ArticlePDF Available

Abstract and Figures

We examine the valuation impact of an employee-friendly (EF) culture. Using a sample of 3446 firms from 43 countries for the period 2003 to 2014, we show that firms with a more EF culture are valued higher and perform better (ROA, ROE). Consistent with the good governance view, the impact is stronger for firms in countries with better investor protection and for firms with better governance and lower agency costs. We further document a positive valuation associated with the enactment of laws aimed at improving parental leave policies. The impact on valuation stems from improved technical efficiency. Using various approaches, our results suggest that the impact of an EF culture on firm value is causal.
Content may be subject to copyright.
Does it Pay to Treat Employees Well? International Evidence on the Value of Employee-
Friendly Culture*
Larry Fauver
Michael B. McDonald
Alvaro G. Taboada§
Abstract
We examine the valuation impact of an employee-friendly (EF) culture. Using a sample of 3,446
firms from 43 countries for the period 2003 to 2014, we show that firms with a more EF culture
have higher valuations (Tobin’s q). The impact is stronger in countries with more competitive
industry structures and with a more productive labor force. We find that the value of an EF culture
is important during periods of crisis. We further document increases in firm value associated with
the enactment of laws aimed at improving parental leave policies. The impact on valuation stems
from improved technical efficiency. Through the use of various approaches our results suggest
that the impact of an EF culture on firm value is causal.
Key Words: Culture, Employee treatment, Corporate Finance, Behavioral Finance, Governance,
Valuation, International, Theory of the Firm
* We have received helpful comments from Seong Byun, Xinxin Li, Sandra Mortal, and Paige Ouimet. We also thank
seminar participants at the Federal Reserve Bank of New York/Journal of Accounting and Economics joint conference
on Economics of Culture: Balancing Norms against Rules, the 2016 Financial Management Association Annual
Conference, the MMM Conference at the University of Mississippi, the 2017 Northern Finance Association
Conference, Mississippi State University, and the University of Tennessee.
Associate Professor of Finance, James. F. Smith, Jr. Professor of Financial Institutions, Research Fellow, Corporate
Governance Center, The University of Tennessee, lafauver@utk.edu,, 865-974-1722.
Assistant Professor of Finance, Fairfield University, mmcdonald8@fairfield.edu, 203-254-4000 x3118.
§ Interim BancorpSouth Assistant Professor of Finance, Mississippi State University, agt142@msstate.edu, 662-325-
6716.
1
I. Introduction
“Train people well enough so they can leave, treat them well enough so they don’t want to.”
Sir Richard Branson
Is there value in creating a more employee-friendly (EF) culture? The quote above by
Virgin Atlantic’s founder signals what could be the start of a global shift in the way firms view
and treat employees, raising important questions about efficiency for financial economists to
consider. While firms in the tech sector (e.g. Google, Yahoo, Netflix, Microsoft) are well known
for offering employees perks that include free meals, generous paid leave packages and in-building
fitness and entertainment amenities, in addition to paying competitive wages, such perks have not
been as prevalent in other industries.1 Yet, the media, government agencies, and corporations are
beginning to pay closer attention to the treatment of employees. For instance, San Francisco
recently became the first city in the United States to pass a law guaranteeing fully paid parental
leave, while Virgin Group made headlines recently with its generous paternity leave policy in
which new dads get up to 12 months paid leave.2 Are these firms making a rational economic
decision when they offer employees perks like free meals and in-building fitness amenities? Or are
these costs simply unnecessary extravagances that come at the expense of shareholders? To date,
the evidence on this issue is limited.
We study whether, in general, there are financial benefits to a firm by having a friendlier
firm-level culture. There are two competing views on whether policies that create a more EF
1 See a recent article by Forbes (http://www.forbes.com/sites/karstenstrauss/2013/05/31/how-to-keep-
employeeshappy-and-to-just-plain-keep-them/).
2 Tuttle, Brad, “Virgin’s New Paternity Leave Policy Puts Google and Facebook to Shame.” Money 10 June 2015.
2
culture are value enhancing, and the evidence to date is relatively scarce and mixed.3 The
reciprocity view (Akerlof, 1982) argues that by treating employees well managers may motivate
workers to exert high effort, which should lead to improved performance and valuation.4 Edmans
(2011), and Edmans, Li, and Zhang (2015), provide evidence that employee satisfaction is
associated with positive long-term returns and higher valuation in countries with flexible labor
markets.5 On the other hand, employee treatment may be driven by ulterior motives due to
misalignment of managerial and shareholder incentives, and thus be value destroying (Jensen and
Meckling, 1976; Pagano and Volpin, 2005).6 Consistent with the latter view, Cronqvist et al.
(2009) find evidence that entrenched managers pay their workers more to enjoy private benefits
(e.g. lower effort wage bargaining). Landier, Nair, and Wulf (2009) also document that geographic
dispersion is inversely related to employee treatment,7 and further find that divisions that are closer
to headquarters are less likely to experience layoffs, and that such layoffs are less sensitive to
divisional performance.8 Based on the conflicting views and mixed evidence, the question of
whether having an EF culture is value enhancing is an empirical matter.
We use data on various aspects of firms’ compensation, training, health and safety, and the
extent to which firms provide equal treatment and opportunities for advancement to examine the
3 Similar views arise within the broader corporate social responsibility (CSR) literature (see e.g. Ferrell, Liang, and
Renneboog, 2016; Liang and Renneboog, 2017).
4 The reciprocity view (Akerlof, 1982) argues that workers may reciprocate being treated well by exerting such high
effort. Consistent with the reciprocity view, several studies document a link between above-market compensation and
worker productivity, which leads to improved performance (e.g. Cappelli and Chauvin, 1991; Holzer, Katz, and
Krueger, 1991; Mas, 2006; Propper and Van Reenen, 2010; Ouimet and Simintzi, 2015). In addition to wages, there
is evidence that nonmonetary gifts may be helpful in motivating employees (Kube, Maréchal, and Puppe, 2012)).
5 These papers do not explore employee policies directly, but measure employee satisfaction based upon survey
responses using the “List of Best Companies to Work For.”
6 “Treating employees well can be expensive. But the company is always looking for more benefits to offer.” This
was from Suzanne McDowell, the VP of Human Resources at King Arthur Flour in The Atlantic (2014).
7 Cronqvist et al. (2009) also document that entrenched managers pay more to employees who are geographically
closer to the headquarters and closer in terms of the corporate hierarchy.
8 Landier et al. (2009) argue that social factors associated with the proximity between employees and managers (e.g.
reluctance to “fire neighbor employees”) may lead to a misalignment of managerial and shareholder incentives.
3
effect of having an EF culture on firm value. Using a broad sample of 3,446 firms in 43 countries
from 2003-2014, we expand on previous studies by exploring the valuation consequences of an
EF culture, by examining the components of an EF culture that are value enhancing, and by
analyzing the underlying mechanisms through which an EF cul ture impacts fir m value. We further
explore the conditions in which an EF culture is value enhancing.9
Our results show that firms that have a higher EF culture are valued higher (higher Tobin’s
q and market-to-book).10 We also find that the valuation impact of an EF culture is more
pronounced in countries with a more competitive industry structure and with a more productive
labor force. These results suggest that the valuation impact of an EF culture stems at least in part,
from employee-friendly firms’ ability to attract and retain productive employees. Further, we
explore the channels through which an EF culture may impact firm value. Our results indicate
that higher EF culture firms have higher sales-to-assets, lower costs, and have a greater number of
patents. These findings support the “reciprocity view” that argues that treating employees well
leads them to reciprocate by exerting high effort. The findings on patents lend support to Chen et
al. (2016) and Mao and Weathers (2015) who document a positive impact of employee treatment
on innovation for a sample of US firms.
Our study faces at least two problems related to identification. First, reverse causality is a
concern because firms that are more profitable may be able to invest more in their employees,
which results in a more EF culture. One aspect that may mitigate such concern is the fact that
economic theories suggest that a firm’s culture is specific to the firm and is largely fixed over long
9 The Edmans (2011) and the Edmans et al. (2015) sample is limited to roughly 800 firms and 1800 firms, respectively
and are only found in the Best Companies list, which is collected by the Great Place to Work Institute. Our results
are robust when we control for these firms in our sample, as well as if we exclude them from our regression analyses.
10In robustness tests we also find a positive impact of an EF culture of profitability (operating ROA).
4
periods (see e.g. Lazear, 1995; Kreps, 1990). Second, there could be endogeneity bias caused by
omitted variables. If the omitted variable impacts both firm value and a firm’s ability to create an
EF culture, our measure of employee-friendliness would not be exogenous to firm value, and the
coefficients from OLS regressions would be biased and inconsistent.
We perform several tests to alleviate these concerns. First, we use an instrumental variables
approach and project our measure of EF culture on two variables that capture a country’s culture,
borrowing from Hofstede (1980). Specifically, we use two cultural dimensions: Masculinity vs
Femininity (Masculinity) and Indulgence vs. Restraint (Indulgence). The identifying assumption
is that cultural values in a country may shape how firms treat employees, but should not have a
direct impact on firm performance, other than through their impact on employee treatment. Using
the Two-stage Least Squares (2SLS) approach, we continue to find that firms with greater EF
culture have higher firm value. Second, we examine the causal effect between changes in Tobin’s
q and changes in employee-friendliness to directly address the reverse causality concerns. The
results show that while there is a causal effect of changes in employee-friendliness on Tobin’s q,
past changes in Tobin’s q have no significant impact on employee-friendliness. Third, we also
explore two quasi-natural experiments to examine the causal effect of employee-friendliness on
firm value. We first test the differential impact on firm value for firms with high and low EF
culture after a shock to economic activity and employment using the global financial crisis as an
exogenous shock. We find that firms with greater EF culture prior to the crisis are valued higher
during and after the crisis. We also assess whether treating employees well creates value by
exploiting the staggered implementation of parental leave laws across several European countries
during our sample period. Using a difference-in differences (DiD) methodology, we find that the
enactment of these parental leave laws is associated with positive valuation effects, especially for
5
firms that are most likely to have been impacted by the enactment of these laws (i.e. the firms with
poor parental leave policies prior to the enactment of the laws).11 In all of ours tests, we continue
to find that employee-friendliness is associated with higher valuation, providing further support to
our main findings.
Our paper adds to the literature on the impact of culture on firm performance (e.g., Guiso,
et al., 2013; Edmans, 2011; Edmans, et al., 2015) by exploring the valuation consequences of
adopting an EF culture and by examining the conditions in which such culture is value enhancing.12
We show that an EF culture can add value, especially in countries with a more competitive industry
structure and a more productive labor force. We further add to this literature by examining the
channels through which an EF culture impact firm performance and value. We show that firms
with a more EF culture have better technical efficiency (higher sales-to-assets and lower costs-of-
goods sold-per employee). Our findings add further support to theories that emphasize the
importance of employees as key assets in organizations (see e.g. Rajan and Zingales, 1998; Berk,
Stanton, and Zechner, 2010; Carlin, and Gervais, 2009). We also contribute to the literature that
examines how employee treatment affects firm’s capital structure (Bae, Kang, and Wang, (2011))
and corporate innovation (Chang, Fu, Low, and Zhang, (2015)), and to studies that analyze the
impact of employee stock ownership programs (Kim and Ouimet, 2014). Bae et al. (2011)
document that firms that treat their employees better tend to have lower debt ratios.
11 The quality of maternity and parental leave policies is a component of the Diversity category that is part of the
ASSET 4 database’s Social score that measures social performance. We thus classify firms as most impacted by
parental leave laws as those with a Diversity index in the bottom quartile in their country as of the year prior to the
implementation of the law in the country.
12 This question ties in with the broader debate about corporate social responsibility (CSR) and whether certain CSR
activities are consistent with value maximization (see e.g. Ferrell, Liang, and Renneboog, 2016; Bénabou and Tirole,
2010; Krüger, 2015).
6
The paper proceeds as follows. In Section II we discuss the data and the methodology used
in our study. In section III we present our main results on the relation between EF culture and
firm value and performance and explore the channel for the valuation gain from having a higher
EF culture. In section IV we provide robustness results and we conclude in section V.
II. Data and Methodology
II.1. Measure of Employee-Friendliness
We measure an EF culture by focusing on how a firm treats its current employees. To do
so, we rely on questions and attributes of social performance using data from the ASSET4
database. Specifically, we focus on the following five categories: 1) Employment quality
measures a company's management commitment and effectiveness towards providing high-quality
employment benefits and job conditions; 2) Health and safety measures a company's
management commitment and effectiveness towards providing a healthy and safe workplace; 3)
Training and Developmentmeasures a company's management commitment and effectiveness
towards providing training and development (education) for its workforce; 4) Diversity measures
a company's management commitment and effectiveness towards maintaining diversity and equal
opportunities in its workforce, and 5) Human Rights − measures a company's management
commitment and effectiveness towards respecting the fundamental human rights conventions. At
first glance, it could be argued that the Human Rights component does not capture a firm’s EF
culture, especially in more developed countries where basic human rights are protected. However,
this may not be the case for firms in less developed countries, where human rights violations are
not uncommon (e.g. Wernau, 2015). In addition, many large companies (e.g. GAP; Walmart) have
been involved in scandals related to human rights (child labor) violations (see e.g. Brown, 2007;
Smith, 2016). We thus keep the human rights component in our primary measure of EF index. In
7
robustness tests, we exclude the human rights component from our EF index measure and obtain
similar results.
While ASSET4 provides its own aggregate scores for each of these categories, we construct
our own firm-level index (EF-Index) using various attributes, although we validate our main
results using the scores from ASSET4. While our choice of variables is admittedly arbitrary, this
approach is more transparent and allows us to more closely examine important questions such as
when and what factors are important determinants of the impact of EF culture on firm value and
performance. In addition, by constructing our own firm-level measure we apply a consistent
standard to all firms in our sample. Our index construction parallels the construction of the firm-
level governance index by Aggarwal et al. (2009). To alleviate concerns about the validity of our
measure, we also use an alternate index that is based on the scores in each of the above five
categories provided by ASSET4. Specifically, our alternate index, EF-Index ASSET4, is the
average of the scores on the five categories from ASSET4.13
We have a total of 32 employee-treatment attributes covering five categories: Employment
quality (seven attributes); Diversity (eight attributes); Training and development (six attributes);
Health and safety (five attributes), and Human rights (six attributes). For each of the 32 attributes,
our index takes the value of one if the company meets the criteria, and zero otherwise. In the case
in which the attribute is a number (e.g. percentage of women managers), the index takes the value
of one if the value is above (or below) the industry median and zero otherwise.14 We create an
index for each of the five categories, expressed as a percentage, with a maximum value of 100%
13 ASSET4 assigns scores (0-100) to each component of the five components of the social score: Employment quality;
Diversity; Training and Development; Health and Safety, and Human Rights. These are based on multiple factors
(questions) within each category. Higher values are associated with better employee treatment.
14 We use the 2-digit SIC code to determine the median industry values.
8
if a firm meets all the available criteria in each category. Similarly, we compute the aggregate
index of employee-friendliness, EF-Index, with a maximum value of 100% if a firm meets all 32
attributes. For firms that have missing attributes, we compute each index based on the percentage
of all nonmissing attributes that a firm satisfies. Appendix B shows the attributes used to create
the index for each category as well as the percentage of firms in our sample that satisfy each
attribute. The indices are computed annually for each firm.
In terms of Employment Quality, from Appendix B we observe that very few firms in our
sample experience strikes that lead to lost working days, and only 10.8% of our firms have been
included in the “Best Companies to Work For” lists. The latter suggests that our index of
employee-friendliness is a broader measure than the one typically used in prior studies, as it covers
additional areas and firms that go beyond the inclusion in the “Best Companies to Work For” lists.
Assessing the Diversity component, about 74% of our firms have a diversity policy, while the
proportion of women managers is higher than the industry median for about 44% of the firms in
our sample. For Training and Development, we observe that about 62% of the companies in our
sample have policies that support skills training of their employees, while only 7.8% of the
companies provide training to its suppliers. In terms of Health and Safety, 55.4% of companies
establish targets or objectives on employee health and safety. Finally, looking at the Human Rights
component, we observe that only 33.1% our firms have a general policy regarding human rights,
and only 11.9% monitor human rights in its suppliers. The proportion of firms meeting the Human
Rights criteria is the lowest among all five categories.
II.2. Sample Description and Descriptive Statistics
9
Our initial sample consists of all firms covered by ASSET4 database from 2002 through
2014 with available data on the five key categories of social performance. The database covers a
subset of firms from Thompson Financial’s DataStream and WorldScope.15 The database
coverage varies by country, with coverage of developed markets starting in 2002, while some
emerging markets begin coverage in 2007 or beyond. Our initial sample consists of 5,006 firms
from 67 countries. We exclude firms with missing values for total assets, as well as those with
negative sales or negative book value of equity. We proceed with our screening by excluding
firms from regulated industries (financials – SIC codes between 6000 and 6999 and utilities – firms
with SIC codes between 4900 and 4949) and those with missing values on our control variables.
Finally, we exclude countries with fewer than three years of available data and those with fewer
than three firms.16 To mitigate the influence of outliers we winsorize all variables at the top and
bottom 1% of the distribution. While the ASSET4 database coverage starts in 2002, our sample
period starts in 2003 because we use lagged measures of the EF index in our analyses. Our final
sample consists of 3,446 firms from 43 countries totaling 21,103 firm-year observations. In
addition to the firm-level data, we collect country-level data from various sources. We obtain data
on financial development and economic growth from the World Bank Development Indicators.
All variables are defined in Appendix A.
Table 1 shows a description of our sample. Our sample is geographically diverse. Firms
from the US (842), Japan (351), Australia (307), and the United Kingdom (298) comprise about
15 The ASSET4 universe covers over 5,000 firms from major indices including MSCI Emerging Markets, MSCI
World, CAC40, DAX, FTSE250, S&P 500, NASDAQ 100, STOXX 600, ASX 300, SMI, and Bovespa.
16 The following countries were dropped from our sample because of data availability: Cayman Islands, Cyprus, Czech
Republic, Gibraltar, Hungary, Iceland, Isle of Man, Jordan, Kazakhstan, Kuwait, Macau, Morocco, Nigeria, Oman,
Panama, Papua New Guinea, Peru, Puerto Rico, Qatar, Saudi Arabia, Sri Lanka, Ukraine, United Arab Emirates, and
Zimbabwe. Firms from these countries (74) represent about 2.1% of our final sample.
10
half of our sample (52.2%). Our sample is comprised of large firms, covering about 87% of the
total market capitalization of all firms (excluding financials and utilities) covered by WorldScope
as of 2014.
[Insert Table 1 Here]
Table 2 shows descriptive statistics of our main firm- and country-level variables. Firms in
our sample are large, with average (median) total assets of $4.7 billion ($4.5 billion). The average
(median) Tobin’s q is 1.80 (1.44). The average (median) EF Index is 38.6 (36.8) with a standard
deviation of 21.2.
[Insert Table 2 here]
Table 3 shows the pairwise correlation coefficients between all our variables of interest.
Notably, the results show a strong correlation between the EF index and many of the other
variables. While there is a negative correlation between EF Index and Tobin’s q, the EF index
displays a positive correlation with firm size, age, percentage of foreign sales, profitability (ROA),
and the cross-listing indicator, and a negative correlation with cash holdings, the percentage of
closely-held shares, and the level of capital expenditures-to-assets. Many of the other variables
also display unsurprising correlations, but none of these correlations is high enough to suggest a
multicollinearity issue.
[Insert Table 3 Here]
III. Results
III.1. Employee-Friendliness and Firm Value
We first examine whether having an EF culture is associated with higher firm value. The
primary regression specification is a standard OLS regression using Tobin’s q (market value of
11
assets-to-book value of assets) as our main proxy for firm value. Our regressions include several
firm-level, country-level, and industry-level control variables used in prior research to explain
Tobin’s q (Aggarwal et al., 2009; Gompers, Ishii, and Metrick, 2010; Doidge, Karolyi, and Stulz,
2004). Specifically, we include the following firm-level control variables: (1) Size, measured as
the log of book value of assets; (2) Age, the log of firm age; (3) Leverage, debt divided by total
assets; (4) Cash, cash divided by total assets; (5) PPE, property, plant, and equipment divided by
sales; (6) Foreign sales, the two-year average foreign sales divided by sales; (7) R&D, the two-
year average research and development expenses divided by sales; (8) Capex, capital expenditures
divided by total assets; (9) ROA, net income divided by book value of assets; (1) Closely-held, the
percentage of a firm’s shares that are closely held, and (11) ADR, a variable indicating firms cross-
listed on U.S. stock exchanges. To control for patterns over time by country and industry, we
include country-year and industry-year fixed effects in our baseline regressions. In specifications
in which we exclude country-year fixed effects, we include the log of annual GDP per capita (Log
GDP per capita) and the growth rate of real GDP (GDP Growth) to control for financial
development and growth. All control variables are lagged one year. We use the following model
to test the effect of EF culture on firm value:
 
, ∑
,

  , (1)
EF refers to our proxies for EF culture, EF index or EF index-ASSET4; Controls refers to the firm-
level control variables, and ct and jt refer to country-year and industry-year fixed effects,
respectively. Per our main hypothesis, our variable of interest is the coefficient on β1 and we
expect this to be positive and significant if an EF culture is associated with positive valuation
consequences. Consistent with the reciprocity view, the results in Panel A of Table 4 show
evidence of a positive and significant coefficient on β1, suggesting that firms with more EF culture
12
have higher Tobin’s q. The results are both statistically and economically significant. For
example, using the coefficient in Model (2), a one-standard-deviation increase in EF index (21.2
– from Table 2) is associated with a 4.8% increase in Tobin’s q.17
[Insert Table 4 here]
We examine the robustness of our results by estimating various specifications of Equation
1 in Panel A of Table 4. In Model (1) we control for country, industry, and year fixed effects and
include Log GDP per capita and GDP growth to control for financial development and growth.
In Model (2) we include country-year and industry-year fixed effects to control for plausible
patterns in employee-friendliness over time by country and industry. In Model (3) we show results
including firm and year fixed effects to better control for time invariant firm-specific
characteristics. The magnitude of the coefficient on EF index is much smaller when using firm
fixed effects, which suggests that the impact on Tobin’s q is driven mostly by cross-sectional
variation in EF index. In Models (4)-(6), we replicate our results using our alternate measure of
EF culture, EF-index ASSET4, derived from the component scores from the ASSET4 database.
The results using the alternate measure of employee-friendliness are similar in statistical
significance, but slightly larger in economic magnitude compared to our main measure, EF index.
From Model (4) in Panel A of Table 4, a one-standard-deviation increase in EF-index ASSET4
(23.71) is associated with a 5.0% increase in Tobin’s q.18 The coefficient on EF-index ASSET4 is
positive but insignificant when we include firm and year fixed effects (in Model (6)). Given that
17 The coefficient on EF index in Model (2) of Panel A of Table 4 is 0.0041. Thus, a one-standard-deviation increase
in EF index (21.18) is associated with a 0.087 (21.18 x 0.0041) increase in Tobin’s q, which represents a 4.8% increase
(0.087/1.80).
18 The coefficient on EF-index ASSET4 in Model (5) of Table 4 is 0.0038. Thus, a one-standard-deviation increase in
EF-index ASSET4 (23.71) is associated with a 0.090 (23.71 x 0.0038) increase in Tobin’s q, which represents a 5.0%
increase (0.09/1.80).
13
Figure 1 shows there is little time-series variation in the broader EF-index ASSET4 relative to our
main index EF index, it is not surprising the lack of (or weak) significance of our results when
using firm and year fixed effects. These indices are proxies for firms’ EF culture; economic
theories suggest that a firm’s culture is specific to the firm and is largely fixed over long periods
(see e.g. Lazear, 1995; Kreps, 1990). While such culture can be changed, this process takes time;
as such, we expect our results to be driven primarily by cross-sectional differences in EF culture.
In Panel B of Table 4, we examine the impact of the individual components of the EF
index, based on: 1) Employment quality; 2) Health and safety; 3) Training; 4) Diversity, and 5)
Human rights. The results in Panel B show that except for the Health and safety index, all other
components of the EF index have a positive and significant impact on Tobin’s q.19 In terms of
economic magnitude, Human rights and Training have the largest impact. A one-standard-
deviation increase in Human rights (36.75) is associated with a 4.7% increase in Tobin’s q, while
a one-standard deviation increase in Training (28.2) is associated with a 3.3% increase in Tobin’s
q.20 These results suggest that our findings are not just a result of firms paying higher wages, nor
are they driven by firms that make the list of the “Best Companies to Work for.” Note that salaries
and the inclusion on the Best Companies to Work for list are subcomponents of Employment
quality. While Employment quality does have a positive impact on firm value, other indices have
a more significant impact on Tobin’s q.21 In column (6) of Panel B Table 4, we show results from
19 One plausible explanation for the lack of significance of the HS index is that health and safety concerns are of critical
importance only in a few industries (e.g. construction; oil exploration). As such, the average effect of the HS index
dissipates when including industries that place less emphasis on this component.
20 Based on the coefficient on Human rights (0.0023) in Model (5) of Panel B of Table 4, a one-standard deviation
increase in Human rights (36.75) is associated with a 0.085 increase in Tobin’s q, which represents a 4.7% increase
(0.085/1.8). Similarly, based on the coefficient on Training (0.0021) in Model (3) of Panel B, a one-standard deviation
increase in Training (28.2) is associated with a 0.059 increase in Tobins q, which represents a 3.3% increase
(0.059/1.8).
21 Based on the coefficient on Employment quality (0.0030) in Model (1) of Panel B, a one-standard deviation increase
in EQ index (14.35) is associated with a 0.043 increase in Tobin’s q, which represents a 2.4% increase (0.043/1.8).
14
regressions including all five of our index components. Only the coefficients on Training and
Human rights remain positive and statistically significant in these regressions; the coefficient on
Health and safety index switches sign and becomes negative and significant. The high correlation
between these variables likely explains the switch in sign of the coefficient on Health and safety
when we include all variables in the same regression. For example, the correlation between the
Human rights and Training is 0.57.
Overall, the results in Panels A and B of Table 4 lend support to the reciprocity view and
suggest that an EF culture is value enhancing.
III.2. Endogeneity in Employee-Friendliness and Firm Value
While our results suggest that an EF culture is associated with higher Tobin’s q, these
results do not establish causality. One potential concern deals with reverse causality; firms with
higher valuations may be able to spend more on their employees to create a more employee-
friendly working environment. In addition, there could be endogeneity bias caused by omitted
variables. If the omitted variable impacts both firm value and a firm’s ability to invest in EF
policies, our measure of employee-friendliness would not be exogenous to firm value, and the
coefficients from OLS regressions would be biased and inconsistent. While there is no perfect
solution to addressing endogeneity, we perform several tests to alleviate these concerns.
III.2.1. Two-stage Least Squares (2SLS) Estimation
In this section, we address endogeneity concerns by employing a 2SLS procedure using
instrumental variables for our measure of employee-friendliness. We use two instruments that
measure a country’s cultural values from Hofstede (1980). 1) Masculinity− a dimension of culture
15
that represents a preference in society for achievement, heroism, assertiveness, and material
rewards for success; its opposite, Femininity, stands for a preference for cooperation, modesty,
caring for the weak and quality of life; 2) Indulgence− stands for a society that allows relatively
free gratification of basic and natural human drives related to enjoying life and having fun. While
no instrument is perfect, our instruments satisfy both conditions of validity: the relevancy
condition and the exclusion restriction (we discuss the tests of validity below). Countries with low
values of Masculinity have a preference for cooperation, modesty, caring for the weak and quality
of life. We expect better employee treatment in countries with lower levels of masculinity. In
contrast, societies that score high on Indulgence value free gratification of basic and natural human
drives related to enjoying life and having fun. We expect better employee treatment in countries
that score high on indulgence. While these cultural norms could influence a firm’s employment
policies, such norms are unlikely to have a direct impact on firm performance. Our tests of validity
suggest that our instruments meet both the relevance and the exclusion restrictions.
Panel A of Table 5 shows results from the instrumental variable (2SLS) regressions. Model
(1) shows results from the first-stage OLS regressions using the EF index as the dependent
variable; we use the predicted values from the first-stage in the second-stage regressions (Model
(2)). Because there is no within country variation in our instruments, we do not use country-year
fixed effects in the first-stage regressions. Instead, we use region-year and industry-year fixed
effects.22 Our instruments exhibit significant explanatory power for firm-level employee-
friendliness. The coefficient on Masculinity is negative and significant, while Indulgence is
positive and highly statistically significant. The first-stage F-statistic (p-value of 0.000) rejects
22 We group countries into regions using the World Bank regions: East Asia and Pacific, Europe and Central Asia,
Latin America and the Caribbean, Middle East and North Africa, North America, South Asia, and Sub-Saharan Africa.
16
the null hypothesis that the instruments are jointly zero. In addition, the Hansen’s J-statistic
overidentification test (2) fails to reject the null hypothesis that the instruments are valid.23 In
addition, we report F-statistics from the Montiel-Olea-Pflueger (2013) robust weak instrument
tests, which clearly reject the null hypothesis that our instruments are weak. In Model (2) we
report results from the second-stage regressions and confirm our prior findings. Firms with higher
EF index tend to have higher value, even after correcting for endogeneity using the instrumental
variable approach. Using the coefficient from Model (2) in Panel A of Table 5, a one-standard
deviation increase in EF index –IV (12.72) is associated with an 8.5% increase in Tobin’s q.
In Models (3) and (4) of Panel A of Table 5 we show results from first- and second-stage
regressions using our alternate proxy, EF index-ASSET4. Results confirm our earlier findings and
are similar in both statistical significance and economic magnitude as those using the EF index.
III.2.2. Change Regressions
As an alternate way to address endogeneity concerns, we examine the causal effect between
changes in Tobin’s q and changes in EF index. To do so, we run OLS regressions using changes
in Tobin’s q (EF index) between t and t-1 as the dependent variable and use lagged changes in EF
index (Tobin’s q) as the key independent variables, along with all controls used in Equation 1
(measured as differences between t and t-1). We show results in Panel B of Table 5. We omit the
control variables to conserve space. In Models (1) and (2) we use ΔTobin’s qt, t-1 as the dependent
variable, while Models (3) and (4) use ΔEF Indext,t-1. The results in Models (1) and (2) show that
past changes in EF index are associated with future changes in Tobin’s q. The p-value of the Wald
23 The p-value from Hansen’s J-test statistic are 0.638 and 0.610 in regressions using EF index and EF index-ASSET4,
respectively.
17
tests rejects the null that lagged values of ΔEF Index are jointly equal to zero.24 This suggests that
lagged changes in EF index have a causal effect on Tobin’s q. In contrast, results in Models (3)
and (4) show that lagged changes in Tobin’s q have no significant impact on the ΔEF index. The
Wald tests fail to reject the null that lagged values of ΔTobin’s q are jointly equal to zero. Overall,
the results show that while there is a causal effect of changes in employee-friendliness on Tobin’s
q, past changes in Tobin’s q have no significant impact on EF index, which mitigates concerns
about reverse causality.
III.3. Exogenous Shocks and the Employee-Friendliness-Firm Value Relation
III.3.1. Employee-Friendliness and Firm Value following a Shock to Labor Markets
While the results thus far corroborate our main finding that more employee-friendly firms
are valued higher, in this section we use an alternate approach to assess the conditions in which an
EF culture adds value. Specifically, we examine whether an EF culture matters in periods of crisis,
when firms may need to demand more from their employees or be tempted to eliminate benefits
provided to them. We use the financial crisis of 2008-2009 as an exogenous shock, since this
event was likely unanticipated by most of the firms in our sample. In addition, the financial crisis
represents a major shock to labor markets around the world. For example, the unemployment rate
in Spain rose from 8.6% to 18.1% between 2007 and 2009. Other countries experienced similar
shocks to unemployment.25 The crisis also directly impacted employees. A 2012 Wharton Study
from the Management group documents a drop in employee-loyalty following the crisis. The
article mentions that “… some employees are clearly feeling disconnected from their work. Among
the reasons cited for this: the recession, during which companies laid off huge swaths of their
24 In Models (1) and (2) of Panel A of Table 5, the p-values of the Wald test are 0.027 and 0.031, respectively.
25 We collect unemployment figures from OECD: https://data.oecd.org/unemp/unemployment-rate.htm and
complement these using data from the International Labor Organization Statistics and Databases (ILOSTATS).
18
employees with little regard for loyalty or length of service; a whittling away of benefits, training
and promotions for those who remain.” The drop in employee-loyalty is likely to be less severe
in firms that treat their employees better (those with a better EF culture). This should be reflected
in improved performance for such firms during and after the crisis relative to firms that place little
value on their employees.
To analyze whether an EF culture matters for firm value during a crisis, we first classify
firms as high (low) employee-friendliness based on their value of EF index as of 2006.26 Firms
with values of EF index as of 2006 in the top third of the distribution in their country are classified
as High EF index. Using this indicator variable, we employ a DiD methodology using a propensity
score- (PSM) matched control group of firms with low values of the EF index as of 2006. To
identify the control group, we first run a probit model to calculate propensity scores using the
indicator variable High EF index and the full sample of firms with available data as of the end of
2006. We use the full set of firm-level control variables in our regressions and include country
and industry fixed effects in our estimation. We then match each treated firm (High EF index=1)
to a control firm using the nearest neighborhood method with replacement.27 To examine whether
the post-financial crisis performance differs between firms with high and low ex-ante employee-
friendliness, we run several specifications of the following regression:
 
     ∑ ∑   (3)
26 We rank firms based on their pre-crisis employee-friendly culture values as of 2006. Although the height of the
crisis happened in 2008, many firms were affected (and possibly made changes) in the latter part of 2007. As such,
using values of culture as of the end of 2007 may pick up adjustments made by firms due to the deteriorating economic
conditions.
27 In our online appendix, we show results from the Probit regressions, along with various tests to assess the quality
of our PSM matching procedure. The results show that after matching the normalized differences (x) for all control
variables between treatment and control firms are all well within the recommended 0.25 threshold (Imbens and
Wooldridge (2009)); the highest value for x is for age (0.10).
19
where High EF is an indicator variable for out treated firms – those with values of our employee-
friendliness proxies as of year-end 2006 in the top 30% of the distribution; Post is an indicator
variable that equals one for years after 2008 and zero otherwise. All of the other control variables
are the same ones used and discussed previously. The results from the estimation of Equation 3
are shown in Panel A of Table 6.
The key assumption in our identification strategy is that High EF index firms and Low EF
index firms follow similar trends prior to the crisis. To test this parallel trends assumption
underlying the DiD design, in Model (2) we include interactions between the High EF index
variable and an indicator variable, Pre, which is equal to one for the two years leading up to the
crisis (2005 and 2006) and zero otherwise. The results show an insignificant coefficient on the
interaction term (Pre x High EF index), which suggests that High and Low EF index firms followed
similar trends prior to the crisis. Importantly, the results in Models (1) and (2) show a positive and
significant coefficient on the interaction term (Post x High EF index), indicating that on average
firms with higher ex-ante EF index have higher Tobin’s q in the post-crisis years. The results in
Model (1) in Panel A of Table 6 show that relative to their matched firms High EF index firms
have Tobin’s q that are on average 5.3% higher in the post-crisis period.28
We use the financial crisis as an exogenous shock to the labor market and the results show
that firms with a more EF culture seem to fare better around this shock. However, the effects of
the crisis on the labor market (i.e. the magnitude of the shock) differs widely across countries.29
Next, we incorporate the magnitude of the shock into our analysis. Specifically, we run regressions
28 The coefficient on the interaction term Post x High EF index in Model (1) is 0.090. The mean Tobin’s q for firms
in the sample is 1.71. Thus, in the post-crisis period, High EF index firms have Tobin’s q that is 0.053 (0.090/1.71)
higher.
29 For example, the unemployment rate more than doubled between 2007 and 2009 in the US, while in Germany, it
went down slightly from 8.6% to 7.7%.
20
using an expanded version of Equation 3 that includes interactions with an indicator variable, High
Impact, that is equal to one for countries with an above-median change in the unemployment rate
between 2007 and 2009, and zero otherwise. Our variable of interest is the coefficient on the triple
interaction term, Post x High EF index x High Impact. We expect this coefficient to be positive
and significant if, as expected, an EF culture is more valuable in periods of large shocks to the
labor markets. The results shown in Models (3) and (4) of Panel A of Table 6 confirm that more
employee-friendly firms tend to outperform other firms during periods of crisis when the labor
markets were impacted more than the median. The results are both economically and statistically
significant. Relative to matching firms, High EF index firms in High Impact countries have a
Tobin’s q that is on average 17.3% higher in the post-crisis period.30 The magnitude and statistical
significance of the impact are significantly larger than those found in Model (1) in which we did
not account for differences in the magnitude of the impact of the crisis on the labor market.
The results in this section suggest that firms with a more EF culture perform better during
periods of crisis. In such periods, treating employees well appears to be beneficial. These results
add further support to our main findings and suggest that having an EF culture is value enhancing,
especially in periods of crisis when firms may need to demand more from their employees, ceteris
paribus.
III.3.2. Employee-Friendliness and Firm Value around Parental Leave Laws
As a final way to assess whether employee-friendliness impacts firm value, we exploit the
staggered implementation of parental leave laws during our sample period across several European
countries. Specifically, we examine the implementation of the Parental Leave Directive 2010/18
30 The coefficient on the interaction term Post x High EF index x High Impact in Model (3) is 0.296. The mean
Tobin’s q for firms in the sample is 1.71. Thus, in the post-crisis period, High EF index firms in High Impact countries
have Tobin’s q that are 17.3% (0.296/1.71) higher.
21
across Europe. The staggered implementation of the directive allows us to better assess the causal
effects of a regulation that aims to improve employee treatment by improving the quality of
parental leave policies.31
Using the enactment of these laws, we employ a DiD design. An advantage of this
approach is that countries enacted parental leave regulations at different points in time, which helps
with our identification strategy.32 This approach implicitly takes as the benchmark group all firms
from countries that did not enact such regulations as of a particular time (Bertrand and
Mullainathan, 1999, 2003). Specifically, we estimate various specifications of the following
regressions:
 
     ∑ ϑ (4)
Treat is an indicator variable that is equal to one for firms in countries that implemented
the Parental Leave Directive and zero otherwise. Post is an indicator that is equal to one starting
the year after the enactment of the Parental Leave Directive in the country and zero otherwise.33
We use the same set of controls employed and discussed preciously. i and t refer to firm and
year fixed effects. We use firm and year fixed effects to identify the within firm and within year
change in valuation between treatment and control firms after the enactment of the parental leave
31 Among others, the directive extended the minimum period that parents can take parental leave from three to four
months “The Parental Leave Directive further provides protection from discrimination for workers on the grounds of
applying for or taking of parental leave and stipulates that, at the end of the leave, workers have the right to return to
the same job or, if that is not possible, to an equivalent or similar job consistent with their employment contract or
employment relationship” (Palma Ramalho, Foubert, and Burri (2015)). Palma Ramalho, et al. (2015) provide details
on the implementation of Parental Directive 2010/18 across 33 European countries.
32 We have 13 countries that implemented the Parental Leave Directive during our sample period: Belgium (2012),
Denmark (2013), Finland (2011), France (2012), Germany* (2007), Greece (2012), Ireland (2013), Italy (2013),
Luxembourg (2013), Netherlands (2011), Norway (2011), Poland (2013), and United Kingdom (2013). *Germany
passed the Federal Law on Parental Allowance and Parental Leave in 2007. The enactment of such law made Germany
compliant with the Parental Leave Directive 2010/18. We thus include Germany in our sample of treated countries
adopting Parental Law. Excluding Germany from our treatment group does not affect our results.
33For countries in the control group we set Post equal to zero. In robustness tests, we set Post equal to one for years
after 2011 for countries in the control group. Results (untabulated) are similar when using this alternate approach.
22
laws. To address the parallel trend assumption underlying our DiD design, we include a variable,
Pre that is equal to one in the two years leading up to the enactment of the law and zero otherwise,
as well as its interaction with Treat. If the parallel trend assumption holds, the interaction term
(Pre x Treat) should be insignificant, which would indicate that in the absence of treatment, the
treatment and control firms follow similar trends.
We show results in Panel B of Table 6. In Models (1) and (2) we assess the average effect
of the enactment of parental leave laws. We include all control variables used and discussed
previously, but do not report them to conserve space. The results suggest that firm value increases
following the enactment of parental leave laws. The results show that Tobin’s q on average
increases by 7% following the enactment of the parental leave laws for firms in our treatment
group of countries.34 In Model (2), we include the Pre indicator, and observe that the interaction
term (Pre x Treat) is insignificant, but our variable of interest, the coefficient of the interaction
term, Post x Treat remains positive and statistically significant.
The average effects of the enactment of parental leave laws on firm value may be driven
by simultaneous regulations and other confounding events in these countries. To more directly
examine the effect of parental leave laws, we next examine changes in Tobin’s q for firms that are
most likely to be impacted by the enactment of parental leave laws. Firms with generous parental
leave policies are unlikely to be affected by the enactment of the Parental Leave Directive.
However, those firms with stingy parental leave policies may need to change their policies to
comply with the reforms. Since the quality of parental leave is a subcomponent of Diversity, we
34 The coefficient on Post x Treat in Model (1) of Panel B of Table 6 is 0.126. This represents an increase of 7.0%
(0.126/1.80) in Tobin’s q.
23
expect that firms with low scores on Diversity prior to the reforms are most likely to be impacted
by the enactment of the directive.
To perform this test, we estimate Equation 4 using our treatment sample of firms and
include an interaction between Post and Most Impacted, which is an indicator that is equal to one
for firms with a Diversity index in the bottom quartile in their county as of year t-1 relative to the
enactment of the parental leave law and zero otherwise. Our sample size is greatly reduced
because of the need to have data as of the year before the enactment of the parental leave law. We
report results in Models (3) and (4) of Panel B of Table 6. The results show a positive and
significant coefficient on the interaction term Post x Most impacted. The results show that the
changes in firm value after the enactment of the parental leave laws is higher for firms that are
most impacted. This adds support to the view that through the improvement in parental leave laws
(and employee treatment in general), is value enhancing.
In Models (5) through (8) of Panel B of Table 6 we replicate our results from Models (1)-
(4) using a PSM-matched sample, to better control for differences between firms in our treatment
group and those in our control group. To identify the control group, we use a PSM matching
approach. We first run a probit model to calculate propensity scores using an indicator variable,
Treat, which is equal to one for firms from firms from the 13 European countries that adopted
parental leave laws during our sample period, as our dependent variable. We show the full set of
results from these probit regressions in Appendix C. Using the propensity scores, we match each
treatment firm with a firm from the control group of countries using the nearest neighborhood
method (we employ a 1:1 matching with replacement).
To assess the quality of our matching approach, we run several tests. First, we rerun the
above probit regression using the matched sample. Model (2) of Appendix C shows these results.
24
The results show that after matching, none of the independent variables are statistically significant.
In addition, the Pseudo R2 drops from 0.28 in Model (1) to 0.01 in the post-match sample (Model
2). To more directly assess the quality of our matching, in Panel B of Appendix C we compare
the values of control variables between our treatment firms and the control firms pre- and post-
match. Following Imbens and Wooldridge (2009) and Focke, Maug, and Niessen-Ruenzi (2017),
we compare firms based on normalized differences (x).35 As Imbens and Wooldridge (2009)
argue, using normalized differences addresses problems associated with t-statistics when there are
large differences in the means of two distributions. The results for the full sample in Panel B of
Appendix C show that firms in our treatment sample of European countries tend to have a larger
proportion of foreign sales and are more likely to be cross-listed. For the matched sample, the
normalized differences (x) are all within the recommended 0.25 threshold (Imbens and
Wooldridge (2009)). Overall, our tests suggest that the PSM matching procedure yields a
comparable set of treatment and control firms.
We show results from the estimation of Equation 4 using the PSM-matched sample in
Models (5)-(8). In Models (5) and ( 6) we use the full sample of treatment firms and their respective
matches. In Models (7) and (8), we obtain matches only for firms that are most likely to be
impacted by the enactment of the parental leave law (Most Impacted).36 Our results show that
after the enactment of parental leave laws in their country, firms experience an increase in value
relative to their PSM-matched firms. Taking the coefficient in Model (6) as an example, following
the enactment of parental leave laws, firms in our treatment sample have Tobin’s q that is 7.8%
35 x = ̅̅/
; where ̅(̅ is the sample mean of the covariates for treatment (control) firms, and
() is the estimate of the variance.
36 We match each Most impacted firm with a firm from the control group of countries. To find matches, we use
propensity scores from probit regressions, as before, using Most impacted indicator as the dependent variable. We
report results from these regressions pre- and post-match in Models (3) and (4) of Appendix C.
25
higher than that of their PSM-matched control firms.37 For Most impacted firms the magnitude of
the impact is significantly larger. From Model (8), following the enactment of parental leave laws,
Most impacted firms have Tobin’s q that is 22% higher than that of their PSM-matched firms.38
Overall, the results in this section suggest that employee treatment is value enhancing.
Reforms aimed to improve employee treatment through enhancing parental leave policies are
associated with subsequent increases in valuation, especially for firms that are most likely to be
affected by the reform.
[Insert Table 6 Here]
III.4. Employee-Friendly Culture and Productivity
In this section we investigate the channels through which an EF culture may impact firm
value. The reciprocity view argues that better employee treatment should encourage workers to
be more productive, which may help explain the observed improvements in firm value. If this
mechanism exists, we should observe that more employee-friendly firms have workers that are
more productive. Firms with more motivated and driven employees should be able to maximize
their earnings potential and improve technical efficiency by making better products, delivering
better services, and potentially lowering costs. This should ultimately impact firm performance
and firm value. To explore this hypothesis, we use two measures of technical efficiency from
previous literature (see e.g. Loderer, Stulz, and Waelchli, 2014): 1) Sales-to-assets and 2) COGS-
37 From Model (6) in Panel B of Table 6, the coefficient on the interaction term Post x Treat is 0.141, which is 7.8%
of the mean Tobin’s q for this subsample (1.81).
38 From Model (8) in Panel B of Table 6, the coefficient on the interaction term Post x Most impacted is 0.413, which
is 21.97% (0.413/1.88) of the mean Tobin’s q (1.88) for this subsample.
26
to-employees (log)−cost of goods sold per employee. We also examine the impact on a proxy for
innovation, the number of patents.39
Per our hypothesis, we expect that firms with a more EF culture to have higher asset
turnover (higher Sales-to-assets), lower costs (lower COGS-to-employees), and more innovation.
We report results from these regressions in Table 7. To address endogeneity concerns, we also
present results from 2SLS regressions in which we instrument EF index using our two instruments
based on country culture: Masculinity and Indulgence. In all of the regressions we control for
various factors that have been shown to affect technical efficiency and innovation including: firm
age, size, capital expenditures, leverage, R&D expenses-to-sales, market-to-book ratio, volatility,
and profitability (ROA). All control variables are lagged one year.
[INSERT TABLE 7]
In Models (1), (3), and (5), and (7) of Table 7 we report results from OLS regressions. In
Models (2), (4), and (6), we report results from our 2SLS regressions. We find results consistent
with the reciprocity view. Firms with higher EF index are associated with improved technical
efficiency and innovation. Taking the coefficients in Model (1), a one-standard-deviation increase
in EF index is associated with a 10.54 (20.95 x 0.503) increase in Sales-to-assets, which represents
an 11.4% increase relative to its mean (92.1). We find similar results, albeit of smaller magnitude
when using costs of goods sold per employee and patents. Results continue to hold when we
instrument our EF index.
39 Unfortunately, data availability for patents is limited to a small subsample of firms (1,544 firm-year observations).
This limits the generalizability of our results. Our findings, however, are consistent with existing evidence in US
studies that document a positive impact of employee treatment on innovation (e.g. Chen et al., 2016).
27
Overall, the results in this section are consistent with our hypothesis that firms with more
EF culture encourage employees to work harder (and thus be more innovative) and this increased
effort appears to improve efficiency, profitability, and ultimately firm value.
III.5. Country Characteristics and Employee-Friendly Culture
In this section, we analyze how the impact of an EF culture on firm value is affected by
country characteristics. We assess how the level of competition in the country as well as the
productivity of the labor force may impact the relation between employee friendliness and firm
value. An EF culture could serve to attract or retain better employees; this would be more
important in more competitive markets. In addition, retaining or attracting employees may be
more important in countries with a more productive labor force, in which all else equal, workers
may have better and more outside opportunities for employment in the country or abroad.
To test this hypothesis we use two measures of competition based on the level of industry
concentration in a country: 1) HHI- Herfindahl index and 2) Concentration- sales of the top three
industries in the country, as a fraction of total sales across all firms in a country as a proxy for
competitiveness.40 We also use Output – the output (GDP) per worker obtained from the
International Labor Organization (ILO) Labor Statistics as a proxy for labor productivity. Using
these variables, we create indicator variable High Competition that is equal to one if the HHI
(Concentration) measure is in the top quartile of the distribution in a given year and zero otherwise.
Similarly, we create an indicator variable (High Productivity) using the Output measure of labor
40 Specifically, the Herfindahl index (HHI) for country c in year t is computed as: , ,,
, .

Industry salesj,c,t refers to total sales for all firms in industry j in country c in year t; Total salesc,t, is the total sales for
all firms in country c in year t. We use the 2-digit SIC code to compute HHI and Concentration.
28
productivity. We then interact these measures with our EF culture variables. We report results
from OLS and 2SLS regressions in Table 8.
The results show that an EF culture is value enhancing in competitive (less concentrated)
countries, and in those countries with a more productive labor force. As an example, taking the
coefficient in Model (1), the impact of an EF culture is not significant in less competitive (highly
concentrated) countries. In contrast, a one-standard deviation increase in EF index is associated
with a 5.9% increase in Tobin’s q in highly competitive markets.41 Results using the alternate
proxy for competition (Concentration) are similar in magnitude, although the impact on less
competitive countries is weakly significant. The results also show that an EF culture is more value
enhancing in countries with a more productive labor force. From the coefficients in Model (3), a
one-standard deviation increase in EF index is associated with a 2.3% increase in Tobin’s q
(marginally significant) in countries with a less productive labor force, but a much larger 5.8%
increase in Tobin’s q in countries with a highly productive labor force.42 The results from 2SLS
regressions continue to hold. The impact of an EF culture on firm value is no longer statistically
significant in less competitive (less productive) countries after we instrument our EF index.
IV. Additional Robustness Tests
In Table 9 we present alternative specifications from our main valuation regression results
found in Panel A of Table 4. Specifically, in Model (1) we report results from regressions in which
41 From the coefficients in Model (1) of Table 8, a one-standard deviation increase in EF index (21.18) is associated
with a 0.106 [(0.002+0.003)x 21.18] increase in Tobin’s q in countries with High competition, which represents a
5.9% increase relative to its mean (1.80).
42 From the coefficients in Model (3) of Table 8, a one-standard deviation increase in EF index (20.9 for this sample)
is associated with a 0.042 (0.002 x 20.9) increase in Tobin’s q in countries with a less productive labor force, or 2.3%
of its mean (1.8). The impact is larger, 0.105 [(0.002+0.003) x 20.9] in countries with a highly productive labor force,
which represents 5.8% relative to its mean.
29
we exclude US firms from the sample, as they account for roughly 24% of the sample. The results
here are very similar in significance and magnitude as those in Table 4, Panel A. To examine
whether our results are driven by firms that are included in the list of “Best Companies to Work
for” (BC firms) used in prior studies, in Model (2) we run regressions excluding BC firms. The
results continue to hold when excluding BC firms. Further, as an additional robustness test we
calculate value using market-to-book value of equity instead of Tobin’s q. We report results using
market-to-book for the full sample, as well as excluding US firms in Models (3) and (4),
respectively. We find similar results to those reported earlier. Finally, to examine whether our
results are affected by the changing composition of our sample of firms, since ASSET4 database’s
coverage improves throughout our sample period, in Models (5) and (6) of Table 9 we show results
in which we restrict the sample to firms with available data for the full sample period using Tobin’s
q and market-to-book, respectively.43 The results using this subsample of firms corroborate our
main findings. In an additional test (reported in our internet appendix), we also use a measure of
profitability, Operating ROAoperating income divided by total assets, as an additional dependent
variable. We find that an EF culture has a positive impact on Operating ROA as well.
Also, in our internet appendix, we replicate results in Panel A using 2SLS regressions in
which we instrument EF index with our instruments Masculinity and Indulgence. All results
continue to show positive and significant coefficients on EF index IV. The one exception is Model
(5); in this model, we limit our sample to firms with available data for the full sample period, the
coefficient on EF index IV is positive, but not statistically significant (t-statistic of 1.39).
[Insert Table 9 Here]
V. Conclusion
43 This also avoids any survivorship bias that may exist in our earlier regression results.
30
Anecdotal observation suggests that some firms are starting to offer more perks to
employees in an attempt to create a more employee-friendly culture. We examine the economic
rationale behind this behavior and explore the conditions in which improving firm culture is value
enhancing for shareholders. Overall, we show that firms with a more EF culture (for example by
providing more benefits and training, and equal opportunities for advancement) have higher
valuations and perform better. Specifically, we find that firms with higher EF index (our proxy
for firm-level culture) have higher value (Tobin’s q). The impact on firm value is stronger in
countries with more competitive industry structures and more productive labor force. Our results
suggest that an EF culture adds value via enhanced employee motivation, which encourages
employees to become more efficient. Quasi-natural experiments suggest that the effect of EF
policies on firm value is causal. Specifically, we find that firms with a more employee-friendly
culture perform better during the global financial crisis.
We also document that the enactment of parental leave laws aimed to enhance parental
leave policies (and thus improve employee treatment) across Europe increase firm value,
especially for firms that are most affected by such laws (i.e. firms most likely to make changes to
improve parental leave policies). Finally, we document that employee treatment adds value
through improved efficiency (i.e. higher sales-to-assets; lower costs) and higher profitability
(Operating ROA).
31
References
Aggarwal, R., I. Erel, R. Stulz, and R. Williamson, 2009, Differences in governance practices
between U.S. and foreign firms: Measurement, causes, and consequences, Review of Financial
Studies 22, 3131–3169.
Akerlof, G.A., 1982, Labor contracts as partial gift exchange, The Quarterly Journal of Economics
97, 543-569.
Bae, K.H., J.K. Kang, and J. Wang, 2011, Employee treatment and firm leverage: A test of the
stakeholder theory of capital structure, Journal of Financial Economics 100, 130-153.
Bénabou, R., J. Tirole, 2010, Individual and corporate social responsibility, Economica 77, 1-19.
Berk, J., R. Stanton, and J. Zechner, 2010, Human capital, bankruptcy, and capital structure,
Journal of Finance 65, 891–926.
Bertrand, M., S. Mullainathan, 1999, Is there discretion in wage setting? A test using takeover
legislation, Rand Journal of Economics 30, 535-554.
Bertrand, M., S. Mullainathan, 2003, Enjoying the quiet life? Corporate governance and
managerial preferences. Journal of Political Economy 111, 1043-1075.
Brown, H. (2007, October 28). Gap Admits Possible Child Labor Problem.” ABC News. Retrieved
from http://abcnews.go.com.
Cappelli, P., K. Chauvin, 1991, An interplant test of the efficiency wage hypothesis,” The
Quarterly Journal of Economics 106, 769-787.
Carlin, B. I., S. Gervais, 2009, Work ethics, employment contracts, and firm value, Journal of
Finance 64, 785-821.
Chang, X., K. Fu, A. Low, and W. Zhang, 2015, Non-executive employee stock options and
corporate innovation, Journal of Financial Economics 115, 168-188.
Chen, C., Y. Chen, P-H. Hsu, and E. J. Podolski, 2016, Be nice to your innovators: Employee
treatment and corporate innovation performance, Journal of Corporate Finance 39, 78-98.
Cronqvist, H., R. Heyman, F., Nilsson, M., Svaleryd, H, Vlachos, J., 2009, Do entrenched
managers pay their workers more? Journal of Finance 64, 309-339.
Dehejia, R., S. Wahba. 2002, Propensity score-matching methods for nonexperimental causal
studies. The Review of Economics and Statistics 84, 151–161.
32
Doidge, C., G.A. Karolyi, and R. Stulz, 2004, Why are foreign firms listed in the U.S. worth
more? Journal of Financial Economics 71, 205–238.
Edmans, A., L. Li, and C. Zhang, 2015, Employee satisfaction, labor market flexibility, and stock
returns around the world, London Business School Working Paper.
Edmans, A. 2011, Does the stock market fully value intangibles? Employee satisfaction and equity
prices, Journal of Financial Economics 101, 621-640.
EU, 2004. The takeover bids directive assessment report. European Commission, Brussels,
Belgium.
Focke, F., Maug. E., and Niessen-Ruenzi, A., 2017, The impact of firm prestige on executive
compensation, Journal of Financial Economics, 123, 313-336.
Gompers, P., J. Ishii, and A. Metrick, 2010, Extreme governance: An analysis of dual-class firms
in the United States, Review of Financial Studies 23, 1051–1088.
Hofstede, G., 1980. Culture’s consequences−international differences in work-related values
(Sage, London).
Holzer, H.J., L.F. Katz, and A. Krueger, 1991, Job queues and wages, The Quarterly Journal of
Economics 106, 739-768.
Imbens, G. W., J. Wooldridge, 2009, Recent developments in the econometrics of program
evaluation. Journal of Economic Literature 47, 5–86.
Jensen, M., W. Meckling, 1976, Theory of the firm: managerial behavior, agency costs, and
ownership structure, Journal of Financial Economics 3, 305-360.
Kim, E.H., and P. Ouimet, 2014, Broad-based employee stock ownership: Motives and outcomes,
Journal of Finance 69, 1273-1319.
Knowledge@Whaton, 2012, Declining Employee Loyalty: A Casualty of the New Workplace.
May 9.
Krüger, P., 2015, Corporate goodness and shareholder wealth, Journal of Financial Economics
115, 304-329.
Kube, S., M.A. Maréchal, and C. Puppe, 2012, The currency of reciprocity: Gift exchange in the
workplace,” American Economic Review 102, 1644-1662.
Landier, A., V. Nair, and J. Wulf, 2009, Trade-offs in staying close: corporate decision making
and geographic dispersion, Review of Financial Studies 22, 1119-1148.
33
Liang H., and L. Renneboog, 2017, On the Foundations of Corporate Social Responsibility,
Journal of Finance 72, 853-910.
Loderer, C., R. Stulz, and U. Waelchli, 2014, Limited managerial attention, innovation, and
corporate aging, NBER Working Paper.
Mao, C. X. and J. Weathers, 2015, Employee treatment and firm innovation, Fox School of
Business Research Paper No. 15-071. Available at SSRN: https://ssrn.com/abstract=2648588 or
http://dx.doi.org/10.2139/ssrn.2648588.
Mas, A., 2006, Pay, reference points, and police performance, The Quarterly Journal of Economics
122, 783-821.
Montiel Olea, J. L., and C. Pflueger, 2013, A Robust Test for Weak Instruments, Journal of
Business & Economic Statistics 31, 358-369.
Ouimet, P., and E. Simintzi, 2015, Wages and firm performance: Evidence from the 2008 financial
crisis, University of North Carolina Working Paper.
Pagano, M., and P.F. Volpin, 2005, Managers, workers, and corporate control, Journal of Finance
60, 841-868.
Palma Ramalho, M. R., Foubert, P., and Burri S., 2015. Report on the implementation of parental
leave directive 2010/18 in 33 European countries. European Network of Legal Experts in the Field
of Gender Equality.
Smith, A (2016, June 1st). Report slams Walmart for 'exploitative' conditions in Asia factories.
CNN Money. Retrieved from http://money.cnn.com/2016/05/31/news/companies/walmart-gap-
hm-garment-workers-asia/index.html.
The Atlantic, 2014, A New Business Strategy: Treating Employees Well. November 26.
Wernau, J. (2015, July 30). Child Labor On The Rise in West Africa as Demand for Cocoa
Grows. Wall Street Journal. Retrived from: https://blogs.wsj.com.
34
Table 1: Sample Distribution across Countries
The table reports the number of firms, total number of observations, and the first year of available data for firms in
the country. Our sample includes all firms covered by Thomson Reuters’ ASSET4 database. We exclude financial
firms and utilities (SIC codes between 6000 and 6999 and between 4900 and 4949) and firms with missing data on
total assets, as well as those with negative sales or negative book value of equity. We require countries to have three
years of data on at least three firms. Our sample consists of 3,446 firms (21,103 firm-year observations) from 43
countries from 2003 through 2014.
Countr
y
First Yea
r
# of Firms # of Observations
Australia 2003 307 1
,
312
Austria 2003 13 104
Bel
iu
2003 18 141
Bermuda 2005 10 49
Brazil 2008 58 201
Canada 2003 240 1
,
233
Chile 2009 12 61
China 2005 119 478
Colombia 2011 5 16
Denmar
k
2003 21 173
E
gypt
2012 8 20
Finlan
d
2003 25 223
France 2003 82 692
German
y
2003 76 485
Greece 2003 15 79
Hon
g
Kon
g
2003 86 511
India 2008 64 257
Indonesia 2009 25 80
Irelan
d
2003 28 212
Israel 2010 11 39
Ital
y
2003 27 219
Ja
p
an 2003 351 2
,
806
Luxembour
g
2005 9 55
Mala
y
sia 2009 34 110
Mexico 2009 27 64
N
etherlands 2003 40 237
N
ew Zealan
d
2005 10 65
N
orwa
y
2003 17 149
Phili
pp
ines 2011 9 24
Polan
d
2010 12 41
Portu
g
al 2003 8 68
Russian Federation 2008 28 135
Sin
g
a
p
ore 2005 40 273
South Africa 2009 95 263
South Korea 2005 86 353
S
p
ain 2003 30 232
Sweden 2003 42 355
Switzerlan
d
2003 63 486
Taiwan 2009 118 473
Thailan
d
2009 20 71
Turke
y
2009 17 67
United Kin
g
dom 2003 298 2
,
102
United States 2003 842 6
,
089
TOTAL 3
,
446 21
,
103
35
Table 2: Descriptive Statistics
The table shows descriptive statistics for our main variables. Our sample consists of 3,446 firms (21,103 firm-year
observations) from 43 countries from 2003 through 2014. Financial and stock market data are obtained from
Thomson’s WorldScope and DataStream. Data on our measures of employee-friendliness are obtained from ASSET4
database. Variable definitions are found in Appendix A.
Descriptive Statistics
N Mean 25th. pctl. Median 75t pctl. Std. dev.
Firm-level variables:
EF Index % 21,103 38.58 21.05 36.84 55.56 21.18
EF index-ASSET4 % 21,103 51.65 31.43 51.71 71.94 23.71
Employment quality (%) 21,103 33.46 25.00 25.00 42.86 14.35
Training and development (%) 21,103 40.47 25.00 50.00 50.00 28.20
Diversity (%) 21,103 49.19 33.33 50.00 66.67 30.03
Health and safety (%) 21,103 45.25 0.00 50.00 75.00 36.95
Human rights (%) 21,103 32.00 0.00 16.67 66.67 36.75
Tobin's q 21,103 1.80 1.11 1.44 2.07 1.12
Operating ROA % 21,089 8.98 4.36 8.04 12.80 9.95
Size 21,103 22.27 21.38 22.22 23.17 1.38
Log Age 21,103 3.03 2.48 2.94 3.56 0.86
Leverage 21,103 23.09 10.56 22.28 33.40 15.98
Cash-to-assets % 21,103 8.05 1.65 5.21 11.37 8.91
PP&E-to-sales % 21,103 113.06 28.95 57.32 122.10 169.42
Foreign sales-to-sales % 21,103 37.31 1.50 33.73 64.05 32.56
R&D expenses-to-sales 21,103 2.49 0.00 0.04 2.40 5.27
Capex-to-assets % 21,103 5.78 2.19 4.16 7.35 5.47
ROA % 21,103 6.82 3.18 6.37 10.49 8.13
Closely-Held % 21,103 25.61 3.46 19.51 42.73 23.46
ADR 21,103 0.19 0.00 0.00 0.00 0.40
Country-level variables:
Log GDP per capita 21,103 10.50 10.49 10.70 10.82 0.65
GDP growth 21,103 1.90 1.12 2.19 2.88 2.69
36
Table 3: Correlations
The table shows correlation among variables used in our analysis. * indicates that the correlation is significant at least at the 10% level. See Appendix A for variable
definitions.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (21)
(1) 1
(2) 0.83* 1
(3) 0.46* 0.38* 1
(4) 0.78* 0.72* 0.32* 1
(5) 0.66* 0.50* 0.17* 0.40* 1
(6) 0.71* 0.55* 0.20* 0.48* 0.50* 1
(7) 0.85* 0.72* 0.28* 0.57* 0.38* 0.46* 1
(8) -0.06* -0.04* 0.03* -0.06* -0.02* -0.12* -0.04* 1
(9) 0.03* 0.06* 0.04* 0.04* 0.02* -0.03* 0.03* 0.50* 1
(10) 0.37* 0.45* 0.11* 0.33* 0.19* 0.20* 0.38* -0.29* -0.06* 1
(11) 0.17* 0.18* 0.04* 0.11* 0.17* 0.12* 0.15* -0.02* 0.05* 0.16* 1
(12) 0.03* 0.05* -0.02* 0.03* 0.03* 0.05* 0.03* -0.24* -0.14* 0.26* 0.00 1
(13) -0.07* -0.10* 0.00 -0.05* -0.08* -0.07* -0.05* 0.21* 0.05* -0.21* -0.10* -0.28* 1
(14) -0.07* -0.10* -0.03* -0.10* -0.01 0.03* -0.10* -0.12* -0.26* -0.07* -0.09* 0.10* -0.10* 1
(15) 0.27* 0.27* 0.10* 0.21* 0.12* 0.17* 0.30* 0.02* 0.01* 0.10* 0.05* -0.05* 0.11* -0.10* 1
(16) 0.01 0.03* 0.01* -0.02* 0.09* -0.03* 0.01 0.22* -0.05* -0.05* 0.02* -0.19* 0.24* -0.09* 0.20* 1
(17) -0.02* -0.04* -0.01* -0.03* -0.01 0.05* -0.04* 0.02* 0.00 -0.09* -0.10* 0.04* -0.09* 0.41* -0.05* -0.16* 1
(18) 0.03* 0.05* 0.06* 0.06* -0.02* -0.03* 0.03* 0.44* 0.61* -0.08* 0.01* -0.19* 0.08* -0.25* 0.02* -0.07* 0.03* 1
(19) -0.12* -0.10* -0.02* -0.01* -0.22* -0.11* -0.08* 0.00 -0.03* -0.05* -0.25* 0.02* 0.03* 0.04* -0.08* -0.13* 0.06* 0.05* 1
(20) 0.20* 0.27* 0.07* 0.20* 0.04* 0.10* 0.22* -0.07* -0.04* 0.25* 0.06* 0.05* -0.01 0.02* 0.19* 0.01 0.02* 0.00 0.09* 1
(21) 0.02* 0.00 0.01* -0.10* 0.21* 0.02* -0.04* -0.04* -0.02* -0.05* 0.13* -0.03* 0.01* 0.03* 0.14* 0.12* -0.06* -0.11* -0.38* -0.10* 1
(22) -0.14* -0.10* -0.05* -0.08* -0.17* -0.13* -0.09* 0.12* 0.05* -0.04* -0.05* -0.02* 0.03* 0.04* -0.06* -0.04* 0.02* 0.07* 0.15* 0.00 -0.35*
(1) EF Index (7) Human rights (%) (13) Cash-to-assets (19) Closely-Held %
(2) EF –Index ASSET4 (8) Tobin's q (14) PP&E-to-sales (20) ADR
(3) Employment quality (%) (9) Operating ROA (15) Foreign sales-to-sales (21) Log GDP per capita
(4) Training and development (%) (10) Size (16) RD-to-sales (22) GDP growth
(5) Diversity (%) (11) Log Age (17) Capex-to-assets
(6) Health and safety (%) (12) Leverage (18) ROA %
37
Table 4: The Relationship between Employee-Friendliness and Firm Value
Panel A presents regression results of the impact of EF policies on Tobin’s q. EF index is an index ranging from 0-100
based on the proportion of 32 attributes of employee-friendliness adopted by a firm. The 32 attributes cover the
following areas from the social score components from ASSET 4 database: 1) Employment quality; 2) Diversity; 3)
Training and development; 4) Health and safety, and 5) Human rights. Panel B reports results using the scores on the
individual components of the EF index. EF-index ASSET4 is the average of the five component scores from the
ASSET4 database. The control variables (not shown in Panel B to conserve space) include: 1) Size; 2) Age; 3) Leverage;
4) Cash; 5) PPE; 6) Foreign sales; 7) R&D; 8) Capex; 9) Closely-held; 10) ADR; 11) ROA; 12) Log GDP per capita,
and 13) GDP growth. In specifications with country-year fixed effects, the country-level variables are subsumed by
the country-year fixed effects. t-statistics, in parentheses, are based on standard errors clustered at the country level. *,
**, and *** indicate significance at the 0.10, 0.05, and 0.01 two-tailed levels, respectively. See Appendix A for variable
definitions.
Panel A
Im
p
act of em
p
lo
y
ee-friendliness on firm value
De
p
endent variable: Tobin’s
q
(
1
)
(
2
)
(
3
)
(
4
)
(
5
)
(
6
)
EF Index
t
-1 0.0033*** 0.0041*** 0.0014**
(
6.83
)
(
5.72
)
(
2.46
)
EF index-
A
SSET4
t
-1
0.0032*** 0.0038*** 0.0004
(
5.87
)
(
4.88
)
(
0.55
)
Size
t
-1 -0.2141*** -0.2164*** -0.4985*** -0.2200*** -0.2226*** -0.4981***
(
-17.25
)
(
-17.68
)
(
-10.59
)
(
-17.14
)
(
-17.39
)
(
-10.33
)
Lo
g
A
g
e -0.0191 -0.0234* -0.1588* -0.0197 -0.0248* -0.1610*
(
-1.43
)
(
-1.87
)
(
-1.96
)
(
-1.45
)
(
-1.96
)
(
-1.99
)
Levera
g
e
t
-1 -0.0027 -0.0032 -0.0039*** -0.0027 -0.0032 -0.0039***
(
-1.47
)
(
-1.51
)
(
-3.51
)
(
-1.46
)
(
-1.50
)
(
-3.52
)
Cash
t
-1 0.0131*** 0.0125*** 0.0058*** 0.0133*** 0.0126*** 0.0059***
(
4.87
)
(
4.52
)
(
4.56
)
(
4.93
)
(
4.60
)
(
4.65
)
PPE
t
-1 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002
(
0.78
)
(
1.13
)
(
1.10
)
(
0.82
)
(
1.16
)
(
1.10
)
F
orei
g
n sales-to-
s
ales 0.0002 0.0004 -0.0016*** 0.0002 0.0004 -0.0016***
(
0.39
)
(
1.08
)
(
-2.80
)
(
0.57
)
(
1.28
)
(
-2.71
)
RD-to-
s
ales 0.0372*** 0.0351*** -0.0024 0.0367*** 0.0346*** -0.0025
(
9.78
)
(
8.72
)
(
-0.61
)
(
9.97
)
(
8.93
)
(
-0.64
)
Ca
p
e
x
-to-assets
t
-1 0.0094** 0.0094*** -0.0023 0.0090** 0.0092*** -0.0023
(
2.19
)
(
3.13
)
(
-0.99
)
(
2.11
)
(
3.09
)
(
-1.00
)
Closel
y
-held 0.0019*** 0.0021*** 0.0005 0.0019*** 0.0021*** 0.0005
(
2.77
)
(
3.20
)
(
0.67
)
(
2.78
)
(
3.27
)
(
0.64
)
A
DR 0.0890** 0.0734** 0.0855** 0.0709**
(
2.21
)
(
2.09
)
(
2.17
)
(
2.04
)
ROA 0.0494*** 0.0512*** 0.0123*** 0.0490*** 0.0510*** 0.0123***
(
14.56
)
(
11.92
)
(
7.87
)
(
14.55
)
(
11.94
)
(
7.84
)
Lo
g
GDP
p
er Ca
p
ita -1.0204*** -0.0761 -0.8995*** -0.0784
(
-6.44
)
(
-0.40
)
(
-6.15
)
(
-0.41
)
GDP Growth 0.0460*** 0.0143 0.0432*** 0.0144
(
5.50
)
(
1.63
)
(
5.27
)
(
1.63
)
Countr
y
fixed effects Yes
N
o
N
o Yes
N
o
N
o
Industr
y
fixed effects Yes
N
o
N
o Yes
N
o
N
o
Firm fixed effects
N
o
N
o Yes
N
o
N
o Yes
Year fixed effects Yes
N
o Yes Yes
N
o Yes
Countr
y
-
y
ear fixed
N
o Yes
N
o
N
o Yes
N
o
Industr
y
-
y
ear fixed
N
o Yes
N
o
N
o Yes
N
o
Observations 21
,
103 21
,
103 21
,
103 21
,
103 21
,
103 21
,
103
Ad
j
usted R2 0.420 0.451 0.777 0.420 0.451 0.777
# countries 43 43 43 43 43 43
38
Table 4: The Relationship between Employee-Friendliness and Firm Value. Continued.
Panel B – Components of EF index
Dependent variable: Tobin’s q
(1) (2) (3) (4) (5) (6)
Employment quality 0.0030*
0.0021
(1.94)
(1.47)
Training 0.0021***
0.0013***
(5.24) (2.97)
Diversity
0.0013**
0.0007
(2.51)
(1.30)
Health and safety
0.0002 -0.0008***
(1.07) (-3.03)
Human rights
0.0023*** 0.0020***
(5.81) (4.45)
Controls Yes Yes Yes Yes Yes Yes
Country-year fixed effects Yes Yes Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes Yes Yes
Observations 21,103 21,103 21,103 21,103 21,103 21,103
Adjusted R2 0.438 0.439 0.438 0.437 0.440 0.442
# countries 43 43 43 43 43 43
39
Table 5: Additional Tests of the Relation between Employee-Friendliness and Firm Value
The table shows results from 2SLS and OLS regressions of Tobin’s q. EF index is an index ranging from 0-100 based
on the proportion of 32 attributes of employee-friendliness adopted by a firm. The 32 attributes cover the following
areas from the social score components from ASSET 4 database: 1) Employment quality; 2) Diversity; 3) Training and
development; 4) Health and safety, and 5) Human rights. EF-index ASSET4 is the average of the five component scores
from the ASSET4 database. See Appendix A for variable definitions. In Panel A, we report results from 2SLS
regressions in which we instrument EF index (EF-index ASSET4) using two measures of country culture from Hofstede
(1980): 1) Masculinity – masculinity versus femininity orientation, and 2) Indulgence captures the extent to which a
society allows relatively free gratification of basic and natural human drives related to enjoying life and having fun.
Panel B shows results from regressions of changes in Tobin’s q (EF index) on lagged changes in EF index (Tobin’s q)
and all control variables (measured as changes from t-1 to t) included in Panel A. The control variables (not shown to
conserve space) include: 1) Size; 2) Age; 3) Leverage; 4) Cash; 5) PPE; 6) Foreign sales; 7) R&D; 8) Capex; 9) ROA;
10) Closely-held, and 11) ADR. t-statistics, in parentheses, are based on standard errors clustered at the country level.
We report F-statistics and p-values from the first-stage regressions, and Hansen’s J-statistic for the test of
overidentifying restrictions. The last row in Panel A reports F-statistics from the Montiel Olea-Pflueger (2013) robust
weak instrument test. The last row in Panel B reports p-values from Wald tests of the significance of lagged values of
EF index (
Tobin's q).*, **, and *** indicate significance at the 0.10, 0.05, and 0.01 two-tailed levels, respectively.
See Appendix A for variable definitions.
Panel A – 2SLS Regressions
First-stage Second-stage First-stage Second-stage
Dependent variable: EF index Tobin's q Social score Tobin's q
(1) (2) (3) (4)
EF index 0.012***
(3.12)
EF index-ASSET4
0.013***
(3.17)
Masculinity -0.118*** -0.111***
(-3.61) (-2.97)
Indulgence 0.237*** 0.211***
(4.87) (3.77)
Controls Yes Yes Yes Yes
Region-year fixed effects Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes
Observations 21,015 21,015 21,015 21,015
Adjusted R2 0.465 0.422 0.407 0.422
# countries 41 41 41 41
1st stage F-stat 22.163 12.046
1st stage F-statistic p-value 0.000 0.000
Hansen J-statistic 0.221 0.261
2 test (p-value) 0.638 0.610
Effective F-statistic (weak instruments) 21.079*** 13.065**
40
Table 5: Additional Tests of the Relation between Employee-Friendliness and Firm Value. Continued
Panel B – Causal Effect of Employee-Friendliness on Firm Value
Dependent variable: Tobin’s qt,-1,
t
EF index t,-1,
t
(1) (2) (3) (4)
EF Index t,
t
-1 0.0010* 0.0007
(1.81) (1.17)
EF Index
t
-1,
t
-2 0.0010* 0.0008
(1.76) (1.38)
EF Index
t
-2,
t
-3 0.0009*
(1.92)
Tobin's q t,
t
-1
0.1540 0.1160
(1.61) (1.17)
Tobin's q
t
-1,
t
-2
0.0379 0.1820
(0.47) (1.52)
Tobin's q
t
-2,
t
-3 -0.0361
(-0.56)
Country-year fixed effects Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Observations 17,582 14,388 17,582 14,388
Adjusted R2 0.206 0.225 0.111 0.128
# countries 43 43 43 43
Wald test - lagged EF index (Tobin's q) are
jointly equal to zero (p-value) 0.027 0.031 0.167 0.141
41
Table 6: The Impact of Exogenous Shocks on the Employee-Friendliness-Firm Value Relationship
The table shows various results from OLS regressions of Tobin’s q. EF index is an index ranging from 0-100 based
on the proportion of 32 attributes of employee-friendliness adopted by a firm. The 32 attributes cover the following
areas from the social score components from ASSET4 database: 1) Employment quality; 2) Diversity; 3) Training and
development; 4) Health and safety, and 5) Human rights. In Panel A we examine the relative performance following
the global financial crisis for firms with High (top third) and Low (bottom third) EF index as of the end of 2006. Post
is an indicator variable that equals one for years after 2008 and zero otherwise. Pre is an indicator that is equal to one
for years 2005 and 2006 and zero otherwise. In Models (3) and (4) of Panel D, we include additional interactions with
an indicator variable, High Impact that is equal to one for countries with above median change in unemployment between
2007 and 2009. In Panel B, we use the enactment of parental leave laws in EU countries as a shock to the diversity
component of the employee-friendly index. In Models (1) and (2) of Panel B we show results using interactions between
Treat- an indicator variable that is equal to one for firms in countries that enacted parental laws during our sample period
and zero otherwise. Our control group includes all firms from countries that did not adopt parental leave laws during
our sample period. Post is an indicator that is equal tone for years starting after the enactment of the parental leave laws
in the country and zero otherwise. We set Post equal to zero for our control group. Pre is an indicator variable that is
equal to one in years t-3, t-2 and t-1 relative to the enactment of the parental leave law, and zero otherwise. In Models
(3) and (4) we include interactions between Post and Most Impacted, an indicator that is equal to one if a firm in our
treatment sample has a Diversity index score in the bottom 25th percent of the distribution in their country in the year
prior to the enactment of the parental leave law. In Models (5)-(8), we show results for regressions using a propensity
score matched (PSM) sample of firms from the control group. We use propensity scores from a Probit regression using
an indicator variable Treat that is equal to one for firms in our treatment sample of countries and zero otherwise. In
Models (7) and (8), we only include Most Impacted firms from our treatment group. We match each treatment (Treat
or Most impacted) firm with a firm from the control group using the nearest neighbor matching technique (1:1) with
replacement. Appendix C has the results from the probit regressions used to obtain the propensity scores. Controls,
which are not shown to conserve space include: 1) Size; 2) Age; 3) Leverage; 4) Cash; 5) PPE; 6) Foreign sales; 7)
R&D; 8) Capex; 9) ROA; 10) Closely-held, and 11) ADR. t-statistics, in parentheses, are based on standard errors
clustered at the country level. *, **, and *** indicate significance at the 0.10, 0.05, and 0.01 two-tailed levels,
respectively. See Appendix A for variable definitions.
Panel A – Performance following the financial crisis
Dependent variable: Tobin’s q
Control group: PSM-matched sample
(1) (2) (3) (4)
Post x High EF index 0.090*** 0.105*** -0.183* -0.208**
(3.16) (3.67) (-1.88) (-2.38)
Post x High EF index x High Impact
0.296*** 0.338***
(2.97) (3.70)
Pre x High EF index 0.039 -0.061
(1.31) (-0.43)
High EF index 0.039 0.024 0.422*** 0.447***
(0.82) (0.45) (4.80) (3.77)
High EF index x High Impact -0.417*** -0.458***
(-4.08) (-3.33)
Pre x High EF index x High Impact 0.104
(0.66)
Controls Yes Yes Yes Yes
Country-year fixed effects Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes
Observations 9,449 9,449 9,449 9,449
Adjusted R2 0.531 0.531 0.532 0.532
42
Table 6: The Impact of Exogenous Shocks on the Employee-Friendliness-Firm Value Relationship. Continued.
Panel B – The Impact of Parental Leave Laws
Dependent variable: Tobin’s q
Full sample Treatment sample PSM matched sample
(1) (2) (3) (4) (5) (6) (7) (8)
Post x Treat 0.126** 0.137***
0.097* 0.141**
(2.63) (3.87)
(1.83) (1.97)
Pre x Treat 0.023
0.080
(0.42)
(0.90)
Post x Most impacted
0.111** 0.105*
0.358** 0.413**
(2.21) (1.76)
(2.42) (2.37)
Post
-0.067** -0.050 -0.034 -0.009 -0.347* -0.370*
(-2.18) (-1.64) (-0.62) (-0.13) (-1.91) (-1.87)
Pre x Most impacted
-0.010
0.116
(-0.20) (1.08)
Pre 0.026 0.016 -0.057
(1.29) (0.22) (-0.56)
Controls Yes Yes Yes Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 21,103 21,103 3,545 3,545 9,616 9,616 1,838 1,838
Adjusted R2 0.777 0.777 0.744 0.743 0.813 0.814 0.782 0.782
# countries 43 43 13 13 43 43 42 42
43
Table 7: Employee-Friendliness, Technical Efficiency, Innovation, and Performance
Table shows results from OLS and 2SLS regressions on the impact of employee-friendliness on firm technical efficiency.. We show results using two measures of
technical efficiency: 1) Sales-to-assets and 2) COGS-to-employees- the natural logarithm of cost of goods sold per employee. In Models (5) and (6) we show results
using the number of patents (log) as a proxy for innovation. Control variables (not shown to conserve space) include: 1) Size; 2) Age; 3) Leverage; 4) Capex; 5) R&D;
6) Market-to-book, and 7) Volatility – the standard deviation of weekly stock returns. EF index is an index ranging from 0-100 based on the proportion of 32 attributes
of employee-friendliness adopted by a firm. The 32 attributes cover the following areas from the social score components from ASSET 4 database: 1) Employment
quality; 2) Diversity; 3) Training and development; 4) Health and safety, and 5) Human rights. In Models (2), (4), and (6)we report results from 2SLS regressions in
which we instrument EF index using two measures of country culture from Hofstede (1980): 1) Masculinity masculinity versus femininity orientation, and 2)
Indulgence captures the extent to which a society allows relatively free gratification of basic and natural human drives related to enjoying life and having fun.
Country-year and industry-year fixed effects are included in all regressions. We use region-year and industry-year fixed effects in the first-stage regressions (not
reported to conserve space) because our instrument varies by country-year. t-statistics, in parentheses, are based on standard errors clustered at the country level. We
report F-statistics and p-values from the first-stage regressions, and Hansen’s J-statistic for the test of overidentifying restrictions. *, **, and *** indicate significance
at the 0.10, 0.05, and 0.01 two-tailed levels, respectively. See Appendix A for variable definitions.
Panel A - Employee-Friendliness, Technical Efficiency and Innovation
Dependent variable: Sales-to-assets COGS-to-employees (log) Ln (patents)
Estimation method: OLS 2SLS OLS 2SLS OLS 2SLS
(1) (2) (3) (4) (5) (6)
EF index
t
-1 0.503*** -0.002* 0.008***
(9.94) (-1.67) (3.43)
EF Index IV 1.351** -0.014** 0.021***
(2.44) (-2.33) (3.23)
Controls Yes Yes Yes Yes Yes Yes
Country-year fixed effects Yes Yes Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes Yes Yes
Observations 20,330 19,963 18,308 17,990 1,544 1,530
Adjusted R2 0.283 0.264 0.865 0.868 0.351 0.344
# countries 43 39 43 39 31 29
1st stage F-stat 31.481 22.811 11.411
1st stage F-statistic p-value 0.000 0.000 0.000
Hansen J-statistic 1.828 0.472 0.225
2 test (p-value) 0.176 0.492 0.635
44
Table 8: Country-Characteristics and the Employee-Friendliness-Firm Value Relation
The table shows results from 2SLS and OLS regressions of Tobin’s q. EF index is an index ranging from 0-100 b
on the proportion of 32 attributes of employee-friendliness adopted by a firm. The 32 attributes cover the follo
w
areas from the social score components from ASSET 4 database: 1) Employment quality; 2) Diversity; 3) Tra
i
and development; 4) Health and safety, and 5) Human rights. We classify countries as highly competitive usin
g
proxies for competition: 1) HFI – a Herfindahl index, and 2) Concentration – the proportion of total sales by fir
m
the top three industries in the country. High competition is an indicator variable that is equal to one for countries
HHI (Concentration) values in the top quartile in a given year. We also use Output – output per worker to class
country’s labor force. High Productivity is an indicator that is equal to one for countries with Output in the top qu
a
of the distribution and zero otherwise. In Models (1)-(3) we report results from OLS regressions, while Models
(6) show results form 2SLS regressions in which we instrument EF index using two measures of country culture
f
Hofstede (1980): 1) Masculinity and 2) Indulgence. The control variables (not shown to conserve space) includ
Size; 2) Age; 3) Leverage; 4) Cash; 5) PPE; 6) Foreign sales; 7) R&D; 8) Capex; 9) ROA; 10) Closely-held, an
d
ADR. t-statistics, in parentheses, are based on standard errors clustered at the country level. The last row repo
r
values from F-tests of the significance of the sum of the coefficient EF index and its interaction term. (
Tobin's
q
**, and *** indicate significance at the 0.10, 0.05, and 0.01 two-tailed levels, respectively. See Appendix
A
variable definitions.
Dependent variable: Tobin’s q
OLS Regressions 2SLS Regressions
(1) (2) (3) (4) (5) (6
)
EF index x High competition-HFI 0.003**
0.010***
(2.17)
(3.05)
EF index x High competition- Concentration 0.002*
0.010***
(1.73) (3.05)
EF index x High productivity 0.002* 0.00
8
(1.67) (2.5
7
EF Index t-1 0.002 0.003* 0.003*** -0.000 -0.001 0.0
0
(1.57) (1.78) (2.84) (-0.05) (-0.11) (0.4
0
Observations 21,103 21,103 21,054 21,015 21,015 21,0
Adjusted R2 0.440 0.440 0.440 0.438 0.438 0.4
3
# countries 43 43 42 41 41 41
F-test [EF index x High + EF index]=0 72.65 70.39 46.55 3.46 3.47 4.2
8
p-value 0.000 0.000 0.000 0.070 0.070 0.0
4
45
Table 9: Robustness Tests
Table shows results from OLS and 2SLS regressions. We use two dependent variables: Tobin’s q or MTB – market-to-
book value of equity. EF index is an index ranging from 0-100 based on the proportion of 32 attributes of employee-
friendliness adopted by a firm. The 32 attributes cover the following areas from the social score components from ASSET
4 database: 1) Employment quality; 2) Diversity; 3) Training and development; 4) Health and safety, and 5) Human rights.
In Model (1), we show results for regressions excluding firms in the US. In Model (2) we run regressions excluding firm
in the list of Best Companies to Work for (BC firms). In Models (3) and (4) we use an alternate measure of firm value,
MTB, for the full sample and the sample excluding the US, respectively. In Models (4) and (6) we restrict the sample to
firms with available data for the entire sample period and use Tobin’s q and MTB, respectively. In Panel B we show
results from 2SLS regressions in which we instrument EF index using two measures of country culture from Hofstede
(1980): 1) Masculinity masculinity versus femininity orientation, and 2) Indulgence captures the extent to which a
society allows relatively free gratification of basic and natural human drives related to enjoying life and having fun. The
control variables (not shown to conserve space) include: 1) Size; 2) Age; 3) Leverage; 4) Cash; 5) PPE; 6) Foreign sales;
7) R&D; 8) Capex; 9) Closely-held, and 10) ADR. t-statistics, in parentheses, are based on standard errors clustered at the
country level. *, **, and *** indicate significance at the 0.10, 0.05, and 0.01 two-tailed levels, respectively. See Appendix
A for variable definitions.
Panel A - Robustness tests
Excludes US
Excludes BC
firms Full sample Excludes US
Firms with available data for
all years
Dependent variable Tobin’s q Tobin’s q MTB MTB Tobin’s q MTB
(1) (2) (3) (4) (5) (6)
EF index t-1 0.004*** 0.004*** 0.016*** 0.013*** 0.006*** 0.018***
(4.44) (4.70) (4.53) (3.39) (3.15) (2.96)
Controls Yes Yes Yes Yes Yes Yes
Country-year fixed effects Yes Yes Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes Yes Yes
Observations 15,014 18,495 21,103 15,100 3,927 3,927
Adjusted R2 0.439 0.316 0.205 0.217 0.405 0.240
# countries 42 43 43 42 41 41
Panel B - Robustness tests – 2SLS Regressions
Excludes US
Excludes BC
firms Full sample Excludes US
Firms with available data for
all years
Dependent variable Tobin’s q Tobin’s q
M
TB
M
TB Tobin’s q
M
TB
(1) (2) (3) (4) (5) (6)
EF index I
V
0.011*** 0.011*** 0.026*** 0.024** 0.018 0.058**
(2.77) (2.99) (2.80) (2.25) (1.39) (2.44)
Controls Yes Yes Yes Yes Yes Yes
Region-year fixed effects Yes Yes Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes Yes Yes
Observations 14,926 18,318 21,015 14,926 3,883 3,883
Adjusted R2 40 41 41 40 39 39
# countries 0.320 0.308 0.158 0.178 0.397 0.180
1st sta
g
e
F
-sta
t
24.320 23.438 22.163 24.320 3.325 3.325
1st sta
g
e
F
-statistic
p
-value 0.000 0.000 0.000 0.000 0.047 0.047
Hansen
J
-statistic 0.069 0.061 4.559 2.469 0.456 0.113
2 test
(
p
-value
)
0.793 0.805 0.033 0.116 0.499 0.737
46
Figure 1. Employee-Friendly Culture (EF) Indices by Year
The figure shows the annual average values for our two proxies of employee-friendly culture: 1) EF index,
and 2) EF index-ASSET4. EF index is an index ranging from 0-100 based on the proportion of 32 attributes
of employee-friendliness adopted by a firm. The 32 attributes cover the following areas from the social
score components from ASSET 4 database: 1) Employment quality; 2) Diversity; 3) Training and
development; 4) Health and safety, and 5) Human rights. EF-index ASSET4 is the average of the five
component scores from the ASSET4 database. See Appendix A for variable definitions.
0
10
20
30
40
50
60
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
%
EFIndexbyYear
EFIndex EFIndex‐ASSET4
47
Appendix A: Variable Definitions
Variable Name Description
ADR Indicator that equals one if the firm is cross-listed on a U.S. stock exchange and zero
otherwise.
Capex Capital expenses scaled by the lagged book value of assets.
Cash Cash divided by total assets.
Closely-held Percentage of closely held shares.
COGS-to-employees Cost of goods sold divided by the total number of employees (log).
EF Index An index ranging from 0-100 based on the proportion of 32 attributes of employee-
friendliness adopted by a firm. The 32 attributes cover the following areas from the
social score components from ASSET 4 database: 1) Employment quality; 2)
Diversity; 3) Training and development; 4) Health and safety, and 5) Human rights.
EF Index – ASSET4 The average of the five social component scores from the ASSET4 database: 1)
Employment quality; 2) Diversity; 3) Training and development; 4) Health and safety,
and 5) Human rights. Each component receives a percentage score by ASSET 4
based on several factors.
Foreign sales Two-year average of the ratio of foreign sales to sales.
GDP Growth Annual growth in real gross domestic product (GDP).
Indulgence Hofstede’s (1980) measure of culture. Indulgence stands for a society that allows
relatively free gratification of basic and natural human drives related to enjoying life
and having fun.
Leverage Total debt divided by book value of assets.
Ln Patents The log of one plus the number of patents.
Log Age Log of firm age. Firm age is the number of years since the firm was incorporated.
When the date of incorporation is unavailable, firm age is calculated as the number of
years since the firm first appeared on the DataStream and WorldScope databases.
Log GDP per capita Annual log of real gross domestic product per capita (constant U.S. dollars).
Operating ROA Operating income divided by the book value of assets.
Masculinity Hofstede’s measure of culture. The Masculinity side of this dimension represents a
preference in society for achievement, heroism, assertiveness, and material rewards
for success. Society at large is more competitive. Its opposite, Femininity, stands for
a preference for cooperation, modesty, caring for the weak and quality of life.
Most impacted An indicator variable that is equal to one for firms with a Diversity index in the bottom
quartile of the distribution in their country as of the year prior to the enactment of the
Parental Leave Directive 2010/18 in the country and zero otherwise.
PPE Property, plant, and equipment, scaled by sales.
R&D The two-year average research and development (R&D) expenses divided by sales
ROA Net income divided by book value of total assets.
Sales-to-assets Sales divided by book value of assets as of the beginning of the year.
Size Log of total assets (US$ 000s).
Tobin’s q Total assets less book value of equity plus market value of equity divided by book
value of total assets.
Treat An indicator variable that is equal to one for firms in European countries that
implemented the Parental Leave Directive 2010/18 during our sample period and zero
otherwise.
48
Appendix B: Employee-Friendly (EF) Index Components
The 32 attributes correspond to five categories of social performance: Employment quality; Diversity;
Training and development; Health and safety, and Human rights. The attributes are based on a subset of
questions used by ASSET to rate each of these components. A firm is assigned a value of one for positive
responses, or if its value is above (below) the industry median. We create an index for each of the five
categories with a maximum value of 100% based on the fraction of all nonmissing attributes that a firm
satisfies. An aggregate index is computed in a similar fashion (as the proportion of all nonmissing attributes
that a firm satisfies). We report the percentage of firms that meet each of the attributes (% meeting). To
do so, we first compute the percentage of firms that meet each attribute each year and report the time-series
average.
% meetin
g
EMPLOYMENT
Q
UALITY:
1 Com
p
an
y
monitors or measures its
p
erformance on em
p
lo
y
ment
q
ualit
y
8.21%
2 Percenta
g
e of em
p
lo
y
ee turnover below industr
y
median. 41.35%
3 Strikes that led to lost workin
g
da
y
s below industr
y
median. 97.21%
4 Avera
g
e salaries and benefits above industr
y
median. 48.54%
5 Company won an award or any prize related to general employment quality "Best Company to
Work For"
10.78%
6 CEO salar
y
-to-avera
g
e wa
g
e below industr
y
median. 47.83%
7
N
umber of la
y
-offs divided b
y
the total number of em
p
lo
y
ees below industr
y
median. 0.08%
DIVERSITY:
8 Com
p
an
y
has a diversit
y
and e
q
ual o
pp
ortunit
y
p
olic
y
73.59%
9 Com
p
an
y
has a wor
k
-life balance
p
olic
y
. 30.17%
10 Company has the appropriate communication tools (whistle blower, ombudsman, suggestion box,
hotline, newsletter, website, etc.
)
to im
p
rove diversit
y
and o
pp
ortunit
y
.
38.68%
11 Com
p
an
y
sets tar
g
ets or ob
j
ectives to be achieved on diversit
y
and e
q
ual o
pp
ortunit
y
. 25.79%
12 Com
p
an
y
sets tar
g
ets or ob
j
ectives to be achieved on em
p
lo
y
ees' wor
k
-life balance. 13.89%
13 Percenta
g
e of women em
p
lo
y
ees above industr
y
median. 46.03%
14 Percenta
g
e of women mana
g
ers above industr
y
median. 44.05%
15 Percenta
g
e of elderl
y
em
p
lo
y
ees above industr
y
median. 31.75%
TRAINING AND DEVELOPMENT:
16 Com
p
an
y
has a
p
olic
y
to su
pp
ort the skills trainin
g
of its em
p
lo
y
ees. 61.63%
17 Com
p
an
y
has a
p
olic
y
to su
pp
ort the career develo
p
ment of its em
p
lo
y
ees. 57.02%
18 Com
p
an
y
monitors its own trainin
g
and develo
p
ment
p
ro
g
rams. 14.06%
19 Avera
g
e hours of trainin
g
p
er
y
ear
p
er em
p
lo
y
ee above industr
y
median. 41.86%
20 Com
p
an
y
p
rovides trainin
g
in environmental
,
social or
g
overnance factors to its su
pp
liers. 7.80%
21 Trainin
g
costs
p
er em
p
lo
y
ee above industr
y
median. 41.14%
HEALTH AND SAFETY:
22 Com
p
an
y
has an em
p
lo
y
ee health & safet
y
team. 33.19%
23 Company has the appropriate internal communication tools (whistle blower, suggestion box,
hotline, newsletter, website, etc.) to improve employee health & safety. 41.21%
24 Com
p
an
y
sets tar
g
ets or ob
j
ectives to be achieved on em
p
lo
y
ee health & safet
y
. 55.42%
25 Total number of in
j
uries and fatalities
p
er one million hours worked is below industr
y
median. 43.71%
26 Number of injuries and fatalities reported by employees and contractors while working for the
company is below industry median. 42.48%
HUMAN RIGHTS:
27 Com
p
an
y
has a
p
olic
y
to ensure the freedom of association of its em
p
lo
y
ees. 22.11%
28 Com
p
an
y
has a
p
olic
y
to avoid child labor. 30.92%
29 Com
p
an
y
has a
p
olic
y
to avoid forced labor. 28.43%
30 Com
p
an
y
has a human ri
g
hts
p
olic
y
that is a
pp
lied to its su
pp
l
y
chain. 26.28%
31 Com
p
an
y
has a
g
eneral
,
all-
p
ur
p
ose
p
olic
y
re
g
ardin
g
human ri
g
hts. 33.13%
32 Com
p
an
y
monitors human ri
g
hts in its or its su
pp
liers' facilities. 11.87%
49
Appendix C. Details of Propensity-Score-Matching (PSM) Procedure
The propensity-score-matching approach involves pairing treatment and control firms (Dehejia and Wahba 2002). The
dependent variable, Treat is an indicator variable that is equal to one for firms in EU countries that enact parental
leave laws implementing EU Parental Leave Directive 2010/18 during our sample period and zero otherwise. The
treatment countries include Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg,
Netherlands, Norway, Poland, and the United Kingdom. We first estimate a probit regression to model the probability
of being a firm in our treatment sample. Next, we estimate the propensity score for each firm using the predicted
probabilities from the probit model. We then match each treatment firm to a control firm from a control country using
the nearest neighbor matching technique (with replacement). Panel A reports the estimation results of the probit
model. In Model (1) we show results from the probit model used to generate the propensity scores. In Model (2), we
run the probit model using only the matched sample to determine whether there are significant differences between
matched and control firms. In Models (3) and (4), we run the pre (post) match probit regressions restricting the
treatment group of firms to those that are deemed Most Impacted by the parental leave law. Specifically, Most
Impacted is an indicator that is equal to one for firms in our treatment sample with a Diversity index in the bottom
quartile of the distribution in their country in the year prior to the enactment of the parental leave law. Panel B show
descriptive statistics of the firm-level variables for our group of treatment firms and the control firms for the full
sample. We report the mean values for each matching characteristic pre- and post-match, along with the normalized
difference (X) to evaluate the quality of the matching, following Imbens and Wooldridge (2009). Z-statistics, in
parentheses, are based on standard errors clustered at the country level. *, **, and *** indicate significance at the
0.10, 0.05, and 0.01 two-tailed levels, respectively. See Appendix A for variable definitions.
Panel A
Probit Re
g
ressions
De
p
endent variable: Trea
t
M
ost Im
p
acted
Pre-match Pos
t
-match Pre-match Pos
t
-match
(
1
)
(
2
)
(
3
)
(
4
)
Size
t
-1 -0.045 -0.006 -0.012** -0.031
(
-1.49
)
(
-0.12
)
(
-2.39
)
(
-0.56
)
Lo
g
A
g
e 0.001 0.026 -0.003 -0.013
(
0.04
)
(
0.52
)
(
-0.53
)
(
-0.21
)
Levera
g
e
t
-1 0.001** -0.001 0.000 -0.001
(
2.11
)
(
-0.74
)
(
1.53
)
(
-0.51
)
Cash
t
-1 -0.003 -0.001 -0.001 0.003
(
-1.52
)
(
-0.33
)
(
-1.63
)
(
0.57
)
PPE-to-Sales
t
-1 -0.000* -0.000 0.000 0.000
(
-1.69
)
(
-0.29
)
(
0.27
)
(
0.59
)
F
orei
g
n sales-to-
s
ales 0.004*** 0.001 0.001*** 0.000
(
4.63
)
(
0.64
)
(
4.04
)
(
0.21
)
RD-to-
s
ales -0.006*** 0.001 -0.000 0.000
(
-2.61
)
(
0.19
)
(
-0.47
)
(
0.03
)
Ca
p
e
x
-to-assets
t
-1 -0.002 0.002 0.000 0.014***
(
-1.22
)
(
0.47
)
(
0.06
)
(
3.60
)
Closel
y
-held 0.000 0.000 0.000 0.001
(
0.04
)
(
0.13
)
(
1.02
)
(
0.23
)
A
DR 0.207** 0.034 0.032* 0.114
(
2.32
)
(
0.24
)
(
1.82
)
(
0.78
)
ROA 0.001 0.001 0.000 0.001
(
0.65
)
(
0.50
)
(
1.39
)
(
0.22
)
Lo
g
GDP
p
er Ca
p
ita 0.049 0.121 0.016 0.227
(
0.69
)
(
0.84
)
(
1.19
)
(
1.50
)
GDP Growth -0.068*** -0.127*** -0.012*** -0.117***
(
-3.26
)
(
-3.51
)
(
-3.80
)
(
-3.27
)
Industr
y
fixed effects Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes
Observations 21
,
060 9
,
597 15
,
767 1
,
776
Pseudo R2 0.252 0.141 0.254 0.188
# of countries 43 43 43 42
50
Appendix C. Procedure to develop propensity-score-matched (PSM) firms. Continued.
Panel B – Descriptive Statistics of Treatment (