Money left on the table: An analysis of
participation in employee stock purchase plans.
ILONA BABENKO and RIK SEN
October 22, 2012
ABSTRACT
We document that over 70% employees at large U.S. …rms fail to take advantage of
a risk-free investment opportunity provided by employee stock purchase plans. In
plans that we consider employees can purchase the company stock at a discount from
the market price and then resell it immediately in the open market. An employee
who does not sign up for the plan forfeits on average $4,627 annually. Using data
on individual employees, we …nd that non-participation is more likely among people
who earn lower wages, are less educated, and make fundamental errors in valuation
of …nancial securities. Further, employees who fail to earn risk-free pro…ts are less
likely to participate in pension plans and the equity market in general. Our results
suggest that individuals do not make the best …nancial choices even when there is no
uncertainty over outcomes of investment.
Babenko is at Arizona State University and Sen is at Hong Kong University of Science and Technology. We are extremely grateful to Joseph Blasi and Doug Kruse for sharing their survey data on
individual stock ownership of employees and to David Yermack for providing many helpful suggestions. We also thank for comments Andrew Ang, Sreedhar Bharath, Doug Diamond, Robert Dittmar,
Daniel Ferreira, Bruce Grundy (the discussant), Harrison Hong, Vikram Nanda (the discussant),
Mark Seasholes, Stephan Siegel, Tyler Shumway, Yuri Tserlukevich, participants of the Indian School
of Business CAF Summer Research Conference, and participants of seminars at Arizona State University, BI Norwegian School of Management, University of Michigan, Purdue University, and Simon
Fraser University. We are grateful to Kirak Kim for able research assistance.
1
Are individuals good at making optimal investment choices? This question has
been hotly debated in a number of di¤erent contexts, such as participation in equity
market and saving for retirement (see e.g., Mankiw and Zeldes (1991), Haliassos and
Bertaut (2005), and Benartzi and Thaler (2007)). However, it is generally di¢ cult
to …nd de…nitive evidence of suboptimal behavior in these contexts since a wide variety of unobserved factors (e.g., individual risk aversion) can determine the optimal
choices of individuals. Additionally, the interpretation of evidence often depends on
the speci…cation of the model of market equilibrium against which the rationality is
judged (Shleifer (2000)). In this paper, we avoid these issues by analyzing a unique
setting provided by employee stock purchase plans (ESPPs) and empirically analyze
the central prediction of economic theory –that individuals should always take up an
investment opportunity with positive pro…ts and zero risk.
Although a large economic literature is now devoted to studying investments by
employees in 401(k) plans (e.g., Benartzi (2001), Cohen (2009), Choi et al. (2011)),
relatively little attention has been paid to employee stock purchase plans, which are
almost as common.1 In essence, ESPPs are company-run programs that allow participating employees to buy company stock at a discount. The typical explicit discount is
set at the 15% of the stock price, but the actual value provided by these plans is often
higher because of a lookback feature, i.e., the option to buy stock at the lower of the
prices at two points of time. An interesting feature of ESPP plans is that they typically
allow participating employees to sell the stock right away following the purchase. This
contractual feature gives employees a choice not to take any risk on this investment and
to secure a pro…t.2 In this paper, we focus only on plans that allow immediate resale
of stock in the open market and analyze whether all eligible employees participate in
such plans.
1
For example, out of 918 large …rms that we consider 460 …rms had some employee stock purchase
plan during the period 1998 to 2007.
2
It is possible that after participating in the ESPP, some employees may prefer to hold the stock
(e.g., for tax saving purposes). However, participating in the ESPP and selling immediately should
always dominate the strategy of not participating at all.
2
Despite the obvious attractiveness of ESPPs, we …nd that most employees fail to
take advantage of this money-making opportunity. In our sample of large publicly
traded U.S. …rms, the average participation rate is below 30%. Moreover, an employee
who does not sign up for the plan leaves a considerable amount of money on the
table, foregoing on average $4,627 each year. The aggregate employee losses due to
non-participation amount to over $10 billion for …rms in our sample (S&P 500, S&P
400 Midcap, and NASDAQ 100 …rms over 1998 to 2007). We argue that transaction
costs are relatively small and are unlikely to explain the wide non-participation by
employees.
We then go on to analyze a variety of factors that could drive non-participation.
We …nd that participation in ESPP tends to be higher in …rms where employees earn
higher salaries and exercise more stock options, which could be taken as evidence that
liquidity constraints a¤ect participation decisions. However, even for …rms with highly
paid employees who are unlikely to be constrained, the average participation rate is
well below 100%. For example, the average participation rate among the …rms with
average annual salary over $100,000 is 38.1%. We also …nd that factors likely correlated
with employee loyalty towards the company appear to be important for participation,
such as employee approval ratings of …rm’s CEO and the …rm being one of the best
100 companies to work for.
In addition to using the aggregate annual …rm-level participation data, we also analyze the individual data from a detailed survey of employees at four companies with
ESPPs, which was conducted in 2004-2005 as a part of National Bureau of Economic
Research (NBER)’s Shared Capitalism Research project. The survey questionnaires
contained 80-100 questions about various stock ownership programs and were administered at 323 di¤erent work sites (see Kruse, Freeman, and Blasi (2010) for a detailed
description). In all four …rms, employees were allowed to sell the stock immediately
after the purchase through ESPP. De…ning participation rate based on whether an
employee ever participated in ESPP to any extent, we …nd that average participation
3
rate in this sample is 58%. Similar to the other sample, we …nd that employees who
report lower salaries and lower household wealth are less likely to sign up for the ESPP.
Non-participation is also particularly common among people of young and old age, people prone to apathy, as measured by their lack of participation in national elections,
and individuals who incorrectly value …nancial securities. For example, employees who
grossly overvalue or undervalue their employee stock options show, respectively, a 3.5%
and 6.9% lower propensity to sign up for the plan and employees without a college degree are 4.5% less likely to participate in ESPP.
The survey data also allow us to examine whether employees tend to sell the stock
acquired through the ESPP. It should be noted that a decision to hold the stock
purchased through ESPP is not necessarily irrational. For example, there may be tax
reasons for holding or employees may have favorable information about their …rm’s
prospects. Nevertheless, an average employee in our sample is undiversi…ed and has a
substantial fraction of wealth invested in company stock. Moreover, between 39 to 59%
of employees who bought the stock through ESPP never sell it over their tenure. We
…nd that females, employees with lower salary and fewer promotions, and employees
that make more mistakes in valuing …nancial securities are all less likely to sell the
company stock acquired through ESPP.
Our results on low participation in ESPP plans have implications for the large
literature that attempts to answer why a substantial fraction of households do not
participate in equity markets (see, e.g., Mankiw and Zeldes (1991), Ang, Bekaert, and
Liu (2005), Campbell (2006), Rooij, Lusardi, and Alessie (2011), among others). Given
that a similar set of factors may be driving decisions to participate in ESPP and invest
in a general stock market, with an important di¤erence that ESPP investment is riskless, our setting allows to examine an interesting question –what factors would matter
for equity market participation, if the risk-return-tradeo¤ aspects were removed?3 This
3
Several non risk-based explanations to the low stock market participation have been o¤ered,
including the large …xed costs associated with participation, lack of trust, unawareness, lack of familiarity, and …nancial illiteracy (see, among others, Vissing-Jorgensen (2002), Hong, Kubik, and Stein
(2004), Brown et al. (2008), and Guiso, Sapienza, and Zingales (2008), Grinblatt, Keloharju, and Lin-
4
question is of particular importance because non-participation can have direct implications for the equity risk premium in the economy and the equilibrium interest rate
(Basak and Cuoco (1998)).
Our paper is related to several recent papers in behavioral economics that document
individuals forfeiting signi…cant …nancial gains. For example, Green and Lacour-Little
(1999) show that many households do not re…nance their mortgages when interest rates
decline. However, the inability to precisely observe the price of a house in their setting
leaves open the possibility that a drop in collateral values could prevent re…nancing.
Scholes and Wolfson (1989) study the discount dividend-reinvestment and stock purchase plans that allowed existing shareholders to purchase stock at a small (typically
5%) discount. They show that investment in these plans generated signi…cant pro…ts,
and yet many shareholders did not participate. However, since one typically had to
wait for shares bought through the plan for several weeks, it was di¢ cult to construct
a strategy that would generate riskless pro…t.4 Moreover, many plans quickly decreased or completely removed the discounts to eliminate this investment opportunity.
In a related paper, Poteshman and Serbin (2003) document that some traders exercise their publicly traded call options suboptimally, violating the basic principle that
people should always prefer more money to less. Similarly, Barraclough and Whaley
(2012) …nd that many exchange-traded put options that should have been exercised
early remain unexercised.
The two studies that are most closely related to ours are Engelhardt and Madrian
(2004) and Choi, Laibson, and Madrian (2011). Engelhardt and Madrian (2004) study
the tax treatment of ESPP plans in general and analyze the participation data at one
large health services company. Our results are broadly consistent with their …ndings,
but in their context, it is not clear whether the ESPP being analyzed allowed employees
nainmaa (2011), Rooij, Lusardi and Alessie (2011)). However, whether these factors apply to settings
when the investment has no risk is not immediately clear. For example, it is conceivable that lack
of trust or past stock market experience is relevant only in conjunction with having to take on some
risk, but not otherwise.
4
Scholes and Wolfson (1989) used expensive hedging strategies to minimize the price risk.
5
to sell their stock immediately. In both our samples, we restrict attention to plans
that allow immediate dispositions of stock by employees. Without this restriction,
participation in ESPP would not be risk-free and would not allow to study factors
driving lack of investments in risk-free pro…table opportunities, which is our aim. In
our study, we also …nd the e¤ects of …nancial literacy and analyze selling decisions by
employees which are not considered by Engelhardt and Madrian (2004).
Choi, Laibson and Madrian (2004) study 401(k) contributions by employees at
seven companies who are able to make penalty-free 401(k) withdrawals because they
are older than 59 12 . They …nd that these employees often do not contribute up to the
employer matching contribution threshold foregoing $507 annually. By the very nature
of their study, they focus on employees who are relatively old and constitute approximately 3% of the total employee population in their sample …rms. Whether failure
to take advantage of risk-free pro…table opportunities generalizes to people of all ages
is an interesting question. For example, Agarwal et al. (2009) provide evidence that
analytic cognitive function declines over the individual’s life-cycle by approximately
one percentile unit for each year of age, starting from the age of 20. Since ESPPs plans
are open virtually to all employees, the non-participation phenomenon we …nd is much
more general. Further, we are able to show that this phenomenon is widespread in a
large sample of …rms and quantify the aggregate employee losses.
The rest of this paper is structured as follows. Section I describes our data sources
and presents summary statistics on ESPP plan characteristics. Section II discusses
the non-participation rates in ESPP and losses of individuals. Section III explores the
determinants of failing to participate at both the …rm and individual levels. Section
IV investigates the decision not to sell the stock following the purchase. Section V
concludes with a brief summary.
6
I. Data and Summary Statistics
A. Firm Level Data
We hand collect …rm level data on employee stock purchase plans from 10-K forms
for all …rms in the S&P 500 index, NASDAQ 100 index, and the S&P 400 midcap
index over the …scal years from 1998 through 2007. If the company has an ESPP, we
also obtain a detailed ESPP contract; such contracts are typically located in the past
SEC …lings. We restrict attention to tax-quali…ed ESPPs since they are open to all
employees (with exception of executives owning more than 5% of …rm’s stock),5 and do
not include any plans that require employees to hold the stock following the purchase.
The sample selection criteria are outlined in Table I.
The contract features of stock purchase plans are given in Table II. On average, the
plan has been adopted by the …rm’s shareholders more than 8 years ago and allows
for contributions not exceeding the lower of $25,000 per year or a speci…ed percentage
of annual salary, commonly 10%, 15%, or 20%. Some of the plans impose additional
restrictions on the number of shares that can be purchased by employees during the year
or specify a lower dollar cap on contributions. Taking into account these restrictions
on participation, the average maximum allowed contribution comes out to $11,347 per
annum.
The average discount stipulated by the plan is 13.8%, with most plans having a
15% discount o¤ the market price.6 In addition to discount, over 80% of ESPPs have
a lookback feature, that allows employees to purchase stock at the discounted price
based on the lower of prices at the beginning and the end of the purchase period. On
the top of that, 8.8% of plans have a reset option that allows to reset the purchase
price to the price in the beginning of the previous purchase period, provided that
5
In addition, a company is allowed (but not required) to exclude employees with less than two years
of tenure, employees working less than 20 hours per week, and “highly compensated employees,” as
de…ned in section 414(q) of the Code.
6
The clustering of discounts is probably due to the nature of the tax code. Speci…cally, one of the
requirements of Section 423 is that, to preserve the tax-quali…ed status, the plan discount cannot be
set higher than 15%.
7
price was lower than the price at the beginning of the current purchase period. The
value of the lookback and reset options is 13.9% of the purchase price on average.
Thus, participation in ESPP provides an average expected discount of 27.7%, which is
equivalent to a return on investment of 38.3% over the purchase period.
To join the ESPP, an employee has to opt in (typically by …lling a small form
indicating what percentage of pay employee wants to contribute to the plan), and the
company then buys the stock for employee at the end of purchase period. Unlike some
of the 401(k) plans (see, e.g., Madrian and Shea (2001)), none of the ESPPs that we
consider automatically enroll employees in the plan. However, if an employee signs up
once and does not take any action to opt out, the default is that employee remains
enrolled in the plan and contributes the previously speci…ed percentage of pay.
As documented in Panel A, over 91% of plans allow employees to withdraw the
contributed funds from the plan up to the date of the actual purchase. Thus, if employees are unexpectedly hit by a liquidity shock during the o¤ering period, in most
cases they can get their accumulated contributions back. Some plans also allow to
decrease or increase the contribution rate during the o¤ering period (61.4% and 43.3%,
respectively) and virtually all plans allow to change the contribution for the next o¤ering period. Finally, we document in Panel A of Table II that majority of ESPP plans
allow employees to sign up starting from the …rst day of their employment, with few
plans requiring a minimum employment of several months (most typical restrictions, if
any, are 30 and 90 days of employment).
Since the maximum contribution to an ESPP is often set as a percentage of an
employee’s annual compensation, we obtain employee salary data from the website
Glassdoor.com. These data are anonymously reported by …rm employees (segregated
by the job title in each …rm) and cover most of …rms in our sample. A disadvantage
of these data is that the average number of respondents per …rm is only 286 people in
our sample, and we have a single cross section of salaries. As an alternative, we also
obtain salary data from the Compustat “sta¤ expense”item that has been used in the
8
prior literature as a proxy for employee salaries (e.g., Hanka (1998)). However, these
data often include non-salary items, such as expenses associated with pension plans,
and are available only for a small part of our sample (less than 15%). For …rms that do
not report sta¤ expense, we use the median value within industry (de…ned by two-digit
SIC code) for that year.7 Overall, we believe that survey salary is a more accurate
measure and use the Compustat salary only for robustness checks (the correlation
between the two measures of participation is 87.5%). We also obtain data on employee
stock option grants and exercises for …rms in our sample, which are available through
the RiskMetrics database.
Panel B of Table I shows that the average annual salary that employees receive is
$76,902, whereas it is $56,927 for imputed Compustat wage. Employees also receive
grants of stock options with an average Black-Scholes value of $27,079 and realize
an average value of $22,754 from option exercises each year. Most of the employees
approve of a …rm’s CEO (Glassdoor.com), with the average approval rate of 61.3%.
Panel B also shows that in 8.0% of …rm-years, the …rms in our sample make the list of
100 best companies to work for, which is maintained by the Fortune magazine and the
Great Place to Work Institute.
B. Individual Level Data
Our second data set on ESPPs comes from the survey of employees at fourteen companies, conducted in 2004-2005 as a part of National Bureau of Economic Research
(NBER)’s Shared Capitalism Research project. Company surveys administered 80-100
questions to …rm employees and were given either over the web or in a paper-based
format at 323 di¤erent work sites. Typically, the companies that have broad-based
ownership programs (employee stock ownership plans, broad-based stock option plans,
7
When sta¤ expense item is non-missing for the …rm, the correlation between Compustat salary
and the survey salary is 67.4% and the survey salary has a lower mean, consistent with Compustat
data overestimating the annual salary. However when we use the industry median values for the sta¤
expense item, the correlation between the two proxies drops to 22.3%.
9
401(k)) were chosen. The survey methodology and the choice of companies are described in more detail in Kruse, Freeman, and Blasi (2010). Since only …ve out of the
fourteen companies had an ESPP plan and one of those companies was not publicly
traded, we restrict our analysis to four companies, that we label for convenience A, B,
C, and D.
Table III Panel A gives simple summary statistics for the four companies. Company A is a large multinational company that has over 30,000 employees and operates
in the high-tech industry. Company B is mid-size …rm that produces food products,
with all employees located in the United States. Company C is a large …nancial …rm
employing over 5,000 people. Finally, Company D is a small high-tech …rm that has
55.2% of employees working in countries other than U.S. The …ve most common countries that enter our sample, other than U.S., are United Kingdom, Canada, Australia,
Netherlands, and India. Since there may be minor di¤erences in terms of the plans for
employees working in di¤erent countries, we use country-…xed e¤ects in regressions and
include only countries with at least one employee participating in ESPP.8 Our results
are also robust to including only U.S. employees in the dataset.
Panel B presents the demographics of employees working in each of four …rms. Employees at company A are more educated, with 83.6% having a college degree, are more
likely to be married (82.0%), and earn higher salaries than employees at other …rms.
The average base salary, including commissions and overtime, is $98,985 in company
A, and an average reported household wealth is $724,533, where wealth includes the
household’s value of house minus the mortgage, stocks and mutual funds, cash, checking accounts, retirement accounts including 401(k) and pension assets. Employees at
company B are the least well-paid, with the average annual salary of $35,518. They
also report the lowest household wealth and are much less likely to have a college degree
(26.1%). Company C has the average salary of $46,335 and has the highest proportion
8
The reason for this is that some country-speci…c laws may not allow the …rm to establish ESPP.
The two countries that do not have any employees participating in ESPP in our sample are Hungary
and Romania.
10
of females (66.2%) and the lowest proportion of whites (8.2%) in the sample. Employees at company D have a relatively high pay (the average salary is $86,188), and are
well-educated (75.0% have a college degree).
II. ESPP Non-Participation and Employee Losses
Having established that stock purchase plans provide on average a 38.3% return on
investment over the purchase period of approximately 6 months, we next turn to examine participation in these plans by employees. We de…ne two measures of ESPP
participation. The …rst is equal to the contributions to ESPP per employee normalized
by the maximum allowed contribution per year, where the latter is the minimum of the
annual dollar limit speci…ed by the …rm and the percent of compensation employee can
contribute to the plan multiplied by the average survey salary. The second measure is
similar, but uses the salaries imputed from Compustat. The non-participation is one
minus participation. Since in each case we use the average salary at the …rm level,
rather than each individual employee salary to calculate the participation rate, it is
possible that the participation rate can be biased. However, as we show in Appendix
A, the participation rate will be overestimated by our procedure if the individual participation in ESPP is positively correlated with the individual’s salary. We show later
in the paper that this correlation is indeed positive in the data.
We start by reporting the average non-participation rate across all …rms in our
sample (Table IV). It turns out that the average non-participation rate is 80.4% in
the full sample when we use survey salary and 70.4% when we estimate salaries from
Compustat. Since none of the …rms in our sample require employees to hold the stock
following the purchase, not participating in ESPP is equivalent to leaving money on
the table since investment is both riskless and pro…table.
We next calculate how much money employees forfeit by not signing up for the plan.
When the maximum allowed dollar contribution is C, employees can buy the number
11
of shares equal to C=Discounted Price, and earn on each purchased share the spread
between the current market price and the discounted price. Thus if the combined
expected discount is D, the discounted price is equal to 1
D multiplied by the market
price, and the average annual loss is
Loss = C
(Market Price-Discounted Price)
D
=C
:
Discounted Price
1 D
(1)
Panel A shows that by not signing up for the ESPP in a …rm, an average employee
forfeits $4,627 per year or 5.7% of her annual salary. Part of this value is guaranteed
for sure because the stock plan provides a …xed discount (on average $1,642), whereas
the rest of the value comes from the built-in options (lookback and reset). The value
of options can never be negative but varies with the stock returns over the purchase
period.
To better assess the real losses of employees, we also provide estimates on the aftertax basis. Speci…cally, we calculate the tax liability of employee assuming she sells all
the shares immediately and thus there is no capital gain or loss. In this case, all the
pro…t from the ESPP purchase and the same-day sale is taxed at the ordinary income
tax rates.9 We assume the individual tax rate of 28%, which is applicable to employees
with the combined annual income of more than $69,000 and less than $144,000 (as
of year 2003). Using this tax rate, we estimate the average after-tax forfeited gain
from non-participation as $3,331 per year. If we were to assume that all employees are
subject to the highest possible marginal tax rate of 35%, we would obtain the after-tax
forfeited gains from non-participation of $3,008 per year, on average.
In Panel C of Table IV, we present the summary statistics on ESPP non-participation
in …rms that allow employees to sign up for the plan during their …rst month of employment. The tax legislation in general allows companies not to extend the ESPP
plan to employees working for the company for less than two years. Most companies,
9
If the employee holds the ESPP stock following the purchase, the tax treatment of ESPP pro…t
is more complicated but generally results in lower tax bill for the employees.
12
however, either make all full-time employees eligible or restrict participation in the
plan to employees working for a …rm for more than 90 days. Since we have the ESPP
contract terms for most …rms in our sample, we can examine whether non-eligibility of
new hires drives our results. We …nd that the non-participation is only slightly lower
in the sample of …rms that allow immediately eligibility for ESPP plan bene…ts, with
the average non-participation rate of 75.4%. However, it should be noted that lower
non-participation is at least partially driven by the higher bene…ts these plans tend to
provide to employees. For example, the average loss due to non-participation is $5,186
per individual in this sample or 6.1% of annual salary.
To see whether there is signi…cant variation in non-participation rates across …rms,
we further split the sample into …rms with high and low employee pay. Since our
sample consists of large public …rms, employees are likely to earn salaries higher than
the national average, so we use the cuto¤s of $50,000 and $100,000. Panel C shows
that in …rms with the average salary less than $50,000 per year, the non-participation
rates are remarkably high at 93.1% or 92.3%, depending on the measure. In contrast,
for …rms where employees make more than $100,000 on average, the non-participation
rates are considerably lower at 62.9% and 39.9%.
Since the …rm-level analysis uses the average salary and because of that imprecisely
estimates the average participation rate, we also use the individual employee data at
four public …rms. The main bene…t of this dataset is that it has detailed information
on individual employees, including their salaries, household wealth, investments in the
general equity market, and investments in 401(k). A disadvantage of these data, however, is that they only provide information on whether an employee ever participated
in company ESPP, and thus cannot be used to infer whether the employee participates
currently or contributes up to the maximum allowed limit. In three out of four …rms,
employees become eligible for the ESPP from the …rst day of their employment, and
in …rm B they have to work for the …rm for at least 30 days to join. Therefore, in …rm
B we do no consider employees with less than one month of tenure. In all four …rms,
13
the employees are allowed to sell the stock following the purchase.
Table V documents that the non-participation rates range from 7.7% (in …rm A) to
59.1% (in …rm D). Although these numbers are considerably lower than the averages
from our …rm-level data, the ESPP participation is measured di¤erently here (based
on whether employee ever participated to any extent in the stock purchase plan) and
the …rm selection for survey may be non-random. For example, …rm A, which is also
a part of our broad …rm sample, turns out to have the highest participation rate in
ESPP among all …rms in S&P 500 and Midcap 400.10
The average annual loss per non-participant is $4,589 in …rm A or approximately
5.1% of salary. Over the employee tenure, the amount of losses adds to $18,714.
Similarly, non-participating employees at …rm D forfeit $3,855 annually or 5.4% of
salary on average. The employee dollar losses are somewhat lower at two other …rms,
in part because of lower average salaries and also because …rm C had an annual limit
of $5,000 on ESPP contributions.
We also calculate the after-tax dollar loss to employees. Since we know the salary of
each individual employee we can obtain their marginal tax rates in the year of the survey
(assuming they do not have other income during the year). The tax treatment of ESPP
sales is somewhat complicated, with di¤erent amount of tax levied on disqualifying and
qualifying dispositions. In general, however, the e¤ective tax rate goes down if employee
holds the stock longer. For our purposes, we are interested in the amount of tax that
would be triggered if employee engages in the same-day sale. In this case, the tax
treatment is simple as there is no capital gain to consider and all income that employee
earns on the ESPP is ordinary income. Calculated in this way, the average annual
after-tax loss ranges from $599 in …rm C to $2,983 in …rm A.
In all four …rms employees are not required to hold the stock for any period of time
after the purchase. Nevertheless, we see that many employees who ever participate in
ESPP never sell the stock over their tenure. The percentage of employees who never
10
Additionally, not all the employees respond to the survey. This is particularly true for employees
of …rm A.
14
sell the ESPP stock varies across …rms from 39.5% (company D) to 58.5% (company
C). This is particularly striking since employees at four …rms already have substantial
holdings in …rm’s stock through employee stock options. For example, employees of
…rm A report the average value of currently owned stock options of over $320,000 and
employees of …rm D of over $140,000. Additionally, as reported in the table, some
employees go even further and buy more company stock on the open market. Overall,
…nding that many employees do not bother to sell the stock and take the path of least
resistance resonates well with evidence in Madrian and Shea (2001). They …nd that
many employees who were automatically enrolled in a 401(k) plan, do not change either
the default contribution rate or fund allocation.
Also reported in Table V are the participation rates in 401(k) plans by employees,
which range from 82.4% (…rm B) to 94.5% (…rm A). Additionally, we report the average values for the variable that captures equity market participation by employees.
Speci…cally, the employees were asked whether they frequently buy and sell the securities on the open market and the respondents who answered negatively were classi…ed
as non-participants in the equity market. Measured in this way, the equity market
participation is likely underestimated since some people may passively hold the stocks
without frequent trading (e.g., through assets in 401(k) plan). Firm B (where employee
pay is low and few people hold a bachelor’s degree) has the lowest fraction of employees participating in the equity market of 7.6%, and Firm A has the highest fraction of
equity market participants of 18.3% of all employees.
Finally, Table V provides the average values of our measures of …nancial sophistication. In the survey of …rm A, there were several additional questions included on the
valuation of employee stock options that can be related to …nancial literacy. Specifically, employees were asked for the least number of shares of stock they would take
in exchange for 10 out-of-the-money stock options. Note that in …rm A, over 99% of
employees received stock options sometime over their tenure. Moreover, because of
unfavorable stock returns over the previous years many employees did own underwater
15
stock options. Amazingly, the survey shows that 5.1% of surveyed employees consider such stock options completely worthless and would exchange them for 0 shares
of stock. On the other extreme, there are also people who would not exchange their
10 underwater stock options for anything less than 10 shares of stock, with frequently
suggested numbers in the range of 11-20 shares of stock. We …nd that overall 15.0% of
all surveyed people grossly overvalue stock options.
We next show some simple univariate comparisons of ESPP participation rates
among employees with di¤erent salaries, household wealth, education, and age. In Figure 1, we show the distribution of participation in ESPP by the reported annual salary.
As can be seen from the …gure, the relation is monotonic, with highly paid employees
more likely to join ESPP. For example, in …rm D there are no employees earning less
than $25,000 per year who signed up for the plan, whereas among people with income
over $75,000 close to 60% of employee participate in the plan. The corresponding
Table VI shows that the di¤erences among the groups with high and low employee
pay (below $50,000 and over $100,000) are statistically di¤erent from zero. We also
present the relation between ESPP participation and the reported household income
(Figure 2). Overall, a similar picture emerges with wealthier employees more likely to
participate in the plan. However, in this case the relation is not always monotonic,
perhaps because of the imprecise estimates of wealth provided by employees. Table VI
shows that indeed the non-participation rates in ESPP are signi…cantly di¤erent for
people with household wealth under $100,000 and those over $200,000.
In Figure 3, we examine the e¤ect of general education on the propensity of employees to participate in ESPP. We report the participation rates across four levels of
education (high school or less; some college, but no degree or associate degree; bachelor’s degree; and master’s, professional, or doctoral degree). Generally, the participation rates tend to increase with employee education, with particularly large di¤erences
observed between groups with and without bachelor’s degree. However, more education does not always translate in more plan sign-ups. For example, in …rms A and B
16
there are no signi…cant di¤erences between rates of participation by employees with a
bachelor’s degree and those with a master’s or Ph.D. The corresponding Panel C of
Table V shows the univariate comparisons of ESPP non-participation by education.
For example, in …rm D over 84% of people without bachelor’s degree have never signed
up for the ESPP, whereas for people with college degree the non-participation is approximately 50%. Figure 4 plots the average participation rate in ESPP as a function
of employee age. As can be seen from the graph, the relation has an inverse U-shape,
with both young and old individuals showing relatively lower propensity to participate.
Finally, we explore in a univariate setting whether the employees who take advantage of riskless investment opportunity through ESPP are also more likely to participate
in the general equity market and make contributions to the 401(k) plan. It can be seen
from Figure 5 that indeed equity market participation rates are consistently higher
among employees who sign up for the ESPP. For example, in …rm A only 10.1% of employees who ignore ESPP plan trade on the market, whereas this number is 19.0% for
ESPP participants. Similarly, Figure 6 shows that there is a positive relation between
401(k) contributions and ESPP participation rates.
III. Determinants of Failing to Participate in ESPP
A. Factors A¤ecting ESPP Participation
In this section, we list potential factors that can a¤ect ESPP participation rates and
describe some of the empirical variables used in our tests. We start with factors that
should matter if employees strictly maximize their wealth, have unlimited power to
understand and weigh multiple options, have perfect self-control, can forecast the consequences of each option, and choose rationally. In particular, we consider plan attractiveness, transaction costs and the opportunity costs of time, and employee liquidity
constraints. We then move on to factors that can be important if employees are boundedly rational. For example, employees may have limits to processing information or
17
derive utility not only from a higher level of wealth. Speci…cally, we consider such factors as awareness of the plan, familiarity in trading stocks, …nancial literacy, experience
and cognitive aging, trust and loyalty of employees to the …rm, and past individual
experiences in the stock market.
A.1. Attractiveness of the Plan
Since plans that we consider provide di¤erent bene…ts, we start by examining whether
participation rates are higher in …rms that o¤er more generous plans. In a broad
sample of …rms, we include the two most important characteristics of the plan: the
discount and the value of lookback and reset options as a percentage of the market price.
The logic for including these two components separately is that the discount provided
by the plan represents a certain bene…t that participating employees receive if they
immediately sell the stock, whereas the value of lookback and reset options cannot
be negative but can have di¤erent realized values depending on the stock returns. If
employees are considerably risk averse and value only the certain bene…ts provided by
the plan, the participation rates should be more sensitive to the discount rate than to
option value.
A.2. Transaction Costs and Opportunity Costs of Time
To see whether transaction costs can explain away the wide non-participation in ESPP,
we estimate the employee out-of-pocket expenses associated with trading the ESPP
stock. From our readings of company plans, many employers who specify how the
transaction costs are to be handled, tend to pay for the ESPP account maintenance
and cover at least part of brokerage costs associated with purchases of ESPP stock.
However, the fees charged for stock sales are almost always the responsibility of employee. To estimate expenses associated with trading ESPP stock, we assume that
employees sell the stock each end of purchase period (e.g., twice a year if the purchase
period is 6 months or four times a year if purchase period is 3 months) and assume an
18
average brokerage fee for a single buy or sale transaction of $25. These assumptions
yield an estimated $140 spent in brokerage fees per annum if employees are responsible
for fees associated with both stock purchases and sales.11
Additionally, we estimate the opportunity costs of time needed to manage the sales
of stock (purchases are automated by the company). We assume that each sale of
stock at the end of the o¤ering period takes one hour of employee time and calculate
the amount of forfeited salary during the year because of the time taken away by
this trading. This gives us an estimate of $101 per year on average, bringing the
total transaction costs and the opportunity costs of time to $241 per annum. The
combined estimated costs are considerably smaller than the after-tax pro…t from ESPP
participation of $3,331.
A.3. Liquidity Constraints
The next factor that we consider is the liquidity constraints faced by employees. In his
2006 presidential address, Campbell argues that binding borrowing constraints is one
of the de…ning characteristics of households. The empirical importance of household’s
liquidity constraints have also been shown in several economic contexts. For example,
Lusardi, Schneider, and Tufano (2011) …nd from the survey data that half of Americans
would probably or certainly be unable to come up with $2,000 in 30 days. Using credit
card data, Gross and Souleles (2002) …nd that increases in credit limits are associated
with an immediate and signi…cant increase in debt, particularly for people who are
close to their credit limit.
In our tests at the …rm level, we proxy for the tightness of liquidity constraints by
the average employee salary and the value realized from option exercises by employees
during the …scal year. At the individual level, we use employee salary and the total
household wealth.
11
Wwe do not …nd any relation between the ESPP participation rate and a dummy variable for
whether the …rm pays part of transaction costs (unreported).
19
A.4. Awareness, Familiarity, and Financial Literacy
Another potential reason for leaving money on the table is unawareness of the signi…cant bene…ts to ESPP participation. For example, in the context of stock market
participation, Guiso and Jappelli (2005) …nd that over 35 percent of Italian households
were simply unaware of the stocks in the late 1990s, and Hong, Kubik, and Stein (2004)
argue that many households do not participate in the general stock market in the U.S.
because they are unaware of its existence or attractiveness. They also show that more
social households show a higher propensity to invest in the stock market.
To proxy for the awareness of the ESPP at the …rm level, we use the number of
years since the plan has been adopted, presuming that over time employees had an
opportunity to learn about the plan’s existence. We also use the value of stock options
grants per employee in this …rm. The rationale for this variable is that employees
who receive stock options are more likely to be familiar with dealing with the stock
in general. At the individual level, we look at whether employee ever received stock
options and whether employee frequently trades other securities.
A related explanation to the low participation is …nancial illiteracy of employees
or their perception of such. For example, Graham, Harvey, and Huang (2005) …nd
that investors who are more comfortable about their ability to understand investment
products tend to trade more frequently. Rooji, Lusardi, and Alessie (2011) show that
many of the Dutch households cannot tell the di¤erence between stocks and bonds and
do not understand the bene…ts of diversi…cation, whereas Lusardi and Mitchell (2006)
provide evidence that approximately 80% of baby boomers cannot correctly compound
interest. They also document that individuals who lack literacy are much less likely to
plan for retirement, whereas Kimball and Shumway (2009) and Calvet, Campbell, and
Sodini (2007) …nd that less sophisticated investors are less likely to participate in the
stock market and/or to diversify their investments. Additionally, Bernheim, Garrett,
and Maki (2001) …nd that …nancial education has e¤ect on asset accumulation, and
Calvet, Campbell, and Sodini (2009) …nd that educated investors tend to rebalance
20
their portfolios more actively.
To proxy for …nancial literacy and employee education, we use the dummy variables
for gross overvaluation or undervaluation by employees of their out-of-the-money stock
options, as well as the employee education level (whether she received a college degree),
and the number of promotions received over tenure.
A.5. E¤ect of Age through Experience and Cognitive Ability
Previous research has documented the lifecycle patterns in …nancial mistakes by individuals. For example, Agarwal et al. (2009) study di¤erent types of individual credit
behavior and conclude that middle-aged adults are less prone to mistakes than both
younger and older adults, with the cost-minimizing performance occurring around the
age of 53. They argue that young people lack necessary experience to make the right
choices, whereas very old individuals can have declining cognitive ability. They present
evidence that prevalence of dementia sharply increases after age of 60 and doubles with
every …ve years of age. Similarly, Korniotis and Kumar (2011) examine the individual
investments in general equity market and document an inverted U-shape age-skill pattern. To analyze the e¤ect of age on employee decisions to sign up for the ESPP, we
include in the individual level regressions employee age and age-squared.
A.6. Individual Past Experiences
We also consider a possibility that individual experiences of stock market ‡uctuations
a¤ect the decision to buy the stock through ESPP. Malmendier and Nagel (2011)
provide evidence that experiencing lower stock returns over the course of one’s life
lowers the individual’s willingness to take on risk. They show that such people have
lower expectations of future stock returns and are less likely to participate in the stock
market. Similarly, Odean, Strahilevitz, and Barber (2010) …nd evidence that investors
repeat actions that previously resulted in pleasure and avoid those that led to pain.
Additionally, past experience in the stock market can be important because employees
21
may extrapolate from the past to the future (Benartzi (2001)).
It is important to point out that in our setting, employees do not need to be
exposed to price risk since they can sell the stock right away. Nevertheless, people
who experienced negative stock returns and saw, for example, large declines in the
value of their employee stock options or retirement accounts, may stay away from any
investments in stocks. In the individual level tests, we use a measure of weighted past
stock returns of …rm over the course of employee tenure, calculated as
Stock Return Experience =
tenure
Xit 1
(2)
wit (k; ) Rt k ;
k=1
wit (k; ) =
(tenureit
tenure
Xit 1
k)
(tenureit
;
(3)
k)
k=1
where we take
= 1:5 as estimated by Malmendier and Tate (2011). At the …rm level,
we test for the in‡uence of past experiences on the decision to sign up for the plan
by using the contemporaneous …rm’s stock returns and the buy-and-hold stock returns
over the previous three years.
A.7. Trust and Loyalty
Finally, we test for two other explanations to non-participation. The literature has
suggested that trust may be an important element needed for an individual to invest
into something (Guiso, Sapienza, and Zingales (2008)). It is an interesting empirical
question whether trust remains an important element when individual can earn riskless
pro…t on investment. To proxy for trust at the individual employee level, we rely
on the survey questions that ask employees to evaluate whether the company keeps
its promises and whether the company is fair to its employees. We also include the
interactive variable between trust and employee education, since it is possible that
trust can have larger e¤ect for employees who lack necessary education to …gure out
the details of the plan.
22
Another possibility is that some employees may not like to participate in ESPP because they do not like to be associated with the company or be considered its investors
(no matter how temporary). For example, Cohen (2009) shows in the context of 401(k)
plans, the importance of loyalty in pension contribution decisions of employees. At the
…rm level, we proxy for loyalty using the dummy variable for whether the …rm makes a
list of 100 best companies to work for during the year and the average approval rate of
a …rm’s CEO by employees. At the individual level, we rely on survey questions that
ask employees whether they feel loyal towards the …rm and whether they have a sense
of sharing a common purpose with their employer.
B. Firm-Level Results
The determinants of ESPP participation at the …rm level are presented in Table VII.
In all speci…cations at the …rm level, we control for such …rm-characteristics as the
amount of research and development normalized by …rm’s assets, …rm size as proxied
by the logarithm of book assets, Tobin’s Q, the average salary of …rm’s employees, the
discount provided by the plan, the value of lookback and reset options, and the number
of years since the plan has been adopted. Controlling for …rm size may be important
since …rms of di¤erent sizes may have di¤erent human resource policies. The amount
of research and development can capture …rms with more educated employees, and
average Q proxies for …rm’s growth opportunities. We also include year and industry…xed e¤ects (Fama-French 17 industries); the standard errors are clustered at the …rm
level.
We …nd that participation rates indeed tend to respond positively to plan bene…ts.
For example, an additional 1% of discount increases participation rate by 0.5-1.1%.
Consistent with employees valuing less the built-in options, we observe that a 1% of
lookback and reset value increases participation by smaller amount (approximately 0.20.3%). This can indicate either employee risk aversion or their inability to understand
23
or evaluate the value of the more complex plan terms.12
The liquidity constraints also appear to be important for employee sign-ups to the
ESPP. For example, the participation rates tend to increase with the average employee
salary (the coe¢ cient is signi…cantly di¤erent from zero in 5 out of 6 speci…cations).
The participation rates are also related to the value obtained by employees from their
option exercises. The economic e¤ect of liquidity constraints is modest. For example,
a one standard deviation increase in value from option exercises increases the participation rate by 2.9%, when the average participation is 20.6%.
We …nd some evidence that awareness of the plan decreases the non-participation.
Speci…cally, as the number of years since plan adoption increases we see higher participation rates, with each additional year increasing participation by approximately
0.5%.13 Firms that make larger option grants tend to have higher participation rates,
perhaps because their stock programs are more broad-based and employees are more
familiar in dealing with stock (column 2).
Our results also provide some support for employee loyalty explanation. Speci…cally,
we …nd that being the best 100 employer during the year is associated with higher
participation rates. The economic magnitude of loyalty is considerable since being a
best employer is associated with an approximately 9.3% higher participation rate. We
observe similar e¤ects on participation if we use employee CEO approval ratings.14
However, a caveat is that CEO ratings are available only at a single date and may be
endogenous. For example, participating in ESPP may make employees more satis…ed
with their jobs and approve of a CEO.
12
We also considered whether the participation rates are a¤ected by other terms of the plan, such
as withdraw contributions option, increase and decrease options, whether the plan pays interest on
contributions, and whether in addition to payroll deductions employees can contribute in the form of
lump-sum payments. Among these, we found only a weak positive e¤ect of the increase option on
participation (signi…cant at the 10% level).
13
If we include both years since adoption and years since adoption squared, we observe the coe¢ cient
on the …rst variable is positive and the coe¢ cient on the second variable is negative. This suggests
that learning pace slows down after the plan has been in place for a number of years.
14
Other measures of loyalty that we considered are employee satisfaction with their jobs and the
number of analysts covering the stock. Both of these variables are positively correlated with ESPP
particicpation.
24
Finally, in the last speci…cation we examine whether employees decision to sign up
for the plan is determined by their past experiences in the stock market. We …nd that
when we control for the …rm’s Q (market-to-book) ratio, the past returns do not add
any explanatory power to the participation rates. However, if Q were dropped from
the speci…cation, the contemporaneous stock returns, as well as stock returns over the
previous three years would be signi…cant predictors of participation rates. Thus, our
results on the importance of past stock returns experience are mixed.
C. Individual Level Results
We next turn to the participation in ESPP at the individual level. In all speci…cations,
we include country- and …rm-…xed e¤ects. Country-…xed e¤ects are included to capture
the e¤ects of distance, language, culture, and any unobserved di¤erences in terms of
the plans. We use …rm-…xed e¤ects to capture the variation due to di¤erences in plan
designs and human resource policies across …rms. The results of probit estimation are in
Table VIII. Overall, we …nd that participation is more common among people holding
college degrees, who have been promoted, have longer tenure at the …rm, and report
higher salaries (column 1). Asian people tend to participate more frequently in ESPP,
and in unreported results we …nd no other race variables to be signi…cant determinants
of participation in the plan. We also …nd a non-linear e¤ect of age on the decision
to sign up for the plan. As people age and become more experienced they are more
likely to take advantage of plan bene…ts, however this e¤ect is reversed at old age. The
inverted U-shape age-skill pattern we …nd in the context of the ESPP participation is
similar to results in Korniotis and Kumar (2011) on individual investments in general
equity market.
Our evidence also points to importance of familiarity in dealing with stock. For
example, we …nd that trading other securities and having ever received stock option
are associated with lower non-participation (column 2). In column 3, we test whether
people who are likely to make mistakes in valuation of stock options are also more
25
prone to not sign up for the ESPP. We …nd that indeed both groups of employees who
grossly overvalue and undervalue underwater stock options are less likely to make use
of ESPP. These results underscore the importance of …nancial literacy.
We also analyze whether the past individual’s experiences of market ‡uctuations,
as measured by the weighted …rm stock returns over the employee tenure, a¤ect the
decision to enroll in ESPP plan. We …nd that a 1% higher annual stock returns
over the course of one’s tenure increases the probability of participation in ESPP by
4.4%. We also test whether general tendency to procrastinate, as proxied by the nonparticipation in the most recent national election, is associated with non-participation
in ESPP. Indeed we …nd support for this explanation (column 4), with people voting
in national elections having a 3.8% higher likelihood to participate in ESPP.
We explore such explanations to non-participation as the lack of trust and loyalty.
Overall, we …nd that people who report to feel loyal to the …rm or have a sense of
common purpose participate more in the plan (column 2). A similar result is observed
for employee trust at the individual level, as proxied by the responses to questions of
whether company keeps its promises and whether it is fair to its employees. However,
the e¤ect of trust is not signi…cant in the full sample. Interestingly, when we interact
the trust with employee education, we …nd that trust indeed is important determinant
of participation by employees who hold no bachelor’s degree. Our interpretation of this
result is that people with less education …nd it more di¢ cult and costly to study the
terms of the plan and thus perhaps cannot determine whether it is indeed a good deal.
Whenever these individuals tend to trust the …rm at which they work, they are more
likely to invest into ESPP.
Finally, to address a possible issue that there may be slight di¤erences in the terms
of plans o¤ered to US and international employees (e.g., international plans may require
employees to exchange the proceeds from ESPP sale into local currency using speci…ed
exchange rates), we rerun all speci…cations for only U.S. employees and …nd similar
results.
26
IV. Determinants of Failing to Sell the ESPP Stock
Our individual dataset also allows us to examine whether employees quickly sell the
stock acquired through ESPP. A large number of papers argue that employees in the
U.S. invest too much into their company stock (see e.g., Benartzi (2001), Huberman
and Sengmueller (2004), Poterba (2003), and Cohen (2009)). For example, Meulbroek
(2005) estimates that under plausible parameters the value of company stock is only
50% on a risk-adjusted basis to undiversi…ed employees. However, the decision not sell
the stock immediately may be justi…ed if employees have favorable information about
future stock returns, or if employees trade o¤ the diversi…cation bene…ts with smaller
tax bill resulting from longer holding period. Overall, given such a large estimate of
the diversi…cation cost to employees, the decision to never sell the stock over employee
tenure is likely suboptimal. Additionally, we …nd in unreported results that people
who report not selling the ESPP stock also report lower pro…t from ESPP.
Here we investigate the determinants of failing to sell the company stock over
employee tenure. Since only employees who decided to participate in ESPP in the …rst
place, can make a decision whether to keep or sell the stock afterwards, we estimate
our model using the Heckman two-stage sample selection method. We use the same
dependent variables for the selection equation (ESPP participation), as for the outcome
variable (decision not to sell the stock), so that our identi…cation comes from the nonlinearity in the Mills ratio term. Table IX reports our results for the outcome equation,
where the dependent variable is equal to one if employee indicates that she has never
sold the company stock acquired through ESPP, otherwise the dependent variable is
equal to zero.
We …nd that employees with more promotions and higher salary are more likely
to sell the stock (column 1). Females are signi…cantly more likely to hold the stock,
which is consistent with results provided by Barber and Odean (2001) that women
tend to trade less than men. However, there are some peculiar results as well. For
27
example, we …nd that people with college degrees tend to hold the company stock
longer. One interpretation of such result, is that highly educated people are more
likely to be overcon…dent and hold the stock in hopes of reaping high rewards.
Since some people who acquire the stock through ESPP perhaps do not know
how to sell it through the broker or are frightened by the complexities of the tax
code associated with stock trading, we test whether previous exposure to trading is
associated with more dispositions of stock acquired through ESPP. Indeed, we …nd that
people who report to trade frequently in the stock market are more likely to dispose
of the ESPP stock (column 2).
We also …nd evidence that …nancial illiteracy, as proxied by over- and undervaluation of underwater stock options, is associated with lower propensity to sell the stock.
This is especially interesting since people who place no value on out-of-the money
stock options must have low expectations of future stock price, yet they tend to hold
the ESPP stock longer. This result underscores that this variable is more likely to
capture the degree of …nancial illiteracy rather than beliefs about future stock returns
of the company.
Table IX also shows how employee loyalty and trust to the company a¤ect employee
decisions to sell the stock. Perhaps non-surprisingly, we …nd that employees who are
more loyal to the …rm and more trusting tend not to dispose of the stock. Thus while
loyalty and trust may allow employees to make use of the money-making opportunity
through ESPP, at the same time they may hurt employees by leaving them subject to
risk.
V. Conclusion
In this paper, we document that a majority of employees at large public …rms do
not invest in ESPP even when it guarantees a riskless positive return on investment.
An average employee forfeits approximately 5.7% of salary or $4,627 per annum by
28
not signing up for the ESPP. Our results suggest that non-participation is at least
partially attributed to …nancial illiteracy and unfamiliarity in dealing with stocks.
Using individual survey data, we document that non-participating employees earn lower
wages, are less educated, and are more likely to incorrectly value …nancial securities.
Employees who fail to take advantage of riskless investment through ESPP are also
less likely to enroll in a 401(k) plan or participate in a general equity market. Our
…ndings suggest that households make mistakes that signi…cantly a¤ect their welfare.
29
References
Agarwal, Sumit, John C. Driscoll, Xavier Gabaix, and David Laibson, 2009, The age of
reason: Financial decisions over the life-cycle with implications for regulation, Brookings Papers on Economic Activity 2, 51-117.
Ang, Andrew, Geert Bekaert, and Jun Liu, 2005, Why stocks may disappoint, Journal
of Financial Economics 76, 471-508.
Barraclough, Kathryn, and Robert E. Whaley, 2012, Early exercise of put options on
stocks, Journal of Finance 67, 1423–1456.
Barber, Brad M., and Terrance Odean, 2001, Boys will be boys: Gender, overcon…dence, and common stock investment, Quarterly Journal of Economics 116, 261-292.
Basak, Suleyman, and Domenico Cuoco, 1998, An equilibrium model with restricted
stock market participation, Review of Financial Studies 11, 309-341.
Benartzi, Shlomo, 2001, Excessive extrapolation and the allocation of 401(k) accounts
to company stock, Journal of Finance 56, 1747-1764.
Benartzi, Shlomo, and Richard H. Thaler, 2007, Heuristics and biases in retirement
savings behavior, Journal of Economic Perspectives 21, 81-104.
Bernheim, B.Douglas, Daniel M. Garrett, and Dean M. Maki, 2001, Education and
saving: The long-term e¤ects of high school …nancial curriculum mandates, Journal of
Public Economics 80, 435–465.
Brown, Je¤rey R., Paul A. Smith, Zoran Ivkovich, and Scott J. Weisbenner, 2008,
Neighbors matter: Causal community e¤ects and stock market participation, Journal
of Finance 63, 1509-1531.
Calvet, Laurent E., John Y. Campbell, and Paolo Sodini, 2007, Down or out: Assessing
the welfare costs of household investment mistakes, Journal of Political Economy 115,
707-747.
Calvet, Laurent E., John Y. Campbell, and Paolo Sodini, 2009, Fight or ‡ight? Portfolio rebalancing by individual investors, Quarterly Journal of Economics 124, 301-348.
Campbell, John Y., 2006, Household …nance, Journal of Finance 61, 1553-1604.
Choi, James J., David Laibson, and Brigitte C. Madrian, 2011, $100 Bills on the
sidewalk: Suboptimal investment in 401(k) plans, forthcoming in Review of Economics
and Statistics.
Cohen, Lauren, 2009, Loyalty-based portfolio choice, Review of Financial Studies 22,
1213-1245.
Englehardt, Gary V., and Brigitte C. Madrian, 2004, Employee stock purchase plans,
National Tax Journal 57, 385-406.
30
Graham, John, Cam Harvey, and Hai Huang, 2005, Investor competence, trading frequency, and home bias, NBER working paper.
Green, Richard, and Michael Lacour-Little, 1999, Some truths about ostriches: Who
never re…nances their mortgage and why they don’t, Journal of Housing Economics 8,
233-248.
Grinblatt, Mark, Matti Keloharju, and Juhani Linnainmaa, 2011, IQ and stock market
participation, Journal of Finance 66, 2121-2164.
Gross, David B., and Nicholas S. Soulels, 2002, Do liquidity constraints and interest
rates matter for consumer behavior? Evidence from credit card data, Quarterly Journal
of Economics 116, 149-185.
Guiso, Luigi, and Tullio Jappelli, 2005, Awareness and stock market participation,
Review of Finance 9, 537-567.
Guiso, Luigi, Paola Sapienza, and Luigi Zingales, 2008, Trusting the stock market,
Journal of Finance 63, 2557-2600.
Haliassos, Michael, and Carol Bertaut, 1995, Why so few hold stocks? Economic
Journal 105, 1110-1129.
Hanka, Gordon, 1998, Debt and the terms of employment, Journal of Financial Economics 48, 245-282.
Hong, Harrison, Je¤rey D. Kubik, and Jeremy C. Stein, 2004, Social interaction and
stock-market participation, Journal of Finance 59, 137-163.
Huberman, Gur, and Paul Sengmueller, 2004, Performance and employer stock in
401(k) plans, Review of Finance 8, 403-443.
Korniotis, George M., and Alok Kumar, 2011, Do older investors make better investment decisions?, Review of Economics and Statistics 93, 244-265.
Kimball, Miles S., and Tyler Shumway, 2010, Investor sophistication and the home bias,
diversi…cation, and employer stock puzzles, University of Michigan working paper.
Kruse, Douglas L., Richard B. Freeman, and Joseph R. Blasi (eds.), 2010, Shared
capitalism at work: Employee ownership, pro…t and gain-sharing, and broad-based
stock options, University of Chicago press.
Lusardi, Annamaria, Daniel J. Schneider, and Peter Tufano, 2011, Financially fragile
households: Evidence and implications, Brookings Papers on Economic Activity, 83134.
Lusardi, Annamaria, and Olivia S. Mitchell, 2006, Baby boomer retirement security:
The roles of planning, …nancial literacy, and housing wealth, working paper.
Madrian, Brigitte C., and Dennis F. Shea, 2001, The power of suggestion: Inertia
in 401(k) participation and savings behavior, Quarterly Journal of Economics 116,
1149-1525.
31
Malmendier, Ulrike, and Stefan Nagel, 2011, Depression babies: Do macroeconomic
experiences a¤ect risk taking? Quarterly Journal of Economics 126, 373-416.
Mankiw, Gregory N. and Stephen Zeldes, 1991, The consumption of stockholders and
nonstockholders, Journal of Financial Economics 29, 97-112.
Meulbroek, Lisa, 2005, Company stock in pension plans: How costly is it? Journal of
Law and Economics 48, 443-474.
Odean,Terrance, Michal A. Strahilevitz, and Brad M. Barber, 2010, Once burned,
twice shy: How naïve learning, counterfactuals, and regret a¤ect the repurchase of
stocks previously sold, UC Berkeley working paper.
Poterba, James, 2003, Employer stock and 401(k) plans, American Economic Review
93, 398-404.
Poteshman, Allen M., and Vitaly Serbin, 2003, Clearly irrational …nancial market
behavior: Evidence from the early exercise of exchange traded stock options, Journal
of Finance 58, 37-59.
van Rooij, Maarten, Annamaria Lusardi, and Rob Alessie, 2011, Financial literacy and
stock market participation, Journal of Financial Economics, 449-472.
Scholes, Myron, and Mark A. Wolfson, 1989, Decentralized investment banking: The
case of discount dividend-reinvestment and stock-purchase plans, Journal of Financial
Economics 23, 7-35.
Shleifer, Andrei, 2000, Ine¢ cient markets: An introduction to behavioral …nance (Oxford University Press, Oxford).
Vissing-Jorgensen, Annette, 2002, Towards an explanation of household portfolio choice
heterogeneity: Non…nancial income and participation cost structures, NBER working
paper.
32
Appendix A. The Bias in Estimation of Participation
in ESPP Using the Average Salary Data
Let the random variable Pe denote the participation rate in the ESPP, i.e. the contri-
bution of an employee divided by the maximum contribution allowed. Let the random
variable Se denote salary of an employee. For each …rm, we would like to measure the
average participation rate, i.e. E(Pe). However, because we observe only the average
employee salary per …rm instead of a set of individual salaries, we are measuring
e
E(Pe S)
:
e
E(S)
To see whether we over- or underestimate the true average participation rate in
each …rm, we have to sign the di¤erence between the two measures, i.e.,
e
E(Pe S)
e
E(S)
E(Pe):
e and using the de…nition of covariance we see
Multiplying the equation by E(S),
e
E(Pe S)
e = Cov(Pe; S):
e
E(Pe)E(S)
Thus if the correlation between the individual employee salary and participation in
ESPP plan is positive, we will tend to overestimate the participation rate.
33
Appendix B. Figures and Graphs
Figure 1. Relation between employee salary and the average ESPP participation rate.
Salary is the employee’s annual base pay (excluding any bonuses and commissions).
Figure 2. Relation between household wealth and the average ESPP participation rate.
Wealth is the employee’s reported estimate of the household’s value of house minus the
mortgage, plus the value of stocks, mutual funds, and bonds owned, cash, checking
accounts, and the value of retirement accounts.
34
Figure 3. Relation between employee education and the average ESPP participation
rate. High school includes employees who completed less than 9th grade, 9th to 12th
grade with no diploma, or high school; some college category includes employees with
some college but no degree and employees with associate’s degrees. Master’s or PhD
category includes employees with master’s degrees, professional school degrees (MD,
JD), and doctorate degrees.
Figure 4. Relation between employee age and the average ESPP participation rate.
35
Figure 5. Relation between equity market participation in each …rm and ESPP
participation. Equity market participation is equal to one for an individual if she
answers “Yes” to survey question “Do you frequently buy and sell securities on the
market?” and is zero otherwise.
Figure 6. Relation between 401(k) participation in each …rm and ESPP participation.
401(k) is equal to one for an individual if she reports to currently contribute to …rm’s
401(k) plan and is equal to zero otherwise.
36
Table I. Sample Selection Criteria.
The table describes the sample selection criteria and the remaining number of …rms. The
sample includes S&P 500, S&P 400 midcap, and NASDAQ 100 …rms that sponsor at least
one employee stock purchase plan over the period 1998 to 2007 with the following features:
(1) the plan is tax-quali…ed, (2) the plan does not require employees to hold the stock for
any amount of time following the purchase, and (3) su¢ cient data are available to calculate
the participation rate in ESPP.
Selection Criteria
Unique …rms in S&P 500, S&P 400, and NASDAQ 100 as of 2007
Non-foreign …rms that …led 10K forms
Firms o¤er an ESPP plan
ESPP plan o¤ered is tax-quali…ed
ESPP plan o¤ered allows for immediate stock resale
Su¢ cient data are available to calculate participation rate
37
Remaining Firms
925
918
460
359
271
237
Table II. Descriptive Statistics of ESPP Plan Characteristics and Other
Variables (Firm Level).
The sample consists of S&P 500, S&P 400 midcap, and NASDAQ 100 …rms with tax-quali…ed
stock purchase plans that allow employees to sell the stock following the purchase. Years since
plan adoption is the number of years since the plan was …rst adopted as disclosed in SEC
…lings; maximum percent of comp. is the maximum percent of compensation an employee can
contribute towards ESPP; annual dollar limit on contributions is the maximum of the dollar
limit on participation, as speci…ed by the company, and the average survey wage multiplied
by the maximum percent of comp.; discount is the percentage discount at which employee can
buy a stock; lookback option is equal to one if the price at which employees can buy the stock
is lower of the price at the beginning and the end of the o¤ ering period minus the speci…ed
discount; reset option is the option to reset the purchase to the price at the beginning of the
previous o¤ering period; lookback and reset value is the total value of the lookback and reset
options as a percentage of purchase price; combined expected discount is the sum of discount,
and lookback and reset value; withdraw contribution option is equal to 1 if the …rm allows
to withdraw contributions during the o¤ering period, and is 0 otherwise; increase (decrease)
contribution option is equal to 1 if the …rm allows to increase (decrease) contributions rate
during the o¤ering period, and is 0 otherwise; minimum employment is the minimum number
of months of employment to qualify for the stock purchase plan; survey salary is the average wage as self-reported by employees on the Glassdoor.com; Compustat salary is equal to
the sta¤ expense (from Compustat) normalized by the number of employees; whenever this
item is missing, it is set to the median value within industry (de…ned by two-digit SIC code)
for that year; option grants/employee is the Black-Scholes value of granted employee stock
options, normalized by the number of employees; value from option exercises/employee is
the number of options exercised by employees multiplied by the di¤erence in stock price and
weighted average exercise price, all normalized by the number of employees; CEO approval
by employees is the percentage of employees that approve of the …rm’s CEO (Glassdoor.com);
100 best company is the dummy equal to 1 if the …rm is listed in the year as one of the 100
best companies to work for by the Fortune magazine, and is 0 otherwise.
Panel A: ESPP Plan Characteristics
Obs.
2,249
2,232
2,282
2,266
2,229
2,262
2,213
2,113
2,113
2,175
2,128
2,128
2,229
Years since plan adoption
Maximum percent of comp. (%)
Annual dollar limit on contributions ($)
Discount (%)
Purchase period (months)
Lookback option (%)
Reset option (%)
Lookback and reset value (%)
Combined expected discount (%)
Withdraw contribution option (%)
Increase contribution option (%)
Decrease contribution option (%)
Minimum employment (months)
38
Mean
8.23
17.83
11,347
13.81
6.27
81.02
8.81
13.85
27.70
91.68
42.90
62.64
3.75
SD
6.09
21.25
6,138
3.20
4.13
39.21
28.35
8.92
12.49
27.63
49.51
48.39
6.67
10%
2
10
5,019
10
3
0
0
0
15.00
0
0
0
0
50%
7
10
9,768
15
6
1
0
12.29
26.89
1
0
1
0
90%
16
20
25,000
15
12
1
0
28.99
43.99
1
1
1
12
Panel B: Other Variables
Compustat salary (industry val. used) ($)
Survey salary ($)
Compustat salary (if sta¤ expense and
survey salary non-missing) ($)
Survey salary (if sta¤ expense and
survey salary non-missing) ($)
Option grants/employee ($)
Value from option exercises/employee ($)
CEO approval by employees (%)
100 best company
Obs.
2,162
2,151
223
Mean
56,927
76,902
86,696
SD
28,078
21,294
56,803
10%
34,681
42,805
35,029
50%
52,324
80,078
66,487
90%
82,130
102,253
196,324
223
67,699
19,414
45,527
69,190
101,221
1,565
1,650
1,117
2,280
27,079
22,754
61.28
0.080
58,897
48,118
22.33
0.272
368
101
25.00
0
5,150
3,466
64.00
0
68,858
61,675
88.00
0
39
Table III. Firm and Employee Characteristics at Four Firms (Individual
Level).
Panel A gives the broad …rm characteristics and Panel B presents the individual employee characteristics. Number of promotions is the number of promotions received over the
employee tenure, where 3 or more promotions are coded as 3; salary is the annual base pay
(excluding any bonuses and commissions) before taxes and deductions; wealth is the employee’s estimate of household’s value of house minus the mortgage, plus value of stocks,
mutual funds, and bonds owned, cash, checking accounts, and value of retirements accounts
including 401(k).
Panel A: Firm Characteristics
Year of survey
Web or paper survey
Response rate (%)
Industry
Total active employees
Number of employees
Fraction of US
employees (%)
Company has ESPP
Company has 401(k)
Company has broadbased stock options
Company A
2004
Web
19
High tech
Over 30,000
6,733
84.5
Company B
2004/2005
Paper
72
Manufacturing
500-1,000
429
100
Company C
2005
Web
63
Financial
5,000-10,000
1,584
100
Company D
2005
Web
77
High tech
Under 500
230
44.8
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Company C
66.2
62.0
84.1
3.2
43.7
38.5
6.1
1.5
$46,335
$37,500
$204,403
$125,000
Company D
24.5
73.1
62.8
32.1
75.0
36.8
2.2
0.5
$86,188
$80,500
$327,571
$200,000
Panel B: Employee Characteristics
Fraction female (%)
Fraction married (%)
Fraction White (%)
Fraction Asian (%)
Fraction with BA degree (%)
Average age (years)
Average tenure (years)
Average number of promotions
Average salary
Median salary
Average wealth
Median wealth
Company A
22.0
82.0
72.3
18.6
83.6
39.4
4.4
1.3
$98,985
$90,000
$724,815
$375,000
40
Company B
44.7
61.3
96.6
1.0
26.1
40.0
5.5
1.1
$35,518
$27,500
$179,064
$87,500
Table IV. ESPP Non-Participation and Employee Losses (Firm Level).
The sample consists of S&P 500, S&P 400 midcap, and NASDAQ 100 …rms with taxquali…ed stock purchase plans that allow employees to sell the stock following the purchase.
Panel A is for the full sample; Panel B is for …rms that allow employees to sign up for the
plan within their …rst month of employment; Panel C is for …rms with average employee
salary of less than $50,000; and Panel D is for …rms with the average employee salary of more
than $100,000. ESPP non-participation is one minus the ESPP contributions per employee
normalized by the minimum of (1) the annual dollar limit, and (2) the maximum percent of
compensation multiplied by the survey salary; ESPP non-participation (Compustat) is one
minus the ESPP contributions per employee normalized by the minimum of (1) the annual
dollar limit, and (2) the maximum percent of compensation multiplied by the Compustat
salary; loss per year is the combined expected discount provided by the ESPP multiplied by
the annual dollar limit on contributions and divided by one minus the combined expected
discount; certain loss per year is the discount provided by the ESPP multiplied by the annual
dollar limit on contributions and divided by one minus the discount; loss per year as % of
salary is the loss per non-participant divided by the survey salary; after-tax loss per year is
the loss per year minus the expected tax liability, which is calculated as the 28% individual
tax rate on pro…t from ESPP trade.
Panel A: Full Sample
ESPP non-participation (%)
ESPP non-participation (Comp.) (%)
Loss per year ($)
Certain loss per year ($)
Loss per year as % of annual salary
After-tax loss per year ($)
Obs.
1,821
1,807
1,690
1,690
1,690
1,690
Mean
80.41
70.35
4,627
1,642
5.70
3,331
SD
20.12
29.86
4,424
869
4.47
3,185
10%
51.34
17.17
1,095
605
1.76
788
50%
88.34
82.26
3,354
1,514
4.62
2,415
90%
98.27
97.91
9,557
2,707
10.95
6,881
Panel B: Firms with no restriction on minimum employment
ESPP non-participation (%)
ESPP non-participation (Comp.) (%)
Loss per year ($)
Certain loss per year ($)
Loss per year as % of annual salary
After-tax loss per year ($)
Obs.
1,163
1,139
1,078
1,078
1,078
1,078
Mean
75.42
62.87
5,186
1,731
6.10
3,474
SD
21.42
31.36
4,771
809
4.72
3,196
10%
45.49
7.16
1,400
847
1.76
938
50%
81.34
72.47
3,751
1,648
4.90
2,513
90%
97.37
96.39
10,501
2,679
11.60
7,036
Panel C: Firms sorted by average survey salary
Mean ESPP non-participation (%)
Mean ESPP non-participation
(Compustat) (%)
salary<$50K
93.09
92.31
41
salary>$100K
62.87
39.91
t-test
18.14
19.90
Table V. Employee Participation in Stock Plans and Open Market at
Four Firms (Individual Level).
Employees never enrolled in ESPP is the fraction of eligible employees in each …rm who
report to never have participated in ESPP; average loss per year is the average across all
non-participants in each …rm of the combined expected discount provided by the ESPP multiplied by the annual dollar limit on contributions and divided by one minus the combined
expected discount; average certain loss per year is the average across non-participants of
the discount provided by the ESPP multiplied by the annual dollar limit on contributions
and divided by one minus the discount; average loss per year as % of salary is the loss per
non-participant divided by the survey salary; average after-tax loss per year is the average
of the loss minus the expected tax liability, which is calculated as the marginal tax rate in
2004/2005 individual tax rate on pro…t from ESPP trade.
Employees never enrolled in ESPP (%)
Number of employees with losses
Average loss per year ($)
Average certain loss per year ($)
Loss over the employee tenure ($)
Average after-tax loss per year ($)
Average loss per year as % of salary ($)
Employees ever enrolled in ESPP
who never sold stock (%)
Employees received ESOs last year (%)
Average intrinsic value of owned ESOs ($)
Employees participate in 401(k) (%)
Employees ever bought company stock on the
open market (%)
Employees frequently buy and sell securities
on the market (%)
Employees who exchange 10 underwater ESOs
for 10 or more shares of stock (%)
Employees who exchange 10 underwater ESOs
for 0 shares of stock (%)
Correlation coe¢ cient between being ever enrolled in ESPP and employee salary
42
Firm A
7.7
491
4,589
1,856
18,714
2,983
5.1
39.5
Firm B
47.7
184
1,149
543
6,204
747
4.0
44.8
Firm C
52.2
711
868
868
4,181
599
2.9
58.5
Firm D
59.1
67
3,855
1,510
8,181
2,506
5.4
38.0
97.8
320,174
94.5
36.3
23.5
7,133
82.4
6.6
75.4
36,617
84.5
5.4
61.1
141,177
85.3
2.1
18.3
7.6
7.9
17.7
15.0
N/A
N/A
N/A
5.1
N/A
N/A
N/A
0.035
(0.005)
0.294
(0.000)
0.270
(0.000)
0.345
(0.000)
Table VI. Employee Non-Participation in ESPP by Salary, Household Wealth,
and Education (Individual Level).
The table displays the percentage of employees at each in four …rms who never enrolled in
ESPP over their tenure at the …rm. Salary is employee’s annual base pay (excluding any
bonuses and commissions). T-test is for the di¤erence in means between the two subsamples.
Panel A: Non-participation by employee annual salary
Firm
Firm
Firm
Firm
A
B
C
D
Salary under $50K
23.6
50.6
59.5
83.3
Salary over $100K
6.7
17.4
29.4
28.1
t-test
10.29
3.12
6.80
4.80
Panel B: Non-participation by household wealth
Firm
Firm
Firm
Firm
A
B
C
D
Wealth under $100K
12.7
54.7
65.0
67.3
Wealth over $200K
6.0
32.6
32.7
46.2
t-test
7.33
3.54
9.06
2.28
Panel C: Non-participation by college education
Firm
Firm
Firm
Firm
A
B
C
D
No BA degree
10.9
54.2
59.2
84.2
BA degree
7.2
31.2
41.8
50.4
43
t-test
4.08
4.19
5.89
3.82
Table VII. Employee Participation in ESPP (Firm Level).
The dependent variable is the ESPP participation. RD/assets is research and development
expenses normalized by the book value of assets; …rm size is the natural logarithm of the
book value of assets; Tobin’s Q is the sum of market value of equity and book value of debt
normalized by the book value of assets; other variables are de…ned in the header of Table II.
The estimation includes year and industry …xed-e¤ects (Fama-French 17). T –statistics based
on robust standard errors clustered by …rm are listed in parentheses. The numbers below the
t-statistics indicate by how much the participation rate increases for a one standard deviation
increase in the independent variable. Signi…cance at the 1%, 5%, and 10% level is denoted
by ***, **, and *, respectively.
Survey salary
Years since plan
adoption
(1)
0.15
(2.32)
3.19%
0.48
(3.40)
2.92%
Option grants/employee
(2)
0.08
(1.48)
1.70%
0.57
(3.84)
3.47%
0.05
(2.55)
2.94%
Value from option
exercises/employee
(3)
0.12
(1.67)
2.56%
0.55
(3.70)
3.35%
(4)
0.15
(2.36)
3.19%
0.49
(3.52)
2.98%
(5)
0.14
(2.53)
2.98%
0.41
(2.15)
2.50%
0.06
(2.83)
2.89%
100 best company
9.30
(3.40)
2.51%
CEO approval
by employees
0.12
(2.11)
2.68%
Contemporaneous stock
return
Past stock return
Discount
Lookback and reset value
RD/assets
Firm size
Tobin’s Q
Adjusted-R2
Observations
(6)
0.11
(2.08)
2.34%
0.47
(3.17)
2.86%
108.8
(5.09)
24.58
(2.63)
103.2
(6.29)
2.13
(2.95)
3.27
(7.30)
0.460
1,688
93.92
(4.44)
27.29
(3.27)
103.6
(5.75)
1.58
(2.37)
2.32
(5.22)
0.480
1,260
44
104.7
(4.17)
21.82
(2.29)
105.3
(5.85)
1.92
(2.58)
2.13
(4.02)
0.462
1,291
105.3
(4.83)
25.43
(2.71)
95.21
(5.81)
1.56
(2.26)
2.98
(6.99)
0.472
1,686
52.97
(2.12)
15.19
(1.62)
116.1
(5.52)
1.64
(1.98)
2.90
(5.67)
0.562
868
0.38
(0.40)
0.05
(0.03)
99.30
(5.17)
31.08
(3.65)
104.4
(5.62)
2.02
(2.81)
3.40
(5.47)
0.562
1,454
Table VIII. ESPP Participation and Employee Characteristics (Individual Level).
The dependent variable is equal to 1 if employee ever participated in ESPP, and is equal
to 0 otherwise. The model is estimated by probit with …rm-…xed e¤ects and country-…xed
e¤ects. Ever received stock options is equal to 1 if employee ever been granted employee stock
options, and is equal to 0 otherwise. Trade other securities is equal to 1 if employee reports
that he/she frequently buys and sells securities in the market, and is equal to 0 otherwise.
Irrational ESO overvaluation is equal to 1 if employee would not exchange 10 underwater
stock options for anything less than 10 shares. Irrational ESO undervaluation is equal to
1 if employee would exchange 10 underwater stock options for 0 shares of stock. Positive
stock return experience is past stock returns of the …rm weighted over the employee tenure,
with more recent years carrying larger weights. Employee loyalty is the sum of how loyal employee feels to the …rm on a scale of 1 to 4 and whether employee feels she shares a common
purpose with …rm on a scale of 1 to 4. Voted in national election is equal to 1 if employee
voted in the most recent country election, and is 0 otherwise. Employee trust is the sum of
whether employee feels company is fair to its employees on a scale of 1 to 4 and whether the
company keeps its promises. Each cell shows the coe¢ cient, t-statistic (in parentheses), and
the economic magnitude, measured as the average marginal e¤ect multiplied by the standard
deviation of the independent variable (if continuous) or by 1 (if a dummy). Signi…cance at
the 1%, 5%, and 10% level is denoted by ***, **, and *, respectively.
45
(1)
-0.01
(-0.30) -0.2%
0.06
(1.33) 1.2%
0.28
(4.01) 5.1%
0.25
(5.14) 4.5%
0.07
(3.91) 10.1%
-0.001
(-3.89) -1.0%
0.13
(4.89) 1.9%
0.17
(8.81) 3.4%
(2)
-0.01
(-0.22) -0.2%
0.05
(1.02) 0.9%
0.27
(3.73) 4.7%
0.26
(5.21) 4.5%
0.06
(3.50) 9.1%
-0.001
(-3.57) -0.9%
0.15
(4.12) 2.1%
0.14
(7.09) 2.8%
(3)
0.09
(1.22) 1.1%
-0.01
(-0.19) -0.2%
0.25
(2.84) 3.0%
0.12
(1.56) 1.4%
0.03
(0.86) 2.6%
-0.00
(-1.04) -0.3%
0.09
(2.46) 0.8%
0.18
(6.27) 2.4%
(4)
-0.02
(-0.51) -0.4%
0.05
(0.98) 0.9%
0.34
(4.48) 6.1%
0.24
(4.69) 4.2%
0.06
(3.25) 8.4%
-0.001
(-3.36) -0.9%
0.13
(4.67) 1.8%
0.17
(8.54) 3.3%
(5)
-0.01
(-0.29) -0.2%
0.07
(1.36) 1.2%
0.28
(3.95) 5.1%
0.57
(2.81) 10.4%
0.06
(3.75) 9.8%
-0.001
(-3.75) -1.0%
0.14
(5.07) 2.0%
0.17
(8.73) 3.4%
0.27
(5.44) 2.9%
0.08
(1.06) 0.6%
0.27
(5.13) 2.7%
0.28
(5.46) 2.9%
Ever received
stock options
0.21
(4.09) 2.2%
0.82
(6.51) 14.5%
Trade other
securities
0.30
(4.68) 5.4%
Positive stock
return experience
0.28
(2.32) 1.2%
Employee loyalty
0.02
(1.57) 0.6%
Female
Married
Asian
BA degree
Age
Age2
Log(tenure)
Number of
promotions
Log(salary)
Irrational ESO
overvaluation
-0.27
(-3.60) 3.5%
Irrational ESO
undervaluation
-0.52
(-4.95) 6.9%
Voted in national
elections
0.16
(3.29) 2.9%
Employee trust
0.03
(1.89) 1.1%
-0.04
(-1.75) -1.4%
Employee trust*
BA degree
Pseudo-R2
Firms included
Observations
0.268
All
7,537
0.280
All
7,453
0.094
A
5,206
46
0.265
A, B, C
7,234
0.268
All
7,453
Table IX. Not Selling Stock Purchased through ESPP and Employee
Characteristics (Individual Level).
The dependent variable is equal to 1 if employee participated in ESPP but never sold
the stock, and is equal to 0 if employee participated in ESPP and sold the stock sometime
in the past. The model is estimated by the two-stage Heckman selection model with …rmand country-…xed e¤ects. The selection equation (whether employee chooses to participate)
uses the same dependent variables as the outcome equation (whether employee holds the
stock). Ever received stock options is equal to 1 if employee ever been granted employee
stock options, and is equal to 0 otherwise. Trade other securities is equal to 1 if employee
reports that he/she frequently buys and sells securities in the market, and is equal to 0
otherwise. Irrational ESO overvaluation is equal to 1 if employee would not exchange 10
underwater stock options for anything less than 10 shares. Irrational ESO undervaluation
is equal to 1 if employee would exchange 10 underwater stock options for 0 shares of stock.
Employee loyalty is the sum of how loyal employee feels to the …rm on a scale of 1 to 4
and whether employee feels she shares a common purpose with …rm on a scale of 1 to 4.
Employee trust is the sum of whether employee feels company is fair to its employees on a
scale of 1 to 4 and whether the company keeps its promises. Each cell shows the coe¢ cient,
t-statistic (in parentheses), and the economic magnitude, measured as the average marginal
e¤ect multiplied by the standard deviation of the independent variable (if continuous) or by
1 (if a dummy). Signi…cance at the 1%, 5%, and 10% level is denoted by ***, **, and *,
respectively.
47
Female
Married
Asian
BA degree
Age
Age2
Log(tenure)
Number of promotions
Log(salary)
(1)
0.30
(7.55) 10.9%
-0.08
(-1.82) -2.8%
-0.04
(-0.82) -1.4%
0.20
(4.33) 7.2%
-0.02
(-1.15) -6.1%
0.00
(1.44) 0.8%
-0.01
(-0.56) -0.4%
-0.15
(-7.73) -6.0%
-0.18
(-3.51) 3.8%
Ever received stock
options
Trade other securities
(2)
0.27
(7.03) 9.9%
-0.08
(-1.85) -2.8%
-0.01
(-0.25) -0.4%
0.22
(4.97) 7.9%
-0.01
(-0.84) -4.2%
0.00
(1.17) 0.6%
-0.01
(-0.63) -0.4%
-0.14
(-8.23) -5.6%
-0.16
(-3.38) -3.4%
-0.11
(-0.38) -3.8%
(3)
0.24
(5.05) 8.8%
-0.10
(-2.14) -3.9%
-0.11
(-2.16) -4.0%
0.13
(2.58) 4.9%
-0.03
(-1.49) -9.1%
0.00
(1.98) 1.2%
-0.01
(-0.37) -0.2%
-0.17
(-9.64) -6.9%
-0.22
(-3.85) -4.7%
-0.14
(-3.29) -5.2%
0.03
(2.48) 1.5%
Employee loyalty
Irrational ESO
overvaluation
0.16
(3.09) 5.9%
Irrational ESO
undervaluation
0.39
(4.74) 14.4%
Employee trust
Inverse Mills ratio
Firms included
Observations
(4)
0.30
(7.67) 11.1%
-0.08
(-1.82) -2.8%
-0.02
(-0.34) -0.6%
0.20
(4.35) 7.2%
-0.02
(-0.95) -5.0%
0.00
(1.23) 0.7%
-0.01
(-0.60) -0.4%
-0.15
(-7.91) -6.1%
-0.18
(-3.57) -3.9%
0.54
(2.15)
All
7,501
0.91
(3.30)
All
7,418
48
-4.67
(-0.02)
A
5,192
0.02
(2.62) 1.7%
0.61
(2.37)
All
7,418