ArticlePDF Available

Employee stock ownership and firm exit decisions: A cross-country analysis of rank-and-file employees

Authors:

Abstract and Figures

While multinational firms invest large amounts of money in employee stock ownership plans (ESOPs) to reduce turnover, there is little evidence regarding ESOPs' effectiveness in retaining rank-and-file employees and none on a global scale. Building on psychological ownership (PO) arguments, we predict that a rank-and-file employee's ESOP participation will be negatively associated with a firm exit decision and that this effect will be stronger in contextual settings that are more conducive to turnover. For our analysis, we used internal data from a large multinational firm covering 190,453 rank-and-file employees and approximately 650,000 employee years. We find that ESOP participation is associated with a lower likelihood of individual firm exit decisions. We also find this effect to be more pronounced in countries with favorable labor market conditions and lower uncertainty avoidance (UA). Additional tests support our argument that PO arising from ESOP participation is particularly important for rank-and-file employees, who often only invest small amounts. Overall, our study provides cross-country evidence regarding the retention effect of ESOP participation for rank-and-file employees.
Content may be subject to copyright.
Employee stock ownership and rm exit decisions: A cross-country
analysis of rank-and-le employees
Jan C. Hennig
a
,
*
, Carolin Ahrens
b
, Jana Oehmichen
c
, Michael Wolff
b
a
Department of Accounting, University of Groningen, Nettelbosje 2, 9747 AE, Groningen, Netherlands
b
Chair of Management and Control, University of Goettingen, Platz der Goettinger Sieben 3, 37073, Goettingen, Germany
c
Chair of Organization, Human Resources, and Management, University of Mainz, Jakob-Welder-Weg 9, 55128, Mainz, Germany
article info
Article history:
Received 8 August 2018
Received in revised form
7 March 2022
Accepted 12 June 2022
Available online xxx
Keywords:
Employee stock ownership plan (ESOP)
Rank-and-le employees
Firm exit decision
Psychological ownership
Global scale
abstract
While multinational rms invest large amounts of money in employee stock ownership plans (ESOPs) to
reduce turnover, there is little evidence regarding ESOPs' effectiveness in retaining rank-and-le em-
ployees and none on a global scale. Building on psychological ownership (PO) arguments, we predict that
a rank-and-le employee's ESOP participation will be negatively associated with a rm exit decision and
that this effect will be stronger in contextual settings that are more conducive to turnover. For our
analysis, we used internal data from a large multinational rm covering 190,453 rank-and-le employees
and approximately 650,000 employee years. We nd that ESOP participation is associated with a lower
likelihood of individual rm exit decisions. We also nd this effect to be more pronounced in countries
with favorable labor market conditions and lower uncertainty avoidance (UA). Additional tests support
our argument that PO arising from ESOP participation is particularly important for rank-and-le em-
ployees, who often only invest small amounts. Overall, our study provides cross-country evidence
regarding the retention effect of ESOP participation for rank-and-le employees.
©2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
1. Introduction
As the estimated turnover cost averages 150% to 175% of a
leaving employee's annual pay (Hansen, 1997), rms are intensely
interested in retaining employees. To provide incentives for em-
ployees to stay with their current employer, rms are increasingly
turning to employee stock ownership plans (ESOPs)
1
(GEO, 2019).
While prior empirical studies rely on economic arguments to
explain the retention effect arising from stock ownership mostly for
higher-level managers (e.g., Aldatmaz, Ouimet, &Van Wesep, 2018;
Balsam &Miharjo, 2007;Oyer &Schaefer, 2005), we question
whether there could be other retention arguments for rank-and-le
employees with signicantly smaller ESOP investments. To answer
this question, we use data on individuals' participation in a broad-
based ESOP from a large, multinational rm that allows its rank-
and-le employees to invest only as much as 5% of their income.
2
From a theoretical point of view, stock-based compensation
provides incentives to align the interests of employees with those
of their employers. Thereby, the extant literature has mostly
focused on executives and explained their behavior through eco-
nomic arguments (e.g., Balsam &Miharjo, 2007;Guay, Core, &
Larcker, 2003). Studies based on the entire workforce (i.e.,
including lower hierarchical levels) have adopted these economic
arguments when examining the effect of ESOPs on overall turnover
behavior (Aldatmaz et al., 2018;Balsam, Gifford, &Kim, 2007). We,
however, provide an alternative explanation based on psychologi-
cal theory to further elucidate the behavior of rank-and-le em-
ployees. In contrast to compensation for executives and senior
managers, stock-based compensation for rank-and-le employees
*Corresponding author.
E-mail addresses: j.c.hennig@rug.nl (J.C. Hennig), carolin.ahrens@wiwi.uni-
goettingen.de (C. Ahrens), j.d.r.oehmichen@rug.nl (J. Oehmichen), michael.wolff@
uni-goettingen.de (M. Wolff).
1
Generally, ESOPs offer stock ownership to employees at privileged terms (e.g.,
Poutsma &de Nijs, 2003). However, ESOPs' characteristics vary substantially be-
tween countries and rms (e.g., Jones &Kato, 1995). The focal rm's ESOP is offered
to employees on all hierarchical levels (exclusions apply, for example, to temporary
employees) and thus qualies as broad-based (Kim &Ouimet, 2014). For a detailed
description of the ESOP, see Section 3.
2
However, on average, rank-and-le employees in the rm invest a lot less than
5% of their income. Based on data for 2018, the average rank-and-le employee
invested only 2.25%. Given the median annual income of V52,083, 5% translates into
V2,604, and 2.25% into V1,172 .
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
https://doi.org/10.1016/j.aos.2022.101390
0361-3682/©2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Accounting, Organizations and Society xxx (xxxx) xxx
Please cite this article as: J.C. Hennig, C. Ahrens, J. Oehmichen et al., Employee stock ownership and rm exit decisions: A cross-country analysis
of rank-and-le employees, Accounting, Organizations and Society, https://doi.org/10.1016/j.aos.2022.101390
makes up signicantly smaller amounts and fractions of their salary
and is far less frequent (Edmans, Gabaix, &Jenter, 2017;GEO, 2019).
Above and beyond the economic arguments, we argue that the
desirable effects of ESOPs for rank-and-le employees are likely to
unfold through the concept of psychological ownership (PO)
(Dawkins, Tian, Newman, &Martin, 2017;Pierce, Rubenfeld, &
Morgan, 1991).
This paper builds on the long-standing literature on PO span-
ning organizational behavior, psychology, and management
research (e.g., Pierce, Kostova, &Dirks, 2003,2001;Chi &Han,
2008;Dawkins et al., 2017;Pierce et al., 1991). PO describes a
state in which individuals feel as though the target of ownership or
a piece of that is theirs(i.e., It is mine!)(Pierce et al., 2003, p. 86).
In contrast to legal ownership, [PO] consists of an emotional
attachment to the targetthat does not depend on the economic
evaluation of the target (Dawkins et al., 2017, p. 164). However, as
the state of PO is reached through feelings of investing in the target,
having power and control over the target, and knowing the target,
the existence of legal ownership (i.e., ESOP participation) can
facilitate and accelerate the emergence of PO (Pierce et al., 2003;
Pierce, O'Driscoll, &Coghlan, 2004). In contrast to rank-and-le
employees, higher-level managers experience these feelings more
often in their daily business and already possess a stronger feeling
of PO even without an ESOP. Hence, ESOP participation should be
particularly effective in building PO and, consequently, retention
among rank-and-le employees, while for senior managers, the
nancial aspects of ESOPs could be more important.
3
Thus, we
expect rank-and-le employees' ESOP participation to be associ-
ated with a strong attachment to the rm through PO, thus
reducing the likelihood of a rm exit.
We further investigate specic contextual settings, to better
understand how increased PO from ESOP participation relates to
retention effects. Thus, we draw on the concept of employees
perceived ease of movement, which has often been used in the
prediction of voluntary turnover (e.g., Gerhart, 1990;March &
Simon, 1958;Trevor, 2001). We argue that ESOP participation is
less effective in retaining employees when the intention to leave
the rm is weak (i.e., low perceived ease of movement). In contrast,
when the perceived ease of movement is high, ESOP participation
should be more effective in retaining employees. Following prior
literature about the characterization of the perceived ease of
movement as general job availability and individual attributes, we
consider measures of (1) the unemployment rate and (2) uncer-
tainty avoidance (UA) (Allen, Weeks, &Moftt, 2005;Trevor, 2001):
First, when labor market conditions are good and employees are
faced with many outside opportunities (i.e., perceived ease of
movement is high), we expect the retention effect of ESOP partic-
ipation through PO to be more pronounced. Second, as job changes
entail uncertainty, we expect the retention effect of ESOP partici-
pation via PO to be stronger for employees with lower UA (i.e.,
perceived ease of movement is high).
To test our predictions, we use internal rm data from a large
multinational industrial rm that is headquartered in Europe and
has subsidiaries worldwide. The rm introduced a broad-based
ESOP on a global scale in the late 2000s, and, since then, it has
offered its employees the possibility of purchasing shares each year.
The data contain detailed information about all eligible employees,
their demographics, and their ESOP participation behavior. The
research design benets from rm-level effects being held
constant, which enables us to better isolate the effects of our
country-level moderators. On the country level, we combine the
data with values for the unemployment rate from the World Bank
and for UA from Hofstede (1984,1994). We consider these country-
level measures as they are closely tied to the theoretical concept of
employeesperceived ease of movement. The nal sample includes
654,706 employee observations for 190,453 unique employees in
39 countries from 2011 to 2015. In our empirical analysis, we
address endogeneity concerns by using both ordinary least squares
(OLS) and two-stage least squares (2SLS) regressions ein the main
model as well as in all additional analyses. The results are consis-
tent with our predictions. In particular, we nd a negative effect of
ESOP participation on rm exit decisions of rank-and-le em-
ployees, which is stronger in countries with favorable labor market
conditions and lower levels of UA. While the Hausman test suggests
the existence of endogeneity, the calculation of the impact
threshold for a confounding variable (ITCV) indicates that it is
rather unlikely that an omitted variable has confounded our results.
We perform a number of sensitivity tests, in which we further
address endogeneity concerns. First, we account for differences in
observable characteristics and conduct a propensity score matching
(PSM) analysis. Second, we exclude employees with specic
knowledge about the ESOP. Third, in an extension of the 2SLS re-
gressions, we control for potentially confounding country-level
effects in the rst- and second-stage regressions. These tests
consistently support our main results.
In additional tests, we substantiate our arguments on the rele-
vance of PO arising from ESOP participation for rank-and-le em-
ployees. First, we exploit survey data and nd signicant
correlations between ESOP participation and PO-related responses.
Second, we conduct subsample analyses stratied by groups that
vary in experiencing PO in their daily business and nd the
retention effect arising from ESOP participation to be more pro-
nounced for employees with less pre-existing PO (e.g., new em-
ployees). Third, we investigate the retention effect of very small
ESOP investments that are hardly likely to be an economic moti-
vator, but that contribute to PO, and nd negative and signicant
relations with rm exit decisions. Fourth, while economic moti-
vations expire with the vesting period, we nd that the ESOP's
retention effect for rank-and-le employees holds beyond the
vesting period, again suggesting PO to be the underlying
mechanism.
Our study contributes to the literature in several ways. First, we
extend prior research investigating the effect of ESOP participation
on employee behavior (Aldatmaz et al., 2018;Balsam et al., 2007;
Pierce et al., 1991) as we provide empirical evidence for ESOPs'
retention effect for individual rank-and-le employees, a concern at
the heart of any rm's management control system. In contrast to
prior studies that either used rm-level data (Aldatmaz et al., 2018),
investigated broader groups of employees, including those on
higher levels (Aldatmaz et al., 2018;Balsam et al., 2007), and that
deducted propositions from conceptual models (Pierce et al., 1991),
we observe the effect of individuals' ESOP participation on the rm
exit decisions of rank-and-le employees while controlling for their
demographics and job characteristics. In this vein, the analysis on
the individual level improves our understanding of actual behavior
as recently called for by Hanlon, Yeung, and Zuo (2021), especially
regarding investment (e.g., Babenko &Sen, 2014) and rm exit
decisions (e.g., Trevor, 2001).
Second, our study ts within a growing body of accounting
research that investigates the incentive effects arising from ESOP
participation among rank-and-le employees (Call, Kedia, &
Rajgopal, 2016;Holderness, Huffman, &Lewis-Western, 2019).
We add to this stream of research by proposing a retention effect
based on behavioral arguments using the theory of PO. PO seems
3
Similarly, studies have shown that relationships between legal ownership and
organizational identication as well as organizational identication and turnover
are particularly pronounced for lower-level employees (Cole &Bruch, 2006;French
&Rosenstein, 1984).
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
2
particularly applicable for rank-and-le employees who only have
minor levels of legal ownership in their rms, thus giving greater
relevance to their psychological attachment to organizations.
Moreover, rank-and-le employees typically invest themselves to a
smaller extent in their rms, exert only little power and control,
and have less information about the rm in contrast to higher-level
managers. This gives rank-and-le employees less experiences in
developing a sense of PO in their daily business routine. Therefore,
minor levels of legal ownership resulting from ESOP participation
can already be effective in building PO among rank-and-le em-
ployees. Thereby, we also extend the literature on PO that previ-
ously primarily focused on rm-specic human capital
(knowledge), employee empowerment (delegation of re-
sponsibility), and employeesinvestment of the self into the job
(e.g., in form of time, energy, creativity, and judgement) (Pierce,
Kostova, &Dirks, 2001;Zhang, Liu, Zhang, Xu, &Cheung, 2021).
Third, we add to the literature on ESOPs by examining their
retention effect across different countries. Thereby, we extend prior
empirical studies in the eld that have exclusively focused on US
rms (Aldatmaz et al., 2018;Balsam &Miharjo, 2007;Kim &
Ouimet, 2014;Oyer &Schaefer, 2005). Based on a cross-country
setting, the results reveal the differing effectiveness of ESOP
participation at retaining employees under varying labor market
conditions. In this vein, we also shed light on the substitutive na-
ture of social control mechanisms, such as cultural norms and PO, in
mitigating employeesturnover intentions. At the same time, the
results demonstrate that the retention effect of ESOP participation
holds on a global scale.
2. Literature review and hypotheses development
2.1. Literature on stock-based incentives and rm exit
In the last few decades, the economic literature has increasingly
studied stock-based compensation as an instrument to resolve
agency problems (Core &Guay, 1999;Haugen &Senbet, 1981;
Prendergast, 1999). The literature has focused on the great impor-
tance of incentivizing executives and senior managers as corporate
decision makers who signicantly affect shareholders' wealth (e.g.,
Dechow &Sloan, 1991). A much studied element of compensation
schemes has been the vesting period that aims to align inter-
estsdand specically time horizonsdby providing incentives to
stay with the rm (Balsam &Miharjo, 2007;Cadman, Rusticus, &
Sunder, 2013;Guay et al., 2003). In order to provide economic in-
centives and inuence behavior (i.e., to create a retention effect),
stock-based compensation linked to the vesting period has reached
substantial amounts and frequently makes up more than half of the
executivescompensation (Edmans et al., 2017;Murphy, 1999).
Given that rms are increasingly offering stock to employees on the
lower hierarchical levels (Core &Guay, 2001), the question arises as
to whether signicantly smaller proportions of stockdand smaller
economic benetsdare able to retain rank-and-le employees.
4
A common challenge for empirical studies investigating lower-
level employees is the availability of adequate data. Thus, as
opposed to studies on stock ownership and turnover at higher
levels, there is little research on the lower levels. A recent empirical
study by Aldatmaz et al. (2018) addresses the data challenge by
approximating the average value of stock option grants for all
employees minus the top ve executives with ExecuComp data,
which they combine with data on total employee turnover counts
per establishment. In their analysis, Aldatmaz et al. (2018) nd
indications of a negative relationship between large ESOPs and
total rm exits during the vesting period that was stronger in up
markets. In their case study of a Fortune 100 rm, Balsam et al.
(2007) examine three broad-based ESOP offerings during the
1990s and how they affected the turnover behavior of a broad
group of employees. Balsam et al. (2007) nd signicant differences
in turnover rates in the six months prior to and after the option
vesting date, while also considering limited individual data.
Our study differs from these studies in fundamental ways. First,
our study analyzes the effect of individuals' ESOP participation on
the individual rm exit decisions of rank-and-le employees, which
is signicantly different from the effect of rm-wide ESOP offerings.
Paired with information on demographics and job characteristics,
this allows us to develop a better understanding of how ESOPs
work by observing effects on an individual level by abstaining from
observations and measurements based on average characteristics
of broader employee groups.
5
For example, we are able to address
concerns that the retention effect of ESOP participation is limited to
certain subgroups with specic characteristics such as those with
ESOP-relevant knowledge (e.g., employees in the Finance and Hu-
man Resources (HR) Departments). Second, the individual-level
data enables us to exclude executives and senior managers in or-
der to precisely narrow our focus down to rank-and-le employees.
In this way, we are able to analyze the behavior of the largest
employee group within rms without confounding effects from
other employee groups. Third, given the exclusive focus on em-
ployees with lower incomes, we propose a psychological motiva-
tion besides an economic motivation as the mechanism behind the
retentive ESOP effect. The focal rm's ESOP supports this mecha-
nism, as it involves only small investments and physical share
ownership, turning participating employees into actual share-
holders. As shareholders, ESOP participants also receive voting
rights, giving them increased levels of perceived power and control
over the rm. This experience is a crucial trigger for the sense of PO.
Fourth, the global scale of our study extends the prevailing focus
from US establishments in previous studies to a cross-country
setting. At the same time, our study benets from potentially
confounding rm effects being held constant, thus resulting in a
thorough cross-country analysis.
2.2. The relevance of psychological ownership for rank-and-le
employeesrm exit decisions
Rank-and-le employees' stock-based compensation accounts
for a signicantly smaller amount as well as proportion of their
compensation in comparison to executives and senior managers
(GEO, 2019). While stock-based compensation commonly makes up
more than 50% of executives' compensation (Edmans et al., 2017),
rank-and-le employees' common annual ESOP investment in the
focal rm is 2.25% of their income, even reaching values as low as
0.01%.
6
This raises the question as to whether there are plausible
alternative explanations for rank-and-le employees beyond the
economic incentives that have been found to affect the rm exit
behaviors of executives as well as broader groups of employees,
4
In a similar vein, Hall and Murphy (2003) question employees' economic in-
centives arising from nancial gains through stock ownership.
5
For rms, it is important to gain knowledge about the ESOP participation
behavior of employee subgroups with specic characteristics, for example, highly
qualied employees or groups who rarely participate. As ESOP participation de-
creases the likelihood of a rm exit, rms could tailor their communication to these
groups to foster their participation and thereby reduce the total number of rm
exits.
6
The rm provides employees with the possibility of acquiring fractional shares
and has not set a minimum threshold for ESOP investments in many countries.
Thus, even employees with very low incomes or those in nancial distress can nd
ways to benet from the ESOP.
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
3
including senior managers. We draw on the concept of PO to offer
an alternative explanation for the potential relationship between
ESOP participation and rank-and-le employeesrm exit
decisions.
PO describes a feeling of an emotional attachment to the target,
where individuals have a state of mind in which they perceive the
target or a piece of it as theirs (Dawkins et al., 2017;Pierce et al.,
2001). In their seminal paper, Pierce et al. (2001) describe how
members of an organization come to feel PO: (1) investing in the
target, (2) controlling the target, and (3) knowing the target. In day-
to-day business, these experiences are more likely to occur for se-
nior managers who can exert power over the rm and hold more
information-intensive positions. In contrast, rank-and-le em-
ployees possess less control over the rm and are frequently
conned to a more specic task that includes less information
sharing and gathering. In addition to that, senior managers and
rank-and-le employees differ substantially in their experiences
regarding their investments in the rm. Frankly, senior managers
are likely to develop a stronger sense of PO based on their daily
business routines, while rank-and-le employees lack these expe-
riences and are less likely to feel PO.
Prior research explicitly points to employee stock ownership as
a driver of rank-and-le employees' PO and argues that stock
ownership leads to psychological ownership, a bonding or inte-
gration of the employee-owner with the organization(Pierce et al.,
1991, p. 121).
7
More precisely, ESOP participation in the focal rm
helps to create all three experiences of how rank-and-le em-
ployees can come to feel PO. First, through ESOP participation, rank-
and-le employees literally invest in the target (i.e., the rm).
Second, as ESOP participation implies the purchase of physical
stock from the employing rm, rank-and-le employees come to
possess controlling shareholder rights (i.e., voting rights in annual
meetings).
8
Third, as shareholders, rank-and-le employees have a
higher demand for business-related information. Empirical studies
indeed nd a positive relation between employee ownership and
PO (Chi &Han, 2008;Chiu, Hui, &Lai, 2007). Chi and Han (2008)
report that employee participation in prot sharing has a positive
effect on PO, while Chiu et al. (2007) document such an effect
arising from employeesESOP participation. While we assume
rank-and-le employees will have less PO-building experiences in
their daily business routines, their feeling of PO arising from ESOP
participation should hence be particularly strong. Similarly, studies
nd the relationship between employee ownership and organiza-
tional identication to be stronger for lower-level employees (Cole
&Bruch, 2006;French &Rosenstein, 1984).
However, as the observed ESOP investments of the rank-and-le
employees are quite small, the question arises as to whether the
amounts are substantial enough to affect behavior. Based on eco-
nomic arguments, we might expect no or a small effect due to the
low nancial gains. Based on psychological arguments, we assume
their sense of PO to depend less on the investment amount
(Dawkins et al., 2017).
9
Derived from our explanations above, we
expect ESOP participation to lead to strong identication with and
attachment to the rm (Pierce et al., 2001). In this sense, ESOP
participation triggers an expression of and contributes to em-
ployees' self-identity (Dawkins et al., 2017), evoking a sense of
belongingness in terms of having a place in which to dwell(Pierce
et al., 2003, p. 88). Where individuals strongly identify themselves
with their rm, they are likely to align their behavior with the rm's
interests (Abernethy, Bouwens, &Kroos, 2017). Similarly, organi-
zational research suggests that PO and organizational identication
are particularly effective in reducing turnover intentions at lower
hierarchical levels (Abrams, Ando, &Hinkle, 1998;Cole &Bruch,
2006). This is in line with the PO literature that suggests a
perceived responsibility and increased accountability of the
employee for the employing rm (Avey, Avolio, Crossley, &Luthans,
2009;Dawkins et al., 2017;Van Dyne &Pierce, 2004). Thus, the PO
consequences emerging from ESOP participation should attach the
rank-and-le employees to the rm and reduce the likelihood of a
rm exit:
10
H1. ESOP participation is negatively related to rank-and-le em-
ployees' individual decision to leave the rm.
2.3. The inuence of country-level factors on employeesperceived
ease of movement
Rank-and-le employeesevaluation of the ESOP investment
and their attachment to the rm is likely to vary depending on
contextual settings, as is the relationship between ESOP partici-
pation and the rm exit decision. To dive deeper into the retention
effect of ESOPs, we draw on the concept of perceived ease of
movement, which is frequently applied to predict voluntary turn-
over (Gerhart, 1990;Trevor, 2001;Wu &Chi, 2020). Accordingly, we
assume that a low perceived ease of movement is associated with a
reduced likelihood of rm exit. Thus, in these situations, employees
are generally less likely to leave the rm, and the need for ESOP
participation to retain employees is low. In contrast, when the
perceived ease of movement is high, employees are more likely to
leave the rm, and the retaining effect of ESOP participation should
be more effective. We follow prior literature that links the
perceived ease of movement to general job availability as well as
individual attributes, and test the moderating effects of (1) labor
market conditions and (2) UA (Allen et al., 2005;Trevor, 2001).
First, we know that employees' perceived ease of movement and
7
The literature on PO provides many similar statements that highlight the
importance of stock ownership to facilitate and speed up the emergence of psy-
chological ownership(Pierce et al., 2003, p. 96). For the focal rm, statements from
a survey among the ESOP participants in 2018 conrm an emerging feeling of PO:
[The ESOP] means a lot for me and my family. My wife, my sons and I, want to let
you know that we feel [name of rm] as ours and will do our best to get our
company to the topand I would like to keep [the stock] for a very, very long time,
to be part of my life.
8
There are even associations of shareholders that consist of employee share-
holders from the rm. These associations of shareholders show an active involve-
ment as representatives of the employees at the rm's annual general meetings
(AGMs) by not only attending, but especially in the form of proposals, counter-
motions, and similar actions. These associations of employee shareholders reect
several characteristics of PO. For example, both associations promote a sense of
togetherness. One is being named We for (name of the rm),while the other
describes the ownership in the rm as an identity-creating measure for employees.
Further, the latter association led a countermotion to not raise the dividend as
planned in one recent year, which also speaks against the notion that the employee
shareholders are purely nancially motivated.
9
In additional tests in Section 6, we provide evidence that even very small ESOP
investments reduce the likelihood of a rm exit, thus suggesting that PO instead of
economic reasons is behind the retention effect. Further, we provide evidence that
for rank-and-le employees, the retention effect of ESOP participation is at least
equally important as for senior managers, despite the enormous differences in the
annual ESOP investment amounts (on average: V1,172 for rank-and-le employees
and V16,422 for senior managers). Thus, we are condent in our assumption that,
for senior managers, the economic aspects of ESOP participation are likely to be
more pronounced in driving their rm exit decisions, while for rank-and-le em-
ployees, the effect arising from PO is likely to be more pronounced in their rm exit
decisions.
10
First, leaving the rm implies forgoing the sense of PO and the associated
benets (e.g., part of one's self-identity, sense of belongingness, and having a place).
Second, rank-and-le employees' perceived responsibility and accountability
should lead to stronger consideration of the consequences of a rm exit (e.g.,
feelings of letting their colleagues and employing rm down). Given the rareness of
employee ownership and it being a strong trigger for PO, we expect these aspects to
represent decisive benets and costs for employees in their rm exit decisions.
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
4
thus their rm exit decisions are contingent on their appraisal of
potential labor market opportunities (Gerhart, 1990;Oyer, 2004;
Trevor, 2001). Negative labor market conditions coincide with less
outside employment opportunities and make quitting less likely, as
it might jeopardize nancial security (Allen, Renn, Moftt, &
Vardaman, 2007). In contrast, favorable labor market conditions
are associated with more outside opportunities, thus increasing the
ease of movement (Munasinghe, 2006). While the likelihood of a
rm exit increases, the potential loss of the PO benets counteracts
this exit likelihood (i.e., a costebenet trade-off). It is in this context
that we expect the psychological benets in the form of a sense of
belongingness in the rm, identication with the employer, or the
feeling of having a home in the current rm to be more effective in
retaining employees. Thus, employeesconsiderations of the psy-
chological benets emerging from ESOP participation should gain
in importance when labor market conditions are favorable and
have a greater impact on the decision as to whether to leave the
rm. Based on the arguments above, we propose:
H2a. The retention effect of ESOP participation is greater when labor
market conditions are favorable.
Second, we argue that the relationship between ESOP partici-
pation and employees' rm exit decisions is contingent on their UA.
UA describes the extent to which people feel threatened by un-
certainty and attempt to avoid ambiguous situations (Hofstede,
1980).
11
Uncertainty-averse individuals prefer predictable events,
while individuals ranking low in UA are more likely to accept and
tolerate uncertainty (Hofstede, 2001). The concept of UA is closely
linked to employees' perceived ease of movement, as leaving the
rm entails unpredictable outcomes arising from the job search
and a new environment. Hence, employees ranking low in UA are
more likely to consider a rm exit, as they perceive the unpre-
dictability of outside opportunities as less detrimental (e.g., higher
ease of movement). It is in these situations of higher ease of
movement that we expect the psychological benets from PO to
weigh stronger in counteracting employeesturnover intentions.
On the other hand, for employees ranking high in UA, who have
fewer intentions to leave, the retention effect of PO is less impor-
tant. Consequently, we assume ESOP participation to unfold a
higher effectiveness in retaining employees through PO for em-
ployees ranking low in UA. Accordingly, the effect of ESOP partici-
pation on rm exit decisions should be stronger for employees
ranking low in UA. This leads to the following hypothesis:
H2b. The retention effect of ESOP participation is greater when un-
certainty avoidance (UA) is low.
3. Research design
3.1. Data and sample
To estimate the effect of ESOP participation on the individual
decisions of rank-and-le employees to leave their rms in a cross-
country setting, we focus our analysis on a large rm based in
Europe with subsidiaries worldwide. The rm operates in a variety
of industries, such as healthcare, mobility management, and energy
solutions (i.e., the rm has huge internal heterogeneity; see
Appendix B for a detailed explanation of the rm's structure and
how we accounted for it in our regressions). We were able to get
access to internal data from 2011 to 2015 on individual employees'
ESOP participation, turnover, and demographic information.
12
The rm offers a broad-based ESOP on a yearly basis, where
employees can decide to invest up to 5% of their income. To enable
employees with very low incomes to access the ESOP, the program
offers them the possibility of acquiring fractional shares, and thus
of investing very small amounts. Moreover, depending on the
country, the focal rm has either set very low minimum thresholds
for ESOP investment amounts (e.g., V10 monthly) or none at all. The
investment amount is deducted from the employee's salary each
month, and physical shares are then added to the employee's se-
curities account. Thus, participating rank-and-le employees get to
exercise control (e.g., participation in annual meetings, voting
rights).
The ESOP represents an economically attractive investment, as
participants receive dividends during the holding period and one
matching share for every three purchased shares after a three-year
vesting period (or a fractional share at a ratio of 1:3 if employees
possess less than three shares). If employees leave the rm before
the shares vest, they keep the purchased shares, benet from div-
idends as well as a potential positive share price development, and
only forfeit the matching shares. Over previous years, the share
price has shown solid development. Even though employees from
all hierarchy levels are eligible to participate in the program, we
nd that a great proportion of the rank-and-le employees do not
participate in the ESOP.
13
Since its introduction in the late 2000s,
the features of the ESOP and its communications have been ho-
mogeneous across countries. Similar to the study by Hofstede
(1980), our research design holds rm effects constant, which en-
ables us to better isolate the effects of country-level factors on the
relationship between ESOP participation and rm exits.
Our initial dataset comprised observations for all employees
outside the rm's home country from the rm's internal HR data-
base. In order to analyze individual rank-and-le employees' de-
cisions to leave the rm, we excluded the following from our
observations: (1) senior managers, (2) short-term employees, (3)
employees entering retirement, (4) employees from divested
organizational units,
14
(5) countries where only senior managers
were eligible and all countries where no ESOP participant had left
the rm, (6) employees who had only been in the rm's database
for one year or who had only been eligible for one year,
15
and (7) all
observations with missing data for the control or country-level
variables.
The resulting sample comprises 654,706 observations for
190,370 individual employees in 39 countries over the period from
2011 to 2015. Table 1 summarizes the sample selection, and Table 2
11
In the questionnaire used for Hofstede's cultural values framework, one of the
three questions for the UA dimension particularly addresses employment stability
by asking, How long do you think you will continue working for this company?
(Hofstede, 1980, p. 162). Moreover, vis-
a-vis Hofstede's other cultural dimensions,
we consider UA to t particularly well with the concept of employees' perceived
ease of movement and thus their rm exit decisions due to the inherent unpre-
dictability of a job change. For example, employees ranking high in UA are likely to
experience a low perceived ease of movement, while employees ranking low in UA
are likely to experience a high perceived ease of movement. In additional tests, we
control for Hofstede's other cultural dimensions and nd that our results hold.
12
Some of the data have been used in other cross-country studies (Ahrens,
Oehmichen, &Wolff, 2018;Oehmichen, Wolff, &Zschoche, 2018). However, pre-
vious studies only used cross-sectional data from 2012 and did not have access to a
panel data set.
13
From an economic perspective, this nding is startling yet conrmed by a large
body of literature examining this issue (e.g., Babenko &Sen, 2014;Bova, Kolev,
Thomas, &Zhang, 2015;Duo&Saez, 2002;Oehmichen et al., 2018).
14
We relied on an approximation by identifying organizational units where more
than or equal to 70% of employees had left, suggesting that these organizational
units were subject to divestitures. We validated the results by collating our iden-
tied divestiture units with publicly available data for major divestments and found
conrmation for the approach. Moreover, we reran our analyses with the exclusion
of organizational units where 80% and 90% of the employees had left. The results
remained unchanged.
15
We did so to reduce potential database errors.
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
5
presents an overview of the absolute and relative number of em-
ployees, ESOP participants, and rm exits per country. Emerging
from this overview, we began our exploration of the relationship
between ESOP participation and rm exit with a scatterplot. Fig. 1
presents a plot of the country-average values of ESOP participa-
tion and rm exit rates. The relationship is strongly negative, sug-
gesting that in countries with higher ESOP participation relatively
fewer employees leave the rm and thus provides a rst indication.
3.2. Empirical models and variables
To test our hypotheses, we follow prior literature (Angrist &
Pischke, 2009; Aobdia &Cheng, 2018; Guo &Masulis, 2015) and
estimate a linear probability model (LPM) using OLS regressions,
where we cluster standard errors by employees.
16
As ESOP partic-
ipation represents an endogenous choice of employees, omitted
factors could drive this relationship. Therefore, we complement
OLS regressions with 2SLS regressions in the main analysis and
supplemental tests where applicable (e.g., Fu, Kraft, &Zhang, 2012;
Liu, 2020).
17
After describing the hypotheses testing using OLS in
the following section, we then discuss the implementation of 2SLS
for the hypotheses testing.
3.2.1. OLS regressions
To test Hypothesis 1 regarding the relationship between ESOP
participation and the rm exit decisions of rank-and-le em-
ployees, we estimate the following regression:
Firm Exit
i;tþ1
¼
a
þ
b
ðESOP ParticipationÞ
i;t
þ
g
ðControlsÞ
i;t
þYear
t
þBusiness Unit
i
þRegion
i
þε
i;t
(1)
where Firm Exit is our dependent variable that measures whether
an employee left the rm in year t þ1. The variable is coded as 1 if
the employee left the rm in the year following his/her decision
and 0 otherwise. Our independent variable ESOP Participation takes
a value of 1 if the employee participated in the offered ESOP in year
t and 0 otherwise. We intentionally created this one-year lag to
separate the point in time when the employee had decided
whether to participate in the ESOP from the decision regarding
whether to leave the rm.
We include several control variables that might inuence rm
exit decisions of rank-and-le employees. On the individual level,
we consider the following variables: Gender is a dummy variable
that takes a value of 1 for males and 0 for females. We measure Age
in years and Tenure as the natural logarithm of years. Education
takes the value of 1 if the employee has a master's degree and
0 otherwise. We include Total Cash as a proxy for employees' in-
come.
18
We code Hierarchy Level as 1 if the employee belongs to the
upper hierarchy levels among the rank-and-le employees and
0 otherwise. Further, we add indicator variables that cluster the job
families within the areas of Administration,Production,R&D, and
Sales, while we omit the area of Procurement. On the organizational
level, we consider Group Behavior as the proportion of employee
rm exits within each organizational unit. We also include Orga-
nizational Unit Size measured as the logarithm of the number of
employees in each organizational unit. Finally, we control for year,
business unit, and region xed effects
19
throughout. Our results are
robust to different ways of clustering the standard errors, such as at
the country level, business-unit level, or organizational-unit level,
and using organizational unit and country xed effects.
To investigate our second set of hypotheses, we include an
interaction term of ESOP Participation and Labor Market Conditions
for Hypothesis 2a, and, for Hypothesis 2b, an interaction term of
ESOP Participation and Uncertainty Avoidance. We add the respec-
tive moderator (M)ofLabor Market Conditions and Uncertainty
Avoidance:
20
Firm Exit
i;tþ1
¼
a
þ
b
ðESOP ParticipationÞ
i;t
þ
b
2
ðESOP Participation x MÞ
i;t
þ
g
1
ðMÞ
i;t
þ
g
2
ðControlsÞ
i;t
þYear
t
þBusiness Unit
i
þRegion
i
þε
i;t
(2)
where Labor Market Conditions is the unemployment rate from the
World Bank, which we reverse to facilitate the interpretation of
outside opportunities and favorable labor market conditions (e.g.,
Kale, Ryan, &Wang, 2019). For our Uncertainty Avoidance measure,
we employ Hofstede's cultural values framework (Hofstede, 1980).
Validated by its cumulative impact on business research(Karolyi,
2016, p. 216), we rely on the UA index to differentiate between
Table 1
Sample selection.
Observations
Observations for all eligible employees for the years 2011e2015 outside the rm's home country 908,031
Remove observations of senior managers 30,897
Remove observations of employees with a temporary contract 100,587
Remove observations of employees who retire 3,410
Remove observations of employees who belong to divested organizational units 36,108
Remove observations of employees in countries where only senior managers are eligible 536
Remove observations of employees in countries where no ESOP participant leaves the rm 3,360
Remove observations of employees who have only been one year in the rm's database 42,420
Remove observations of employees who have only been ESOP eligible for one year 846
Remove observations with missing data for control or country-level variables 35,161
Final Sample 654,706
16
We chose an LPM over a nonlinear logit or probit model for two reasons. First,
an LPM facilitates the interpretation of the interaction terms that we include to test
Hypotheses 2a and 2b (Norton, Wang, &Ai, 2004). Second, given our inclusion of a
large number of xed effects, an LPM helps to avoid the incidental parameters
problem (Wooldridge, 2002).
17
In their paper on the use of 2SLS, Larcker and Rusticus (2010) also recommend
comparing OLS and 2SLS regressions results.
18
As we were not able to get individual income data for all employees, we
assumed that employees' income per sub-job family and country was comparable
and calculated the respective average income. The sub-job family is a more precise
description of an employee's function, while the job family describes a broader
functional area that includes various sub-job families. For example, the job family
salesincludes the sub-job family key account management,while R&Din-
cludes software development.See Appendix B for a more detailed explanation of
the rm's structure.
19
Given the distribution of countries in our sample, we use the following global
regions: Africa, North America, South America, Mesoamerica, East Asia, West and
Central Asia, Europe, and Oceania.
20
We mean-centered the variables in the interaction terms to reduce issues of
multicollinearity (Aiken &West, 1991).
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
6
employees (and their beliefs) across cultures (e.g., Brochet, Miller,
Naranjo, &Yu, 2019;Debus, Probst, K
onig, &Kleinmann, 2012;
Kwok &Tadesse, 2006). As we conduct our study within one rm,
there are no rm effects that confound the effects on the country
level. Appendix A outlines denitions and data sources for all var-
iables. Appendix B explains the focal rm's structure. Appendix C
provides the values for our moderator variables.
3.2.2. 2SLS regressions
The instrumental variable (IV) approach is an effective and
commonly used method in accounting research to tackle endoge-
neity concerns, such as an omitted variable bias (Larcker &Rusticus,
2010). In our case, ESOP participation is a choice variable and not
likely to be random. Thus, we face the potential of an omitted factor
driving both an employee's ESOP participation and rm exit deci-
sion. More precisely, in our study we argue that an employee's ESOP
participation leads to PO, which then results in a reduced likelihood
of rm exit. However, there is the possibility of pre-existing PO (a
factor we cannot capture) that affects the employee's ESOP
participation and rm exit decision at the same time. If this were
the case, the omitted factor (i.e., pre-existing PO) would strengthen
the positive relationship between ESOP participation and rm exit.
In general, IVs must satisfy two conditions to be valid. First, the
relevance condition requires that instruments are correlated with
the endogenous variable. Second, the exclusion condition requires
that instruments should only indirectly relate to the outcome var-
iable through its effect on the endogenous variable. Thus, it is
important to select an IV that relates to ESOP participation, but
must not be a determinant of rm exit. Moreover, in line with our
arguments of the construct of PO, the IV should also not be related
to pre-existing PO. The success of 2SLS depends on the quality of
the IV to fulll these conditions. Larcker and Rusticus (2010)
particularly point to the issue of weak instruments as they are
more likely to cause wrong statistical inference than standard OLS
estimates that make no correction for endogeneity.
We select nancial literacy as the instrument for ESOP partici-
pation. The rationale behind this instrument is that individuals
with higher nancial literacy face lower barriers when dealing with
stock markets. However, it is a widespread issue that people lack
basic economic knowledge (Hilgert, Hogarth, &Beverly, 2003;
Lusardi &Mitchell, 2007). In this vein, studies point out that stocks
are complex assets, and many households may not know or un-
derstand stocks and the workings of the stock market.(van Rooij,
Lusardi, &Alessie, 2011, p. 460) We expect nancial literacy to
enable employees to better understand the investment in ESOPs
and to recognize its benets. Empirically nancial literacy has also
been linked to investment decisions as those with high literacy are
more likely to invest in stocks (van Rooij et al., 2011) and ESOPs in
Table 2
Overview of the number (percentage) of employee-years, ESOP participation employee-years, and rm exit employee-years per country for the years 2011e2015.
Country No. of employee-years No. (Perc.) of ESOP participations No. (Perc.) of rm exits
Australia 7,821 1,036 (13.25%) 535 (6.84%)
Austria 10,692 3,882 (36.31%) 412 (3.85%)
Belgium 7,917 713 (9.01%) 356 (4.50%)
Brazil 30,502 731 (2.40%) 3,635 (11.92%)
Canada 17,517 1,887 (10.77%) 1,084 (6.19%)
Chile 562 62 (11.03%) 32 (5.69%)
China 41,575 4,315 (10.38%) 3,012 (7.24%)
Czech Republic 38,762 1,876 (4.84%) 2,005 (5.17%)
Denmark 16,513 3,750 (22.71%) 853 (5.17%)
Egypt 1,254 47 (3.75%) 87 (6.94%)
Finland 1,215 369 (30.37%) 67 (5.51%)
France 32,068 2,170 (6.77%) 1,474 (4.60%)
Hong Kong 2,052 219 (10.67%) 108 (5.26%)
Hungary 7,038 209 (2.97%) 424 (6.02%)
India 28,859 765 (2.65%) 2,207 (7.65%)
Indonesia 7,134 37 (0.52%) 765 (10.72%)
Ireland 1,355 51 (3.76%) 52 (3.84%)
Israel 2,040 218 (10.69%) 181 (8.87%)
Italy 11,030 153 (1.39%) 323 (2.93%)
Malaysia 7,042 717 (10.18%) 214 (3.04%)
Mexico 27,037 444 (1.64%) 2,505 (9.27%)
Netherlands 11,742 1,091 (9.29%) 366 (3.12%)
New Zealand 414 92 (22.22%) 39 (9.42%)
Norway 3,223 542 (16.82%) 159 (4.93%)
Philippines 1,391 171 (12.29%) 46 (3.31%)
Poland 3,407 378 (11.09%) 131 (3.85%)
Romania 4,231 121 (2.86%) 189 (4.47%)
Singapore 6,823 964 (14.13%) 593 (8.69%)
Slovakia 6,554 396 (6.04%) 279 (4.26%)
South Africa 2,513 130 (5.17%) 165 (6.57%)
South Korea 7,382 1,132 (15.33%) 236 (3.20%)
Spain 7,212 1,583 (21.95%) 202 (2.80%)
Sweden 4,515 1,622 (35.92%) 198 (4.39%)
Switzerland 21,443 3,717 (17.33%) 1,078 (5.03%)
Taiwan 2,495 651 (26.09%) 125 (5.01%)
Thailand 5,106 162 (3.17%) 410 (8.03%)
Turkey 11,981 947 (7.90%) 668 (5.58%)
United Kingdom 26,804 4,979 (18.58%) 1,530 (5.71%)
United States 227,485 32,932 (14.48%) 14,316 (6.29%)
Total 654,706 75,261 (11.50%) 41,061 (6.27%)
Notes: This table reports the number (percentage) of employee-year observations, the number (percentage) of ESOP participation employee-year observations, and the
number (percentage) of rm exit employee-year observations per country for the years 2011e2015.
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
7
particular (Babenko &Sen, 2014). Thus, nancial literacy satises
the relevance condition. At the same time, it is difcult to imagine
how nancial literacy should relate to pre-existing PO or rm exit
decisions and therefore likely satises the exclusion condition.
21
While our data for ESOP participation varies at the employee
level, there was no individual-level data for nancial literacy
available, especially, on a global scale. Instead, we obtained nan-
cial literacy data aggregated at the country level from the Standard
&Poor's Ratings Services Global Financial Literacy Survey (S&P
Global FinLit Survey). The survey is based on over 150,000 in-
terviews with representative adults in more than 140 countries. We
follow Klapper, Lusardi, and van Oudheusden (2014) and Meoli,
Rossi, and Vismara (2021) in measuring nancial literacy as the
percentage of adults who are nancially literate in one country and
then apply this value to the employees in the country.
22
In the rst-stage regression, the dependent variable is ESOP
Participation and the independent variables include Financial Lit-
eracy as the IV as well as the control variables discussed in the OLS.
In the second-stage regressions, the tted values of ESOP
Participation are used as the independent variable. The other vari-
ables remain the same as in the OLS.
4. Empirical results
4.1. Summary statistics
Table 3 reports the descriptive statistics and the correlation
matrix for our nal sample of 654,706 employee-year observations.
The individual employee turnover variable Firm Exit has a mean of
0.06 and is similar to the data from Aldatmaz et al. (2018). Providing
an indication consistent with our expectation, the matrix shows a
negative correlation between ESOP participation and rm exit
decisions.
4.2. First-stage regression results
Table 4 presents the rst-stage regression results of the 2SLS, in
which we estimate employees' ESOP Participation using our IV
(Financial Literacy) and control variables. We observe that, consis-
tent with theory and prior research, Financial Literacy loads posi-
tively and signicant (
b
¼0.003, p<0.01). Following Larcker and
Rusticus (2010), we further evaluate the strength of our instru-
ment with a partial F-statistic to assess its incremental explanatory
power over the control variables. The partial F-statistic greatly ex-
ceeds the Stock and Yogo (2005) critical value of 16.38 based on a
10% Stock-Yogo maximum IV size and thus rejects the null that the
instrument is weak. Moreover, we nd that an employee's partic-
ipation is positively (negatively) related to Gender,Age,Tenure,
Education,Total Cash,Hierarchy Level,Administration,R&D, and Sales
(Production,Group Behavior, and Organizational Unit Size).
4.3. OLS and 2SLS (second-stage) regression results
Table 5 presents the regression results of the multivariate
analysis with Panel A reporting the OLS and Panel B the 2SLS re-
gressions. In both panels, Model 1 shows the results for Hypothesis
Fig. 1. Scatterplot of average country-level ESOP participation and rm exit rates.
21
To address concerns of reverse causality, nancial literacy is supposed to be
only related to ESOP participation, but not directly to PO or rm exit. As PO could
also stem from controlling or knowing the target, we have to rule out that nancial
literacy relates to these two sources of PO. First, nancial literacy should have no
direct effect on controlling the target (i.e., power). Even if employees in the nance
department would rank systematically higher and thus possess more power, we
would capture this through the set of control variables. Second, nancial literacy is
a proxy for general knowledge and not rm-specic knowledge, which makes
knowing the target not more likely. Moreover, we cannot think of a conceptual
expectation regarding an obvious relation between nancial literacy and rm exit.
22
Given that nancial literacy varies at the country-level, it may relate to other
country-level determinants of ESOP. Indeed, our data show some high correlations
between nancial literacy and country-level factors, such as culture, legislation, or
the economic situation. In sensitivity tests, we therefore include potential country-
level determinants of ESOP in both the rst- and second-stage regressions. The
results show that after controlling for a bunch of country-level factors the relation
between nancial literacy and ESOP participation in the rst-stage remains positive
as well as highly signicant and the results of the second-stage regressions also
remain quantitatively and qualitatively similar.
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
8
1, testing whether ESOP participation is negatively related to an
employee's exit. Consistent with our hypothesis, we nd negative
and signicant coefcients for ESOP Participation in the OLS
(
b
¼0.024, p<0.01) and the 2SLS (
b
¼0.103, p<0.01). For the
assessment of the economic effects, we focus on the results of the
OLS as employees can only choose to participate (1) or not to
participate (0) in the ESOP. Therefore, the binary variable of ESOP
Participation in the OLS facilitates a more straight-forward inter-
pretation that better reects reality, whereas ESOP Participation in
the 2SLS is the expected value in form of a continuous variable.
Nonetheless, when we compare the economic effects of the two
models in terms of increases of one standard deviation from the
mean in ESOP Participation,wend almost identical effect sizes.
23
Based on the results from OLS, we nd rank-and-le employees
who participate in the ESOP to be 2.41 percentage points less likely
to leave the rm within the next year. While the average likelihood
of a rm exit is 6.27%, it is 6.55% for non-participants and 4.14% for
participants. To put this result into perspective, a survey study by
Bryson and Freeman (2019) found a gap of 6 percentage points in
the intention to leave between non-participants (8%) and partici-
pants (2%) of a rm's employee ownership plan. Given that these
survey results might be subject to the common rater bias
(Podsakoff, MacKenzie, Lee, &Podsakoff, 2003), the more moderate
economic magnitude of our results seems plausible.
In both panels, Model 2 tests Hypothesis 2a, which proposes
that the effect of ESOP participation on rm exit is stronger under
favorable labor market conditions. We continue to nd negative
and signicant coefcients for ESOP Participation. We also nd
negative and signicant coefcients for the interaction term be-
tween ESOP Participation and Labor Market Conditions in the OLS
(
b
¼0.002, p<0.01) and the 2SLS (
b
¼0.018, p<0.01), which
support our hypothesis. The OLS results suggest that an increase of
one standard deviation in Labor Market Conditions augments the
retention effect of ESOP Participation to 2.99 percentage points. To
put it differently, when labor market conditions are favorable
(mean þ1SD), the likelihood of a rm exit is 6.89% for non-
participants and 3.90% for participants. Further, we nd that the
coefcient for Labor Market Conditions is positive and signicant,
consistent with employees being more likely to leave when
encountering favorable outside employment opportunities.
In both panelsModel 3, we test Hypothesis 2b, which posits a
stronger retention effect from ESOP participation in cultures with
lower levels of UA. Again, we nd that the coefcients for ESOP
Participation are negative and signicant. In line with our hypoth-
esis, the results document positive and signicant coefcients for
the interaction term between ESOP Participation and Uncertainty
Avoidance in both the OLS (
b
¼0.0002, p<0.01) and the 2SLS
(
b
¼0.003, p<0.01). The OLS results indicate that a decrease of one
Table 3
Summary statistics.
No. Variables Mean SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 Firm Exit 0.063 0.242 0.053 0.025 0.036 0.005 0.023 0.036 0.050 0.004 0.021 0.014 0.011 0.006 0.007 0.012 0.386 0.042
2 ESOP Participation 0.115 0.319 0.053 0.107 0.017 0.083 0.045 0.089 0.066 0.090 0.184 0.132 0.037 0.121 0.049 0.038 0.073 0.122
3 Financial Literacy 50.128 13.814 0.031 0.123 0.099 0.214 0.027 0.280 0.102 0.019 0.487 0.042 0.035 0.011 0.020 0.003 0.024 0.103
4 Labor Market Conditions 20.250 3.078 0.030 0.013 0.079 0.335 0.031 0.103 0.063 0.008 0.116 0.013 0.031 0.004 0.052 0.033 0.095 0.071
5 Uncertainty Avoidance 54.658 19.271 0.005 0.089 0.229 0.347 0.018 0.040 0.041 0.015 0.216 0.049 0.011 0.099 0.108 0.006 0.039 0.055
6 Gender 0.740 0.439 0.023 0.045 0.045 0.018 0.011 0.050 0.037 0.020 0.056 0.056 0.251 0.012 0.148 0.120 0.010 0.064
7 Age 42.773 10.629 0.032 0.083 0.332 0.057 0.083 0.050 0.581 0.020 0.317 0.183 0.008 0.012 0.069 0.073 0.066 0.100
8 Tenure 2.219 0.774 0.042 0.065 0.156 0.073 0.020 0.039 0.576 0.030 0.159 0.172 0.002 0.002 0.041 0.043 0.099 0.070
9 Education 0.134 0.341 0.004 0.090 0.012 0.020 0.023 0.020 0.018 0.031 0.094 0.153 0.065 0.174 0.158 0.045 0.039 0.128
10 Total Cash 0.561 0.358 0.015 0.178 0.544 0.016 0.289 0.050 0.299 0.146 0.103 0.243 0.132 0.413 0.108 0.179 0.107 0.330
11 Hierarchy Level 0.096 0.295 0.014 0.132 0.058 0.007 0.063 0.056 0.177 0.167 0.153 0.269 0.127 0.148 0.002 0.019 0.072 0.223
12 Administration 0.237 0.425 0.011 0.037 0.034 0.039 0.019 0.251 0.007 0.000 0.065 0.148 0.127 0.285 0.306 0.346 0.045 0.273
13 Production 0.208 0.406 0.006 0.121 0.048 0.026 0.114 0.012 0.014 0.002 0.173 0.384 0.148 0.285 0.281 0.318 0.103 0.427
14 R&D 0.232 0.422 0.007 0.049 0.022 0.051 0.101 0.148 0.063 0.041 0.158 0.086 0.002 0.306 0.281 0.341 0.013 0.016
15 Sales 0.278 0.448 0.012 0.038 0.042 0.037 0.007 0.120 0.070 0.041 0.045 0.161 0.019 0.346 0.318 0.341 0.040 0.087
16 Group Behavior 0.064 0.123 0.479 0.045 0.037 0.051 0.019 0.034 0.048 0.082 0.013 0.023 0.018 0.020 0.009 0.011 0.023 0.303
17 Organizational Unit Size 3.483 1.213 0.052 0.127 0.158 0.089 0.088 0.027 0.116 0.075 0.130 0.331 0.217 0.260 0.470 0.039 0.116 0.106
Notes: This table presents descriptive statistics and the correlation matrix for the regression variables of our sample of 654,706 employee-year observations. The Pearson (Spearman) correlations are below (above) the diagonal.
All correlations greater than or equal to 0.004 are signicant at the 1 percent level.
23
To compare the effects sizes resulting from the OLS and 2SLS, we assess in-
creases in ESOP Participation of one standard deviations from the mean. For the OLS,
an increase in ESOP Participation results in a shift of rm exit likelihood from 6.27%
at the mean to 5.50% at the mean plus on standard deviation. For the 2SLS, an
increase in ESOP Participation results in a shift of rm exit likelihood from 6.27% at
the mean to 5.54% at the mean plus on standard deviation. Thus, the difference in
rm exit likelihood between the OLS and 2SLS results for increased ESOP partici-
pation is smaller than one percent. We note that the results for the following hy-
potheses vary more strongly between the regression models. The results obtained
from the 2SLS consistently translate into larger effect sizes than the ones obtained
from the OLS and discussed in the text. Thus, we are condent that the effect sizes
discussed in the text rather represent conservative estimates. In particular, we nd
that an increase (mean þ1 SD) in ESOP Participation in favorable labor market
conditions (mean þ1 SD) decreases the rm exit likelihood by 15% in the OLS and
by 20% in the 2SLS. We further nd that an increase (mean þ1 SD) in ESOP
Participation for rank-and-le employees ranking low in UA (mean e1SD) de-
creases the rm exit likelihood by 13% in the OLS and by 21% in the 2SLS.
J.C. Hennig, C. Ahrens, J. Oehmichen et al. Accounting, Organizations and Society xxx (xxxx) xxx
9
standard deviation in Uncertainty Avoidance increases the retention
effect of ESOP Participation to 2.69 percentage points. More spe-
cically, among rank-and-le employees ranking low in UA (mean
e1SD), the likelihood of a rm exit is 6.77% for non-participants
and 4.08% for participants. The coefcients for Uncertainty Avoid-
ance are negative and signicant, as leaving the rm involves
uncertainty.
Following Larcker and Rusticus (2010), we next assess the
severity of the endogeneity problem. First, we perform the Haus-
man test to evaluate the difference between the OLS and 2SLS re-
sults. We nd that the 2SLS results are signicantly different from
the OLS results (p<0.01) thus suggesting the existence of endo-
geneity. Second, given the conrmation of the existence of endo-
geneity in our study, we use an alternative approach to assess how
large the endogeneity problem has to be to overturn our OLS re-
sults. In line with Larcker and Rusticus (2010), we use the approach
by Frank (2000) to calculate how strongly an omitted variable
would have to be correlated with both the independent and
dependent variable to make the coefcient statistically insigni-
cant. Specically, we calculate the impact threshold for a con-
founding variable (ITCV). The ITCV is 0.0412, implying that the
omitted variable would have to be correlated at 0.203 with Firm Exit
and 0.203 with ESOP Participation (signs are interchangeably) to
overturn our results. We follow Larcker and Rusticus (2010) and
compare the ITCV with the impact of our controls. The ITCV is 2.77
times higher than the impact of Group Behavior (i.e., the control
variable with by far the highest impact). Given the assumption that
we have selected a reasonable set of control variables this indicates
that it is rather unlikely that an omitted variable has confounded
our OLS results. Paired with the results from the 2SLS, this provides
some condence in the inferences drawn from the regressions.
5. Sensitivity tests
5.1. Propensity score matching (PSM) procedure
ESOP participation is an endogenous choice, and thus unob-
served factors could drive the employeesdecision to participate in
the ESOP and the decision to leave the rm. The PSM procedure
accounts for differences in observable characteristics between
ESOP participants as well as non-participants and thus mitigates
concerns about unobservable characteristics that may be linked to
these observable characteristics (Chang, Chung, &Moon, 2013;
Chen, Luo, Tang, &Tong, 2015). The PSM matches each treatment
observation (i.e., participant) with a control observation (i.e., non-
participant) that has similar observable characteristics. We follow
the procedure suggested by Shipman, Swanquist, and Whited
(2017), and regress ESOP Participation on our previously used con-
trol variables to estimate the probability (i.e., the propensity score)
that an employee will participate in the ESOP. Based on the pro-
pensity score, we perform a one-to-onematching without
replacement and set the caliper to be within ±0.001% to make the
matched employees as similar as possible. Our matched sample
consists of 146,204 employee years.
We assess the matching quality using both univariate and
multivariate tests (see Internet Appendix,Table IA1). The tests
reveal that the post-match regression explains almost none of the
observed variance and that no differences in the propensity score
remain in the matched sample. Moreover, for the individual control
variables, we observe some statistically signicant differences be-
tween the two samples, which, however, are much smaller in the
matched sample than in the unmatched sample. Based on the
matched sample, we reexamine our hypotheses. Table 6 reports the
results, which substantiate our previous ndings.
5.2. Subsample analyses excluding employees with specic
knowledge about the ESOP
To control for the impact of different degrees of familiarity with
the ESOP, we rerun our tests excluding specic employee groups.
First, we exclude employees from the Finance and HR Departments,
as they are deeply involved in the rm's communication to foster
participation. Second, we exclude employees who hold a master's
degree, as employees with higher levels of education might have
better access to stock-related information. Third, we exclude em-
ployees who have been with the rm for a long time (i.e., above the
median tenure), as these employees should be very familiar with
the rm, including the ESOP. Our results (see Internet Appendix,
Tables IA3-IA.5) show that for all three modications in the OLS as
well as the 2SLS regressions, the results hold.
5.3. Alternative country-level variables
In this section, we evaluate our results' robustness regarding the
country-level context. In our main tests, we controlled for the
countriesafliation to specic regions. This enabled us to account
for regional economic factors and cultural covariates without
specifying individual country-level variables. In order to
Table 4
First-stage regressions results.
VARIABLES (1)
ESOP Participation
Financial Literacy 0.003***
(19.035)
Gender 0.020***
(13.459)
Age 0.000***
(3.779)
Tenure 0.007***
(7.178)
Education 0.048***
(19.536)
Total Cash 0.081***
(23.639)
Hierarchy Level 0.072***
(23.246)
Administration 0.014***
(4.600)
Production 0.008***
(-3.035)
R&D 0.032***
(10.384)
Sales 0.021***
(7.298)
Group Behavior 0.071***
(-19.059)
Organizational Unit Size 0.006***
(-10.159)
Intercept 0.169***
(-13.120)
Year xed effects Yes
Business unit xed effects Yes
Region xed effects Yes
N 654,706
Adjusted R
2
0.10
Partial F-statistic 1510.93***
Notes: This table reports the rst-stage regression of the 2SLS regression
analysis using nancial literacy as the instrument variable. ***,**, and *
indicate signicance at the 1%, 5%, and 10% levels (two-tailed), respectively.
Robust standard errors are clustered at the employee level and t-statistics are
provided in parentheses. Ap