Non-Executive Stock Options and Firm Performance*

Yael V. HochbergKellogg School of Management

Northwestern University

Laura LindseyW. P. Carey School of Business

Arizona State University

This Version: August 31, 2007

PRELIMINARY AND INCOMPLETECOMMENTS WELCOMEPLEASE DO NOT CITE

We examine whether options granted to rank and file employees affect the performanceof the firm by exploring the link between option portfolio implied incentives and firm operatingperformance. We employ an instrumental variables approach that combines information about thelabor market characteristics in which firms compete with information on firm option programsfrom the IRRC to identify causal effects. Firms whose non-executive employee option portfolioshave higher implied incentives exhibit higher subsequent operating performance. Consistent witheconomic theories, the incentive-performance effect is larger in smaller firms and in firms withhigher growth opportunities. Additionally, the incentive-performance effect is concentrated solelyin firms that grant options broadly to non-executive employees. Our results suggest that free-riding does not dominate incentives for options programs and are consistent with non-executiveoptions inducing mutual monitoring among co-workers.

* Hochberg is with the Kellogg School of Management, Northwestern University, 2001 Sheridan Road,Evanston, IL 60208-2001, y-hochberg@kellogg.northwestern.edu. Lindsey is with the W.P. CareySchool of Business, Arizona State University, P.O. Box 873906, Tempe, AZ 85207-3906,laura.lindsey@asu.edu. The authors thank Jeff Coles, Carola Frydman, Stuart Gillan, Yaniv Grinstein,Cami Kuhnen Spencer Martin, Lalitha Naveen, Mark Nelson, Paul Oyer, Michael Waldman andseminar participants at Arizona State University for helpful discussions and comments. The authors aregrateful for financial support from the Financial Services Exchange and from the Searle Center atNorthwestern University Law School. All errors are our own.

1

1. Introduction

Over the last two decades, the granting of company stock options to non-

executive employees has become an important and increasingly common component of

compensation policy1 (Mehran and Tracy (2001), Murphy (2003)). While there is a vast

academic literature addressing option compensation for executive employees, there have

been comparatively few papers on the use of option grants as compensation for non-

executive employees. Most existing work focuses on the rationales for use of these

options in the workplace2. In this study, we address an open question in the literature:

whether stock options granted to rank and file employees affect the resulting performance

of the firm.

Typically, accepted economic wisdom is that while options may be granted due to

financial constraints or labor market considerations, they are unlikely to have an effect on

performance beyond these channels. Most theoretical treatments of non-executive option

programs argue that free-riding among employees will outweigh any incentive effects, as

the option grants align the incentives of the worker with increasing the value of the whole

firm rather than with individual performance (Core and Guay (2001), Oyer (2004), Oyer

and Schaefer (2005a)). When employees are compensated for joint performance

improvements, they share the rewards from higher effort, resulting in dilution of worker

incentives and mitigation of additional effort (Alchian and Demsetz (1972)). Therefore,

as non-executive options are granted across the firm, free-riding would be expected to

preclude any performance effects. 3 An alternative literature, however, suggests that non-

executive options may induce mutual monitoring among co-workers (Baker, Jensen and

Murphy (1988)). If employees jointly agree to exert high effort, and monitor their

colleagues in order to enforce such collusion, the incentive for the individual worker to

exert effort will increase. In this mutual monitoring scenario, while it is clear that hiring

and retention, as well as firm financial constraints, may still motivate option grants, using

1 The National Center for Employee Ownership (NCEO) estimates that as of the year 2005, 4000 U.S.companies had broad-based stock option plans, defined as plans that grant options to 50% or more ofcompany employees.2 Oyer and Schaefer (2005a) and Ittner, Lambert and Larcker (2003), for example, argue that sorting andretention motives appear to be consistent with empirical data. Others, such as Core and Guay (2001), arguethat firms grant options due to cash and other financial constraints.3 In fact, under conventional methods for measuring such programs, we document that non-executiveoptions are granted quite broadly, with 44% of firms in our sample demonstrating broad-based programs.

2

equity based compensation may then have incentive effects, especially if such plans are

broad-based.

Exploring the impact of options on firm performance empirically is nontrivial. To

isolate the effect of non-executive stock options on firm performance, we must carefully

control for the endogenous nature of the existence and scope of these plans. Our models

employ an instrumental variables (IV) approach that centers on determinants of option

grants and incentives that are unlikely to be related to our measures of operating

performance.

Much of the prior empirical literature on non-executive employee stock options

has employed measures of annual stock option grants to rank-and-file employees. We can

more directly capture the incentives implied by a non-executive option plan by looking at

the total sensitivity of the value of the firm’s outstanding non-executive options to an

increase in the underlying value of a firm’s stock. One of the constraints in the prior

literature in this area has been that ExecuComp, a typical source for options data, only

allows the researcher to infer grants to non-executive employees, rather than the basic

characteristics of the entire portfolio of outstanding options. Using new data from the

Investor Responsibility Research Center’s (IRRC) Dilution Database, we are able to

observe the total options outstanding for a broad panel of firms. This allows us to look

directly at the entire portfolio of outstanding rank-and-file employee options, and to

obtain a measure of implied incentives, rather than relying on proxies for grants.

For our measure of implied rank-and-file incentives from outstanding options, we

calculate the total sensitivity of the firm’s non-executive options to an increase in the

underlying value of a firm’s stock, i.e. the cumulative delta of the firm’s non-executive

options. We examine the relationship between this incentive measure and subsequent

firm operating performance, as measured by the firm’s return on assets (ROA) and cash

flow to assets (CFA). Controlling for the potential endogeneity of non-executive stock

option incentives, we find that these incentives exert a positive effect on firm

performance. This positive effect of the implied option incentives on subsequent firm

performance suggests that free-riding stemming from an individual employee’s inability

to substantially affect firm value or profits on his own may not outweigh the incentive

3

effect provided by stock options in this setting, and is consistent with mutual monitoring

among co-workers overcoming the free-riding problem.4

Next, we attempt to ascertain where the positive relationship between option

incentives and firm performance is most effective. Free-riding is aggravated in large

firms (Holmstrom (1982)). In smaller firms, employees share the rewards for their efforts

with fewer colleagues, thus reducing the free-riding problem. At the same time, mutual

monitoring, which may counteract free-riding, is likely to be less effective in large firms,

as employees in large groups may be less able to observe each other’s efforts and less

willing to incure the costs associated with monitoring and sanctioning their fellow

workers (Heckathorn (1988), Kandel and Lazear (1992)). Thus, we expect any incentives

from non-executive options to be most effective in smaller firms, and particularly so if

free-riding is a dominant factor. When we segment our sample into smaller and larger

firms, however, we find that the relation between incentives and performance is not

confined to smaller firms, though the effect is larger for smaller firms than for larger

ones.

Similarly, incentives from options should be most effective in firms where

employee effort is more likely to have a significant effect on creating real value. In firms

with higher growth opportunities, the impact of any one worker’s effort is greater, and the

rewards to be split with colleagues from obtaining a performance improvement are

greater, thus strengthening the incentive to monitor and sanction colleagues. We segment

our sample into firms with high and low growth opportunities per employee (Core and

Guay (2001)) and repeat our analysis. We find that non-executive option incentives exert

a significant positive influence on performance in firms with higher growth opportunities,

but not in firms with lower growth options.

The above findings suggest that mutual monitoring may outweigh free-riding.

The incentive for workers to exert effort should be higher when the joint performance

goal on the basis of which they will be compensated is dependent on other workers who

do receive similar incentive compensation. In many firms, options are granted broadly, to

most, if not all employees (Oyer and Schaefer (2005)). In other firms, option grants to

4 Labor market considerations might result in more sensitive pay for performance attracting higher qualityworkers. In this case, delta may not capture a pure incentive effect, but the results still represent a causalrelation between delta and subsequent performance.

4

non-executive employees are targeted towards specific workers or groups of workers.

When options are targeted, rather than granted broadly, workers may be less likely to

work hard, since the individual worker cannot expect to have a significant effect on the

performance of the firm as a whole, and the worker is aware that other employees may

not have similar incentives. Similarly, mutual monitoring is more likely to occur when

workers know all have similar incentives, and therefore can jointly decide to maximize

total gains by exerting effort and sanctioning those who deviate from the group

agreement. Thus, mutual monitoring is more likely to obtain in firms where options are

granted broadly. Taken together, we expect to see a stronger relationship between option

incentives and performance in firms that broadly grant options. When we examine the

relationship between operating performance and the existence of broad-based option

grants in the cross-section of firms, we find that both a naïve OLS modeling of the

relationship and the IV approach observe a positive relationship between broad option

grants to non-executive employees and subsequent operating performance. We then

interact option portfolio incentives with an indicator for broad-based grants to non-

executive employees. We find that incentive effects are significantly and positively

related to performance only for the group of firms with broad-based plans. We consider it

unlikely that the existence of a broad-based program is merely a proxy for high

incentives, for two reasons. First, having a broad-based program is determined by a

firm’s grants in a particular year, whereas the option portfolio incentives are calculated

off the entire outstanding portfolio of non-executive options. Second, we observe no

statistical difference between the incentives implied by the options outstanding for firms

segmented by whether they have broad-based plan.

Our findings provide new insights into this expanding form of non-executive

compensation. To the best of our knowledge, this is the first study to document a causal

relationship between non-executive option portfolio incentives and firm performance.

Furthermore, our findings suggest that free-riding is unlikely to be the dominant effect in

this setting, and that option compensation can have a causal incentive effect on firm

performance. While free-riding may mitigate the incentive effects of option

compensation to some degree, an incentive effect remains, consistent with option

compensation inducing mutual monitoring among co-workers.

5

The remainder of the paper is organized as follows. Section 2 reviews the related

literature on non-executive stock options. In Section 3, we describe the data used in the

study and present some descriptive statistics. In Section 4, we present our empirical

analysis of the relationship between the implied incentives of non-executive stock option

portfolios and firm performance. In Section 5, we more closely examine free-riding and

mutual monitoring. In Section 6, we examine the incentive-performance relationship for

targeted and broad-based option plans. Section 7 concludes.

2. Related Literature

While there is a vast academic literature addressing option compensation for

executive employees (see Baker, Jensen and Murphy (1988), Murphy (1999) for an

overview), there have been comparatively few papers on the use of option grants as

compensation for non-executive employees. Early literature on broad based equity

compensation for non-executive workers examined stock price responses to the adoption

of employee stock purchase programs, and found positive responses to such

announcements (Bhagat, et al., 1985). In these cases, however, it is difficult to separate

any positive effects related to the adoption of the plan from potential signaling effects

about future company performance.

Kedia and Mozumdar (2002) examine the relationship between the total level of

firm stock option grants and firm stock market performance using a sample of 200 large

NASDAQ firms in 1999. However, in an efficient market, there is no particular reason to

expect a positive relationship between option grants and abnormal returns if option

granting behavior remains relatively constant over time. A more direct approach to

studying whether non-executive option programs affect performance is to look directly at

firm operating performance. By tying options measures to objective measures that are

not influenced by future expectations, we can obtain a clearer portrait of the effect that

non-executive option plans option programs have on the performance of firms and the

economy.

Sesil, Kroumova, Blasi and Kruse (2002) study differences in financial outcomes

for companies that do and do not grant stock options broadly. Ittner, Lambert and Larcker

(2003) study the determinants of grants in a sample of companies that employ option

6

plans and measure the success of these plans against the company’s stated objectives.

Both these studies treat the existence of the stock option plan as given. In contrast, our

work accounts for the endogenous nature of stock option plans. Oyer and Schaefer

(2005a) consider three possible economic justifications for broad-based option programs.

They conclude that sorting and retention appear to be most consistent with empirical data.

Core and Guay (2001), argue that firms grant options due to cash and other financial

constraints. Our approach takes advantage of the fact that grants may be related to sorting

and retention needs or firm financial constraints, and looks at the residual effects of these

options on firm performance.

Other papers on non-executive options plans include Oyer and Schaefer (2005b),

who compare observed stock option grant programs to hypothetical cash-only and

restricted stock-based plans, and reject tax-based favorable accounting explanations for

the granting of non-executive stock options. Oyer (2004) considers a model of broad-

based stock option grants that can explain the use and recent rise of broad-based stock

option plans even if these pay plans have no effect on employees’ on-the-job behavior.

Mehran and Tracy (2001) examine the effect of stock option exercises on employee

compensation in the 1990’s. Zhang (2002) investigates a market-valuation-based

hypothesis for employee stock options, and Liang and Weisbenner (2001) examine the

relationship between firm market valuation and option grants. Bergman and Jenter (2005)

analyze the relationship between option grants to non-executive employees and employee

optimism. Hand (2005) argues that compensating too few employees with options may

more negatively affect performance than granting too deeply. Landsman, Lang and Yeh

(2005) examine the determinants and consequences of the split of options between

executive and non-executive employees.

3. Description of Data

Our primary source of options data is the Investors Responsibility Research

Center (IRRC) Dilution Database. The IRRC Dilution database contains company option

plan information collected from public filings for firms in the S&P Super 1500,

composed of the S&P 500, S&P midcap 400, and the S&P small cap 600. Coverage

begins in 1997, and extends through 2004, with each year of coverage providing

7

information on the prior year’s stock option plans.5 The IRRC collects information on

year-end outstanding grants, weighted average exercise price of options outstanding and

weighted average contractual life of outstanding options, as well as information on new

grants and option exercises.

To isolate information on the option portfolios of non-executive employees, we

match information on the option plan as a whole with information on grants and options

outstanding for top executives contained in the Compustat’s ExecuComp Database.

Previous studies of non-executive options have employed either small samples of hand-

collected data or data from Compustat’s ExecuComp. ExecuComp, however, only

provides information on new grants to executives and the percentage of total grants to all

employees represented by these grants to executives, and does not provide data on the

total option portfolio outstanding. Thus, data from ExecuComp alone can provide

information on new grants to non-executive employees, but does not allow the researcher

to examine measures related to the total portfolio of options outstanding for rank and file

employees. The IRRC Dilution database, on the other hand, allows us to track the entire

portfolio of outstanding options.

We obtain data on firm operating performance and other financial characteristics

from Compustat. We use CRSP data to obtain risk-free rates. Data on industry level

labor turnover is obtained from the Bureau of Labor Statistics and data on education

levels by geographic region is obtained from the Census Bureau 2000 Census.

3.1 Dependent Variables

As a measure of incentives implied by a firm’s portfolio of non-executive stock

options, we compute the cumulative option delta, the change in employee wealth for a

1% change in stock price, for each firm, for the firm’s non-executive option portfolio

outstanding at the end of the year. We use the one-year estimation method for portfolio

incentives outlined in Core and Guay (2002).6 We calculate the incentive measure for the

total portfolio of options for all employees using the aggregate number of options

5 Because each year of data from IRRC contains two years of lagged data when available for variables ofinterest, we purchase each 3rd year of data beginning in 1997. As such, our sample approximates the S&P1500.6 Core and Guay (2002)’s ‘One-year Approximation’ (OA) method values stock options using the BlackScholes (1973) model, as modified by Merton (1973) to account for dividend payouts.

8

outstanding and their associated characteristics at the end of each year.7 Similarly, we

calculate the incentive measure for the portfolio of options held by the top five

executives. We then subtract the executive incentives measure from the total incentives

measure to obtain the portfolio incentives measure, or DELTA, for the option portfolio of

non-executive employees.8

Presumably, free-riding among employees will be less of a concern, and mutual

monitoring more likely, in smaller firms, firms where growth are opportunities are larger,

and where the growth of the firm is more likely to be affected by any individual worker.

We interact our measure of incentives with indicator variables based on above or below

median measures of these characteristics. The number of employees is used to capture

this aspect of firm size. We employ firm market-to-book ratio as a measure of overall

growth opportunities. To proxy for the influence an individual worker may have, we

calculate growth options per employee as in Core and Guay (2001), defined as the market

value of equity minus the book value of equity, divided by the number of employees.

To determine if a firm’s non-executive stock option program is broad-based, i.e.

grants options to over 50% of employees, we follow the criterion described in Oyer and

Schaefer (2005). Because many firms have more than five very high ranking employees,

which is the required threshold for detailed option compensation reporting, defining non-

executive options as all options granted to employees other than the five most highly

compensated executives tends to overstate the level of option grants and incentives for

non-executive employees, particularly those in large firms. Oyer and Schaefer (2005),

calibrating from a data set where they can observe option program eligibility directly,

assume that the top 10% of employees receive an option grant one-tenth as large as the

grants received by the 2nd through 5th most highly paid employees in the firm. They

classify a program as broad-based if the residual grants to employees after this

7 Our data source does not separate aggregate options information by exercisable and non-exercisableoptions. In essence, we assume all are exercisable at the average time to expiration.8 We compute two such measures for portfolio incentives: one that is the cumulative incentives for allemployees other than the top five executive officers, and one allowing an adjustment for other executivesbeyond the top five. The adjustment for high ranking employees follows the logic described in Oyer andSchaefer (2005); here, we assume the top 10% of employees hold a portfolio one tenth as large as theportfolios of the second through fifth most highly paid employees. For brevity, we report results using onlythe first measure, however we obtain similar results using the second measure as well.

9

adjustment exceed 0.5% of the shares outstanding. We define our variable BroadPlan

accordingly, and also interact it with our measure of portfolio incentives.

We examine two measures of operating performance. Our first measure of

performance is operating return on assets before depreciation. Barber & Lyon (1996)

argue that this measure is the preferred measure of operating performance because it is

unaffected by leverage, extraordinary items, discretionary expenditures, or depreciation

policy. Our second measure is cash flow deflated by assets. For each firm, we compute

both industry-adjusted ROA (CFA), which is the difference between the firm’s ROA

(CFA) in a given year and the industry median (defined using the Fama-French 30

industry classification), and normalized industry-adjusted ROA (CFA), which is the

percentage difference between the firm’s ROA (CFA) in a given year and the industry

median. A natural concern might be that these measures are mechanically higher for

firms employing option compensation given the accounting treatment of such plans. We

recalculate these measures subtracting out the Black-Scholes value of the option grants

from the numerator and our results are robust to this alternative treatment.

3.2 Instruments

To draw causal inferences about the impact of non-executive options on firm

performance, we must treat the incentives implied by the options as endogenous. The

extent to which firms grant options to non-executive employees is affected by

characteristics of the firm and the marketplace that may also affect operating

performance. Failing to correct for this fact in our models could lead to inconsistent

estimation and misleading inference. We therefore consider a number of instruments for

the implied incentives of rank and file option plans. Our instruments are positively

correlated with the extent of options grants to non-executive employees (and their

implied incentives), but are unlikely to affect year over year industry-adjusted operating

performance directly. 9

One of the rationales offered for the granting of non-executive options is the

retention of employees. Therefore, we employ a measure of annual average industry

turnover by two-digit SIC code taken from the Bureau of Labor Statistics Job Openings

9 Ultimately, this is an empirical question, and our tests suggest the instruments are exogenous.

10

and Labor Turnover Survey. This aggregate measure can be taken as exogenous to the

individual firm. Additionally, if options for non-executives are thought be a means for

companies to attract and retain employees, especially in jobs where human capital is

important, option plans may be correlated with education levels in the geographic region

where the company competes for employees. As a measure of labor market

competitiveness along this dimension, we compute using 2000 Census Data the natural

log of the number of masters degrees held by the male population in the region of the

company’s headquarters location, defined by two-digit zip code.10 Also, since the

number of employees will likely increase the aggregate incentives offered (and, later,

likely decrease the probability of a broad plan), we include the natural log of the number

of employees as an instrument.

Note that our objective in this exercise is to identify exogenous instruments that

are correlated with the incentives implied by the outstanding non-executive option

portfolio. Our goal is not to identify (nor do we econometrically require) all factors that

may influence broad granting of options or option portfolio incentives. Rather, we seek to

build an instrumental variables model that will allow us to properly identify causal effects

of broad-based plans or option incentives on performance.

3.3 Control Variables

To isolate the performance effects of non-executive stock options, we must first

control for a variety of firm characteristics that may affect the operating performance of

the firm, and in some cases, the granting of options as well. Larger firms and older firms

may have different operating performance characteristics than smaller or younger firms.

We control for firm size using the natural log of market value of assets, defined as the

market value of equity plus the book value of assets. We control for the age of a firm

using the natural log of the number of years the firm has been publicly traded. Marginal

corporate tax rates may also affect the tendency to grant options (Yermack (1995),

Dechow, Hutton and Sloan (1996), Hall and Liebman (2000)) and likely affect our after-

tax measure of performance. Option compensation should be more costly for firms with

10 Results are robust to using a variety of other measures of educational attainment as well, such as thenumber or percent of the population (male or total) with college or post-college degrees.

11

high marginal tax rates. Firms receive an immediate tax deduction for cash compensation,

as opposed to the future tax deduction from deferred compensation instruments such as

options. We define indicator variables for firms facing high (HMT) or low (LMT)

marginal tax rates as in Core and Guay (2001). HMT takes the value of one if the firm

has positive income and no net operating loss carryforwards in any of the previous three

years, and zero otherwise; LMT takes the value of one if the firm has negative taxable

income and net operating loss carryforwards in each of the previous three years, and zero

otherwise. To capture the effects of the tax shields provided by debt, we include an

indicator variable for the existence of long-term debt. Further, option grants and

performance are both influenced by a company’s research intensity. We therefore

include the three-year average of R&D expenditures as a control variable. Bushman,

Indjejikian, and Smith (1995) suggest that interdependencies among operating units, and

therefore aggregate performance measures, increase with firm inter-segment sales, and

decrease with greater product market and geographic diversification. We control for the

ratio of inter-segment sales to total net sales for the firm. Using data from the Compustat

segments file, we define product diversification as ∑i Pi*ln(1/Pi), where Pi is the dollar

sales of product i divided by total firm sales. Geographic diversification is similarly

defined, as ∑i Gi*ln(1/Gi), where Gi is the dollar value of sales for geographic region i

divided by total firm sales.

While it is unlikely that non-executive employees can effectively shift the risk

profile of the firm, recent studies suggest that executive incentive contracts may have

such an effect in addition to providing incentives (Coles, Daniel, and Naveen (2007)). We

therefore include a measure for the portfolio delta for the top five executives (the

sensitivity of executive wealth to changes in stock price) and vega (the sensitivity of

executive wealth to changes in stock price volatility). Finally, we include year fixed

effects.

An important aspect of our research design is that we do not ask whether broad-

based plans or non-executive option portfolio incentives are priced. Instead, we focus on

the question of how these option plan characteristics affect the operating performance of

the firm. Accordingly, we do not include stock performance related variables in our

analysis. If we were to include such variables, which may already incorporate market

12

pricing of these instruments, we may not discern a performance effect simply because it

is already priced into market-related controls.

Table 1 presents the mean and standard deviation of each of the above variables

for the full sample. Of the 8606 firm-year observations, 43.7% are from firms that grant

options broadly to employees. We can calculate the non-executive option portfolio

incentives measure for 5553 of these firm-year observations. Our three instrumental

variables are available for 7970 of the firm-year observations.

4. Non-Executive Option Incentives and Firm Performance

We begin by examining the effect of the incentives implied by the total

outstanding non-executive employee option portfolio on realized firm operating

performance. To draw causal inferences, we address the fact that non-executive option

grants are endogenously determined by other factors that may be related to firm

performance.

To provide a benchmark, we first estimate a naïve OLS model of the relationship

between firm performance and the implied incentives of the outstanding option portfolio.

The dependent variable is one of four measures of firm operating performance: industry-

adjusted ROA, normalized industry-adjusted ROA, industry-adjusted CFA, and

normalized industry-adjusted CFA. The main explanatory variable of interest is the

existence of a broad-based stock option plan. Our dataset is an unbalanced panel.

Standard errors are White (1980) heteroskedastic-consistent, clustered at the firm level.

As controls, we include firm operating performance in year t-1, prior 3 year average

R&D expenditures, firm long term debt in year t-1 , firm size, firm product

diversification, firm geographic diversification, firm inter-segment relatedness, the

indicator for low marginal tax rate, the indicator for high marginal tax rate, portfolio delta

of the option portfolio of the top five executives in the firm, and portfolio vega of the

option portfolio of the top five executives in the firm11. We also include year fixed

effects. In all four models, we observe a non-negative relationship between option

incentives and firm performance, though pnly the CFA specifications produce a

11 We note that the coefficients on the last 2 variables should be regarded as correlations throughout theanalysis.

13

statistically significant coefficient. As they can not be interpreted in a causal fashion, for

the sake of brevity, we do not report these estimates in table form.

While our benchmark estimates in the naïve OLS model suggest a possible

positive relationship between the existence of a broad-based non-executive stock option

plan and subsequent firm operating performance, before drawing causal inferences, we

must address the concern that option grants are endogenous. To do this, we employ an IV

approach, where the first stage model predicts the implied option incentive measure using

the exogenous instruments described in Section 3.3 and all explanatory variables from the

performance equation, and the second stage model predicts firm performance.

Table 2 presents the estimates from our IV models for operating performance. In

Panel A, we present OLS regressions of our incentives measure on our three proposed

instruments one at a time and all together. Our instruments appear to be well-correlated

with our endogenous variable. Our proposed instruments have differential power for

predicting the implied incentives of the option portfolio. Both the natural logarithm of the

number of employees in the firm and the geographic education level are significantly and

positively related to our incentives measure, both when used as the sole dependent

variable and when used in concert with our other instruments and industry controls. In

contrast, industry turnover does not appear to be significantly related to implied non-

executive option plan incentives.12

Staiger and Stock (1997) note that having valid instruments that meet the

exclusion restriction is not sufficient to ensure consistent two-stage estimators in finite

samples. The instruments also have to be ‘strong’ in the sense that they correlate

‘strongly’ with the endogenous first-stage variable. Staiger and Stock recommend a

critical value of 10 in an F-test for the joint significance of the instruments in the first

stage. An F statistic for the joint significance of all instruments exceeds the Staiger and

Stock (1997) critical value, suggesting our instruments are collectively strong.

We can now examine the determinants of firm operating performance. The

dependent variables are our four measures of operating performance used in the naïve

OLS models above, namely industry-adjusted ROA, normalized industry-adjusted ROA,

12 Our results are robust to the exclusion of this instrument, but we keep it in order to have a consistent setof instruments across specifications.

14

industry-adjusted CFA and normalized industry-adjusted CFA. (We will refer to these

variables as normalized and non-normalized ROA and CFA for the remainder of the

paper.)

Panel B of Table 2 presents the results of our full two-equation IV model

estimations. The dependent variables are our four measures of operating performance,

described in Section 2.2. The main explanatory variable of interest is NON-EXECUTIVE

DELTA, the delta of the firm’s non-executive option portfolio. The control variables are

as outlined for the benchmark OLS above.13 We estimate the four IV models using

GMM. Standard errors are White (1980) heteroskedastic-consistent, clustered by firm.

All four models appear to be well-specified. The uncentered R-squared ranges

from 31.84% for the model for normalized CFA, to 62.45% for the model for ROA.

Hansen J test statistics for over-identification of all instruments fail to reject the null of

valid instruments across all four models. Anderson canonical correlations likelihood ratio

tests for under-identification, which test that the instruments are relevant, reject the null

of irrelevance at the 1% level. The Shea partial R-square test (unreported) indicates the

excluded instruments increase explanatory power at over 99% confidence.

The estimates from the models support the hypothesis that the implied incentives

of the option portfolio affect firm operating performance. The coefficient on non-

executive option portfolio delta is positive and statistically significant. The economic

effect is non-negligible: a one-standard deviation change in our incentives measure

implies an increase of 12.34% in normalized ROA, and an increase of 1.48% in non-

normalized ROA. The magnitudes of the effects on CFA are larger: 26.11% increase in

normalized CFA and 1.98% increase in non-normalized CFA for a one standard deviation

increase in the incentives measure. The coefficients on the control variables are primarily

as expected. Average R&D expenditure is significantly and negatively related to firm

operating performance, while the existence of long term debt is significantly and

positively associated with performance. Larger firms have lower performance, as do

firms with greater geographic diversification. The delta of the top five executives option

portfolio is significantly and positively related to performance, while the vega of said

13 We also include industry fixed effects, which will be redundant in the industry-adjusted performanceequation, but not in the non-executive delta equation.

15

portfolio is significantly and negatively related to performance. Other controls, such as

product market diversification, intersegment relationships, marginal tax rates and the age

of the firm, observe no significant relationship to performance in the models.

The estimates from the four models are consistent with a positive, causal effect of

option portfolio incentives on firm operating performance, suggesting that free-riding

may not be the dominant force in determining the effect of non-executive stock options.

To further explore this finding, we now turn to a closer examination of the free-riding and

mutual monitoring hypotheses.

5. Free-Riding and Mutual Monitoring

The fact that the implied incentives of the non-executive option portfolio

positively affects firm operating performance could be considered surprising. Non-

executive stock options may be granted to hundreds or thousands of employees, each of

whom individually is likely to have a negligible impact on overall firm performance. In

accepted economic wisdom, such conditions are generally expected to favor free-riding,

and to limit mutual monitoring. To further explore this issue, we segment our sample on

measures that are likely to be related to the incentives to free-ride or to encourage mutual

monitoring: the size of the firm, the growth opportunities of the firm, and the growth

options per employee at the firm.

In general, the incremental likelihood that an individual worker’s wealth will

increase through his option compensation if the worker exerts effort is expected to be a

decreasing function of firm size, as overall performance is less sensitive to the actions of

individual workers in larger firms. The larger is the firm, the greater the free-rider

problem. Furthermore, compensation based on firm-wide performance introduces

externalities between the efforts of individual workers and the welfare of their colleagues.

If a worker exerts low effort, he not only reduces the likelihood that he will receive an

increase in his wealth from his options, but also the likelihood that other employees will

receive a wealth increase. This creates incentives for employees to monitor their

colleagues and encourage them to exert more effort. However, as noted by Knez and

Simester (2001), mutual monitoring induces a second order free-riding problem (wherein

workers free-ride on their colleagues monitoring efforts), and the larger the firm, the

16

greater this second-order free-riding problem. Hence, mutual monitoring may be less

effective in larger firms.

As the free-rider problem is greater in larger organizations, and mutual

monitoring may be less effective in large firms, we may expect the incentive-

performance effect documented in the previous section to be confined to smaller firms.

To test this hypothesis, we segment our sample at the median number of employees, and

re-estimate our models, allowing the coefficient on our measure of non-executive delta to

differ for firms with above- and below-median labor force size.

The results of our estimation are presented in the first two columns of Table 3. In

Panel A the dependent variables are ROA and normalized ROA, and in Panel B the

dependent variables are CFA and normalized CFA. All estimates are taken from IV

models, with the first stage model for incentives as described in Section 4. Once again,

the models appear to be well-specified. Uncentered R-squared is higher in the ROA

models than in the CFA models, ranging from 11.77% for non-normalized CFA to

57.58% for ROA. Similarly, IV diagnostics support our choice of instrumental variables.

Hansen J test statistics for over-identification of all instruments fail to reject the null of

valid instruments in both sets of models. Anderson canonical correlations likelihood ratio

tests for under-identification reject the null at the 1% level in both models, and the partial

R-square test for lack of explanatory power rejects at the 5% level.

The results in both sets of models are similar: while our incentive measure has a

positive, statistically significant effect in both smaller and larger firms, this effect is much

greater in smaller firms. A one standard deviation change in implied incentives for small

firms produces a 2.38% increase in industry-adjusted ROA and a 21.02% increase in

normalized ROA. In contrast, a similar change in the incentives measure for larger firms

produces only a 1.45% increase in ROA and an 11.69% increase in normalized ROA.

Economic significance of similar magnitudes is observed for the CFA models. Tests of

differences in the coefficients show the coefficient on incentives for smaller firms to be

statistically significantly different than for larger firms at the 5% and 1% levels

respectively for ROA and CFA, and at the 10% and 1% levels respectively for

normalized ROA and CFA.

17

In similar fashion, one might expect that free-riding would be weaker, and co-

worker monitoring more likely, in firms where the ability of any individual worker to

influence the overall success of the firm is higher. The idea is that individual effort may

have a more significant effect on creating real value. To explore this hypothesis, we

segment our sample based on high and low growth opportunities, as well as high and low

growth options per employee. We proxy for growth opportunities using the firm’s

market-to-book ratio. Growth options per employee are calculated as in Core and Guay

(2001). We allow the coefficient on our measure of implied incentives to differ for firms

with above-and below-median growth options, or growth options per employee, and re-

estimate our models.

The results are presented in the final four columns of Panels A and B of Table 5.

All estimates are taken from IV models, with the first stage model for incentives as

described in Section 4. As in our previous estimations, the models appear to be well-

specified. Uncentered R-squared ranges from 28.57% for normalized CFA to 61.53% for

ROA. IV diagnostics again support our choice of instrumental variables. Hansen J test

statistics for over-identification of all instruments fail to reject the null of valid

instruments in both sets of models. Anderson canonical correlations likelihood ratio tests

for under-identification reject the null at the 1% level in both models, as does the partial

R-square test for lack of explanatory power.

The results from the estimations are consistent with the notion that free-riding

may be weaker and co-worker monitoring may be more likely in firms with higher

growth opportunities and with higher growth options per employee. Columns three and

four of Panels A and B show that the performance effect of option incentives is

concentrated in firms with high growth opportunities. For firms with above-median

market-to-book ratios, a one-standard deviation change in incentives implies a 15.58%

increase in normalized ROA and a 1.87% increase in ROA, both statistically significant

at the 5% level. The marginal effects for high growth firm option incentives are slightly

higher in the CFA models, with a one-standard deviation change in incentives implying

an 27.4% increase in normalized CFA and a 2.37% increase in CFA for high growth

firms. In contrast, the coefficients on incentives for low growth firms are in fact slightly

negative in both the ROA and CFA models, and are not statistically significant. A test for

18

differences in coefficients finds the coefficients on the incentives measure for high

growth firms to be statistically significantly larger than for low growth firms at the 10%

level for both ROA models and the non-normalized CFA model.

A similar picture is seen in columns five and six, where the sample is segmented

based on growth options per employee. Here, a one-standard deviation change in

incentives for firms with high growth options per employee implies an 1.75% increase in

normalized ROA and a 1.97% increase in ROA, both significant at the 5% level. The

corresponding marginal effects for the CFA models are 32.81% and 2.86%, both

significant at the 1% level. Once again, the coefficients on incentives for low growth

options per employee firms are slightly negative and not statistically significant. A test

for differences in coefficients finds the coefficients on the incentives measure for high

growth firms to be statistically significantly larger than for low growth firms at the 5%

level for both ROA models and at the 1% level for both the CFA models. The partial R-

square test for lack of explanatory power rejects at the 1% level for all of the models.

6. Targeted and Broad-Based Option Plans

In many firms, options are granted broadly, to most, if not all employees (Oyer

and Schaefer (2005)). In other firms, option grants to non-executive employees are

targeted towards specific workers or groups of workers. Mutual monitoring is more likely

to occur when workers know all have similar incentives, and therefore can jointly decide

to maximize total gains by exerting effort and sanctioning those who deviate from the

group agreement. Thus, mutual monitoring is more likely to obtain in firms where

options are granted broadly. Taken together, we expect that any relationship between

non-executive options and subsequent firm performance may be concentrated in firms

that broadly grant options, and we expect to see a stronger relationship between the

incentives implied by these options and performance in firms with broad-based option

plans.

Table 4 presents the mean and standard deviation of each of the dependent,

independent and control variables for firm-years with and without BroadPlans, as well as

the p-value from a difference of means test across the two sub-samples. In the

univariate, industry-adjusted performance measures (ROA and CFA) are higher for firms

19

with broad-based option plans. As one might expect, these firms are generally smaller, as

measured by both assets and number of employees, are younger, have higher R&D

expenditures, and lower incidences of long-term debt. There is no difference in

incidences of high or low marginal tax rates. Also, these firms have lower measures of

product diversification, higher measures of inter-segment relatedness, but no difference in

geographic diversification.

Further, the BroadPlan sub-sample is characterized by lower average industry

turnover and the firms are located in regions with slightly higher levels of education as

measured by the population with masters degrees. There is no difference in implied

incentives for the top-five executive as measured by option delta, but option vega is

higher for firms with broad-based plans. Interestingly, the total incentives implied by

non-executive options outstanding are not statistically different across the two sub-

samples.

We then turn to examining the effect of the existence of a broad-based stock

option program for non-executive employees on realized firm operating performance. To

draw causal inferences, we again address the fact that the extent of option programs (in

addition to the incentives implied by the outstanding portfolio) are endogenously

determined by other factors that may be related to firm performance.

To provide a benchmark, we first estimate a naïve OLS model of the relationship

between firm performance and the existence of a broad-based option program. The

dependent variable is one of four measures of firm operating performance. The main

explanatory variable of interest is the existence of a broad-based stock option plan. Once

again, our dataset is an unbalanced panel. Standard errors are White (1980)

heteroskedastic-consistent, clustered at the firm level. As controls, we include firm

operating performance in year t-1, prior 3 year average R&D expenditures, firm long

term debt in year t-1, firm size, firm product diversification, firm geographic

diversification, firm inter-segment relatedness, the indicator for low marginal tax rate, the

indicator for high marginal tax rate, portfolio delta of the option portfolio of the top five

executives in the firm, and portfolio vega of the option portfolio of the top five executives

in the firm. We also include year fixed effects. We estimate models for our four measures

20

of operating performance. Overall, the models appear to be well-specified; the R2 range

from 39.7% for normalized CFA to 63.8% for ROA.

For brevity, we do not report the results in table format. In all four models, we

observe a statistically and economically significant and positive relationship between the

existence of a broad-based option program and firm operating performance. All else

equal, having a broad-based plan increases firm industry-adjusted ROA by 0.25

percentage points (versus an unconditional mean of 0.42%), and increases firm

normalized industry-adjusted ROA by 3.39% (versus an unconditional mean of 2.5%).

The relationship between performance and the control variables is primarily as expected.

Lagged performance enters positively, with firms with higher performance in year t-1

observing higher performance in year t. R&D expenditures are negatively related to

performance. Larger firms observe higher performance after controlling for other factors.

Geographic diversification is statistically negatively related to performance in three out

of the four models, but with little economic significance. Intersegment relatedness is

negatively related to performance, but is statistically significant only in the ROA

specifications. The delta of the top five executives’ option portfolio is positively related

to performance, while the vega of the portfolio observes a negative relationship to

performance. More mature firms observe lower performance. The indicator variable for

low marginal tax rate is negative and significant in the CFA models specification, but

insignificant in the ROA specifications. Similarly, the indicator variable for high

marginal tax rate is positive and significant in the CFA specification, but insignificant in

the ROA specifications. Other control variables such as long term debt and product

diversification observe no significant relationship to operating performance.

While our results in the naïve OLS model suggest a positive relationship between

the existence of a broad-based non-executive stock option plan and subsequent firm

operating performance, as in our previous analyses, we must address the concern that

option grants are endogenous. To do this, we once again employ an IV approach, where

the first stage model is a probit predicting the binary variable for broad-based option

21

plans using the same set of exogenous instruments used to predict option incentives, and

the second stage model predicts performance using a fitted value from the first stage.14

The first stage probit in our IV approach predicts the existence of a broad-based

non-executive option plan as a function of the instruments presented in Section 3.3 and

all explanatory variables from the performance equation. Panel A of Table 5 presents the

estimates for each of our proposed instruments separately and together. Once again, our

instruments appear to be well-correlated with our endogenous variable, however, the

instruments with the greatest predictive power differ somewhat from the specification for

predicting option incentives. Both industry turnover and the natural logarithm of the

number of employees in the firm are negatively and significantly related to the existence

of a broad-based plan, both when inserted alone or in concert with the other instruments.

In contrast, our third proposed instrument, which measures the education levels in the

firm’s geographical area, appears to be unrelated to the existence of a broad-based plan,

regardless of specification.

As our first stage in nonlinear, the joint test of significance for the instruments in

the first stage proposed by Staiger and Stock (1997) is distributed Chi-square. Our Chi-

square measure for the specification with the three instruments is 165, implying p-values

analogously small to pass the critical value threshold for the F-test and suggesting our

instruments are collectively strong. Further, Bound et al. (1995) argue that high partial R-

square, defined as the difference in R-square between the first stage regression with and

without the excluded instruments, ensures instruments are sufficiently strong. We find

that the inclusion of the instruments in a first stage Probit with all exogenous variables

increases the pseudo R-square from 0.067 to 0.107, approximately 60%.15

In each of the four naïve OLS models, we observed a positive and statistically

significant relationship between the existence of a broad-based option program and our

measures of firm operating performance, consistent with the existence of a positive

performance effect for broad-based non-executive option plans. What happens to this

positive relation once we account for the endogeneity of the presence of a broad-based

14 Wooldridge (2001) explains that robust standard errors for the fitted values in the second stage equationsare valid without adjustment when the first stage equation is a probit.15 We note that one drawback of using a more intuitive, non-linear, specification (such as a probit) in thefirst stage for the existence of a broad-based plan is the lack of standard over-identification tests for ourinstruments.

22

plan? Panel B of Table 5 presents the results of our IV models, instrumenting our broad-

based plan indicator variable as described above. The signs and magnitudes of the

coefficients on the control variables remain similar to those observed in the naïve OLS

model estimation. For all four measures of operating performance, we continue to find a

strong, positive and statistically significant relationship between the presence of a broad-

based stock option plan for rank and file employees and subsequent firm operating

performance. Having a broad-based plan is associated with a subsequent 1.33 percentage

point increase in non-normalized ROA (versus an unconditional mean of .42%), and an

increase of 9.99% in normalized ROA (versus an unconditional mean of 2.5%). The

existence of a broad-based plan has a similar effect on subsequent firm CFA, resulting in

a 1.1 percentage point increase in non-normalized CFA (versus an unconditional mean of

.33%) and an 11.67% increase in normalized CFA (versus an unconditional mean of

3.4%). Collectively, the results from the four models suggest that even after accounting

for the endogenous nature of broad-based stock option plans, granting options broadly to

rank and file employees has a positive effect on firm performance.

We now turn to the interaction of option portfolio incentives and broad-based

grants. Mutual monitoring effects and incentive effects in general should be stronger in

firms where options are granted broadly, to most employees; hence, we would expect to

see a stronger relationship between option incentives and performance in firms where

options are granted more broadly.

To explore this hypothesis, we allow the coefficient on our measure of option

portfolio incentives to differ for firms with and without broad-based option plans, and re-

estimate our models. The estimates for our models are presented in Table 6. All models

are estimated instrumenting for option incentives as described in Section 3. Uncentered

R-squared ranges from 20.95% for normalized CFA to 59.40% for ROA. Once again,

Hansen J test statistics for over-identification of all instruments fail to reject the null of

valid instruments in all four of models. Both the partial R-square test for lack of

explanatory power and the Anderson canonical correlations likelihood ratio tests for

under-identification reject the null at the 1% level in all four models.

The results in Table 6 are striking. In all four models, the coefficient on option

incentives for firms with broad-based option plans roughly doubles relative to the

23

estimate in the full sample. In three of the models, the coefficient is significant at the 1%

level; in the fourth model, it is significant at the 5% level. In contrast, the coefficient on

incentives for firm without broad-based plans is slightly negative, and is not statistically

significant in any of the four models. A test for differences in coefficients finds the

coefficients on the incentives measure for high growth firms to be statistically

significantly larger than for low growth firms at the 10% level for the ROA and CFA

models, and at 5% for the normalized CFA specification.16

The models in Table 6 suggest that incentive effects are significantly and

positively related to performance only for the group of firms with broad-based plans. We

consider it unlikely that the existence of a broad-based program is merely a proxy for

high incentives, for two reasons. First, having a broad-based program is determined by a

firm’s grants in a particular year, whereas the option portfolio incentives are calculated

off the entire outstanding portfolio of non-executive options. Second, in examining the

summary statistics presented in Table 4 for firms with and without broad plans, we

observe no empirical relationship between having a broad-based plan for rank and file

employees and the incentives implied by the options outstanding in that plan. Thus, our

results appear to be consistent with the notion that broad-based plans are more likely to

induce monitoring among co-workers.

7. Conclusion

Whether options granted to rank and file employees have causal effects on the

performance of the firm is an open question in the existing economic literature. This

paper examines this relationship by exploring the link between option portfolio implied

incentives and firm operating performance. To the best of our knowledge, this paper is

the first to address in depth the question of these programs’ effects on objective measures

of firm performance while carefully accounting for the endogenous nature of these option

plans.

Common economic wisdom holds that non-executive stock options are unlikely to

affect the performance of the firm due to free-riding. Competing theories argue that

mutual monitoring among employees may overcome the free-riding problem. A new,

16 The estimates for normalized ROA are noisier, with significant differences only at 82% confidance.

24

previously unemployed dataset allows us to obtain a measure of option portfolio

incentives which is unobtainable from previous datasets such as Execucomp. We

examine the total sensitivity of the firm’s non-executive options to an increase in the

underlying value of a firm’s stock. Controlling for other likely determinants of firm

operating performance, as well as the endogenous nature of option programs for non-

executive employees, we find a positive, causal relationship between the implied

incentives of the portfolio of outstanding non-executive options and subsequent firm

operating performance. The magnitude of the incentive effect is economically large, and

suggests that free-riding is not the dominant effect for these programs.

Consistent with economic theory, we find that the incentive-performance effect is

larger in smaller firms, and in firms with higher growth opportunities and higher growth

options per employee. Finally, we find that firms that broadly grant options to non-

executive employees exhibit higher operating performance than firms that do not grant

options broadly. Our models suggest that this performance effect is concentrated solely in

firms that grant options broadly to non-executive employees, supporting the argument

that options may induce monitoring among co-workers.

Our findings contribute to the emerging literature on the use of employee stock

options as effective compensation tools. Our approach employs a wide cross-section of

firms, and attempts to carefully account for the endogeneity of option granting behavior

so as to be able to draw conclusions about causation rather than correlations alone. In

addition, our results provide insight as to when the granting of such options is likely to

have the strongest incentive effects, and find evidence suggesting that targeted incentives

may be less effective in providing incentives than the use of broad-based plans. Our

findings suggest that while free-riding may be present to some extent, mutual monitoring

by co-workers may be the stronger force.

25

References

Alchian, Arman A. and Harold Demsetz, 1972. Production, Information Costs and Economic Organization.

American Economic Review, 62, 777-795.

Baker, George P., Michael C. Jensen and Kevin J. Murphy, 1988. Compensation and Incentives: Practice

vs. Theory. Journal of Finance, 43 (3), 593-616.

Barber, Brad and John Lyon, 1996. Detecting Abnormal Operating Performance: The Emprical Power and

Specification of Test Statistics. Journal of Financial Economics 41 (3), 359-400.

Bergman, Nittai and Dirk Jenter, 2005. Employee Sentiment and Stock Option Compensation. Journal of

Financial Economics, forthcoming.

Bhagat, S., Brickley, J. and Lease, R., 1985. Incentive effects of stock purchase plans, Journal of Financial

Economics 14, 195-216

Bound, J., D. A. Jaeger, R. M. Baker, 1995. Problems with Instrumental Variables Estimation When the

Correlation Between the Instruments and the Endogenous Explanatory Variable is Weak. Journal of

the American Statistical Association, 90, 443-450.

Bushman, Robert M., Raffi J. Indjejikian, and Abbie J. Smith, 1995. Aggregate Performance Measures in

Business Unit Manager Compensation: The Role of Intrafirm Interdependencies. Studies on

Managerial Accounting, 33, 101-128.

Coles, Jeffrey, Naveen Daniel and Lalitha Naveen, 2007, Executive Compensation and Managerial Risk-

Taking, Journal of Financial Economics, forthcoming.

Core, John and Wayne Guay, 2001. Stock Option Plans for Non-Executive Employees. Journal of

Financial Economics 61, 253-287.

Core, J. E. and W. R. Guay, 2002. Estimating the value of employee stock option portfolios and their

sensitivities to price and volatility. Journal of Accounting Research 40, 613-630.

Demsetz, Harold and Kenneth Lehn, 1985. The Structure of Corporate Ownership: Causes and

Consequences. Journal of Political Economy, 93, 1155-1177.

Hall, Brian and Jeffery Liebman, 2000. The Taxation of Executive Compensation. Tax Policy and the

Economy. James Poterba, ed. MIT Press.

Hand, John R.M., 2005, Give Everyone a Prize? Employee Stock Options in Private Venture-backed Firms,

Working Paper, University of North Carolina.

Heckathorn, Douglas D., 1988. Collective Sanctions and the Creation of Prisoner’s Dilemma Norms.

American Journal of Sociology, 94, 535-562.

Ittner, Christopher D., Richard A. Lambert and David F. Larcker, 2003. The Structure and Performance of

Equity Grants to Employees of New Economy Firms. Journal of Accounting and Economics 34, 89-

127.

26

Kandel, Eugene, and Edward P. Lazear, 1992. Peer Pressure in Partnerships. Journal of Political Economy,

100, 801-817.

Kedia, Simi and Abon Mozumdar, 2002. Performance Effect of Employee Stock Options. Working Paper,

Harvard University.

Knex, Marc and Douglas Simester, 2001. Firm-Wide Incentives and Mutual Monitoring at Continental

Airlines. Journal of Labor Economics, 19(4), 743-772.

Landsman, Wayne, Mark Lang and Shu Yeh, 2005. Governance and the Split of Options between

Executive and Non-Executive Employees. Wokring Paper, University of North Carolina.

Lazear, Edward P., 2005. Output-Based Pay: Incentives or Sorting? Research in Labor Economics Volume

23, Accounting for Worker Well-Being. Solomon W. Polachek, ed.

Liang, Nellie and Scott Weisbenner, 2001. Who Benefits from a Bull Market? An Analysis of Employee

Stock Option Grants and Stock Prices. FEDS Working Paper 2001-57.

Mehran, Hamid and Joseph Tracy, 2001. The Effect of Employee Stock Options on the Evolution of

Compensation in the 1990s. FRBNY Economic Policy Review, December.

Merton, R., 1973. Theory of rational option pricing. Bell Journal of Economics and Management Science 4,

141-183.

Murphy, Kevin J., 1999. Executive Compensation. Handbook of Labor Economics, Vol 3. Orly Ashenfelter

and David Card, ed. North Holland.

Murphy, Kevin J., 2003. Stock-Based Pay in New Economy Firms. Journal of Accounting and Economics,

January 2003.

Oyer, Paul, 2004. Why Do Firms Use Incentives That Have No Incentive Effects? Journal of Finance 59,

1619-1649.

Oyer, Paul and Scott Schaefer, 2005a. Why Do Some Firms Give Stock Options to All Employees? An

Empirical Examination of Alternative Theories. Journal of Financial Economics 76, 99-103.

Oyer, Paul and Scott Schaefer, 2005b. Costs of Broad-Based Stock Option Plans. Journal of Financial

Intermediation, forthcoming.

Sesil, James C., Maya K. Kroumova, Joseph R. Blasi and Douglas L. Kruse, 2002. Broad-Based Employee

Stock Options in U.S. ‘New Economy’ Firms. British Journal of Industrial Relations 40, 273-294.

Staiger, D. and J. H. Stock, 1997, “Instrumental Variables Regression with Weak Instruments,”

Econometrica, 65, 557-586.

Wooldridge, J., 2001, Econometric Analysis of Cross Section and Panel Data, MIT Press, Boston, MA.

Zhang, Ge, 2002, Market Valuation and Employee Stock Options. Working Paper, Duke University.

27

Table 1Descriptive Statistics

The unit of analysis is a firm-year. BROADPLAN is an indicator variable that takes the value 1 if the company had a broad-based employeeoption plan, 0 otherwise; ADJUSTED ROA is industry-adjusted return on assets before depreciation; ADJUSTED CFA is industry-adjusted cashflow deflated by assets; NONEXEC DELTA is the estimated wealth increase for the employees other than the top 5 executives from a 1% changein stock price; R&D is the three-year average for R&D expenses; LONG TERM DEBT is an indicator variable equal to 1 if the company hadlong-term debt, 0 otherwise; MARKET VALUE ASSETS is the market capitalization of the firm’s equity plus the book value of debt;PRODUCT DIV measures product diversification using the number of firm segments; GEOGRAPHIC DIV measures geographic diversificationusing the number of geographic segments; INTERSEGMENT REL measures the relatedness of the firm’s segments; LMT is an indicator variablethat takes a value of 1 if the company had a low marginal tax rate, 0 otherwise; HMT is an indicator variable that takes a value of 1 if thecompany had a high marginal tax rate, 0 otherwise; TOP5 DELTA is the estimated wealth increase for the top 5 executives from a 1% change instock price; TOP5 VEGA measures the convexity of compensation of the top 5 executives; AGE is the number of years the firm has been public;EMPLOYEES is the number of employees; INDUSTRY TURNOVER is average employee turnover for the company’s 4-digit SIC code;MASTERS DEG is the number of masters degrees for the male population in the company’s 2-digit zip code.

Variable Numberof

Observations

MeanTotal

Sample

Std.Dev.Total

SampleBROADPLAN 8606 0.4377 0.4961

ADJUSTED ROA 8606 0.0037 0.0012

ADJUSTED ROA (normalized) 8606 0.0223 0.8194

ADJUSTED CFA 8606 0.0029 0.0999

ADJUSTED CFA (normalized) 8606 0.0268 1.1453

NONEXEC DELTA 5553 0.6147 3.0289

R&D 8606 0.0355 0.0654

LONG TERM DEBT 8606 0.8551 0.3520

MARKET VALUE ASSETS (LN) 8606 7.6828 1.5512

PRODUCT DIV 8606 0.4656 0.5737

GEOGRAPHIC DIV 8606 -0.0048 46.281

INTERSEGMENT REL 8606 0.0053 0.0315

LMT 8606 0.0324 0.1771

HMT 8606 0.3129 0.4637

TOP5 DELTA 8606 9.3480 5.3067

TOP5 VEGA 8606 8.2399 4.8024

AGE 8606 21.900 19.412

NUMBER OF EMPLOYEES 8535 16.772 39.557

INDUSTRY TURNOVER 8581 0.0973 0.0553

MASTERS DEG 7970 95905 46856

28

Table 2Incentives and Firm Performance: GMM Estimation

This table presents the results of GMM estimation of IV regressions of industry-adjusted performance (ROA, ROS) on the delta of the outstanding non-executive option portfolio. The unit of analysis is a firm-year. Panel A present the results of OLS models where the dependent variable is the delta of thenon-executive option portfolio, and the dependent variables are instruments. Panel B presents the estimates of the main model, instrumenting for optionportfolio delta using the instruments from Panel A. NONEXEC DELTA is the estimated wealth increase for the employees other than the top 5 executivesfrom a 1% change in stock price; EMPLOYEES is the number of employees; INDUSTRY TURNOVER is average employee turnover for the company’s 4-digit SIC code; MASTERS DEG is the number of masters degrees for the male population in the company’s 2-digit zip code. ROA is industry-adjustedreturn on assets before depreciation; CFA is industry-adjusted cash flow deflated by assets; R&D is the three-year average for R&D expenses; LONGTERM DEBT is an indicator variable equal to 1 if the company had long-term debt, 0 otherwise; MARKET VALUE ASSETS is the market capitalizationof the firm’s equity plus the book value of debt; PRODUCT DIV measures product diversification using the number of firm segments; GEOGRAPHIC DIVmeasures geographic diversification using the number of geographic segments; INTERSEGMENT REL measures the relatedness of the firm’s segments;LMT is an indicator variable that takes a value of 1 if the company had a low marginal tax rate, 0 otherwise; HMT is an indicator variable that takes a valueof 1 if the company had a high marginal tax rate, 0 otherwise; TOP5 DELTA is the estimated wealth increase for the top 5 executives from a 1% change instock price; TOP5 VEGA measures the convexity of compensation of the top 5 executives; AGE is the number of years the firm has been public; Standarderrors are White heteroskedasticity-adjusted and are clustered for the same company (Rogers, 1993). In parenthesis we report z-scores.

Panel A: Option incentives and Instruments

Dependent Variable: Non-Executive Delta

INDUSTRY TURNOVER 2.3477(1.13)

1.1319(0.58)

LN NUMBEREMPLOYEES

0.3718 ***(5.17)

0.3724 ***(5.36)

LN MASTERS DEGREE 0.3960 ***(4.03)

0.4235 ***(4.28)

Intercept 0.4061 **(2.40)

-0.0872(-1.17)

-3.8556 ***(-3.72)

-4.9842 ***(-4.10)

Number of observations 5252 5252 5252 5252F(1,1102) = 1.28 Prob > F = 0.25R-squared = .002

F(1,1102) = 26.68 Prob > F = 0.00

R-squared = .0343

F(1,1102) = 16.24 Prob > F = 0.00R-squared = .006

F(3,1102) = 10.17 Prob > F = 0.00R-squared = .042

29

Table 2Incentives and Firm Performance: GMM Estimation

Panel B: “Second-Stage” Estimation

Dependent Variable:PERFORMANCE

IROA (normalized)

IIROA

IIICFA (normalized)

IVCFA

Delta of Portfolio 0.0399 *(1.85)

0.0048 *(1.89)

0.0844 **(2.18)

0.0064 **(2.04)

LAG PERFORMANCE 0.7371 ***(33.88)

0.7633 ***(37.37)

0.5336 ***(19.74)

0.5371 ***(25.89)

R&D -0.6002 **(-2.01)

-0.0796 **(-2.15)

-1.2476 **(-2.37)

-0.0968 **(-2.23)

LONG TERM DEBT 0.1046 ***(2.77)

0.0133 ***(2.76)

0.0761(1.26)

0.0045(0.83)

MKT VALUE ASSETS -0.0322 **(-2.53)

-0.0039 **(-2.51)

-0.0449 *(-1.83)

-0.0034 *(-1.71)

PRODUCT DIV -0.0135(-0.58)

-0.0008(-0.27)

-0.0236(-0.63)

-0.0032(-0.93)

GEOGRAPHIC DIV -0.0000 ***(-4.04)

-0.0000 ***(-4.76)

-0.0000 *(-1.85)

-0.0000(-1.58)

INTERSEGMENT REL -0.0992(-0.63)

-0.0117(-0.61)

-0.1208(-0.36)

-0.0121(-0.41)

LMT -0.0453(-0.48)

-0.0079(-0.73)

-0.4159 ***(-2.91)

-0.0341 ***(-3.11)

HMT -0.0229(-1.28)

-0.0041 *(-1.80)

-0.0044(-0.15)

-0.0001(-0.05)

TOP5 DELTA 0.0159 ***(2.90)

0.0022 ***(3.15)

0.0146(1.59)

0.0018 **(2.01)

TOP5 VEGA -0.0172 ***(-2.94)

-0.0023 ***(-3.12)

-0.0142(-1.43)

-0.0016 *(-1.76)

LNAGE 0.0143(1.58)

0.0015(1.40)

0.0091(0.55)

0.0006(0.44)

Intercept 0.0997(1.18)

0.0087(0.85)

0.2289(1.45)

0.0197(1.53)

Year and Industry Controls Included but not Reported“Excluded” Instruments: Masters Degree, Industry Turnover, Natural Log of Number of Employees

Number of observations 5252 5252 5280 5280Anderson corr. LR stat Chi2~31 95.665

0.0096.007

0.0090.144

0.0089.856

0.00Hansen J Chi2~30 32.022

0.36633.478.3022

26.7090.638

28.9660.519

F(20,1102) = 88.59 Prob > F = 0.00

UC R-squared = .5872

F(20,1102) = 99.86 Prob > F = 0.00

UC R-squared = .6245

F(20,1104) = 44.54 Prob > F = 0.00

UC R-squared = .3184

F(20,1104) = 69.23 Prob > F = 0.00

UC R-squared = .3558*, ** or *** mean the coefficient is significant at 10%, 5% or 1% level respectively

30

Table 3Incentive Effects split by Number of Employees, Growth Option Measures

This table presents the results of GMM estimation of IV regressions of industry-adjusted performance (ROA, ROS) on the delta of the outstanding non-executive option portfolio, split at the median value for a variety of measures. The unit of analysis is a firm-year. NONEXEC DELTA is the estimatedwealth increase for the employees other than the top 5 executives from a 1% change in stock price; EMPLOYEES is the number of employees;INDUSTRY TURNOVER is average employee turnover for the company’s 4-digit SIC code; MASTERS DEG is the number of masters degrees for themale population in the company’s 2-digit zip code. ROA is industry-adjusted return on assets before depreciation; CFA is industry-adjusted cash flowdeflated by assets; R&D is the three-year average for R&D expenses; LONG TERM DEBT is an indicator variable equal to 1 if the company had long-term debt, 0 otherwise; MARKET VALUE ASSETS is the market capitalization of the firm’s equity plus the book value of debt; PRODUCT DIVmeasures product diversification using the number of firm segments; GEOGRAPHIC DIV measures geographic diversification using the number ofgeographic segments; INTERSEGMENT REL measures the relatedness of the firm’s segments; LMT is an indicator variable that takes a value of 1 if thecompany had a low marginal tax rate, 0 otherwise; HMT is an indicator variable that takes a value of 1 if the company had a high marginal tax rate, 0otherwise; TOP5 DELTA is the estimated wealth increase for the top 5 executives from a 1% change in stock price; TOP5 VEGA measures the convexityof compensation of the top 5 executives; AGE is the number of years the firm has been public; Market to book is calculated as market value of assets overbook value of assets. Growth options per employee is calculated as in Core and Guay (2001) as the difference between market value of assets and bookvalue of assets divided by the number of employees. Standard errors are White heteroskedasticity-adjusted and are clustered for the same company(Rogers, 1993). In parenthesis we report z-scores.

Panel A: ROA measures

Incentive EffectsNumber of Employees

(X=Small)

Incentive EffectsGrowth Options

(X=High Market to Book)

Incentive EffectsGrowth Options per Employee(X=High Mkt-Book per Empl.)

Dependent Variable:

ROA(normalized)

ROA ROA(normalized)

ROA ROA(normalized)

ROA

NON-EXEC INCENTIVE(X=1)

0.2212 **(2.04)

0.0251 *(1.77)

0.0507 **(2.23)

0.0061 **(2.21)

0.0570 **(2.41)

0.0064**(1.72)

NON-EXEC INCENTIVE(X=0)

0.0395 **(2.34)

0.0049 **(2.44)

-0.0393(-0.62)

-0.0064(-0.80)

-0.0777(-1.24)

-0.0097(-1.24)

LAG ADJUSTED ROA 0.7497 ***(34.68)

0.7736 ***(38.41)

0.7246 ***(32.12)

0.7500 ***(35.14)

0.7240 ***(32.63)

0.7510 ***(35.72)

R&D -0.6058 ***(-2.14)

-0.0813 **(-2.26)

-0.7181 **(-2.32)

-0.0929 **(-2.44)

-0.7562 **(-2.43)

-0.0944 **(-2.45)

LONG TERM DEBT 0.0893 **(2.34)

0.0118 **(2.43)

0.0924 **(2.53)

0.0113 **(2.39)

0.0839 **(2.28)

0.0106 **(2.24)

MKT VALUE ASSETS -0.0215 *(-1.82)

-0.0027 *(-1.88)

-0.0262 **(-2.02)

-0.0028 *(-1.79)

-0.0269 **(-2.07)

-0.0030 *(-1.89)

PRODUCT DIV 0.0073(0.32)

0.0016(0.52)

-0.0108(-0.47)

-0.0006(-0.22)

-0.0146(-0.65)

-0.0009(-0.31)

GEOGRAPHIC DIV -0.0000 ***(-3.08)

-0.0000 ***(-3.62)

-0.0000 ***(-4.12)

-0.0000 ***(-4.71)

-0.0000 ***(-3.04)

-0.0000 ***(-3.60)

INTERSEGMENT REL -0.1834 *(-1.42)

-0.0221(-1.43)

-0.0295(-0.18)

-0.0024(-0.12)

-0.1430(-0.95)

-0.0173(-0.93)

LMT -0.0173(-0.20)

-0.0048(-0.50)

-0.0424(-0.46)

-0.0077(-0.73)

-0.0543(-0.58)

-0.0092(-0.87)

HMT 0.0076(0.44)

0.0004(-0.17)

-0.0206(-1.19)

0.0036 *(-1.65)

-0.0081(-0.46)

-0.0023(-1.01)

TOP5 DELTA 0.0113 **(2.00)

0.0016 **(2.26)

0.0127 **(2.19)

0.0018 **(2.49)

0.0136 **(2.41)

0.0020 ***(2.88)

TOP5 VEGA -0.0154 ***(-2.57)

-0.0021 ***(-2.73)

-0.0140 **(-2.57)

-0.0019 **(-2.53)

-0.0153 **(-2.55)

-0.0022 ***(-2.95)

LNAGE 0.0111(1.29)

0.0013(1.20)

0.0116(1.27)

0.0010(0.90)

0.0108(1.15)

0.0010(0.82)

Intercept 0.0455(0.58)

0.0026(0.27)

0.0788(0.92)

0.0052(0.51)

0.1010(1.18)

0.0075(0.73)

Year and Industry Controls Included but not Reported“Excluded” Instruments: Masters Degree, Industry Turnover, Natural Log of Number of Employees

Number of observations 5252 5252 5252 5252 5252 5252

Anderson LR StatisticProb > Chi-sq (7)

56.9760.000

57.6040.002

86.4440.000

86.6970.000

86.7450.000

86.9760.000

Hansen J 33.9660.241

35.1540.200

31.530.341

31.3070.351

31.9570.322

32.6510.292

F(21,1102)Prob > FUncentered R-square

96.34 0.00.5214

110.62 0.00.5758

87.90 0.00.5761

97.51 0.00.6153

89.06 0.00.5651

99.75 0.00.6104

*, ** or *** mean the coefficient is significant at 10%, 5% or 1% level respectively

31

Table 3Incentive Effects split by Number of Employees, Growth Option Measures

Panel B: CFA Measures

Incentive EffectsNumber of Employees

(X=Small)

Incentive EffectsGrowth Options

(X=High Market to Book)

Incentive EffectsGrowth Options per Employee(X=High Mkt-Book per Empl.)

Dependent Variable:

CF/A(normalized)

CF/A CF/A(normalized)

CF/A CF/A(normalized)

CF/A

NON-EXEC INCENTIVE(X=1)

0.4618 ***(2.74)

0.0483 ***(2.99)

0.0892 **(2.26)

0.0077 **(2.33)

0.1066 ***(2.67)

0.0093 ***(2.74)

NON-EXEC INCENTIVE(X=0)

0.0786 **(2.34)

0.0064 **(2.45)

-0.0129(-0.13)

-0.0067(-0.74)

-0.1542(-1.57)

-0.0180 *(-1.87)

LAG ADJUSTED CF/A 0.5356 ***(20.37)

0.5436 ***(27.25)

0.5295 ***(19.22)

0.5287 ***(24.81)

0.5216 ***(19.04)

0.5249 ***(24.78)

R&D -1.4374 ***(-2.78)

-0.1249 ***(-2.90)

-1.3372 **(-2.48)

-0.1138 **(-2.53)

-1.5748 ***(-2.85)

-0.1368 ***(-2.94)

LONG TERM DEBT 0.0633(1.06)

0.0050(0.92)

0.0666(1.11)

0.0032(0.60)

0.0469(0.80)

0.0024(0.46)

MKT VALUE ASSETS -0.0309(-1.34)

-0.0025(-1.35)

-0.0372(-1.48)

-0.0025(-1.20)

-0.0329(-1.36)

-0.0025(-1.18)

PRODUCT DIV 0.0149(0.38)

0.0012(0.35)

-0.0143(-0.38)

-0.0021(-0.59)

-0.0146(-0.39)

-0.0026(-0.75)

GEOGRAPHIC DIV -0.0000(-0.88)

-0.0000(-0.52)

-0.0000 *(-1.83)

-0.0000(-1.58)

-0.0000(-0.68)

-0.0000(-0.31)

INTERSEGMENT REL -0.3057(-0.94)

-0.0367(-1.30)

-0.0948(-0.28)

-0.0069(-0.23)

-0.2509(-0.76)

-0.0276(-0.95)

LMT -0.4287 ***(-3.19)

-0.0347 ***(-3.44)

-0.4133 ***(-2.92)

-0.0036 ***(-3.14)

-0.4219 ***(-3.01)

-0.0341 ***(-3.19)

HMT 0.0376(1.26)

0.0040(1.57)

-0.0061(-0.21)

0.0002(-0.09)

0.0090(0.31)

0.0009(0.35)

TOP5 DELTA 0.0075(0.80)

0.0010(1.08)

0.0117(1.20)

0.0013(1.42)

0.0108(1.15)

0.0016 *(1.71)

TOP5 VEGA -0.0106(-1.05)

-0.0012(-1.27)

-0.0114(-1.10)

-0.0012(-1.26)

-0.0110(-1.10)

-0.0015(-1.56)

LNAGE 0.0111(1.29)

0.0009(0.66)

0.0053(0.32)

-0.0010(-0.90)

-0.0007(-0.04)

-0.0004(-0.25)

Intercept 0.1167(0.77)

0.0109(0.88)

0.1986(1.25)

0.0171(1.31)

0.2253(1.43)

0.0205(1.54)

Year and Industry Controls Included but not Reported“Excluded” Instruments: Masters Degree, Industry Turnover, Natural Log of Number of Employees

Number of observations 5280 5280 5280 5280 5280 5280

Anderson LR StatisticProb > Chi-sq (7)

56.9990.000

58.1110.002

81.1950.000

81.5490.000

82.6860.000

82.8940.000

Hansen J 25.3770.659

24.8360.687

27.4980.545

28.1550.510

24.0790.725

24.4340.707

F(21,1104)Prob > FUncentered R-square

46.39 0.00.1653

75.05 0.00.1177

43.25 0.00.3151

67.38 0.00.3413

44.46 0.00.2857

68.43 0.00.3075

*, ** or *** mean the coefficient is significant at 10%, 5% or 1% level respectively

32

Table 4Targeted versus Broad Plans: Descriptive Statistics

The unit of analysis is a firm-year. BROADPLAN is an indicator variable that takes the value 1 if the company had a broad-based employeeoption plan, 0 otherwise; ADJUSTED ROA is industry-adjusted return on assets before depreciation; ADJUSTED CFA is industry-adjusted cashflow deflated by assets; NONEXEC DELTA is the estimated wealth increase for the employees other than the top 5 executives from a 1% changein stock price; R&D is the three-year average for R&D expenses; LONG TERM DEBT is an indicator variable equal to 1 if the company hadlong-term debt, 0 otherwise; MARKET VALUE ASSETS is the market capitalization of the firm’s equity plus the book value of debt;PRODUCT DIV measures product diversification using the number of firm segments; GEOGRAPHIC DIV measures geographic diversificationusing the number of geographic segments; INTERSEGMENT REL measures the relatedness of the firm’s segments; LMT is an indicator variablethat takes a value of 1 if the company had a low marginal tax rate, 0 otherwise; HMT is an indicator variable that takes a value of 1 if thecompany had a high marginal tax rate, 0 otherwise; TOP5 DELTA is the estimated wealth increase for the top 5 executives from a 1% change instock price; TOP5 VEGA measures the convexity of compensation of the top 5 executives; AGE is the number of years the firm has been public;EMPLOYEES is the number of employees; INDUSTRY TURNOVER is average employee turnover for the company’s 4-digit SIC code;MASTERS DEG is the number of masters degrees for the male population in the company’s 2-digit zip code.

Variable MeanBROADSample

MeanNON-BROAD

Sample

P-value forDifference

BROADPLAN 1 0 N/A

ADJUSTED ROA 0.0101 -0.0014 0.000

ADJUSTED ROA (normalized) 0.0691 -0.0142 0.000

ADJUSTED CFA 0.0101 -0.0027 0.000

ADJUSTED CFA (normalized) 0.0997 -0.0299 0.000

NONEXEC DELTA 0.6176 0.6100 0.9195

R&D 0.0400 0.0320 0.000

LONG TERM DEBT 0.8320 0.8731 0.000

MARKET VALUE ASSETS (LN) 7.5069 7.8198 0.000

PRODUCT DIV 0.4362 0.4886 0.000

GEOGRAPHIC DIV -0.5969 0.4561 0.356

INTERSEGMENT REL 0.0062 0.0045 0.020

LMT 0.0329 0.0320 0.818

HMT 0.3186 0.3085 0.321

TOP5 DELTA 9.3130 9.3752 0.5874

TOP5 VEGA 8.3776 8.1327 0.019

AGE 20.217 23.211 0.000

NUMBER OF EMPLOYEES 8.465 23.335 0.000

INDUSTRY TURNOVER 0.0890 0.1037 0.000

MASTERS DEG 97719 94468 0.002

33

Table 5Broad Plan and Firm Performance: Fitted Value IV Models

This table presents the results of fitted value instrumental variables regressions of industry-adjusted Performance (ROA, ROS) on an indicator variable for abroad-based option plan. The unit of analysis is a firm-year. Panel A present the results of probit models where the dependent variable is an indicator forhaving a broad-based plan, and the dependent variables are instruments. Panel B presents the estimate of the second stage regression, instrumenting from theexistence of a broad-based plan using the instruments from Panel A. BROADPLAN is an indicator variable that takes the value 1 if the company had abroad-based employee option plan, 0 otherwise; EMPLOYEES is the number of employees; INDUSTRY TURNOVER is average employee turnover forthe company’s 4-digit SIC code; MASTERS DEG is the number of masters degrees for the male population in the company’s 2-digit zip code. ROA isindustry-adjusted return on assets before depreciation; CFA is industry-adjusted cash flow deflated by assets; R&D is the three-year average for R&Dexpenses; LONG TERM DEBT is an indicator variable equal to 1 if the company had long-term debt, 0 otherwise; MARKET VALUE ASSETS is themarket capitalization of the firm’s equity plus the book value of debt; PRODUCT DIV measures product diversification using the number of firm segments;GEOGRAPHIC DIV measures geographic diversification using the number of geographic segments; INTERSEGMENT REL measures the relatedness ofthe firm’s segments; LMT is an indicator variable that takes a value of 1 if the company had a low marginal tax rate, 0 otherwise; HMT is an indicatorvariable that takes a value of 1 if the company had a high marginal tax rate, 0 otherwise; TOP5 DELTA is the estimated wealth increase for the top 5executives from a 1% change in stock price; TOP5 VEGA measures the convexity of compensation of the top 5 executives; AGE is the number of years thefirm has been public; Standard errors are White heteroskedasticity-adjusted and are clustered for the same company (Rogers, 1993). In parenthesis wereport z-scores.

Panel A: Broad Plan and Instruments

Dependent Variable: BROADPLAN

INDUSTRY TURNOVER -3.0766 ***(-8.45)

-2.9402 ***(-8.19)

LN NUMBEREMPLOYEES

-0.1481 ***(-9.87)

-0.1433 ***(-9.38)

LN MASTERS DEGREE 0.0522(1.29)

0.0402(1.00)

Intercept 0.1607 ***(3.73)

0.0955 ***(2.79)

-0.7257(-1.59)

-0.0848(-0.19)

Number of observations 7890 7890 7890 7890Wald Chi2 (1)= 71.44

Prob > Chi2= 0.00 R-squared = .0134

Wald Chi2 (1)= 97.36 Prob > Chi2= 0.00 R-squared = .025

Wald Chi2 (1)= 1.67 Prob > Chi2= 0.20 R-squared = .000

Wald Chi2 (3)= 165.49 Prob > Chi2= 0.00 R-squared = .037

*, ** or *** mean the coefficient is significant at 10%, 5% or 1% level respectively

34

Table 5Broad Plan and Firm Performance: Fitted Value IV Models

Panel B: Second Stage Estimation

Dependent Variable:PERFORMANCE

IROA (normalized)

IIROA

IIICFA (normalized)

IVCFA

BROAD PLAN 0.0999 ***(4.77)

0.0133 ***(4.48)

0.1167 ***(3.39)

0.0119 ***(3.52)

LAG PERFORMANCE 0.7423 ***(24.04)

0.7440 ***(26.75)

0.5528 ***(16.61)

0.5726 ***(15.43)

R&D -1.0158 ***(-4.22)

-0.1374 ***(-3.86)

-1.3158 ***(-3.24)

-0.1260 ***(-3.27)

LONG TERM DEBT 0.0039(0.15)

-0.0002(-0.07)

-0.0548(-1.31)

-0.0059 (-1.58)

MKT VALUE ASSETS 0.0401 ***(6.55)

0.0050 ***(5.93)

0.0601 ***(6.35)

0.0053 ***(5.88)

PRODUCT DIV 0.0222(0.96)

0.0031(0.97)

0.0302(0.85)

0.0029(0.89)

GEOGRAPHIC DIV 0.0000 **(2.79)

0.000 ***(2.69)

0.0001 ***(3.15)

0.0001 ***(3.20)

INTERSEGMENT REL -0.2557 **(-2.31)

-0.0312 **(-2.15)

-0.4903(-1.44)

-0.0402(-1.55)

LMT 0.0083(0.10)

-0.0004(-0.04)

-0.4367 ***(-3.27)

-0.0318 ***(-2.88)

HMT 0.0163(1.11)

0.0007(0.37)

0.0704 ***(3.29)

0.0052 ***(2.58)

TOP5 DELTA 0.0329 ***(4.78)

0.0045 ***(5.02)

0.0347 ***(3.30)

0.0036 ***(3.55)

TOP5 VEGA -0.0405 ***(-5.14)

-0.0055 ***(-5.35)

-0.0409 ***(-3.48)

-0.0042 ***(-3.72)

LNAGE -0.0296 ***(-4.10)

-0.0041 ***(-4.15)

-0.0431 ***(-3.65)

-0.0043 ***(-3.91)

Intercept -0.1222 **(-2.51)

-0.0114 *(-1.81)

-0.1749 **(-2.27)

-0.0131 **(-1.97)

Year and Industry Controls Included but not ReportedNumber of observations 7888 7888 7924 7924

F(20,1249) = 101.29 Prob > F = 0.00

R-squared = .6068

F(20,1249) = 128.32 Prob > F = 0.00

R-squared = .6405

F(20,1251) = 39.94 Prob > F = 0.00

R-squared = .3986

F(20,1251) = 50.31 Prob > F = 0.00

R-squared = .4166*, ** or *** mean the coefficient is significant at 10%, 5% or 1% level respectively

35

Table 6Targeted versus Broad Incentives

This table presents the results of GMM estimation of IV regressions of industry-adjusted performance (ROA, ROS) on the delta of theoutstanding non-executive option portfolio for firms with and without broad-based option plans. The unit of analysis is a firm-year.BROADPLAN is an indicator variable that takes the value 1 if the company had a broad-based employee option plan, 0 otherwise;NONEXEC DELTA is the estimated wealth increase for the employees other than the top 5 executives from a 1% change in stockprice; ROA is industry-adjusted return on assets before depreciation; CFA is industry-adjusted cash flow deflated by assets; R&D isthe three-year average for R&D expenses; LONG TERM DEBT is an indicator variable equal to 1 if the company had long-term debt,0 otherwise; MARKET VALUE ASSETS is the market capitalization of the firm’s equity plus the book value of debt; PRODUCTDIV measures product diversification using the number of firm segments; GEOGRAPHIC DIV measures geographic diversificationusing the number of geographic segments; INTERSEGMENT REL measures the relatedness of the firm’s segments; LMT is anindicator variable that takes a value of 1 if the company had a low marginal tax rate, 0 otherwise; HMT is an indicator variable thattakes a value of 1 if the company had a high marginal tax rate, 0 otherwise; TOP5 DELTA is the estimated wealth increase for the top5 executives from a 1% change in stock price; TOP5 VEGA measures the convexity of compensation of the top 5 executives; AGE isthe number of years the firm has been public; Standard errors are White heteroskedasticity-adjusted and are clustered for the samecompany (Rogers, 1993). In parenthesis we report t-statistics.

Dependent Variable: ROA

(normalized)

ROA CFA

(normalized)

CFA

BROADPLAN x delta 0.0767 ***(2.25)

0.0081 **(1.97)

0.1539 ***(2.83)

0.0119 ***(2.51)

NO BROADPLAN x delta -0.0082(-0.27)

-0.0001(-0.04)

-0.0341(-0.58)

-0.0032(-0.59)

LAGGED PERFORMANCE 0.7313 ***(32.94)

0.7576 ***(36.00)

0.5173 ***(18.21)

0.5309 ***(24.42)

R&D -0.7500 **(-2.44)

-0.0909 **(-2.39)

-1.5802 ***(-2.89)

-0.1191 ***(-2.62)

LONG TERM DEBT 0.0919 **(2.43)

0.0120 **(2.47)

0.0594(0.99)

0.0035(0.63)

MKT VALUE ASSETS -0.0273 **(-2.10)

-0.0032 **(-2.04)

-0.0340(-1.38)

-0.0020(-0.97)

PRODUCT DIV 0.0037(0.15)

0.0009(0.27)

0.0109(0.28)

-0.0005(-0.15)

GEOGRAPHIC DIV -0.0000 ***(-4.15)

-0.0000 ***(-4.82)

-0.0000 *(-1.67)

-0.0000(-1.38)

INTERSEGMENT REL -0.1700(-1.17)

-0.0185(-1.01)

-0.4024(-1.21)

-0.0379(-1.26)

LMT -0.0343(-0.37)

-0.0073(-0.70)

-0.4083 ***(-2.95)

-0.0316 ***(-3.01)

HMT -0.0152(-0.84)

-0.0033(-1.44)

0.0006(0.02)

0.0008(0.29)

TOP5 DELTA 0.0253 ***(3.21)

0.0033 ***(3.47)

0.0282 **(2.34)

0.0031 **(2.54)

TOP5 VEGA -0.0270 ***(-3.53)

-0.0035 ***(-3.73)

-0.0278 **(-2.37)

-0.0030 ***(-2.56)

LNAGE 0.0136(1.57)

0.0014(1.31)

0.0105(0.66)

0.0003(0.25)

Intercept 0.0761(0.89)

0.0053(0.51)

0.1625(1.02)

0.0118(0.85)

Year and Industry Controls Included but not Reported“Excluded” Instruments: Masters Degree, Industry Turnover, Natural Log of Number of EmployeesNumber of observations 5252 5252 5280 5280Anderson LR StatisticProb > Chi-sq (7)

80.9660.00

80.1170.00

79.8030.00

78.5150.00

Hansen J 28.9280.469

30.8070.375

20.9330.862

24.4430.707

F(21,1102 or 1104)Prob > FUncentered R-square

70.83 0.0000 0.5312

95.68 0.0000 0.5940

43.16 0.0000 0.2095

64.80 0.0000 0.2697

*, ** or *** mean the coefficient is significant at 10%, 5% or 1% level respectively