOP firms, wage gains are greater and shareholder gains are smaller. These results are not

due to ESOP firms’ own relocation of establishments. Furthermore, they are unique to ESOP

firms and do not reflect an industry-location wide relation between worker mobility and wages.

Since there is no reason to believe the selection story would predict such findings, these results

support our causal interpretation.

2 These gains in Tobin’s Q are high and do not imply any firm can increase shareholder value by 17% by adopting a

small ESOP. Our results are estimated using a set of firms which chose to adopt ESOPs – presumably a set of firms

which expected greater valuation effects than non-ESOP firms. As such, our estimation likely represents the upper tail

of possible average valuation impacts.

3

However, not all ESOPs are associated with both shareholder value and cash wage gains.

We find no significant changes in firm value or cash wages following the adoption of large

ESOPs, defined as ESOPs controlling more than 5% of the firm’s outstanding common stock at

any point in time. These estimation results do not include the value of the ESOP shares granted,

which is not trivial.3 However, even when the value of the ESOP shares are taken into account,

the average gains associated with large ESOPs are substantially less than those with small ESOPs.

The difference between large and small ESOPs does not necessarily imply that large

ESOPs cannot achieve productivity gains similar to those observed following small ESOPs.

Instead, a substantial portion of large ESOPs appear implemented to achieve objectives unrelated

to improving team effort and co-monitoring, clouding the average estimation results. We identify

two alternative motives, one predicting wage cuts and the other predicting wage increases.

Wage cuts should follow if ESOPs are implemented to conserve cash by substituting cash

wages with ESOP shares. Such a motive is likely to apply mainly to large ESOPs. Employees

value ESOP shares less than their market value because the sales restriction on ESOP shares

limits employees’ ability to diversify. (Employees cannot sell ESOP shares until they leave the

company or are close to retirement age.) Diversification restrictions matter because the share

prices are correlated with employee job security and wages. Thus, issuing shares through ESOPs

is more costly than issuing shares in the open market, unless the issuer has very limited (costly)

access to external equity due to asymmetric information. Since the information gap tends to be

greater for smaller and younger firms, we construct a cash constrained index using firm size

(small) and age (young). Hadlock and Pierce (2010) find size and age are the two key firm

characteristics significantly correlated with financial constraints. Our data show large ESOP

adoptions by high index firms (small and young) are followed by significant cash wage cuts. In

contrast, large ESOPs by low index (large and old) firms are followed by increases in cash wages.

3 The average market value of shares granted through large ESOPs in our sample is $26,796 in 2006 dollars per

employee, equal to 5.06% of annual wages if the shares were allocated equally over 10 years.

4

A motive predicting wage increases unrelated to productivity gains is the worker-

management alliance hypothesis suggested by Pagano and Volpin (2005), wherein management

bribes employees possessing control rights through ESOPs with above-market wages. The

purpose is to garner worker support in thwarting hostile takeover bids. Large ESOPs become an

especially effective anti-takeover device when the state of incorporation enacts business

combination statutes (BCS), which impose a temporary moratorium on takeover bids if a

sufficiently large block of investors, such as a large ESOP, vote against it. Thus, according to this

hypothesis, wages at firms with large ESOPs in place should increase following the passage of

BCS in their state of incorporation. Using a difference-in-difference test with BCS as an

exogenous shock, this is indeed what we find. Wages at these firms increase by an average of

7.3%, as compared to a group of control firms.

This paper contributes to the literature on several fronts. First, it fills an important void in

the employee stock ownership literature: Are employees financially better off as the result of

owning company shares? The answer seems affirmative, except for cases involving cash-

constrained firms using large ESOPs as a means to conserve cash.

Second, our wage results add to the literature on how firm surplus is shared between

suppliers of labor and capital (e.g., Akerlof and Yellen, 1990; Bhaskar, Manning, and To, 2002;

Manning, 2003). We measure employee mobility by the Herfindahl index of employer

concentration at the industry-location level. Using this measure as the proxy for employee

bargaining power vis-à-vis shareholders, we demonstrate the bargaining power affects employee

share of the surplus arising from their own productivity gains.

Third, previous studies document that ESOPs reduce the probability of takeovers

(Chaplinsky and Neihaus, 1994; Rauh, 2006), depriving shareholders of the opportunity to earn

large takeover premiums. Our evidence on wages identifies a more direct, previously

undocumented cost to shareholders when ESOPs are used as an anti-takeover device—higher

wages. This finding is also related to the post-BCS state-wide wage increases documented by

5

Bertrand and Mullainathan (2003), who attribute the wage increases to managers’ pursuit of

―quiet life‖. Our firm-level evidence identifies large ESOPs as an important channel through

which the pursuit of quiet life leads to post-BCS state-wide wage increases.

Finally, we shed some light on the controversy over whether cash constrained firms use

equity-based compensations as a means to conserve cash.4 The wage reductions following large

ESOP initiations at small and young firms render credence to the argument that cash conservation

with equity-based compensations is a relevant motive for firms with limited access to external

equity market.

The rest of the paper is organized as follows. Section I provides more details on ESOPs,

develops hypotheses, and briefly describes test outlines. Section II describes the data. Empirical

results concerning employee compensation and firm valuation are presented in Section III and IV,

respectively. Section V concludes.

I. Hypotheses Development

In this section we provide institutional details about ESOPs and review the relevant

literature to develop hypotheses. While the primary objective of this paper is to understand

employee welfare implications of ESOPs, we interpret our evidence in the context of different

motivations for ESOP implementation. The initial focus is on the motivation most often stated by

firms initiating ESOPs--improving incentives and team efforts to enhance worker productivity.

We also review studies on two other motivations: worker-management alliances to thwart hostile

takeover threats and substituting cash wages with ESOP shares to conserve cash.

A. Institutional Details

ESOPs are a type of tax-qualified pension plan.5 To qualify for tax benefits, the plan must

allocate shares in a broad-based manner to its employees. Usually, all full-time employees must

4 For example, Core and Guay (2001) argue firms implementing broad based stock option (BBSO) plans are more cash

constrained, whereas Ittner, Lambert, and Larcker (2003) find evidence that more cash constrained firms are less likely

to adopt BBSO plans. 5 ESOPs provide tax-advantages because for tax purposes qualified firms can deduct dividends paid on stock held by an

ESOP and dividends used to repay ESOP loans.

6

participate in the plan. Shares must be allocated equally to all employees or based on relative

compensation, seniority, or some combinations thereof, which helps prevent discriminatory share

allocation.

ESOPs allocate both cash flow and control rights to employees, creating a new class of

stakeholders, owner-employees. In public corporations, ESOPs must allow employees to vote

their shares. These rights are typically held by employees for an extended length of time because

employees cannot sell their ESOP shares until they leave the firm, although some exceptions are

made to allow employees to diversify their holdings as they get near retirement age.

B. Productivity Gains

Requiring employees to hold ESOP shares for an extended period can help align

employee incentives with shareholder value. How the alignment effect works for a typical

employee may differ from how equity-based incentives work for top executives, who have more

direct influence on overall firm performance than a typical employee. A vast majority of

employees may feel they have little impact on stock price, casting doubt on the ability of personal

incentives through ESOP shares to alter behavior in tasks requiring additional effort or sacrifice.

This also makes ESOPs different from other pay-for-productivity schemes such as piece

rate compensation, which can be effective when individual output is easily observed and

quantified (Lazear, 2000). With an ESOP, workers are not compensated for their individual

productivity; instead, they are rewarded when the productivity of the firm as a whole increases.

This introduces a free-rider problem (Kruse, 1996), diluting the incentives for employees to put

forth greater individual effort.

Kandel and Lazear (1992) argue the free-rider problem can be overcome through peer

pressure. Workers covered by an ESOP can profit as a whole if everyone agrees to work harder;

that is, solve the free-rider problem through collective collaborative activities such as peer

pressure and co-monitoring. If employees as a group can observe one another’s work and

influence behavior through peer pressure, cooperation becomes more likely, resulting in a high-

7

work regime. Although an ESOP alone may be insufficient to increase worker productivity, an

ESOP implemented in a corporate culture promoting group cooperation and co-monitoring may

have a substantially positive impact on productivity.

Survey evidence confirms that ESOPs encourage co-monitoring. Kruse, Freeman, and

Blasi (2010) survey over 40,000 employees from 14 companies with employee ownership plans

as part of the NBER Shared Capitalism research project. This survey, conducted between 2001

and 2006, asks employees how they would respond if they observe a co-worker underperforming.

They were given three options: 1) talk directly to the employee; 2) speak to a supervisor; or 3) do

nothing. An anti-shirking index based on the answers provided by the employees reveals that

those with company stock are much more likely to actively respond to shirking by a co-worker.

The team incentives and improved co-monitoring through ESOPs also may mitigate

inefficiencies in monitoring. Milgrom (1988) shows workers over-invest in tasks that are easily

observed by their bosses. Thus, even if workers do not shirk, their effort may be sub-optimally

allocated across tasks. By aligning team incentives and the focus of co-monitoring with

shareholder value, ESOPs may help reduce inefficiencies in how employees allocate their efforts.

B.1. Related Evidence

Jones and Kato (1995) estimate changes in productivity at Japanese firms following the

adoption of an ESOP by a Cobb-Douglas production function. They find that an ESOP adoption

in Japan leads to a 4-5% increase in productivity, starting about three years after the adoption.

However, the typical Japanese ESOP in their sample is allocated 1% or less of outstanding shares,

quite different from the typical ESOP observed in the U.S. Faleye, Mehrota, and Morck (2006)

look at only large employee share ownership blocks and disagree with the view that ESOPs are

conducive to productivity gains. They argue when employees control five percent or more of

voting stocks, firms have lower Tobin’s Q, invest less, grow more slowly, and create fewer new

jobs. These arguments are based on cross-sectional evidence.

B.2. Hypotheses and Test Outline

8

We examine small and large ESOPs separately, using 5% as the demarcation point. For

each group, we estimate how wages and firm valuation change following ESOP adoptions,

relative to a control group of matched firms that never adopt ESOPs. Since employees and

shareholders are the two main claimants of firm surplus, we infer productivity gains or losses

based on their combined changes associated with ESOP initiations. Although the inference is

indirect, this approach also helps identify whether gains to employees or shareholders arise at the

expense of the other claimant. In addition, directly estimating productivity would greatly reduce

our sample size because such estimates are feasible only for manufacturing firms.

If productivity indeed increases following ESOP adoptions, how are the gains shared

between employees and shareholders? Akerlof and Yellen (1990) argue that workers will demand

a ―fair‖ wage. One definition of a fair wage could include a share of firm profits. As profits

increase, workers will demand higher wages (a ―fair‖ share of the surplus) and be more inclined

to shirk or even quit if wages fall short. Empirical evidence consistent with the rent-sharing view

include Hildreth and Oswald (1997) who find that wages increase following shocks to firm-level

productivity, suggesting employees capture a share of productivity gains.

Employees’ share of productivity gains is likely to depend on workers’ bargaining power,

which may depend on worker mobility. When worker mobility is low and employees have few

alternative employment opportunities, the threat to shirk or quit matters less, giving employers

greater bargaining power (Bhaskar, Manning, and To, 2002; Manning, 2003), which in turn

enables employers to retain more of the surplus for shareholders. Conversely, when worker

mobility is high, employers may have to share a larger fraction of the surplus with employees.

To test these predictions, we develop a Herfindahl index of employer concentration

within each industry and location of work place and use it as a measure of worker mobility within

that industry-location. Because greater employer concentration tends to offer less outside

employment opportunities and makes employees less mobile, we predict that wages at ESOP

firms with productivity shocks will increase more following a decrease in employer concentration,

9

compared to a set of control firms. Since this prediction is unique to the causal interpretation, it

allows us to address the causality issue.

C. Worker-Management Alliance

Pagano and Volpin (2005) develop a theoretical model showing employee control rights

through ESOP shares give rise to a worker-management alliance.6 Management bribes employees

with above-market wages, which give employees incentives to vote with the management in the

event of a takeover contest. Shareholders pay for this alliance in two ways: (1) higher wages and

(2) a lower probability of receiving a takeover premium.

The second prediction of the alliance hypothesis is consistent with earlier findings by

Gordon and Pound (1990), Chang and Mayers (1992), and Rauh (2006). In addition, Chaplinsky

and Neihaus (1994) and Rauh (2006) show that ESOPs significantly reduce the probability of a

takeover.

C.1. Hypothesis and Test Outline

Our test focuses on the first prediction: wage increases. To distinguish this wage increase

from wage increases due to sharing productivity gains, we examine the impact of an exogenous

shock on wages at firms with large ESOPs in place. The shock is the enactments of business

combination statutes (BCS) by the firms’ states of incorporation, which Romano (1987) and

Bertrand and Mullainathan (2003) argue are mostly exogenous. In BCS states, if a sufficiently

large block of investors unaffiliated with management, such as a large ESOP, vote against a

hostile takeover bid, the bidder must wait three to five years before pursuing the target again.

Because courts have established holders of ESOP shares as ―outside‖ investors, BCS makes

ESOPs an especially effective anti-takeover device. Indeed, Rauh (2006) finds a significantly

lower takeover probability for firms with employee ownership incorporated in states with BCS in

effect. Since BCS makes employee support against hostile takeover bids more valuable, we

6 See Atanassov and Kim (2009) for evidence in support of the worker-management hypothesis in the context of

corporate restructuring around the world.

10

predict management’s incentives to bribe workers covered by large ESOPs will increase

following the passage of BCS, leading to higher wages.

D. Cash Conservation

Firms may seek to conserve cash by offering stocks to employees in exchange for lower

cash wages. While this motivation has received limited attention in the ESOP literature, Core and

Guay (2001) argue that cash conservation is a key motivation for implementing broad based stock

option (BBSO) plans, providing evidence that these plans are more common at firms with large

financing needs and high costs of external finance. However, subsequent studies provide

contradictory evidence. Using a different sample, Ittner, Lambert, and Larcker (2003) find more

cash constrained firms appear less likely to adopt BBSO plans. Oyer and Schaefer (2005) also

report mixed evidence: BBSO plans are less common at firms with positive cash flows but are

less likely to be adopted by firms making large investments requiring substantial cash outlays.

Underlying this empirical debate is a theoretical issue: Is substituting equity-based

compensation for cash wages an efficient cash conservation strategy for a publicly traded firm?

To be efficient, the costs associated with compensating risk adverse employees for additional risk

must outweigh the costs associated with accessing more traditional sources of external finance

(Oyer and Schaefer, 2005).

D.1. Hypothesis and Test Outline

ESOP shares impose risk on employees, some of which can be diversified by other

investors but not by employees. Because the value of the ESOP shares is correlated with firm

performance, which affects employee job security and wages, an ESOP share is worth less to risk

averse employees. This makes issuing shares to employees more costly than issuing shares to the

public. Thus, we do not expect firms to issue shares to employees to conserve cash unless their

access to external equity is limited (very costly). Young and small firms tend to have more costly

(difficult) access to external equity (Hadlock and Pierce, 2010). Thus, we conjecture when

11

younger and smaller firms adopt ESOPs, the cash conservation motivation is more likely to be at

work and wages fall subsequent to the adoption of ESOPs.

II. Data Construction, Empirical Design, and Summary Statistics

Our data on ESOPs cover US public firms from 1982 through 2001. We end the sample

period in 2001 because a change in the reporting method of establishment-level identifiers used

by the Census in 2001 makes it difficult to link post-2001 observations to earlier observations. To

identify firms with ESOPs and the first year of ESOP implementation, we conduct a two-step

Factiva search. In the first step, we search Factiva using the terms ―ESOP‖ and ―employee stock

ownership plan.‖ We read all articles and note the first date a firm is mentioned as having an

ESOP.7 In the second step, we run additional Factiva searches using each identified firm’s name

and ―employee stock‖ to obtain further information on its ESOP.8 With this additional search, we

are able to accurately identify the year of ESOP initiation for 410 unique firms. 9

We determine the size of ESOPs by reading annual proxy statements. In most cases,

ESOP share ownership is reported only if the plan has more than 5% of the firm’s common equity.

We assume the ESOP controls less than 5% of the firm’s outstanding shares if the proxy

statement does not report specific numbers concerning ESOP size. The ESOP database is then

matched to Compustat and Center for Research in Security Prices (CRSP) databases for

accounting and stock market variables.

7 This process yields 739 unique public firms with ESOPs over the sample period. Of these firms, we drop 35 firms

with total assets less than $10 million in 2006 dollars. The lack of press coverage on such small firms makes it likely

that we missed other similar-sized firms with ESOPs, wrongly identifying them as non-ESOP firms. This potential

error is important as our control group is obtained from firms in Compustat without identified ESOPs. 8 The purpose of the first search is to identify all firms with ESOPs and to identify the first year of ESOP

implementation, which we assume is the first year the firm is mentioned in an article as having ESOPs. In the second

search we obtain more articles because we use less restrictive search terms. We also read the articles more carefully. In

doing so, we occasionally find ESOPs that were established prior to the date the article was published. For example, the

first article mentioning the firm and an ESOP refers to the ESOP as having been implemented 5 years ago. We exclude

these cases out of concern that their inclusion would introduce a survivorship bias. The failure to observe any news

articles at the time the firm implemented the ESOP may have occurred because, at that time, the firm was too small to

receive adequate press coverage. As the firm grows over time, media attention increases, and we later observe an article

about an ESOP implemented in earlier years. Had the firm failed to grow, the firm would not have attracted media

attention and we never would have observed the ESOP. Thus, inclusion of those ESOPs would bias toward finding a

positive correlation between ESOP implementations and firm performance. 9 If a firm underwent a bankruptcy or was dropped from Compustat for a year or more, we assume the ESOP was

terminated unless other information is present.

12

The ESOP database is also matched to the Longitudinal Business Database (LBD)

maintained by the U.S. Bureau of Census, a panel dataset that tracks all U.S. business

establishments with at least one employee or positive payroll from 1975 to the present. An

establishment describes any facility with a separate physical address, such a factory, service

station, restaurant, and so on. The database is formed by linking years of the Business Register

(formally called the Standard Statistical Establishment List). The Business Register is a Census

Bureau construct based primarily on information from the Internal Revenue Service of the U.S.

Treasury Department.10

It contains information on the number of employees working for an

establishment and total annual establishment payroll. The LBD links the establishments contained

in the Business Register over time and can be matched to Compustat using a bridge file provided

by the US Census.

This Census data is an improvement over the wage and employment data reported in

Compustat. For one, the Census data is available at the establishment level which allows us to

identify changes at each facility instead of relying on firm-level aggregate data. Second, we are

able to observe the state of location for each facility, making it possible to control for geography-

dependant mean wages. Finally, the majority of active firms in Compustat do not report employee

compensation, and the reported employee count and wage data can be unreliable because

personnel information is subject to looser reporting and auditing requirements than financial

variables.

To construct a set of control firms, first we calculate for each ESOP firm 1) total assets, 2)

the average firm-level wage per employee, and 3) the change in average firm-level wage per

employee prior to the ESOP initiation. The change in wages is used to ensure that ESOP firms

and matched firms exhibit a similar trend in wages prior to the year of matching. The change in

wages is defined as (wagest-1 – wagest-2)/wagest-2, where t is the year of ESOP initiation.

10 See Jarmin and Miranda (2002) for more information.

13

We estimate the same variables for a set of potential control firms, which includes all

firms in Compustat that belong to the same industry in the same year that never issued an ESOP.

Industry is classified by the 3-digit SIC code. We estimate the absolute percent difference

between the potential control firm and the ESOP firm on all three variables. Specifically, for

each variable, we calculate a percent difference as (valuecontrol – valueESOP) / valueESOP. These

percent differences are summed up, and the control group is chosen as those firms with the

smallest total differences. For each ESOP firm, we identify the three nearest neighbor matches.

Because we identify the match from the same pool of firms, in some cases a control firm is

matched to multiple ESOP firms. To maintain a sample of independent observations, we require a

match firm appear in the control group as a unique firm by dropping the ―repeats‖.

ESOP firms are included in our sample for the five years before and the ten years after

the ESOP is initiated. We begin five years prior to capture the most current information and

extend to ten years afterward because ESOP shares must be granted to individual employee

accounts within ten years. Observations after ten years are excluded to reduce the impact of

changes unrelated to the ESOP occurring well after the initiation. We also exclude observations

after an ESOP termination to ensure that our baseline is not picking up post-termination effects.11

The same time series is used for the matched group. We keep matched firms in our sample for the

five years before and the up to ten years after the match.

Table 1 lists the number of new ESOP adoptions and observation counts in our ESOP

database by year. It identifies 4,594 firm-year observations between 1982 and 2001 with the

11 There are 56 ESOP terminations (138 plant-year observations) in our ESOP database. Terminating an ESOP is a

complex legal procedure. The firm must be able to legally justify why the ESOP was value-increasing for the firm in

the past but is now value-decreasing; otherwise, it is open to lawsuits from ESOP holders and shareholders. Thus, it is

more common to ―freeze-out‖ an ESOP. A freeze-out usually is not announced officially, so is difficult to identify. In

our sample, firms that elect to freeze-out their ESOP are still recorded as having an ESOP, which is literally true

because the ESOP still exists. There are also some firms that have rolled up their ESOP into a 401-K plan. Such 401-K

plans may still be recorded in our database as an ESOP, which is not completely off-base because they still represent

employee stock ownership.

14

median ESOP controlling 6.70% of shares outstanding.12

Samples used in the subsequent tables

are smaller than the full sample because a number of observations could not be matched to the

Census LBD database. 13

Table 2, Panel A, provides summary statistics of the relevant firm level variables. The

first column details firms that later initiate an ESOP, over the pre-ESOP period of up to five years.

The second column describes firms over the post-ESOP period of up to ten years. The third

column details firms with large ESOPs over the post-ESOP period. An ESOP is considered large

if, at any point during the lifetime of the plan, it controls more than 5% of the outstanding

common shares. We choose this demarcation point because proxy statements only detail the size

if the ESOP controls more than 5% of the firm’s equity. In addition, 5% is often used as a

threshold for various disclosure requirements, presumably because it signifies an important

source of control rights. Of the 410 ESOPs in the sample, 225 achieve a size of 5% or greater at

some point during their lifetime. The median and the mean ownership of these large ESOPs is

12.20% and 18.35% of shares outstanding, respectively. The fourth column summarizes the set of

matched firms over the up-to-16-year period from five years before to ten years after the matched

date.

The summary statistics indicate ESOP firms are more profitable and have higher leverage

as compared to control firms. Furthermore, ESOP firms are larger and valued lower as measured

by (industry-adjusted) Tobin’s Q. The lower valuation is most noticeable for large ESOP firms.

Financial leverage increases following ESOP initiations because they are often debt financed.

Panel B of Table 2 provides summary statistics of relevant payroll information at the

establishment level. We include all establishments owned by either our ESOP group or the

control group. Both pre- and post-ESOP firms have more employees per establishment than the

12 We cannot estimate the mean because proxy statements do not report the size for ESOPs with less than 5% of

outstanding shares. 13 Census confidentiality policy limits the disclosure of the exact sample which was matched to the Census LBD

database.

15

control group. Wages are significantly higher at pre-ESOP firms than at the control sample, and

the difference between these firms increases substantially after the adoption of ESOPs.

III. Employee Compensation

Our empirical investigation begins with the estimation of the relation between employee

cash compensation and the presence of an ESOP, followed by an investigation of the relation

between firm value and ESOPs. Since employees and shareholders are the two main claimants of

a firm’s total surplus, these estimates allow us to draw inferences on how the adoption of an

ESOP is related to firm productivity.

The Census compensation data provides establishment-level annual payroll, which

includes all forms of compensation taxed as ordinary income, such as salaries, wages,

commissions, and bonuses. However, the compensation data do not include the value of ESOP

shares given to employees and, hence, underestimates total employee compensation. Our measure

of wages per employee is the ratio of annual payroll (in thousand dollars, normalized to 2006

dollars) to the number of employees and excludes the value of ESOP shares granted to employees.

This variable is calculated at the establishment level.

A. Univariate Analysis

We begin by examining how wages change around the ESOP initiation. Table 3, Panel A

provides average log wages per employee (in thousands) over a five-year window surrounding

the year of ESOP initiation (year 0), separately for small ESOP firms, large ESOP firms, and the

control group. The five-year window covers -2 to +2 years surrounding the year of ESOP

initiation. We use the log of wages per employee because wages are highly skewed. We do not

consider years beyond the five-year window because the high rate of establishment entry and exit

leads to significant changes to the sample over time. That is, the samples used to calculate the

year 0 average and the year + 5 averages are quite different. However, this is not a concern in the

next section where regressions control for establishment-fixed effects.

16

Because wages are affected by location- and industry-specific factors, Panel B of the

Table adjusts for mean wages in the state of location and in the same industry in the same year.

Unexplained wages are residuals from the following regression:

Wagesit= α0 + 1state-year mean wagesit + 2industry-year mean wagesit + μit (1)

Subscripts i and t indicate establishment i and year t, and wages are the log of the average wage

per employee. State-year mean wage is the log mean wage per employee in the state of location

of the establishment in the same year. Industry-year mean wage is the log mean wage per

employee matched to the establishment’s industry and by year.

Both panels reveal wage decreases prior to the ESOP initiation. Both raw and

unexplained wages show declines from year -2 to year -1 for both small and large ESOP firms.

One concern upon seeing this pre-initiation trend is that it may predict higher wages post-ESOP if

wages follow a mean-reverting process. This is why we require that firms included in our control

group have similar trends in raw wages prior to the year of matching. The last column of Panel A

shows that the control group also experiences a decline in raw wages from year -2 to year -1.

This decline in wages is reversed starting the year of ESOP initiation. For ESOP firms,

both the raw and abnormal wages increase sharply in years 0 and +1, and year +2 wages stay

substantially higher than that in year -1. The raw wages also increase for the control group;

however, unlike the ESOP firms, the control group shows no improvement in unexplained wages,

which remain negative throughout the five-year window. This difference between the treatment

and control groups suggests that workers enjoy higher wages following ESOP adoptions.

B. Multivariate Analysis

The univariate analyses do not control for relevant establishment and firm characteristics.

In this section, we estimate the relation between ESOPs and employee compensation at the

establishment level in a differences-in-differences like approach using all treatment and control

firms meeting our sample construction criteria over 1982 to 2001. The baseline regression is:

Wagesit= ηt + θi + α0 + α ESOPit + Zit + μit (2)

17

Subscripts i and t indicate establishment i and year t, and ηt and θi are year- and establishment

fixed effects. Year fixed effects capture economy-wide effects on wages, and establishment fixed

effects control for time-invariant establishment characteristics. ESOPit includes ESOP and

ESOPg5 indicators: ESOP is equal to one if the firm has an ESOP; and ESOPg5 is equal to one if,

at its maximum, employees control more than 5% of outstanding shares through the ESOP. The

regression estimates the mean difference in wages before and after ESOP adoptions, after

controlling for other firm characteristics.

The set of control variables, Zit, include state-year mean wages measured for the state of

location of an establishment in the same year, excluding the establishment itself from the mean.

This variable controls for changing local conditions affecting wages over time (Bertrand and

Mullainathan, 2003). We also control for industry-year mean wages, the mean wages of all

establishments operating in the same industry as the establishment in the same year, excluding the

establishment itself from the mean. To control for changing establishment and firm characteristics,

we include establishment age and total sales at the firm level. We also control for financial

leverage because some ESOPs are debt financed.14

Berk, Stanton, and Zechner (2010) argue that

higher leverage imposes greater risk on employees, resulting in higher wages in equilibrium.

Wages at establishments owned by the same firm may be correlated, understating the

standard errors. We correct standard errors for clustering at the firm-level. In addition, we

exclude transition periods, the year of the announcement of ESOP adoption and the year after

because ESOP implementation takes time. Jones and Kato (1995) show that it takes about three

years before ESOPs show effects on worker productivity.

B.1. Basic Results

The first column in Table 4 reports estimation results of the baseline regression. As

expected, establishment-level wages are highly correlated with the concurrent average wages in

the same state and the same industry. However, the coefficient on ESOP is insignificant. This is

14 In a leveraged ESOP, the firm purchases the shares to be allocated to the ESOP using a new debt issuance.

18

because treating all ESOPs alike with a single ESOP indicator masks important heterogeneity

across ESOPs of different size. When Column (2) adds an indicator for large ESOPs, the

estimation result reveals a sharp difference between small and large ESOPs. The coefficient on

ESOP indicates a significant wage increase following the adoption of a small ESOP.15

The

magnitude of the coefficient implies a 6.1% average wage increase following the initiation of

small ESOPs. Since the mean wage at pre-ESOP firms is $40,522 (Table 2, Panel B), the 6.1%

increase amounts to $3,114 (in 2006 dollar). Because the value of the ESOP shares allocated to

employees is not accounted for, these wage increases underestimate total compensation increases.

This positive relation between ESOP adoptions and wages applies to only small ESOPs.

The coefficient on ESOPg5 is negative. Because the coefficients on the ESOP indicator variables

are cumulative, the combined coefficient on ESOPg5 is 0.061 – 0.077, or -0.016. Column (3)

enters ESOPg5 alone and shows an insignificant coefficient, indicating no changes in wages

following large ESOPs.

B.2. Value of Shares Granted through Large ESOPs

Looking at wage changes alone understates compensation changes because the value of

ESOP shares is unaccounted for, and the underestimation is greater for large ESOPs. The average

size of shares granted to employees through a large ESOP is 18.35% of the firm’s market

capitalization.16

The average market capitalization of a firm with a large ESOP is $3.5 billion in

2006 dollars and has 23,959 employees. Thus, the average large ESOP has a total value of $642

million, which translates into $26,796 per employee. Given that the average annual wage for

workers at these large ESOP firms is $52,981, the value of the ESOP shares allocated would be

equal to 5.06% of annual wages if the shares were allocated equally over ten years.

15 In a typical down-sizing, low wage workers are fired first, which may lead to an increase in mean establishment

wages due to the shift in the worker composition towards higher-paid employees. This is an unlikely explanation for

our finding because the mean employment at ESOP firms over the sample period increases by 4.1%, which is

comparable to the mean employment increase of 4.2% at the group of control firms. 16 We do not have the equivalent statistic for small ESOPs because, as mentioned earlier, proxy statements do not

provide the percentage held in ESOP accounts if it is below 5%.

19

Risk-averse workers will value ESOP shares less than cash wages because of the sales

restrictions on ESOP shares and their associated risk. Thus, even when we account for the value

of shares granted, the average compensation changes associated with large ESOPs are less than

those associated with small ESOPs. However, the value of the shares granted through large

ESOPs seems substantial, suggesting that employees are better off, on average, following the

adoption of large ESOPs.

B.3. Robustness

Bertrand, Dufflo, and Mullainathan (2004) point out standard errors from difference-in-

differences tests can be biased in the presence of serial correlation in the dependant variable.

Thus, we follow an alternative estimation procedure they suggest. In step 1, we estimate the

following regression:

Wagesit= ηt + θi + α0 + Zit + μit.

Subscripts i and t indicate firm i and year t, and ηt and θi are year- and firm fixed effects. Each

observation reflects the equally-weighted mean of all establishment observations for that firm-

year. Zit is the same set of control variables as in the baseline regression. This first step regression

is estimated on the full sample of both ESOP and control firms.

In step 2, we use the estimated residuals from step 1, but retain these values only for

ESOP firms, thus, our sample in step 2 is limited to firms which either currently have an ESOP or

will adopt an ESOP in the future. In addition, we use only one observation for each firm for the

pre-ESOP period and one observation for each firm for the post-ESOP period, where each

observation reflects the equally-weighted mean of all establishment observations for that firm-

year. We then regress the residuals on the ESOP dummy variables.

Table 5, Column (1) shows a positive and statistically significant coefficient on ESOP,

illustrating the robustness of the significant wage gains following the adoption of ESOPs. In

column (2), we add the indicator for large ESOPs. The coefficient on ESOP hardly changes and

remains significant, while the coefficient on ESOPg5 becomes insignificant. As noted by

20

Bertrand et al. (2004), this type of test has limited power to reject the null. Nevertheless, the

insignificant coefficient for large ESOPs calls for further investigation of large ESOPs, which is

conducted in the later section. There we identify an important sub-group of large ESOP adoptions

followed by wage cuts, while the rest of large ESOPs are associated with wage gains.

C. Worker Mobility

The main inference that can be drawn from our results so far is that employees are

generally better off financially following ESOP adoptions. In this and the following sections, we

provide additional evidence suggesting the relation is causal – that ESOP adoptions cause wage

changes. Because we control for establishment fixed effects, we are controlling for time-invariant

establishment characteristics. Thus, our concern is that a time-varying omitted variable may

predict both the timing of ESOP establishment and future wage changes. For example, a CEO

may expect strong profits in the future and decide to establish an ESOP and increase cash wages

to share these expected gains with his employees. In this case, we would expect to observe higher

wages at ESOP firms even though the ESOP itself did not cause the wage increases.

By contrast, our primary hypothesis is that ESOPs increase productivity through

improved team incentives and co-monitoring and that labor captures a fraction of these gains in

the form of higher wages. To separate this casual interpretation from the non-causal story, we

investigate how shocks in employee mobility affect the division of the gains. We hypothesize that

if ESOPs increase productivity, then the surplus will be shared between employees and

shareholders according to their relative bargaining power. We proxy employees’ bargaining

power by their mobility and predict that employees will capture more of the surplus as their

mobility increases.

C.1. Measuring Mobility

To construct a measure of worker mobility, we assume (1) labor markets are

geographically constrained due to pecuniary and non-pecuniary costs of moving and (2) workers

prefer to remain employed in the same industry due to industry-specific human capital. These

21

assumptions allow us to measure worker mobility within an industry-location pair. We classify

industry by three-digit SIC code and geographic location by Metropolitan Statistical Area (MSA).

Then, we calculate a Herfindahl index of employer concentration for each industry-MSA pair.

This index is created in a manner similar to a traditional sales-based Herfindahl index, except the

measure is based on the fraction of the industry’s workforce employed at establishments owned

by a firm, rather than the fraction of industry sales.

In the first step, we sum the total number of workers at all establishments linked to a firm

for each industry-MSA pair.17

The firm’s share of employees is the number of its employees

divided by the total employees in that industry-MSA pair. The Herfindahl index is then calculated

as the sum of the squares of the employee share of all firms in that industry-MSA pair. This

measure is estimated on an annual basis. Employees at an establishment are defined as having

high (low) mobility if the establishment’s industry and MSA combination has a Herfindahl value

in the bottom (top) sample quartile. We assign indicator variables for high- and low worker

mobility.

C.2. Specification

The specification is the baseline regression with two additional independent variables: the

worker mobility indicator and its interaction with the ESOP dummy. This specification allows us

to interpret the coefficient on the interaction term as the effect on wages at ESOP firms following

an exogenous shock to worker mobility. To do so, we need to 1) ensure the result is not driven by

a firm’s endogenous decision to move and 2) control for any direct relation between worker

mobility and wages affecting all firms in a given industry-location pair.

17 We start by using all establishments included in the Standard Statistical Establishment List (SSEL) available through

the US Census. The SSEL identifies the total count of employees at all public and private U.S. business establishments.

It also includes information on the establishment industry (SIC code) and location (state and county Federal

Information Processing Standard (FIPS) codes). Using the state and county FIPS codes, we map each establishment

located in an MSA to its appropriate MSA. An MSA is defined by the US Census as a geographic area that consists of a

core urban area (which can be as small as 10,000 people) and all surrounding but integrated counties. The link between

FIPS and MSA is provided by the US Census for 1981, 1983, 1990, 1993, and 1999 at:

http://www.census.gov/population/www/metroareas/pastmetro.html. We use the most recent link for each match to

allow for changes to MSA definitions over time.

22

Excluding endogenous changes to worker mobility is important because a firm may

choose to move to locations where worker mobility is higher at times when the firm is better able

to afford higher wages, causing a positive correlation between worker mobility and wages. Our

specification avoids this endogenous change because the Census assigns a new identification

number to any establishment changing location. Consider an establishment moving from a low

worker mobility location to a high worker mobility location. In our data, this is identified as an

establishment in the low-worker mobility location for the years it is located in the first MSA and

then as a separate establishment in the high worker-mobility location in the new MSA for the

years the establishment is located in the second MSA. Thus, assuming no further changes to

worker mobility for this establishment, this establishment will not be reflected in the coefficient

of the high worker mobility dummy. To be reflected in this coefficient, the same establishment

must be observed with both a high and a non-high worker mobility classification. When a firm

moves an establishment, the establishment will be assigned a new identifier after the move, which

means the post-move establishment will have no pre-move observations. Thus, although moves

by firms will affect the local worker mobility, it will not affect our estimate of the coefficient on

the worker mobility dummy variable.

Controlling for the direct relation between worker mobility and wages is also important

because worker mobility may affect wages at all firms, as employees with more bargaining power

demand higher wages. This effect should be stronger at firms recently experiencing a productivity

shock. The fair wage hypothesis of Akerlof and Yellen (1990) suggests employees’ estimation of

a fair wage is made relative to other groups. In our case, shareholders are the benchmark group.

Thus, if ESOPs increase firm surplus and shareholders gain, employees may demand their share.

Such demands are more likely to be met when employees’ external employment opportunities

increase, giving them greater mobility and greater bargaining power. Thus, we expect to observe

wage increases at ESOP firms following an increase in worker mobility. At control firms, we may

also observe a positive relation between wages and increases in worker mobility, however,

23

without a recent productivity shock, there may be little to no new surplus that can be shared. Thus,

this specification is a joint hypothesis test. We expect to observe a significant positive coefficient

on the interaction term of high worker mobility and ESOP if 1) ESOP firms, on average,

experience a productivity shock greater than the average productivity shock at non-ESOP firms

and 2) increased worker mobility enables employees to take a greater share of productivity gains.

C.3. Estimation Results

Table 4, Column 4, includes the indicator for high worker mobility and interacts it with

the ESOP indicator. The interaction term shows a positive and significant coefficient, indicating

that wages at ESOP firms increase as worker mobility increases. Notice that because of

establishment fixed effects in the panel regression, identification is obtained when a non-high-

mobility establishment’s worker mobility increases sufficiently enough to be classified as high-

mobility. Hence, the switches to the top quartile of worker mobility are likely to represent large

increases in employee bargaining power.

This result is consistent with a causal relation between the ESOP and wage increases.

The ESOP leads to an increase in productivity at the firm, creating a surplus. This surplus is then

shared between employees and shareholders according to their relative bargaining power. Such a

finding is not predicted by a selection story in which generous managers time an ESOP

implementation when anticipating greater productivity in the future.

The coefficient on the high worker mobility dummy is insignificant, which may be due to

the presence of offsetting relations between worker mobility and wages for control firms.

Endogenous location decisions may predict a negative relation between our measure of worker

mobility and wages. Firms may choose to enter local labor markets where they anticipate

downward wage pressure in their industry. However, an increase in the demand for industry-

specific skills and worker mobility due to new entries may offset the possible negative relation

due to the location choice.

24

Column 5 includes the indicator for low worker mobility. The interaction term between

the low mobility indicator and the ESOP indicator is insignificant. This implies that ESOP

employers do not take advantage of a decrease in worker mobility to reduce the fraction of

surplus given to employees. The asymmetric results between Columns (4) and (5) could be due to

stickiness in wages. Although employers have to grant higher wages to retain workers with

improved mobility, reducing employees’ share of productivity gains may make it difficult to

sustain the productivity gains, especially if they arise from improved incentive- and team effects.

D. Worker-Management Alliance

Another causal scenario developed in the hypotheses section is that the top management

bribes employees with above-market wages to garner worker support in thwarting hostile

takeover bids (Pagano and Volpin, 2005). This anti-takeover motive is more relevant to large

ESOPs because the number of voting rights matters in takeover battles.

Large ESOPs become more effective anti-takeover devices when states of incorporation

of the ESOP firms enact BCS, providing the incentive for management to strengthen their

alliance with employees. BCS were enacted at the state level in a staggered fashion during our

sample period; New York was the first state to pass it in 1985. Romano (1987) and Bertrand and

Mullainathan (2003) argue that the enactment of BCS is exogenous to most firms incorporated in

the affected states. Using a difference-in-differences like approach, they document significant

increases in wages following the enactment of BCS, which they attribute to management’s pursuit

of quiet lives after BCS relieves them of hostile takeover threats.

To check whether the wage increases accompanying ESOPs are reflecting the state-wide

BCS effect, we estimate the BCS effect on wages in Table 6, column 1. The regression

specification is similar to that in Bertrand and Mullainathan (2003), although data and some

control variables differ. Table 6, Column (1) shows a positive but insignificant coefficient on the

BCS indicator. This result is weaker than that reported by Bertrand and Mullainathan (2003),

perhaps due to different samples. They examine all firms in manufacturing industries over the

25

period 1976-1995, while we cover all industries but limit the sample to ESOP firms and our

matched control firms over the period 1982-2001.

The worker-management alliance hypothesis predicts greater wage gains following the

passage of BCS at firms with large ESOPs in place. Thus, we compare the difference between

large ESOP firms and control firms by adding an interaction term between large ESOPs and the

BCS indicator. The result is reported in Column (2).18

It shows a positive and significant

coefficient, implying that the effect of BCS passage is greater at firms with large ESOPs in place,

in comparison to the control firms or firms with small ESOPs. The coefficients indicate that, on

average, worker wages at firms with large ESOPs increase by 7.3% after the adoption of BCS. In

an unreported regression, we add an interaction term between ESOP and BCS. The result shows

an insignificant coefficient, indicating the positive effect of BCS on wages is unique to large

ESOPs.

These wage increases at large ESOP firms post-BCS imply that management strengthens

its alliance with employees when BCS makes their voting rights through ESOP shares especially

effective in thwarting hostile bids. The results also identify a firm-level channel for the state-wide

BCS effect on wages documented by Bertrand and Mullainathan (2003). In addition, our result is

inconsistent with the non-causal story, which says nothing about possible impacts the passage of

the BCS may have on wages of firms with large ESOPs.

E. Cash Conservation

The third causal scenario developed in the hypotheses section is that some large ESOPs

are motivated to conserve cash by substituting cash wages with company shares. Such ESOPs

18 Column (2) also contains a new control variable, Yeardifg5, which is the number of years since the initiation of a

large ESOP and is set to 0 for firms without large ESOPs. The variable is added to control for possible dilution of the

effect of a large ESOP over time as employees leave the firm with their shares. Because BCS were never repealed, the

average year distance from ESOP initiation for the sample captured in the BCS*ESOP interaction dummy variable will

always be greater than the average year distance from ESOP initiation for the sample captured by the ESOPg5 dummy

variable, raising the possibility that the interaction term could be picking up the dilution effect. That is why Yeardifg5

is added as a control variable.

26

will lead to wage cuts and are most likely to be observed at firms that are cash constrained and

that have difficulty raising external equity through regular channels.

Since small and young firms tend to have more difficulty accessing the market for

external equity, we construct CCindex, a continuous variable measuring how small and young a

firm is relative to the rest of the sample. First, we calculate the distance in firm age and size (total

assets in 2006 dollars) from the sample means. Then, both of these distances are normalized by

the sample standard deviation.19

These two variables are then summed to create a cash

constrained score, which gives higher values to younger and smaller firms. We find the score

highly skewed; thus, we rank firms based on the score. We divide the ranked value by 100 and

define it as CCindex, with the numerical values ranging from 0.01 (the least cash constrained) to

slightly under 10 (the most cash constrained). This variable is calculated for ESOP firms at the

time the ESOP is initiated; for control firms, at the time the firm is matched to an ESOP firm.

The results using this cash constrained score are reported in the last two columns of Table

6. Column (3) adds CCindex to the baseline regression and interacts it with ESOPg5. The

interaction term is negative and significant, indicating that post-ESOP wages are lower, the more

cash constrained an ESOP initiating firm is at the time of an ESOP adoption. The coefficients on

ESOP, ESOPg5, and CCindex*ESOPg5 indicate substantial wage cuts following large ESOPs

implemented by cash constrained firms. However, those implemented by least cash constrained

firms (i.e., those with the lowest ranking in CCindex) are followed by wage increases.20

To investigate whether the cash constraint hypothesis also applies to small ESOPs,

Column (4) adds the interaction term CCindex*ESOP, which shows an insignificant coefficient. It

appears the cash conservation motive applies mainly to large ESOPs. If a firm chooses to use

19 For firm age, this variable is calculated as: [Firm age – mean age]/sample standard deviation of age. The same

normalization process is used for firm size. 20 The 100 least constrained firms represent roughly the bottom deciles out of 946 ESOP and control firms, with

CCindex values ranging from 0.01 to 1. For a CCindex value equal 1, the coefficients imply about a 0.9% average

wage increase.

27

ESOPs as a means to conserve cash, it is unlikely to do it through small ESOPs, which by

definition means issuing less than 5% of equity.

In sum, lower post-ESOP wage gains associated with large ESOPs appear to be driven by

the substitution of cash wages with ESOP shares by cash constrained firms with limited access to

external equity. Large ESOP adoptions by the least constrained firms are followed by positive

cash wage increases. Although the magnitude is smaller than the wage gains associated with

small ESOPs, employees receive more company shares through large ESOPs than small ESOPs.

Therefore, we conclude that ESOPs are followed by significant increases in employee

compensation, except for large ESOPs by small and young firms.

IV. Firm Valuation

Our investigation of wage changes shows that employees are in general better off with

ESOPs. This is consistent with our hypothesis that ESOPs increase worker productivity through

improved team incentives and co-monitoring and workers capture a share of the productivity

gains. This causal interpretation is buttressed by the evidence that worker wages at ESOP firms

increase as worker mobility increases. However, our evidence does not rule out the possibility

that employees gain at the expense of shareholders through their control rights newly bestowed

through ESOPs. In this section, we address this issue by investigating how ESOPs affect

shareholder value.

Firm valuation is proxied by industry-adjusted Tobin’s Q. Q is measured as fiscal

yearend market value of equity plus market value of preferred stock plus total liabilities divided

by the book value of total assets. Following Bebchuk and Cohen (2005), we measure industry-

adjusted Q by subtracting the median Q matched by industry (three-digit SIC code) and year. The

baseline panel regression is:

Ind-Adj Qit= ηt + θi + α0 + α ESOPit + Zit + μit (3)

Subscripts i and t indicate firm i and year t, and ηt and θi are year- and firm fixed effects. ESOPit

includes ESOP and ESOPg5, and Zit is a set of control variables. We include firm fixed effects to

28

control for time-invariant firm characteristics. We also control for year fixed effects. The initial

set of control variables include the log of total assets and the log of sales (both normalized in

2006 dollars). We use the same ESOP sample and control group as with the wage regressions.

The only difference is the use of firm-level data instead of the establishment-level data.

A. Firm Value and Productivity Gains

Table 7, Column (1) shows that the presence of an ESOP is associated with a positive but

insignificant increase in industry adjusted Q. In Column (2), we add ESOPg5, which shows a

significant negative sign. As in wage regressions, separating ESOPs by small- and large size

increases both the magnitude and statistical significance of the ESOP coefficient. The positive

coefficient on ESOP in Column 2 implies that firms establishing small ESOPs realize about a

17% increase in firm valuation relative to the sample mean. When this evidence is considered

together with the 6.1% average wage gains associated with small ESOPs, they imply substantial

productivity gains associated with small ESOPs.

Because the coefficients on the ESOP indicator variables are cumulative, the combined

coefficient on ESOPg5 is 0.174 – 0.175, or -0.001. Column (3) enters ESOPg5 alone and shows

an insignificant coefficient. The average large ESOPs are associated with no significant changes

in firm value or in wages.

To check robustness, Columns (4)-(5) add different combinations of additional time-

varying observable firm-level variables as controls. They include: R&D/Sales, the ratio of R&D

expenditures to sales; CapEx/Assets, the ratio of capital expenditures to total assets; adjusted

sigma, firm idiosyncratic risk measured as the standard deviations of the residuals from a CAPM

model estimated over the previous fiscal year; and adjusted sigmaDum, a dummy variable which

takes the value of one if the data to estimate adjusted sigma is available, and zero otherwise. 21

We also allow for non-linear firm size controls. Previous studies document significant

21 The is the same method used by Himmelberg, Hubbard, and Palia (1999) to avoid reducing the sample size due to

missing data.

29

correlations between these variables and Q (e.g., Himmelberg, Hubbard, and Palia, 1999; Kim

and Lu, 2011). As in previous regressions, all independent variables are lagged by one year.

Column (4) includes these control variables and the log of total assets and its square.

Column (5) does the same with sales instead of total assets. Both columns show significantly

positive coefficients on ESOP and significantly negative coefficients on ESOPg5, with the sum of

the two coefficients netting out to zero.

B. Worker Mobility

Our wage analyses show that wages increase when employee mobility increases at ESOP

firms, implying that increases in bargaining power due to increased mobility enable employees to

capture a greater share of productivity gains from ESOP adoptions. A corollary of this inference

is that shareholders’ share of the productivity gains declines as employee mobility increases. To

test this prediction, Table 8 estimates how changes in worker mobility affect shareholder value.

Column (1) includes the high worker-mobility indicator and interacts it with ESOP. The

estimation result shows a negative and significant coefficient on the interaction term, indicating

that stockholders’ share of the productivity gains declines as worker mobility increases. Column

(2) includes the low worker-mobility indicator and interacts it with ESOP. As in the wage

regression, the interaction term shows an insignificant coefficient.

These results on Q are consistent with the wage results. This consistency between the

wage and Q regression results provides further support for our hypothesis that ESOPs enhances

worker productivity and the gains are shared between shareholders and workers, with the

divisions decided by employees’ bargain power stemming from their mobility.

C. Time-varying, Non-causal Interpretations

Our evidence of small ESOPs increasing firm value is subject to alternative

interpretations based on time-varying firm characteristics. Maybe small ESOPs are established by

managers who have inside information that firm value will increase more than the market

anticipates. However, the result on worker mobility casts doubt on this interpretation. It begs the

30

question of why firm value increases only with small ESOPs, because if the information story is

valid, the effect should be stronger with larger ESOPs. Yet we find no value increases associated

with large ESOPs. Moreover, Figure 1 provides further evidence inconsistent with this non-causal

interpretation of our findings.

Figure 1 plots the mean sample industry-adjusted Q for firms with ESOPs around the

year of ESOP initiation (year 0), separately for small and large ESOPs. For either small or large

ESOPs, we observe no discernable pattern of changes in Q prior to the ESOP initiation. However,

starting the third year after initiation, we observe a steady increase in Q over five years for small

ESOPs, but not for large ESOPs.22

This rise in Q could be due to survivorship bias: bad firms

disappear from the sample over time and only good (high-Q) firms are left in the later years. To

check this possibility, we create the ―Constant Sample,‖ which is a subset of the full sample. To

be included in the Constant Sample, a firm must survive the full 16 years from year -5 through

year +10. The Constant Samples follow patterns similar to the full samples, rejecting the

survivorship bias in favor of the interpretation that it takes about three years after the adoption of

small ESOPs to start realize gains in Q. This three-year lag helps reject the story that small

ESOPs are established by managers who have inside information that firm value will increase

more than the market anticipates. It is unlikely for management to possess such information three

years ahead of the market’s realization. If management had the favorable information, it is

implausible that it would refrain from sharing the information with the market for three years.

Another time-varying alternative interpretation is that management is concerned about

possible takeover bids and implements small ESOPs to deter hostile takeovers; hence, the

announcement of a small ESOP initiation reveals private information on the likelihood of a

hostile takeover bid. Such a scenario is more plausible for large ESOPs because if ESOPs are

22 Previous studies also provide evidence consistent with our causal interpretation. Jones and Kato (1995) control for

potential endogeneity by repeating their tests using lagged values of labor and capital as instrumental variables for

current labor and capital and find a three-year lag before productivity gains are realized.

31

used as an anti-takeover device, large ESOPs are more effective than small ESOPs because they

have more voting rights. Furthermore, revelation of such information is likely to be reflected

immediately in stock prices, not three years later.

Figure 1 also helps reject a ―leverage‖ interpretation of our results. One may argue the

increase in Q is due to an increase in leverage often accompanying an ESOP—e.g., leveraged

ESOPs. If leverage is driving the value increase, we should observe a jump in Q at year 0 and a

gradual decline afterward as the ESOP debt gets paid off.

The three-year lag, however, raises another puzzle: Why didn’t the market anticipate the

gains at the announcement of ESOP adoptions? The evidence that it took three years for small

ESOPs to experience positive effects on firm value does not necessarily imply informational

inefficiency in the stock market. The initiation and implementation of an ESOP takes time and its

effect is not immediate. The complexity of an ESOP makes it difficult to assess its full impact in

the earlier years, because many details of ESOP share allocation are unknown at the

announcement. It is also uncertain how individual workers’ behavioral reaction to becoming an

owner-employee or the prospect of receiving more shares will be sorted out in the group

dynamics of the work place. These complexities and uncertainty require time for resolution.

Furthermore, as Table 1 indicates, a large majority of ESOPs in our sample were initiated

in the late 1980s and early 1990s. At that time, ESOPs were relatively new, offering few

examples of the long-term consequences of these plans. The market may have misjudged the full

value implications of these plans. It was also a period when the media reaction to ESOPs focused

heavily on the anti-takeover implications of these plans. Since hostile takeovers are made at

substantial premiums, when ESOPs are viewed as an anti-takeover device, they may not have

favorable impact on firm value. For example, Gordon and Pound (1990) and Chang and Mayers

(1992) find that ESOPs adopted as a takeover defense are met with negative stock price reactions.

V. Conclusion

32

This paper investigates whether adopting a broad-based employee stock ownership plan

enhances productivity by improving team incentives and co-monitoring. That is, does employee

capitalism work? If so, how are gains divided between shareholders and employees?

We find that small ESOPs increase productivity. Unlike Jones and Kato (1995) on

Japanese ESOPs, our evidence of productivity gains is based on the effects on two main

beneficiaries of such gains: employees and shareholders. Because our evidence indicates both

stakeholders gain from adopting small ESOPs, we infer employee share ownership increases the

size of the economic pie by improving worker productivity.

This causal interpretation is substantiated by our evidence on how the division of

productivity gains is related to employee mobility within an establishment’s industry and location

of work place. We find that when labor mobility increases, increasing workers’ bargaining power

vis-à-vis shareholders’, employees’ share of gains increases and stockholders’ share decreases.

Large ESOPs, defined as those controlling more than 5% of shares outstanding, have a

more or less neutral effect on both cash wages and shareholder value, suggesting that the average

gains are limited to the value of ESOP shares granted to employees. The lower average gains are

due to heterogeneity in motives for establishing large ESOPs. We identify two subgroups of large

ESOPs with motives unrelated to improving group incentives and co-monitoring: cash

conservation and entrenchment against threats from the market for corporate control.

Large ESOPs motivated to conserve cash are followed by wage cuts. By contrast, those

motivated to thwart hostile takeover bids provide substantial wage gains, a previously

undocumented cost of managerial entrenchment to shareholders. With these exceptions

notwithstanding, the overall implication we draw from this study is that employee capitalism

works through group incentives and co-monitoring when ESOPs are designed for that purpose.

References

Akerlof, G. and Yellen, J., 1990. The fair wage-effort hypothesis and unemployment. Quarterly

Journal of Economics, 105:225-283.

33

Atanassov, J. and Kim, E. H., 2009. Labor and corporate governance: International evidence from

restructuring decisions. Journal of Finance, 64:341-374.

Beatty, A., 1995. The cash flow and informational effects of employee stock ownership plans.

Journal of Financial Economics, 38:211-240.

Bebchuk, L. and Cohen, A., 2005. The costs of entrenched boards. Journal of Financial

Economics, 78:409-433.

Berk, J., Stanton, R., and Zechner, J., 2010. Human capital, bankruptcy, and capital structure.

Journal of Finance, 65:891-926.

Bertrand, M. and Mullainathan, S., 2003. Enjoying the quiet life? Corporate governance and

managerial preferences. Journal of Political Economy, 111:1043-1075.

Bertrand, M., Duflo, E., and Mullainathan, S., 2004. How much should we trust differences-in-

differences estimates? Quarterly Journal of Economics, 119:249-275.

Bhaskar, V., Manning, A., and To, T., 2002. Oligopsony and monopsonistic competition in labor

markets. Journal of Economic Perspectives, 16:155-174.

Blasi, J., Conte, M., and Kruse, D., 1996. Employee stock ownership and corporate performance

among public companies. Industrial & Labor Relations Review, 50:60-79.

Chang, S. and Mayers, D., 1992. Managerial vote ownership and shareholder wealth: Evidence

from employee stock ownership plans. Journal of Financial Economics, 32:103-132.

Chaplinsky, S. and Niehaus, G., 1994. The role of ESOPs in takeover contests. The Journal of

Finance, 49:1451-1470.

Core, J. and Guay, W., 2001. Stock option plans for non-executive employees. Journal of

Financial Economics, 61:253-287.

Cramton, P., and Mehran, H., Tracy, J., 2007. ESOP fables: The impact of employee stock

ownership plans on labor disputes. Federal Reserve Bank of New York:

www.newyorkfed.org/research/stafreports/sr347.pdf.

Faleye, O., Mehrotra, V., Morck, R., 2006. When labor has a voice in corporate governance.

Journal of Financial and Quantitative Analysis, 41:489-510.

Gordon, L. and Pound, J., 1990. ESOPs and corporate control. Journal of Financial Economics,

27:525-555.

Hadlock, C. and Pierce, J., 2010. New evidence on measuring financial constraints: Moving

beyond the KZ index. Review of Financial Studies, 23:1909-1940.

Hildreth, A. and Oswald, A., 1997. Rent-sharing and wages: Evidence from company and

establishment panels. Journal of Labor Economics, 15:318-337.

34

Himmelberg, C., Hubbard, R., and Palia, D., 1999. Understanding the determinants of managerial

ownership and the link between ownership and performance. Journal of Financial Economics,

53:353-384.

Ittner, C., Lambert, R., and Larcker, D., 2003. The structure and performance consequences of

equity grants to employees of new economy firms. Journal of Accounting and Economics, 34:89-

127.

Jarmin, R. and Miranda, J., 2002. The Longitudinal Business Database. CES Working Paper No.

02-17.

Jones, D. and Kato, T., 1995. The productivity effects of employee stock-ownership plans and

bonuses: Evidence from Japan. American Economic Review, 85:391-414.

Kandel, E. and Lazear, E., 1992. Peer pressure and partnerships. Journal of Political Economy,

100:801-817.

Kim, E. and Lu, Y., 2011. CEO ownership, external governance, and risk-taking. Journal of

Financial Economics, forthcoming.

Kruse, D., 1996. Why do firms adopt profit-sharing and employee ownership plans? British

Journal of Industrial Relations, 34:515-538.

Kruse, D., Freeman, R., and Blasi, J., 2010. Shared Capitalism at Work: Employee Ownership,

Profit and Gain Sharing, and Broadbased Stock Options. In Shared Capitalism at Work:

Employee Ownership, Profit and Gain Sharing, and Broad-Based Stock Options. Edited by

Douglas Kruse, Richard Freeman and Joseph Blasi. University of Chicago Press: Chicago.

Lazear, E., 2000. Performance pay and productivity. American Economic Review, 90:1346-1361.

Manning, A., 2003. Monopsony in Motion: Imperfect competition in labor markets. Princeton

University Press. Princeton, NJ.

Milgrom, P., 1988. Employment contracts, influence activities and efficient organization design.

Journal of Political Economy, 96:42-60.

Oyer, P. and Schaefer, S., 2005. Why do some firms give stock options to all employees ?: An

empirical examination of alternative theories. Journal of Financial Economics, 76:99-133.

Pagano, M. and Volpin, P. F., 2005. Managers, workers, and corporate control. Journal of

Finance, 60:841-868.

Rauh, J., 2006. Own company stock in defined contribution pension plans: A takeover defense?

Journal of Financial Economics, 81:379-410.

Romano, R., 1987. The political economy of takeover statutes. Virginia Law Review, 73:111-199.

35

Table 1: Counts of Employee Stock Ownership Plan (ESOP) Initiations and Firm-year

Observations by Year.

Fiscal Year

ESOP Initiations Count of ESOP

firm-year

observations

1982 2 6

1983 5 13

1984 8 22

1985 13 38

1986 14 50

1987 24 72

1988 36 105

1989 82 189

1990 53 247

1991 16 262

1992 22 275

1993 10 314

1994 24 332

1995 15 349

1996 26 388

1997 18 396

1998 16 393

1999 17 396

2000 7 381

2001 2 362

Total 410 4,594

36

Table 2, Panel A: Summary Statistics for ESOP Firms and Matched Firms.

Accounting variables are from Compustat. All variables are winsorized at 1%. Assets and sales

are normalized to $2006. Qit is fiscal year-end market value of equity plus market value of

preferred stock plus total liabilities divided by total assets. We industry-adjust Q by subtracting

the median Q matched by industry (3-digit SIC code) and year. Means are reported with median

in parenthesis and standard deviations in brackets. Pre-ESOP Firms include firm-year

observations before the ESOP is adopted for firms that eventually adopt an ESOP. Post-ESOP

Firms include all firm-year observations for firms with ESOPs. Firms with ESOPg5 include all

firm-year observations for firms with large ESOPs, defined as ESOPs that control 5% or more of

the firm’s outstanding common stocks at its maximum point. Matched firms are a group of

control firms, which do not adopt ESOPs, matched to the ESOP firms.

Pre-ESOP Firms Post-ESOP Firms Firms with

ESOPg5

Matched Firms

Operating

Income/Assets

0.129

(0.1)

[0.094]

0.117

(0.1)

[0.091]

0.110

(0.1)

[0.082]

0.099

(0.1)

[0.122]

Leverage 0.169

(0.1)

[0.157]

0.209

(0.2)

[0.172]

0.217

(0.2)

[0.173]

0.188

(0.1)

[0.180]

Assets (millions) 5,377.72

(563.3)

[11,953.69]

7,175.55

(1,529.1)

[13,545.77]

6,418.75

(1,242.2)

[13,039.51]

3,525.10

(327.7)

[9,243.48]

Sales (millions) 3,124.70

(663.1)

[6,233.29]

4,452.52

(1,172.8)

[7,728.11]

4,255.22

(1,187.0)

[7,610.72]

1,569.64

(318.8)

[3,902.00]

Capex/assets 0.070

(0.1)

[0.054]

0.063

(0.1)

[0.048]

0.062

(0.1)

[0.048]

0.063

(0.0)

[0.058]

Q 0.972

(0.8)

[0.760]

1.023

(0.8)

[0.884]

0.868

(0.8)

[0.576]

1.029

(0.8)

[0.949]

Industry-

Adjusted Q

0.082

(0.0)

[0.561]

0.098

(-0.0)

[0.699]

-0.029

(-0.0)

[0.489]

0.129

(-0.0)

[0.763]

N 1480 1884 1136 8265

37

Table 2, Panel B: Summary Statistics for Establishments Owned by ESOP Firms and

Matched Firms.

All variables are winsorized at 1%. Annual payroll and wages per employee are normalized to

$2006. Means are reported with median in parenthesis and standard deviations in brackets. Pre-

ESOP Firms include firm-year observations in the sample years before the ESOP is adopted for

firms that eventually adopt an ESOP. Post-ESOP Firms include all firm-year observations for

firms with ESOPs. Firms with ESOPg5 include all establishment-year observations for firms with

large ESOPs, defined as ESOPs that control 5% or more of the firm’s outstanding common stocks

at its maximum point. Matched firms are a group of control firms, which do not adopt ESOPs,

matched to the ESOP firms.

Pre-ESOP

Firms

Post-ESOP

Firms

Firms with

ESOPg5

Matched

Firms

Annual payroll (thousands) 2,490.32

(371.6)

[6,807.09]

2,479.67

(321.3)

[6,783.98]

2,220.18

(279.4)

[6,407.07]

2,112.14

(298.4)

[6,087.46]

Number of Employees 58.406

(12.0)

[136.42]

52.416

(9.0)

[130.07]

48.049

(8.0)

[126.05]

47.362

(10.0)

[117.93]

Wages per employee

(thousands)

40.522

(33.9)

[30.79]

51.893

(41.1)

[42.11]

52.981

(38.1)

[45.66]

39.693

(31.5)

[30.43]

N 206,433 364,820 232,664 671,504

38

Table 3: Log Wages and Unexplained Wages per Employee.

Panel A reports average log wages per employee (in thousands). Panel B reports average

unexplained wages. Unexplained wages is the residual from the following regression: log wages

per employee = a0 + a1 state-year mean wages + a2 industry-year mean wages + ε. State-year

mean wages is the log mean wage per employee in the state of location of the establishment and

matched by year. Industry-year mean wage is the mean log wage per employee matched to the

establishment’s industry and by year. ESOPg5 captures ESOPs that control 5% or more of the

firm’s outstanding common stock at its maximum point. For the ESOP samples, relative year

represents the year relative to when the ESOP was initiated (year 0). The matched sample is

created at the time the ESOP is initiated and then the matched firms are followed over time. For

the matched sample, the relative year represents the year relative to when the firm was matched to

an ESOP firm initiating an ESOP (year 0.)

Panel A: log wages per employee (in thousands)

Relative Year Small ESOP ESOPg5 Matched Firms

-2 3.462 3.521 3.410

-1 3.203 2.910 2.928

0 3.448 3.564 3.211

1 3.597 3.669 3.406

2 3.604 3.773 3.409

Panel B: unexplained wages

-2 0.020 -0.010 -0.025

-1 0.010 -0.020 -0.019

0 0.045 0.030 -0.014

1 0.090 0.044 -0.016

2 0.055 0.026 -0.025

39

Table 4: Wage Changes Following ESOP Initiations and Changes in Employee Mobility.

The dependant variable is log wages per employee. ESOP is a dummy variable which takes the value of 1 if the firm

has an ESOP. ESOPg5 is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5%

of the firm's outstanding common stocks at any given time. State-year mean wages is the log mean wage per employee

in the state of location of the establishment and matched by year. Industry-year mean wage is the mean log wage per

employee matched to the establishment’s industry and by year. Establishment age and sales are log-transformed. Sales

is normalized to $2006. Worker mobility high (low) is a dummy variable which takes the value of 1 if an

establishment’s industry and MSA combination has a Herfindahl value of employer concentration in the bottom (top)

sample quartile. All regressions include establishment and year fixed effects. The sample includes both ESOP firms and

the matched control sample. Coefficients are reported with standard errors in parentheses. All standard errors are

corrected for clustering at the firm level. "*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%,

respectively.

1 2 3 4 5

ESOP 0.015

(0.016)

0.061

(0.031)

**

0.055

(0.031)

*

0.060

(0.032)

*

ESOPg5 -0.077

(0.043)

*

-0.024

(0.025)

-0.079

(0.043)

*

-0.077

(0.044)

*

State-year mean

wages

0.606

(0.115)

***

0.608

(0.115)

***

0.615

(0.115)

***

0.609

(0.115)

***

0.626

(0.111)

***

Industry-year mean

wages

0.367

(0.110)

***

0.368

(0.110)

***

0.368

(0.110)

***

0.367

(0.110)

***

0.372

(0.110)

***

Establishment age 0.011

(0.015)

0.008

(0.015)

0.014

(0.015)

0.009

(0.015)

0.009

(0.014)

Sales -0.004

(0.014)

0.000

(0.013)

-0.004

(0.013)

*

0.000

(0.013)

-0.003

(0.012)

Leverage -0.040

(0.044)

-0.025

(0.045)

-0.023

(0.046)

-0.025

(0.045)

-0.032

(0.043)

Worker mobility high -0.003

(0.010)

Worker mobility high

* ESOP

0.055

(0.023)

**

Worker mobility low 0.003

(0.009)

Worker mobility low

* ESOP

0.002

(0.014)

N clusters 946 946 946 946 946

N observations 1,023,258 1,023,258 1,023,258 1,023,258 1,023,258

R-squared 0.492 0.493 0.492 0.4927 0.4925

40

Table 5: Difference-in-Differences Control Tests.

The dependent variable is the residual from the following regression: Wagesit= ηt + θi + α0 + Zit

+ μit. Subscripts i and t indicate establishment i and year t, and ηt and θi are year- and

establishment fixed effects. Zit is a set of control variables. The control variables include state-

year mean wages, industry-year mean wages, establishment age, leverage, and sales. The

residuals are estimated using the full sample of both ESOP firms and control firms. The

regressions below are estimated using only the set of establishments owned by firms that adopt an

ESOP during the sample period. ESOP is a dummy variable, which takes the value of 1 if the firm

has an ESOP. ESOPg5 is a dummy variable, which takes a value of 1 if the firm has an ESOP that

controls at least 5% of the firm's outstanding common stocks at any given time. The sample

consists of one observation per firm for the pre-ESOP period and one observation per firm for the

post-ESOP period. Each observation reflects the equally-weighted mean of all establishment

observations for that firm. Coefficients are reported with standard errors in parentheses. "*",

"**", and "***" reflect statistical significance at 10%, 5%, and 1%, respectively.

1 2

ESOP

0.020

(0.010)**

0.019

(0.010)**

ESOPg5

0.007

(0.010)

N observations

495 495

R-squared

0.009 0.010

41

Table 6: Wage Changes Following Enactments of Business Combination Statutes (BCS) and

ESOP Initiations by Cash-Constrained Firms

The dependant variable is log wages per employee. ESOP is a dummy variable that takes the

value of 1 if the firm has an ESOP. ESOPg5 is a dummy variable, which takes a value of 1 if the

firm has an ESOP that controls at least 5% of the firm's outstanding common stocks at any given

time. BCS is a dummy variable which takes the value of 1 if business combination statutes apply

to the establishment. Yeardifg5 is the number of years since the initiation of a large ESOP and is

set to 0 for firms without large ESOPs. CCindex reflects the relative ranking of cash constraints

where a higher value of CCindex implies a more cash constrained firm as proxied by a

combination of smaller and younger firms. Control variables include state-year mean wages,

industry-year mean wages, establishment age, leverage, sales, and establishment- and year fixed

effects. The coefficients of the control variables are not reported. Coefficients are reported with

standard errors in parentheses. All standard errors are corrected for clustering at the firm level.

"*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%, respectively.

1 2 3 4

ESOP 0.061

(0.031)

**

0.057

(0.030)

*

0.058

(0.029)**

0.032

(0.031)

ESOPg5 -0.077

(0.043)

*

-0.083

(0.039)

**

-0.026

(0.038)

-0.001

(0.040)

BCS 0.003

(0.016)

-0.000

(0.017)

BCS * ESOPg5 0.073

(0.027)

***

Yeardifg5 -0.012

(0.005)

**

CCindex -0.003

(0.009)

-0.003

(0.009)

CCindex * ESOP 0.013

(0.009)

CCindex * ESOPg5 -0.020

(0.010) **

-0.033

(0.016)**

N clusters 946 946 946 946

N observations 1,023,258 1,023,258 1,023,258 1,023,258

R-squared 0.4925 0.4929 0.493 0.493

42

Table 7: Industry-adjusted Tobin’s Q Following ESOP Initiations.

The dependant variable is industry-adjusted Q, windorized at 1%. Qit is fiscal year-end market

value of equity plus market value of preferred stock plus total liabilities divided by total assets.

Industry-adjusted Q is obtained by subtracting the median Q matched by industry (3-digit SIC

code) and year. ESOP is a dummy variable, which takes the value of 1 if the firm has an ESOP.

ESOPg5 is a dummy variable, which takes a value of 1 if the firm has an ESOP that controls at

least 5% of the firm's outstanding common stocks at any given time. All regressions include firm

and year fixed effects. Log total assets and log sales are normalized to $2006. The ratio of R&D

to sales is set to 0 if the firm does not report R&D investments. Adjusted sigma is estimated as

the standard deviation of the residuals from a CAPM model estimated over the previous fiscal

year. Adjusted sigmaDum is a dummy variable which takes the value of 1 if the data to estimate

adjusted sigma is available, and zero otherwise. The sample includes full sample of ESOP firms

and the matched control sample of non-ESOP firms. Coefficients are reported with standard

errors in parentheses. "*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%,

respectively.

43

1 2 3 4 5

ESOP 0.074

(0.048)

0.174

(0.085)

**

0.181

(0.079)

**

0.173

(0.079)

**

ESOPg5 -0.175

(0.093)

*

-0.019

(0.047)

-0.179

(0.087)

**

-0.171

(0.087)

**

Log assets -0.164

(0.046)

***

-0.165

(0.046)

***

-0.163

(0.046)

-0.521

(0.097)

***

Log assets squared 0.035

(0.007)

***

Log sales 0.065

(0.053)

0.063

(0.053)

0.064

(0.053)

***

-0.256

(0.066)

***

Log sales squared 0.022

(0.006)

***

R&D / Sales -0.001

(0.000)

***

-0.004

(0.001)

***

CapEx /Assets 1.205

(0.218)

***

1.172

(0.244)

***

Adjusted sigma -0.838

(0.189)

***

-0.771

(0.191)

***

Adjusted sigmaDum -0.372

(0.063)

***

-0.325

(0.061)

***

N clusters 903 903 903 903 903

N observations 9524 9524 9524 9524 9524

R-squared 0.019 0.021 0.018 0.063 0.048

44

Table 8: Industry-adjusted Tobin’s Q Following ESOP Initiations—Interactive Effects of

Changes in Worker Mobility.

The dependant variable is industry-adjusted Q, windorized at 1%. Qit is fiscal year-end market

value of equity plus market value of preferred stock plus total liabilities divided by total assets.

Industry-adjusted Q is obtained by subtracting the median Q matched by industry (3-digit SIC

code) and year. ESOP is a dummy variable, which takes the value of 1 if the firm has an ESOP.

ESOPg5 is a dummy variable, which takes a value of 1 if the firm has an ESOP that controls at

least 5% of the firm's outstanding common stocks at any given time. Worker mobility high (low)

is a dummy variable, which takes the value of 1 if an establishment’s industry and MSA

combination has a Herfindahl value of employer concentration in the bottom (top) sample quartile.

All regressions include log assets, log sales, and firm- and year fixed effects, but their coefficients

are not reported. Log total assets and log sales are normalized to $2006. The sample includes the

full sample of ESOP firms and the matched control sample of non-ESOP firms. Coefficients are

reported with standard errors in parentheses. "*", "**", and "***" reflect statistical significance at

10%, 5%, and 1%, respectively.

1 2

ESOP 0.206

(0.084)

***

0.159

(0.087)

*

ESOPg5 -0.182

(0.092)

**

-0.171

(0.093)

*

Worker mobility high -0.024

(0.035)

Worker mobility high * ESOP

-0.122

(0.083)

*

Worker mobility low 0.036

(0.033)

Worker mobility low * ESOP 0.062

(0.074)

N clusters 903 903

N observations 9524 9524

R-squared 0.022 0.0216

45

Figure 1. Changes to Mean Industry-adjusted Q around ESOP Initiation.

This figure charts the sample mean industry-adjusted Q for firms which adopted ESOPs relative

to the year of the ESOP initiation. Year 0 is the year of the ESOP initiation. A small (large)

ESOP is an ESOP that never (does) control more than 5% of the firm’s outstanding common

stock during the lifetime of the ESOP. The Full Sample includes all available firm-year

observations. The Constant Sample is restricted to firms that have the complete 16 year data to

correct for a possible survivorship-bias.

Changes to Mean Industry Adjusted Q around ESOP Initiation

-0.2

0

0.2

0.4

0.6

0.8

1

-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10

Year Relative to ESOP Initiation

Ind

ustr

y A

dju

ste

d Q

Full Sample Small ESOP

Full Sample Large ESOP

Constant Sample Small ESOP

Constant Sample Large ESOP