Economic Modelling 120 (2023) 106156

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.journals.elsevier.com/economic-modelling

Board diversity and outward FDI: Evidence from europe

Valeria Gattai a, *, Piergiovanna Natale a, Francesca Rossi b
a
Dipartimento di Economia, Metodi Quantitativi e Strategie di Impresa and Center for European Studies (CefES), Università Degli Studi di Milano-Bicocca, Piazza Ateneo
Nuovo 1, 20126, Milano, Italy
b
Dipartimento di Scienze Economiche, Università Degli Studi di Verona, Via Cantarane, 24, 37129, Verona, Italy

A R T I C L E I N F O A B S T R A C T

Handling editor: Sushanta Mallick Employing firm-level panel data from 2011 to 2015, we investigate the relationship between board diversity—in
terms of gender and nationality—and outward foreign direct investment (OFDI) in Europe. Previous studies
JEL classification: suggest that best-performing firms self-select into OFDI and that board diversity affects firm performance and
F23 strategic decisions. Controlling for endogeneity using instrumental variables and control functions, we find that
G30
firms with more diverse boards are less likely to open foreign subsidiaries. Furthermore, we explore the role of
J16
performance as modifier and mediator of the effect of board diversity on OFDI. Relying on suitable interaction
Keywords:
terms, we show that the effect of board diversity on OFDI is stronger for more productive firms. We then
Board diversity
decompose the effect of board diversity on OFDI into a negative direct effect and a positive performance-
Outward foreign direct investment (OFDI)
Foreign direct investment (FDI) mediated effect, with the former outweighing the latter. Our findings are consistent with tough management
Firm performance monitoring by more diverse boards.
Europe

1 Introduction board diversity on outward foreign direct investment (OFDI).2 This is

somewhat surprising as OFDI is a strategic decision that requires board
During the last few decades, European countries have experienced approval, in which board composition plays a role (Levi et al., 2014). It
remarkable changes in board diversity, which are not observed else­ is even more surprising for Europe, which features prominently on the
where (Ferreira and Kirchmaier, 2013). Female directors accounted for OFDI stage. In 2019, outflows from European countries accounted for
no more than 10% of board members in large companies across Italy, 35% of total OFDI flows, and 50% of those originated from developed
France, Germany, Spain, and the United Kingdom in the early 2000s; economies; in the same year, outstocks from European countries
however, this figure more than doubled by the end of 2016. Similarly, amounted to 36% of total OFDI stocks, and 48% of those originated from
the percentage of foreign directors on the boards of major listed com­ developed economies (UNCTAD, 2020).
panies showed a 100% increase in most European countries from 2010 This study explores the relationship between OFDI and board di­
to 2016 (Ciavarella, 2017). versity at the firm level.3 Specifically, we estimate the effect of board
The effect of board diversity on firm performance and strategic diversity, in terms of gender and nationality, on the probability of
decision-making has been extensively investigated.1 However, little opening a foreign subsidiary by firms headquartered in Europe.4 To this
attention has been directed to the effect of board diversity on firm end, we employ firm-level longitudinal data from Orbis, an adminis­
internationalisation decisions and there is no evidence on the effect of trative database issued by Bureau van Dijk. Our sample comprises 1283

* Corresponding author.
E-mail address: [email protected] (V. Gattai).
1
See Section 2 for a literature review.
2
According to the IMF (1993), OFDI is an investment in a foreign company by which the investor owns at least 10% of the ordinary shares and it is undertaken with
the objective of establishing a lasting interest in the country and significant influence on the firm’s management.
3
The choice to study OFDI at the firm level has both theoretical and empirical motivations. See Section 2 for a literature review and Section 4 for a discussion of
our empirical methodology.
4
In this paper, we use ‘opening a foreign subsidiary’ and ‘initiating OFDI’ as synonymous.

https://doi.org/10.1016/j.econmod.2022.106156
Received 21 June 2020; Received in revised form 17 December 2022; Accepted 17 December 2022
Available online 22 December 2022
0264-9993/© 2022 Elsevier B.V. All rights reserved.
V. Gattai et al. Economic Modelling 120 (2023) 106156

Fig. 1. Conceptual framework and research question.

listed industrial companies from 21 European countries, belonging to 15 spectrum, suggesting tougher monitoring by more diverse boards as the
NACE 1- digit industries and continuously operating from 2011 to driving force.7 Further, by means of the so-called causal steps strategy
2015.5 (Judd and Kenny, 1981a,b; Baron and Kenny, 1986; Preacher and Hayes,
The conceptual framework behind our research question is derived 2008; MacKinnon and Dwyer, 1993), we decompose the effect of board
from two strands of literature: one studies the relationship between diversity on OFDI decisions into direct and mediated effects, with per­
board diversity and firm performance as well as the relationship be­ formance being the mediating variable. Our findings reveal that board
tween board diversity and strategic decisions; the other addresses the diversity exerts a large direct negative impact on OFDI, which is atten­
relationship between OFDI and firm performance.6 In the first strand, uated by a (smaller in magnitude) mediated positive effect. More spe­
boards act as monitors and advisors for management. If the board’s cifically, we find that variations in performance explain a percentage of
composition affects its effectiveness in either role, board diversity has an the total effect of board diversity on OFDI that does not exceed 6%,
impact on strategic decisions and firm performance (Adams and Fer­ showing that the predominant effect of board diversity on the proba­
reira, 2007). The second strand of literature shows that firms self-select bility of starting OFDI is negative and direct, i.e. not mediated by
into internationalisation. As long as fixed costs are incurred in serving performance.
foreign markets, only the most productive firms, which command a large The identification of our direct and mediated effects faces two
market share, can successfully enter foreign markets through exports or econometric challenges related to endogeneity issues. The first relates to
OFDI (Melitz, 2003; Helpman et al., 2004). the potential reverse causality between OFDI and performance due to a
In light of the above, the reason why we expect board diversity to learning-by-OFDI mechanism that raises firm-level productivity through
influence OFDI is twofold (see Fig. 1). First, OFDI is a strategic decision foreign exposure. The second concerns the potential endogeneity in
requiring board-level approval. Shaping the board’s effectiveness as board diversity, as directors may self-select into boards according to firm
monitor and advisor for the management, diversity may affect OFDI performance and internationalisation strategies. The first issue is
decisions. Second, firm performance acts as ‘modifier’ and ‘mediator’ of addressed by a convenient definition of our dependent variable such that
the board diversity-OFDI relationship. As only the most productive firms the learning-by-OFDI mechanism is excluded from our model. The sec­
undertake OFDI, the magnitude of the effect of board diversity on OFDI ond source of endogeneity is addressed through a suitable instrument
may differ according to performance. Furthermore, as board diversity and instrumental variables/two-stage least-squares (IV/2SLS) and con­
affects firm performance, board diversity may influence the probability trol function (CF) estimation methods. The latter is not often employed
of opening foreign subsidiaries. to address endogeneity issues in the literature on firm internationalisa­
Using binary dependent variable models and suitable interaction tion.8 Even though a small percentage of firms started to engage in OFDI
terms, we find that board diversity has a strongly significant negative during our survey period, our results are robust to different econometric
impact on the probability of engaging in OFDI and that such negative models and specifications.
effect is even larger for firms at the high-end of the productivity Our study contributes to literature on OFDI and performance as it is
the first to show explicitly the effect of board diversity on OFDI, con­
trolling for performance. As board diversity plays a statistically
5
Lack of data prevents us from considering a longer time span; although data
on subsidiaries reference the period spanning 2007 to 2015 in our database,
7
data on boards are only available from 2011 to 2015. See Section 4 for a detailed discussion.
6 8
See Section 2 for a detailed review of both strands of literature. See Section 4 for a discussion of these techniques’ suitability in our context.

2
V. Gattai et al. Economic Modelling 120 (2023) 106156

significant role in determining OFDI, this paper complements previous Castellani and Disdier (2010), Borin and Mancini (2016), and Baiardi
findings on the influence of ethnic diversity in the workforce (Parrotta et al. (2021) use panel data and regress past performance on a dummy
et al., 2016; Moriconi et al., 2020), the intensity of female ownership for future OFDI starters to observe that firm performance has a positive
and management (Marques, 2015), and board composition (Dou et al., and statistically significant effect on OFDI.9 This methodology allows
2019; Pisani et al., 2019; Rivas, 2012) on internationalisation. In addi­ identifying the performance differentials between firms engaged and not
tion, our study contributes to literature on board diversity and strategic engaged in OFDI before their first involvement in OFDI. Wagner (2007)
decisions as it is the first to consider explicitly the effect of board explains that this is the proper econometric approach to address endo­
composition on OFDI. In doing so, our work complements previous geneity and to provide conclusive evidence of self-selection. Considering
findings on the influence of board gender diversity on acquisitions de­ a dummy for future OFDI firms—rather than starters—would not
cisions (Huang and Kisgen, 2013; Levi et al., 2014; Chen et al., 2016). exclude a learning-by-OFDI mechanism, by which previous OFDI expe­
From a methodological point of view, it is worth emphasising that we rience positively affects performance. The literature provides evidence
decompose the effect of board diversity on OFDI into direct and of this learning-by-OFDI effect (Barba Navaretti et al., 2010; Barba
performance-mediated effects, applying the causal steps strategy. To the Navaretti and Castellani, 2004; Castellani et al., 2008; Borin and Man­
best of our knowledge, this is an original contribution of the present cini, 2016; Huang and Zhang, 2017; Baiardi et al., 2021).
study. Refinements of these simple analyses investigate the mapping of
The remainder of this paper is organised as follows. Section 2 reviews heterogeneous firms into different sourcing strategies, including OFDI
the two strands of literature that motivate our research question. Section (Tomiura, 2007; Federico, 2010; Kohler and Smolka, 2011, 2012; Gattai
3 introduces the data employed for empirical purposes and descriptive and Trovato, 2016), and dissecting OFDI according to the destination
statistics. Section 4 discusses our econometric models and results, and and ownership structure (Aw and Lee, 2008; Damijan et al., 2007; Raff
Section 5 presents several robustness checks. Section 6 concludes by et al., 2012).
providing policy implications as well as by presenting the main limita­
tions of the analysis. 2.2. Board diversity, firm performance, and strategic decisions

2. Literature review Spurred by changes in board composition due to the global debate on
the roles of women and ethnic minorities (Adams and Ferreira, 2009), a
This section briefly reviews the two strands of literature that inspired growing body of literature examines the impact of board diversity on
our research question. firm performance and, to a lesser extent, on strategic decisions.

2.1. OFDI and firm performance 2.2.1. Board diversity and firm performance
As comprehensively reviewed by Adams et al. (2010), Ferreira
Following Bernard and Jensen (1995) seminal work, many studies (2010), Rhode and Packel (2014), and Kirsch (2018), the relationship
have investigated the relationship between internationalisation and firm between board diversity and firm performance is the subject of vast
performance. Regardless of the year and country of analysis, empirical empirical economics and management literature, comprising an array of
evidence suggests that internationalised firms outperform domestic en­ countries, institutional settings, and demographic characteristics.
terprises based on several economic, innovation-related, and financial For our study, it is particularly compelling to provide a systematised
variables. As comprehensively reviewed in the surveys by Lopez (2005), view of the relevant evidence from European countries on the impacts of
Wagner (2007, 2012, 2016), Greenaway and Kneller (2007), Singh gender and nationality diversity on performance.
(2010), and Hayakawa et al. (2012), this result is robust to different Correlation analyses performed on large samples of European firms
internationalisation strategies and performance measures. reveal a positive, strong association between accounting measures of
For this study, our focus is on papers investigating the relationship firm profitability and the proportion of female directors (Christiansen
between OFDI and firm performance. et al., 2016; Ferreira and Kirchmaier, 2013). This association persists,
Castellani and Zanfei (2007), Castellani and Giovannetti (2010), albeit not as strongly, when the endogeneity of board appointments is
Wagner (2006), Arnold and Hussinger (2010), Engel and Procher addressed and alternative measures of firm performance are considered.
(2012), Girma (2005), and Casaburi et al. (2007) present basic corre­ Relying on the IV identification strategy, Green and Homroy (2018)
lations. In these studies, performance variables are regressed on a report that gender diversity has a positive but modest effect on the
dummy for OFDI status or vice versa. Empirical evidence largely sup­ profitability of large European firms. Similar econometric techniques
ports the existence of a positive, statistically significant correlation be­ have been adopted, and similar results have been obtained for French
tween OFDI and performance. Firms with at least one subsidiary abroad (Sabatier, 2015; Dang et al., 2018), Spanish (Campbell and
are larger, more productive, more capital-intensive, and more innova­ Minguez-Vera, 2010), and Italian (Amore et al., 2014) firms. Conversely,
tive than domestic enterprises. However, causality is not addressed Gregory-Smith et al. (2014) find that gender diversity has no effect on
econometrically, which is a major disadvantage of these studies. the profitability of large British firms.
From a theoretical perspective, the existence of a positive correlation Joecks et al. (2013) suggest that overlooked non-linearities can ac­
between OFDI and performance can be explained in terms of self- count for conflicting evidence regarding the effect of gender-diverse
selection. If fixed costs are incurred in serving foreign markets, only boards on firm performance. According to critical mass theory
the most productive firms that command a large market share can suc­ (Kanter, 1997), when there are relatively few female directors, they are
cessfully enter foreign markets through exports or OFDI. The theoretical perceived as representatives of their gender rather than effective board
foundation of the self-selection mechanism can be traced back to members. Joecks et al. (2013) sample 151 German-listed firms for the
Helpman et al. (2004), Head and Ries (2003), and Grossman et al. period between 2000 and 2005 to observe that gender diversity posi­
(2006). These studies extend Melitz’s (2003) benchmark framework to tively affects performance but only for boards with at least one-third
analyse the intra-industry effects of foreign direct investment. female directors. Similar results have been reported for Norway
Consistent with this theoretical framework, Kimura and Kiyota (Torchia et al., 2011) and Italy (Bruno et al., 2018).
(2006), Barba Navaretti and Castellani, 2004, Barba Navaretti, Led by Norway in 2003, many European countries introduced a

9
OFDI firms are those engaged in OFDI in a certain year, OFDI starters are
those engaged in OFDI for the first time in that year.

3
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 1 gender diversity in shaping a firm’s strategy. Employing World Bank

Sample of firms, OFDI firms and noOFDI firms by country. survey data from developing countries, Marques (2015) reports that
Firms OFDI firms NoOFDI firms firms’ export propensity and intensity vary with the gender of top
managers and entrepreneurs. Pisani et al. (2019) extend our under­
Country Number % Number % Number %
standing of the effects of board diversity on firm internationalisation by
Austria 180 3% 174 97% 6 3% showing that gender-diverse boards are more likely to appoint foreign
Belgium 250 4% 232 93% 18 7%
Bulgaria 40 1% 21 53% 19 48%
members in top management teams. As self-selection into a career
Czech Republic 20 0% 19 95% 1 5% abroad is associated a with a higher cognitive tolerance for foreignness
Denmark 225 4% 197 88% 28 12% (Levy et al., 2007), firms with foreign members in top management
Estonia 25 0% 20 80% 5 20% teams exhibit greater internationalisation.10 Conversely, the relation­
Finland 305 5% 290 95% 15 5%
ship between board nationality diversity and firms’ internationalisation
France 1195 19% 1071 90% 124 10%
Germany 1060 17% 925 87% 135 13% strategies has been investigated extensively. Boards act as advisor and
Great Britain 1560 24% 1169 75% 391 25% monitor for the management (Adams and Ferreira, 2007).
Ireland 100 2% 92 92% 8 8% Nationality-diverse directors can affect board’s effectiveness in the
Latvia 15 0% 6 40% 9 60% advising role. Employing U.S. data, Masulis et al. (2012) show that firms
Lithuania 45 1% 21 47% 24 53%
Luxembourg 70 1% 65 93% 5 7%
make better cross-border acquisitions when operating in the directors’
Poland 255 4% 169 66% 86 34% country of origin. Stroup (2017) reports that the presence of directors
Portugal 125 2% 108 86% 17 14% with international experience increases the likelihood of a firm’s first
Romania 45 1% 15 33% 30 67% cross-border acquisition and its success. In contrast, Hahn and Lasfer
Slovakia 20 0% 14 70% 6 30%
(2016) and Frijns et al. (2016) report that nationality-diverse boards
Slovenia 35 1% 35 100% 0 0%
Spain 330 5% 296 90% 34 10% face communication and coordination problems and may fail to impose
Sweden 515 8% 472 92% 43 8% managerial discipline.
Total 6415 100% 5411 84% 1004 16%
3. Data and descriptive statistics
Source: Authors’ elaboration from Orbis (2017)
The present research employs firm-level longitudinal information
mandatory gender quota to address the inadequate female presence on from Orbis, an administrative dataset issued by Bureau van Dijk.11
boards. Such quotas provide quasi-natural experimental settings, in Our dataset is the result of thorough data mining and cleaning pro­
which changes in board diversity can be considered exogenous. Litera­ cesses. As most subsidiary and board information was not readily
ture relying on the gender quotas in their identification strategy high­ available for download from the Orbis website, we retrieved it from
lights that gender diversity has a negative or zero effect on firm Bureau van Dijk under a special research agreement. Moreover, the
performance. Ahern and Dittmar (2012) and Matsa and Miller (2013) retrieved information was not suitable for use and required processing to
note that increased gender diversity negatively influences profitability build consistent measures for OFDI and board diversity.
among Norwegian firms, whereas Dale-Olsen et al. (2013) and Eckbo Ultimately, our balanced panel covers 1283 industrial companies
et al. (2016) find no such effect. Comi et al. (2020) report that gender listed on the stock market that were continuously headquartered in the
diversity has a negative or zero effect on an array of performance European Union from 2011 to 2015. We focus on industrial companies
measures in Spain and France, while it has a positive effect on Italian to study the behaviour of heterogeneous firms within a relatively ho­
firms’ productivity. However, these findings should not be considered mogeneous class, which includes all the available NACE two-digit in­
conclusive, as mandatory changes in board diversity could induce an dustries.12 Although Orbis collects information on both listed and
inefficient selection of board members that results in the negative unlisted companies, we restrict our focus to the former, for which more
impact of gender diversity on boards (Schmid and Urban, 2016). detailed information is available. Finally, we consider a large set of
As mentioned in Section 1, nationality is an increasingly important European countries to exploit country-level heterogeneity. As
source of heterogeneity in European boards. However, empirical evi­ mentioned in Section 1, Europe is the ideal locus to study board diver­
dence on the effect of nationality diversity on firm performance is sity’s effect on OFDI because it is a leading actor in the OFDI stage and
scarce. Green and Homroy (2018) and Estelyi and Nisar (2016) report board diversity has significantly increased across European countries
that firms with more nationality-diverse boards are more profitable, over the last two decades.
whereas Garcia-Meca et al. (2015), Frijns et al. (2016), and Hahn and From a geographical perspective, Table 1 reveals that most of our
Lasfer (2016) note that diverse nationalities negatively affect firm (pooled) sample firms are from Great Britain (24%), France (19%),
performance. Germany (17%), and Sweden (8%). Spain, Finland, Poland, Belgium,
and Denmark are equally represented, each accounting for 4% of the
2.2.2. Board diversity and strategic decisions total. Portugal, Ireland, Luxembourg, Romania, Lithuania, Bulgaria,
The literature on board diversity and strategic decisions investigates Slovenia, Estonia, Slovakia, the Czech Republic, and Latvia have lower
the impact of board composition on several corporate strategy compo­
nents. For this study, we consider contributions on the relationship be­
tween board diversity and a firm’s acquisition policy, along with
contributions on board diversity and internationalisation. 10
As for alternative determinants of internationalisation, Rivas (2012) in­
Ahern and Dittmar (2012) show that changes in board composition vestigates the effects of diversity in age, tenure, and functional background of
following the adoption of the mandatory gender quota led to more ac­ board and top management team members in a sample of U.S. and European
quisitions among Norwegian firms. In contrast, based on large samples firms. Diversity in functional background only has a positive and significant
of U.S. firms, Levi et al. (2014) and Chen et al. (2016) report that effect on internationalisation.
11
gender-diverse boards are less likely to undertake acquisitions, and Kalemli-Ozcan et al. (2015) and Ribeiro et al. (2010) provide detailed in­
formation on the Orbis database.
when they do so, they select smaller targets and pay lower bid pre­ 12
The Statistical Classification of Economic Activities in the European Com­
miums. Similarly, Huang and Kisgen (2013) show that female executives
munity, commonly referred to as NACE (for the French term ‘nomenclature
undertake fewer acquisitions than male executives, and the market re­
statistique des activités économiques dans la Communauté Européenne’), is the
acts more favourably to acquisitions initiated by female executives. industry standard classification system used in the European Union. The current
As for internationalisation, there is little evidence on the role of version is revision 2 and was established by Regulation (EC) No 1893/2006.

4
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 2 As for industry, manufacturing (55%) and information and

Sample of firms, OFDI firms and noOFDI firms by industry. communication technologies (13%) industries account for the largest
Firms OFDI NoOFDI firms share. Construction, transportation, and professional activities each
constitute 5% of the sample, followed by administrative activities (4%),
Industry Number % Number % Number %
mining (3%), electricity (3%), accommodation activities (2%), arts
Accommodation 120 2% 66 55% 54 45% (2%), and agriculture (1%). Although various other industries are rep­
activities
Administrative 235 4% 213 91% 22 9%
resented, their share is negligible (Table 2).
activities OFDI involvement in our sample is remarkable, as 84% of the firms
Agriculture 65 1% 46 71% 19 29% have at least one subsidiary abroad during the study period. This evi­
Arts 95 2% 60 63% 35 37% dence is robust when switching to a country-wise analysis. As shown in
Construction 295 5% 219 74% 76 26%
Table 1, the percentage of OFDI firms surpasses 50% in all countries with
Education 5 0% 5 100% 0 0%
Electricity 190 3% 166 87% 24 13% Lithuania, Latvia, and Romania as the only exceptions.
Financial activities 10 0% 3 30% 7 70% OFDI involvement remains notable when controlling for industry.
Healthcare 25 0% 15 60% 10 40% Generally, more than 50% of the firms had at least one subsidiary abroad
ICT 765 13% 659 86% 106 14% from 2011 to 2015 except for the financial activities industry. As Table 2
Manufacturing 3160 55% 2806 89% 354 11%
Mining 195 3% 170 87% 25 13%
reveals, OFDI firms are particularly relevant in the manufacturing
Other service 20 0% 13 65% 7 35% (89%), administrative activities (91%), and education (100%) in­
activities dustries, whereas the financial activities, accommodation activities,
Professional 260 5% 228 88% 32 12% healthcare, and arts industries have the fewest firms with at least one
activities
subsidiary abroad between 2011 and 2015.
Transportation 270 5% 218 81% 52 19%
Fig. 2 demonstrates that our sample firms’ foreign subsidiaries are
Total 5710 100% 4887 86% 823 14% spread worldwide, and most are concentrated outside the European
Source: Authors’ elaboration from Orbis (2017) Union. Moreover, the number of firms with at least one subsidiary inside
the European Union (OFDI_EU firms) and those with at least one sub­
sidiary outside the European Union (OFDI_Extra_EU firms) tend to be
quite balanced (Fig. 3). Contrary to the general notion that European
countries are open to each other (Baldwin and Wyplosz, 2019), our
evidence suggests that European firms’ OFDI is not restricted to the
European Union where legislation is harmonised; rather, it crosses Eu­
ropean borders despite the higher degree of complexity implied by
opening subsidiaries in less familiar regions.13
It should be noted that despite European enterprises’ deep involve­
ment in OFDI, only 5.2% of our sample firms undertook OFDI for the
first time during the study period. Specifically, although we have a large
share of OFDI firms, our sample comprises only a small percentage of
OFDI starters. This poses certain constraints in the empirical analysis,
addressed in Section 4.
Board diversity is relevant in terms of both gender and nationality:
81% of our firms have at least one female director and 48% have at least
Fig. 2. Average number of foreign subsidiaries by destination. one foreign director. As displayed in Table 3, the results are consistent
Source: Authors’ elaboration from Orbis (2017) after controlling for country. Slovenia has the maximum share of firms
with at least one female director (100%), whereas Slovakia has the
minimum share (45%). Controlling for industry does not deliver any
remarkable differences; most firms in all countries have at least one
female director, with the maximum (100%) in the financial activities
industry and the minimum (63%) in the agricultural industry (Table 4).
Tables 3 and 4 compare the percentage of firms with at least one
foreign director by country and industry, respectively. As for country,
Luxembourg has the highest share (73%), whereas Lithuania has the
lowest (13%). Regarding industry, a high percentage of firms have at
least one foreign director in the healthcare (72%), mining (67%), and
transportation (53%) industries, while the percentages are particularly
low in the agriculture (37%), financial activities (20%), and other ser­
vice (20%) industries.

4. Econometric analysis

To address the effect of board diversity on OFDI, we start by esti­

mating the probit model reported in Equation (1), where Φ is the cu­
mulative distribution function of a standard normal distribution:
Fig. 3. The OFDI firms by destination.
Source: Authors’ elaboration from Orbis (2017) P(StartOFDIit = 1) =

shares, while a few countries are not represented in our data because the
firms headquartered there lack information regarding either their sub­
sidiaries or board composition. 13
We thank an anonymous reviewer for this remark.

5
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 3
Firms with at least one female director and firms with at least one foreign director, by country.
At least one female director No female directors At least one foreign director No foreign directors

Country Number % Number % Number % Number %

Austria 126 70% 54 30% 104 58% 76 42%

Belgium 200 80% 50 20% 135 54% 115 46%
Bulgaria 23 58% 17 43% 18 45% 22 55%
Czech Republic 12 60% 8 40% 14 70% 6 30%
Germany 685 65% 375 35% 477 45% 583 55%
Denmark 191 85% 34 15% 113 50% 112 50%
Estonia 15 60% 10 40% 6 24% 19 76%
Spain 238 72% 92 28% 151 46% 179 54%
Finland 299 98% 6 2% 158 52% 147 48%
France 1150 96% 45 4% 557 47% 638 53%
Great Britain 1239 79% 321 21% 774 50% 786 50%
Ireland 91 91% 9 9% 58 58% 42 42%
Lithuania 28 62% 17 38% 6 13% 39 87%
Luxembourg 38 54% 32 46% 51 73% 19 27%
Latvia 12 80% 3 20% 6 40% 9 60%
Poland 166 65% 89 35% 114 45% 141 55%
Portugal 106 85% 19 15% 67 54% 58 46%
Romania 23 51% 22 49% 12 27% 33 73%
Sweden 488 95% 27 5% 260 50% 255 50%
Slovenia 35 100% 0 0% 22 63% 13 37%
Slovakia 9 45% 11 55% 12 60% 8 40%

Total 5174 81% 1241 19% 3115 49% 3300 51%

Source: Authors’ elaboration from Orbis (2017)

Table 4
Firms with at least one female director and firms with at least one foreign director by industry.
At least one female director No female directors At least one foreign director No foreign directors

Industry Number % Number % Number % Number %

Accommodation activities 105 88% 15 13% 58 48% 62 52%

Administrative activities 192 82% 43 18% 111 47% 124 53%
Agriculture 41 63% 24 37% 24 37% 41 63%
Arts 78 82% 17 18% 36 38% 59 62%
Construction 229 78% 66 22% 136 46% 159 54%
Education 5 100% 0 0% 1 20% 4 80%
Electricity 168 88% 22 12% 113 59% 77 41%
Financial activities 10 100% 0 0% 2 20% 8 80%
Healthcare 21 84% 4 16% 18 72% 7 28%
ICT 655 86% 110 14% 360 47% 405 53%
Manufacturing 2489 79% 671 21% 1527 48% 1633 52%
Mining 171 88% 24 12% 131 67% 64 33%
Other service activities 19 95% 1 5% 4 20% 16 80%
Professional activities 212 82% 48 18% 113 43% 147 57%
Transportation 211 78% 59 22% 142 53% 128 47%

Total 4606 81% 1104 19% 2776 49% 2934 51%

Source: Authors’ elaboration from Orbis (2017)

Φ(α0 + α1 diversityit− 1 + α2 log(tangible fixed assetsit− 1 ) + α3 controlsit− 1 ) experience positively affects firm-level variables.15 Thus, the estimates
(1) would be biased due to endogeneity issues.16
Our main variable of interest, diversity, measures the proportion of
On the left-hand side of Equation (1), our dependent variable,
female and foreign directors on firm i’s board in year t and captures
StartOFDI, is a dummy that equals one if firm i has no subsidiary abroad
overall board diversity. We also account for distinct gender and
in years (t – 2) and (t – 1) and has at least one subsidiary in year t, and
zero otherwise.14 Wagner (2007) indicates that this is a suitable
dependent variable to develop causal inferences when analysing
self-selection mechanisms, as this removes reverse causality arising from 15
Borin and Mancini (2016) and Baiardi et al. (2021) adopt the same
the learning-by-OFDI effect. Considering either a dummy for OFDI status approach to investigate the productivity differentials among firms before they
or the actual number of OFDI rather than an OFDI-start variable, would invest abroad.
16
not exclude the learning-by-OFDI mechanism by which previous OFDI It is also noteworthy that we replicated our analysis by imposing an addi­
tional restriction, in which a firm without subsidiaries is removed from the
sample after subsidiaries are gained, i.e. after the switch from StartOFDI = 0 to
StartOFDI = 1. The resulting sample was an unbalanced panel dataset with
14
Our panel’s five-year span prevents us from adopting longer time spans in significantly fewer observations. Inference based on this restricted sample ex­
constructing StartOFDI. hibits inflated standard errors, and therefore, the results (available upon
request) were not particularly informative. Given the above, we opted to retain
the advantages of a larger dataset compared to those implied by the previously
mentioned stricter definition of StartOFDI.

6
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 5 Table 6
Summary statistics of the independent variables. The effect of board diversity on the probability of initiating outward foreign
Variable Obs. Mean Std. Dev. Min Max
direct investment.
StartOFDI (1) (2)
gender diversity 6371 0.0776 0.0790 0 1
nationality diversity 6371 0.0540 0.0934 0 0.8571 gender diversity − − 0.2220***
diversity 6371 0.1316 0.1303 0 1 (− ) (1.7736/1.26)
labour productivity 5739 220.0409 2067.641 0.8122 78373.84 nationality diversity − − 0.1282**
tangible fixed assets 6310 2,221,080 9,363,192 0 1.67e+08 (− ) (1.3505/1.32)
log(TFP) 5724 5.1983 0.7391 − 0.3326 10.0427 diversity − 0.3606*** −
board size 6415 27.6689 16.1448 0 344 (1.4291/1.33) (− )
dual regime 6415 0.3256 0.4687 0 1 log(tangible fixed assets) 0.0048** 0.0009
OFDI memory 6415 0.8536 0.3531 0 1 (0.0372/1.91) (0.0244/1.91)
IFDI memory 6415 0.9641 0.1859 0 1 board size − 0.0014*** − 0.0009***
OFDI Extra EU 6415 0.7415 0.4378 0 1 (0.0052/2.05) (0.0064/2.08)
high_labour 6415 0.3291 0.4699 0 1 dual regime 0.0104 0.0129*
productivity (0.2323/3.48) (0.1915/3.49)
low_labour 6415 0.2237 0.4167 0 1
productivity country-controls yes yes
high_TFP 6415 0.3308 0.4705 0 1 year-controls yes yes
low_TFP 6415 0.2231 0.4163 0 1 industry-controls yes yes

observations 3098 3098

Notes: This table reports number of observations, means, standard deviations,
correct classifications 98.52% 96.00%
minimum and maximum values of the independent variables used in our study. Wald test of exogeneity 5.96** 20.66**
In the Appendix, we provide a detailed definition of all the variables. log-likelihood − 211.62 − 221.25

Notes: This table reports probit regression results for the effects of diversity
nationality effects on OFDI by considering two additional independent (column 1), and gender and nationality diversity (column 2) on the probability of
variables, gender diversity and nationality diversity, which capture the initiating outward foreign direct investment. All results are estimated from
proportion of female and foreign directors on firm i’s board in year t, Equation (1). Average marginal effects are displayed. Robust, firm-clustered
respectively. We note that consistent with empirical evidence presented standard errors/VIF coefficients are in round brackets. Superscript *, **, and
in Section 1, board heterogeneity in our sample increases over the period *** corresponds to statistical significance at one-, five- and ten-percent level,
between 2011 and 2015. On average, across the sample firms, the pro­ respectively.
portion of female directors is approximately 0.043 in 2011 and increases
to 0.101 in 2015. The proportion of foreign directors increases from second stage nonlinear regression of the binary outcome StartOFDI on
0.0006 to 0.082. our measure of diversity. Therefore, the first stage includes a linear
In addition to our measures of board diversity, on the right-hand side regression of our endogenous variable(s) on the available instruments,
of Equation (1), we include a proxy for firm size (tangible fixed assets) in similar to the first stage of a standard 2SLS procedure, to obtain the
year (t – 1) and additional controls, such as a dummy for the corporate corresponding residuals. The second step includes these residuals as an
governance regime, which is either a dual- or unitary-board system; a additional regressor in our main nonlinear model of interest.
measure for board size at (t – 1); and industry-, country-, and year-fixed The literature on board diversity and strategic decisions does not
effects.17 At this stage, it is worth mentioning that we do not include firm commonly employ CF methods, although they have been suggested by
fixed effects in our model, as it would pose incidental parameter prob­ Dang et al. (2020), to address endogeneity when assessing board
lems in light of our limited time span. Table 5 displays the summary diversity’s impact on firm performance. Specifically, Dang et al. (2020)
statistics of the independent variables used in our analysis, while their use CFs to address endogeneity concerns in the context of a correlated
definitions and pairwise correlations are reported in the Appendix. random-effects method, whereas our model requires CFs to account for
Based on the literature on board diversity, performance, and stra­ endogenous regressors in a nonlinear (probit) model with a binary
tegic decisions, we expect diversity, gender diversity, and national diversity dependent variable. Although the fixed-effects and within-group esti­
to be endogenous (Adams et al., 2010); thus, we estimate Equation (1) mations of Equation (1) are, in principle, an obvious course of action to
with a probit specification using a CF method.18 mitigate endogeneity due to, for instance, omitted time-invariant vari­
The CF method is suitable for estimating a nonlinear model with a ables, the resulting estimates may not be particularly informative. This is
binary dependent variable in the presence of continuous endogenous because the main variables of interest, the variables capturing diversity,
regressors (Rivers and Vuong, 1988). Generally, CF-based techniques do not vary sufficiently over time across our time dimension.
rely on the same types of identification conditions as IV/2SLS, and are We construct our instrument for diversity as the sum of male and
used to obtain consistent estimates in nonlinear models, given the native directors serving on other boards within the same country that
presence of endogenous independent variables. In summary, in its contain at least one female or foreign director. Similarly, the instrument
simplest formulation for our setup, the two-step CF technique involves for gender (nationality) diversity is defined as the number of male (native)
residuals from a first-stage linear model entered as a regressor in a directors of firm i in year t serving on other boards, within the same
country, that contain at least one female (foreign) director.19 Our choice
of instruments relies on Adams and Ferreira’s (2009) methodology and
17
Depending on national regulations, European firms are governed by a subsequent works by Levi et al. (2014), Green and Homroy (2018), and
unitary system with a single board, or by a dual system with two separate Estelyi and Nisar (2016). The rationale behind the proposed instruments
boards. In the unitary system, the single board is comprised of executive and is that board members are recruited from directors’ professional and
non-executive directors and performs both advisory and monitoring functions. social networks. Thus, the broader the directors’ connections with het­
The monitoring role in the dual system is assigned to a supervisory board, while erogeneous peers, the more likely the appointment of diverse board
the management board is only comprised of executives and addresses man­
members.
agement issues (Fedorets et al., 2019; Ferreira and Kirchmaier, 2013). We
We satisfy the exclusion restriction by ensuring that our instrument
compute the board size in dual-board companies as by the union between su­
pervisory and management boards as in Ferreira and Kirchmaier (2013) and
Green and Homroy, 2018.
18 19
Corresponding results for a linear specification of Equation (1) obtained by Standard tests for exogeneity as reported in each table confirm the need for
IV/2SLS are available from the authors upon request. IV regressions.

7
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 7 round brackets) and the Variance Inflation Factor (VIF) coefficients
The effect of board diversity on the probability of initiating outward foreign (second figure in round brackets) to evaluate whether multicollinearity
direct investment with performance as a modifier. among regressors casts doubts on the reliability of our results.21 We
StartOFDI (1) (2) apply the multicollinearity rule of thumb, which states that multi­
gender diversity − − 0.1616**
collinearity is not an issue as long as VIF <10.
(− ) (0.0681/2.02) The results in Table 6 reveal a strong effect of board diversity on
nationality diversity − − 0.0716** OFDI. After controlling for firm and board size, the existence of dual-
(− ) (0.0288/2.53) board systems, industry, country, and year, diversity, gender diversity,
diversity − 0.1946** −
and nationality diversity are found to be negative and statistically sig­
(0.0764/2.04) (− )
log(labour productivity) 0.0067* 0.0050* nificant. Thus, European firms with diverse boards are less likely to
(0.0037/1.62) (0.0029/1.63) initiate OFDI. Specifically, an increase of one percentage point in di­
gender diversity* − − 0.0395* versity is associated with an average decrease of 0.0036 in the proba­
high_labour productivity (− ) (0.0229/1.91) bility of initiating OFDI. Furthermore, column 2 in Table 6 reveals that,
nationality diversity* high_labour productivity − 0.0622
on average, female directors have larger-magnitude, negative impacts
−
(− ) (0.0447/2.15)
diversity* − 0.0330*** − than foreign directors.
high_labour productivity (0.0111/1.72) (− ) Results in Table 6 reveal a strong negative correlation between di­
gender diversity* − − 0.0370 versity and the probability of initiating OFDI. However, these findings
low_labour productivity (− ) (0.0750/1.75)
might be unreliable since the role of performance is not accounted for.
nationality diversity* low_labour productivity − 0.0355
(− ) (0.0799/1.97) In what follows, we distinguish between two different roles poten­
diversity* 0.0090 − tially played by performance in our context; accordingly, we investigate
low_labour productivity (0.0335/1.57) (− ) the role of performance as ‘modifier’ and ‘mediator’ of the board
log(tangible fixed assets) 0.0018* 0.0004 diversity-OFDI relationship.
(0.0011/1.98) (0.0008/1.98)
As for the modifying role of performance, it should be noted that
board size − 0.0007*** − 0.0005**
(0.0002/2.06) (0.0002/2.10) opening subsidiaries abroad entails fixed costs that only the most pro­
dual regime 0.0068* 0.0081** ductive firms, able to command a large market share, find profitable to
(0.0041/3.60) (0.0038/3.61) incur. Previous literature documents that firm performance has a posi­
country-controls Yes Yes tive and statistically significant effect on OFDI, confirming that firms
year-controls Yes Yes self-select into internationalisation.22 Therefore, we suggest that the
industry-controls Yes Yes magnitude of the board diversity’s effect on OFDI may differ according
observations 4071 4071 to firm performance. As for the mediating role of performance, it should
root MSE 0.9997 0.0985 be considered that board diversity is likely to affect firm performance.
test of exogeneity 2.69* 5.95***
Board composition shapes board’s effectiveness as monitor and advisor
Notes: This table reports linear regression results for the effect of diversity for management and thus it is likely to affect firm performance. Since
(column 1), and gender and nationality diversity (column 2) on the probability of firms self-select into internationalisation according to performance, we
initiating outward foreign direct investment, controlling for performance. The expect board diversity to influence OFDI through performance. Previous
interaction terms between diversity (column 1), gender and nationality diversity literature suggests that board diversity mostly has a significant positive
(column 2) and the dummies high_labour productivity and low_labour productivity
effect on performance, while performance has a significant positive ef­
are the explanatory variables of interest. All results are estimated from Equation
fect on OFDI.23 Thus, we expect firm performance to act as mediator of
(2). Regression coefficients are displayed. Robust, firm-clustered standard er­
board diversity on OFDI.
rors/VIF coefficients are in round brackets. Superscript *, **, and *** corre­
sponds to statistical significance at one-, five- and ten-percent level, respectively. We start from investigating whether the negative observed rela­
tionship between OFDI and board diversity varies according to the
correlates with the probability of opening a foreign subsidiary only performance level. To test the role of firm performance as modifier
through exogenous variables that are already included in the regression (moderator or amplifier) of board diversity, we rely on a linear model
model. We might expect the degree of connections among directors with with suitable interaction terms.24
heterogeneous peers, which forms the basis of our instruments, to relate The adoption of a linear probability model has the advantage of
to performance and/or the probability of opening a foreign subsidiary preserving a straightforward implementation and interpretation of co­
through industry- or country-specific effects. By controlling for both efficients even in presence of (potentially endogenous) interactive
industry- and country-fixed effects in all models, we expect that our terms. To evaluate explicitly whether firm performance modifies the
instruments will satisfy the exclusion restriction, while the results of the effect of board diversity on OFDI, Equation (2) is set as follows:
first-stage regression confirm that our instruments also satisfy the rele­
vance condition.20

P(StartOFDIit = 1) =
β0 + β1 diversityit− 1 + β2 log(labour productivityit− 1 )+ (2)
β3 diversityit− 1 ∗ high labour productivityit− 1 + β4 diversityit− 1 ∗ low labour productivityit− 1 +
β5 log(tangible fixed assetsit− 1 ) + β6 controlsit− 1

Table 6 reports results for Equation (1). Specifically, we display

average marginal effects, firm-clustered standard errors (first figure in 21
For probit specifications, regression coefficients are available from the au­
thors upon request.
22
See Section 2.1.
23
See Section 2.2.1 and Section 2.1.
20 24
Results are available from the authors upon request. See Bose et al. (2020) and Bose et al. (2021).

8
V. Gattai et al. Economic Modelling 120 (2023) 106156

The regressors in Equation (2) include measures of board diversity when diversity increases by 0.1, the probability of initiating OFDI de­
and firm performance. We measure performance by means of creases by 0.023 when a firm is in the top quartile in the productivity
labour productivity computed as the ratio of value added to employees. spectrum, compared to a decrease of approximately 0.020 for firms with
An alternative measure of performance is explored in Section 5. We note lower productivity. The figures in column (2) of Table 7 confirm the
that, as discussed above, the choice of the dummy variable StartOFDI as overall pattern of column (1), even though the interaction term
dependent variable ensures that labour productivity remains exogenous involving nationality diversity and high_labour productivity is not statisti­
in Equation (2). cally significant.
In addition to measures of board diversity and performance, the In brief, the effect of board diversity on OFDI is sensitive to firm
right-hand side of Equation (2) contains the interactions of diversity with performance, such that OFDI by the most productive firms responds the
the binary variables high labour productivity and low labour productivity, most to changes in board diversity. Thus, firm performance can be
which take the value of one for observations in the fourth and first considered a modifier (amplifier) of the effect of board diversity on OFDI
quartiles, respectively, of the spectrum of labour productivity and zero for top-performing firms.
elsewhere. Finally, the proxies for firm size and controls are the same as Having established that firm performance acts as modifier of the
those in Equation (1). effect of board diversity on OFDI, we now assess whether it acts as
As we expect diversity, gender diversity, and national diversity to be mediator of board diversity. We identify the direct and performance-
endogenous, we estimate Equation (2) using the IV/2SLS methodology. mediated effect of board diversity on OFDI by the causal steps strategy
We construct our instrument for diversity, gender diversity, and national (Judd and Kenny, 1981a,b; Baron and Kenny, 1986; Preacher and Hayes,
diversity, as described above. Table 7 reports the coefficient estimates of 2008) and quantify it by following the procedure described by MacK­
the parameters in Equation (2), firm-clustered standard errors (first innon and Dwyer (1993).25
figure in round brackets), and the VIF coefficients (second figure in Firm performance (labour productivity in our analysis) can be defined
round brackets). The exogeneity test confirms that IV/2SLS is needed, as as mediator of board diversity if: i) board diversity is a significant
opposed to the standard ordinary least squares approach. determinant of OFDI; ii) board diversity significantly accounts for
The results in Table 7 reveal a strong negative effect of board diversity variability in performance; iii) performance is a significant determinant
on OFDI, as opposed to a significant positive impact of labour produc­ of OFDI and remains significant when controlling for board diversity;
tivity. The first column of Table 7 shows that the negative effect of di­ and iv) the effect of board diversity on OFDI is different in magnitude
versity is the strongest for firms in the fourth quartile of our performance when performance and board diversity enter the model. The approach is
measure distribution. Specifically, controlling for labour productivity, therefore based on Equation 1, and 3-5, in which, as previously dis­
cussed, the definition of StartOFDI ensures exogeneity of
labour productivity:
log(labour productivityit ) =
(3)
γ0 + γ 1 diversityit + γ 2 log(tangible fixed assetsit ) + γ3 controlsit + εit

P(StartOFDIit = 1) =
Φ(δ0 + δ1 log(labour productivityit− 1 ) + δ2 log(tangible fixed assetsit− 1 )+
δ3 controlsit− 1 )
(4)

Fig. 4. Illustration of the causal steps strategy applied to our context. P(StartOFDIit = 1) =
Φ(θ0 + θ1 diversityit− 1 + θ2 log(labour productivityit− 1 )+ (5)
θ3 log(tangible fixed assetsit− 1 ) + θ4 controlsit− 1 )
Table 8
As for Equation (1), we adopt a probit specification to obtain the
The effect of board diversity on performance.
estimates of Equations (4) and (5).26 To identify the mediated effect, we
log(labour productivity) (1) (2) require the statistical significance of coefficient α1 in Equation (1), co­
gender diversity − 2.5047** efficient γ1 in Equation (3), coefficient δ1 in Equation (4), and co­
(− ) (1.1995/1.30) efficients θ1 and θ2 in Equation (5). Moreover, we require the magnitude
nationality diversity 1.0081*
of coefficients α1 and θ1 to be different. Fig. 4 graphically summarises
−
(− ) (0.5979/1.29)
diversity 1.9995** –− the causal steps strategy.
(0.8897/1.34) (− ) Alternative identification strategies for direct and mediated effects
log(tangible fixed assets) 0.0181 0.0242* have been suggested in the literature, such as a joint estimation of
(1.10/1.86) (0.0136/1.86)
Equations (3) and (5) by the seemingly unrelated regression estimator
board size 0.0069*** 0.0068***
(0.0020/1.82) (0.0020/1.84)
(SURE).27 However, SURE would not allow a probit specification for
dual regime − 0.0147 − 0.0141 Equation (5) together with the linear model in Equation (3) and,
(0.0471/3.73) (0.0472/3.73) crucially, it would not be consistent as board diversity is endogenous.
country-controls yes yes Instead, the causal steps strategy allows us to adopt alternative methods
year-controls yes yes to separately estimate each equation and quantify the direct and medi­
industry-controls yes yes ated effects ex-post. After estimating the parameters of Equations (3)–(5)
observations 5085 5085 and testing their significance, the quantification of direct and mediated
R2 0.1878 0.2050 effects follows the procedure described by MacKinnon and Dwyer
root MSE 0.7250 0.7173
Robust score test for endogeneity 4.5141** 4.6159*

Notes: This table reports linear regression results for the effect of diversity
(column 1), and gender and nationality diversity (column 2) on firm performance. 25
See Colombo et al. (2018) and Li et al. (2021).
All results are estimated from Equation (3). Regression coefficients are dis­ 26
A standard linear probability model has also been estimated and results are
played. Robust, firm-clustered standard errors/VIF coefficients are in round available from the authors upon request.
brackets. Superscript *, **, and *** corresponds to statistical significance at one- 27
For a literature review of the available strategies, see Preacher and Hayes
, five- and ten-percent level, respectively. (2008).

9
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 9 in round brackets) confirm that multicollinearity among independent

The effect of performance and board diversity on the probability of initiating variables does not cast doubt on the reliability of the regression esti­
outward foreign direct investment. mates in Equation (5).
StartOFDI (1) (2) (3) In view of the estimates of Equations (1) and (3)–(5), we quantify the
gender diversity − − − 0.2544***
direct and mediated effects of our variable of interest, diversity (or gender
(− ) (− ) (1.3219/1.28) diversity/nationality diversity), following MacKinnon and Dwyer (1993).
nationality diversity − − − 0.0862 First, we scale the estimates of Equation (5) by the standard deviation of
(− ) (− ) (1.7041/1.31) the predicted probabilities to account for the fact that the scale in probit
diversity − 0.4089***
− −
regressions depends on the extent of the prediction, which in turn de­
(− ) (1.2029/1.33) (− )
log(labour productivity) 0.0044** 0.0061* 0.0049** pends on the variables in the model.29 The estimates of Equation (3) do
(0.0654/1.34) (0.0567/1.34) (0.0591/1.34) not require scaling since they are a standard linear model. We then
log(tangible fixed assets) − 0.0021*** 0.0056*** 0.0006 define the performance-mediated effect of board diversity on OFDI as
(0.0302/1.85) (0.0220/1.97) (0.0237/1.97) the product of γ1 in Equation (3) and the standard deviation-scaled θ2 in
board size − 0.0004** − 0.0015*** − 0.0008***
(0.0068/2.00) (0.0050/2.06) (0.0062/2.10)
Equation (5). On the other hand, the standard deviation-scaled θ1 in
dual regime 0.0151** 0.0111 0.0123** Equation (5) captures the direct (not mediated by performance) effect of
(0.2275/3.53) (0.1551/3.60) (0.1409/3.60) diversity on OFDI. We remark that estimates of Equation (1) and
country-controls yes yes Yes Equation (4) do not provide causal effects on their own, as the statistical
year-controls yes yes Yes significance of θ1 and θ2 in Equation (5) suggests the presence of
industry-controls yes yes Yes endogeneity due to omitted variables in both Equations (1) and (4).
observations 2739 2732 2732 Equations (1), (3) and (4) should be regarded as auxiliary models to
correct classifications 98.54% 98.54% 98.52% identify the mediated effect of diversity on OFDI, with performance
Wald test of exogeneity − 7.78** 13.34*** being the mediating variable. We nevertheless retain results of Equa­
log-likelihood − 183.2 − 180.44 − 180.57
tions (1), (3) and (4) in our tables to clearly display the magnitude and
Notes: This table reports probit regression results for the effects of performance significance of such coefficients, on which the causal steps methodology
on the probability of initiating outward foreign direct investment as estimated heavily relies, even though we can only attribute a correlation inter­
from Equation (4) (column 1) and for the effects of diversity (column 2), and pretation to estimates of Equations (1) and (4).30
gender and nationality diversity (column 3) on the probability of initiating out­ Quantitatively, the standard deviation of the predicted probabilities
ward foreign direct investment, controlling for performance, as estimated in
in Equation (3) is 0.0982. Hence, the magnitude of the direct impact of
Equation (5). Average marginal effects are displayed. Robust, firm-clustered
diversity on OFDI is − 7.292/0.0982 = − 74.18, while the mediated effect
standard errors/VIF coefficients are in round brackets. Superscript *, **, and
*** corresponds to statistical significance at one-, five- and ten-percent level,
is 1.999*0.1085/0.0982 = 1.61. The magnitudes of the latter figures are
respectively. difficult to interpret in the context of a nonlinear model. Intuitively, we
can conclude that variations in performance explain (in absolute value)
about 2%, i.e. {[1.61/(1.61–74.18)]*100%}, of the effect of diversity on
(1993) to accommodate the nonlinear and binary nature of Equations
OFDI, and that direct and mediated effects have opposite signs, as ex­
(2), (4) and (5). Specifically, we estimate Equation (3) using standard
pected. Similarly, we find that variation in performance explains about
IV/2SLS and maximum likelihood estimators, while adopting a CF
5% and 6% of the effect of board diversity on OFDI due to gender and
technique to estimate Equations (2) and (5).28
nationality, respectively, with direct and mediated effects of opposite
Table 8 reports the estimates of the mediating variable model in
signs.
Equation (3). We address endogeneity of board diversity by relying on
It is worth mentioning that our result on the direct effect of gender-
the same instruments for diversity, gender diversity, and nationality di­
diverse boards on initiating OFDI is consistent with previous evidence
versity introduced above. Our results confirm that board diversity
on the relationship between board gender diversity and mergers and
strongly and positively affects firm performance. An increase of one
acquisition (M&A) activities. Huang and Kisgen (2013), Levi et al.
percentage point in diversity is associated with a 1.99% increase in
(2014), and Chen et al. (2016) report that gender-diverse boards are less
performance when controlling for firm characteristics and industry-,
likely to undertake acquisitions; and when they do so, they select smaller
country- and year-fixed effects.
targets and pay lower bid premiums.
Table 9 reports the estimates of Equations (4) and (5), and our results
OFDI is a strategic decision that requires board-level approval and
confirm the statistical significance of δ1 and, crucially, of θ2 . Thus,
thus it reflects the effectiveness of the board in its dual role as monitor
performance remains significant when controlling for board diversity.
and advisor for the management (Adams and Ferreira, 2007; Levi et al.,
Consistent with the results in Table 6, Table 9 confirms that perfor­
2014). The literature on board diversity, performance and strategic
mance and diversity variables sort effects of opposite sign on the prob­
decisions shows that diversity affects the board’s effectiveness in either
ability of initiating OFDI even though column (3) of Table 9 shows that
role through its members’ skills and expertise (Kim and Starks, 2016;
nationality diversity does not produce a statistically significant effect.
Masulis et al., 2012; Stroup, 2017), their independence (Adams and
From column (2) in Table 9, an increase of one percentage point in di­
Ferreira, 2009), industry and political connections (Ferreira, 2010;
versity is associated with an average decrease of 0.0041 in the proba­
Sherman et al., 1998), and risk-taking attitude (Bernile et al., 2018).
bility of initiating OFDI, and this effect is strongly significant. In
Our findings on the effect of board diversity on the probability of
addition, the estimates of α1 in Equation (1) and θ1 in Equation (5) differ
initiating OFDI could be framed either in terms of tough monitoring or in
in the expected direction. This implies that the effect of board diversity is
terms of risk aversion.
larger in absolute value when we control for performance as in Table 9,
On the one hand, empirical analyses reveal that gender-diverse
as opposed to the corresponding coefficients in Table 6, in which the
boards are tough management monitors (Adams and Ferreira, 2009;
(positive) mediating effect of performance is not controlled for.
We outline that the VIF coefficients reported in Table 9 (second digit

29
Not reported in Table 9 for the sake of space, the estimated values of co­
efficients of diversity and log(labour productivity) of Equation (5) are equal to
28
Given our definition of the dependent variable, Equation (4) is free from − 7.292 and 0.1085, respectively.
30
endogeneity and parameters can be estimated by maximum likelihood, as it is In Section 5, we show that our results are robust to a change in the measure
customary for standard probit models. of performance as mediating variable.

10
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 10
The effect of diversity and performance on the probability of initiating outward foreign direct investment, controlling for OFDI memory.
StartOFDI (1) (2) (3) (4) (5)

gender diversity − − 0.1775*** − − − 0.2232***

(− ) (1.5931/1.26) (− ) (− ) (1.2968/1.28)
nationality diversity –− − 0.0839 − − − 0.0427
(− ) (1.6250/1.32) (− ) (− ) (2.2116/1.32)
diversity − 0.2872*** − − − 0.3668*** −
(1.6376/1.34) (− ) (− ) (1.2173/1.33) (− )
log(labour productivity) − − 0.0039** 0.0058** 0.0043**
(− ) (− ) (0.0612/1.34) (0.0528/1.34) (0.0646/1.34)
log(tangible fixed assets) 0.0041** 0.0009 − 0.0012 0.0055*** 0.0008
(0.0365/1.92) (0.0202/1.92) (0.0289/1.86) (0.0181/1.98) (0.0238/1.98)
board size − 0.0010*** − 0.0005*** − 0.0001 − 0.0012*** − 0.0005***
(0.0041/2.08) (0.0052/2.11) (0.0052/2.03) (0.0033/2.10) (0.0050/2.13)
dual regime 0.0092 0.0119 0.0131*** 0.0092 0.0107**
(0.2415/3.49) (0.2010/3.49) (0.1628/3.54) (0.1537/3.60) (0.1564/3.60)
OFDI memory − 0.0232 − 0.0283** − 0.0314*** − 0.0213 − 0.0303**
(0.4087/1.23) (0.3320/1.23) (0.4013/1.21) (0.4501/1.21) (0.4652/1.21)

country-controls Yes Yes Yes Yes Yes

year-controls Yes Yes Yes Yes Yes
industry-controls Yes Yes Yes Yes Yes

observations 3098 3098 2739 2732 2732

correct classifications 97.97% 98.55% 98.54% 98.54% 98.50%
Wald test of exogeneity 5.00** 17.45*** − 8.93*** 23.68***
log-likelihood − 192.75 − 191.42 − 164.15 − 162.11 − 161.31

Notes: This table reports probit regression results for the effects of diversity (column 1), and gender and nationality diversity (column 2) on the probability of initiating
outward foreign direct investment, controlling for past OFDI experience, as estimated in Equation (1); for the effects of performance on the probability of initiating
outward foreign direct investment, controlling for past OFDI experience, as estimated from Equation (4) (column 3); for the effects of diversity (column 4) and gender
and nationality diversity (column 5) on the probability of initiating outward foreign direct investment, controlling for performance and past OFDI experience, as
estimated in Equation (5). Average marginal effects are displayed. Robust, firm-clustered standard errors/VIF coefficients are in round brackets. Superscript *, **, and
*** corresponds to statistical significance at one-, five- and ten-percent level, respectively.

Table 11
The effect of diversity and performance on the probability of initiating outward foreign direct investment, controlling for IFDI memory.
StartOFDI (1) (2) (3) (4) (5)

gender diversity − − 0.2180*** − − − 0.2473***

(− ) (1.7064/1.26) (− ) (− ) (1.2927/1.28)
nationality diversity − − 0.1255** − − − 0.0811
(− ) (1.4090/1.32) (− ) (− ) (1.7808/1.32)
diversity − 0.3592*** − − − 0.4056*** −
(1.4665/1.34) (− ) (− ) (1.2445/1.33) (− )
log(labour productivity) − − 0.0043* 0.0060* 0.0048*
(− ) (− ) (0.0702/1.34) (0.0561/1.34) (0.0703/1.34)
log(tangible fixed assets) 0.0048** 0.0009 − 0.0019** 0.0055*** 0.0006
(0.0374/1.91) (0.0243/1.91) (0.0238/1.86) (0.0222/1.98) (0.0243/1.98)
board size − 0.0014*** − 0.0009*** − 0.0003** − 0.0015*** − 0.0008***
(0.0051/2.06) (0.0064/2.10) (0.0053/2.01) (0.0048/2.08) (0.0062/2.11)
dual regime 0.0104 0.0133* 0.0157*** 0.0111 0.0128**
(0.2370/3.49) (0.1921/3.50) (0.1497/3.54) (0.1605/3.60) (0.1451/3.61)
IFDI memory 0.0077 − 0.0020 − 0.0099 0.0080 − 0.0047
(0.1255/1.08) (0.1464/1.09) (0.1851/1.09) (0.1327/1.09) (0.1560/1.09)

country-controls Yes Yes Yes Yes Yes

year-controls Yes Yes Yes Yes Yes
industry-controls Yes Yes Yes Yes Yes

observations 3098 3098 2739 2732 2732

correct classifications 96.68% 98.52% 98.54% 95.35% 98.50%
Wald test of exogeneity 5.82** 14.71*** − 7.52*** 14.08***
log-likelihood − 211.30 − 210.50 − 182.6102 − 180.04 − 180.10

Notes: This table reports probit regression results for the effects of diversity (column 1), and gender and nationality diversity (column 2) on the probability of initiating
outward foreign direct investment, controlling for past presence of foreign shareholders, as estimated in Equation (1); for the effects of performance on the probability
of initiating outward foreign direct investment, controlling for past presence of foreign shareholders, as estimated from Equation (4) (column 3); for the effects of
diversity (column 4) and gender and nationality diversity (column 5) on the probability of initiating outward foreign direct investment, controlling for performance and
past presence of foreign shareholders, as estimated in Equation (5). Average marginal effects are displayed. Robust, firm-clustered standard errors/VIF coefficients are
in round brackets. Superscript *, **, and *** corresponds to statistical significance at one-, five- and ten-percent level, respectively.

Green and Homroy, 2018). In light of this, we expect a more severe 1980; Lewellen et al., 1989). However, gender-diverse boards would
scrutiny of international investment projects by diverse boards. Man­ impose discipline on the management and curb ‘empire building
agers may exploit their superior information on the profitability of OFDI through OFDI’ by rigorously monitoring managerial activities and
to increase the number of assets they control (Hannan and Mavinga, projects.

11
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 12
The effect of diversity and performance on the probability of initiating outward foreign direct investment, controlling for the presence of OFDI Extra_EU.
StartOFDI (1) (2) (3) (4) (5)

gender diversity − − 0.2144*** − − − 0.2499***

(− ) (1.6821/1.26) (− ) (− ) (1.3053/1.28)
nationality diversity − − 0.1173* − − − 0.0830
(− ) (1.5965/1.32) (− ) (− ) (1.9676/1.32)
diversity − 0.3621*** − − − 0.4159*** −
(1.3991/1.34) (− ) (− ) (1.2217/1.34) (− )
log(labour productivity) − − 0.0042* 0.0062* 0.0047*
(− ) (− ) (0.0673/1.35) (0.0615/1.34) (0.0695/1.34)
log(tangible fixed assets) 0.0050*** 0.0010 − 0.0019** 0.0057*** 0.0007
(0.0311/1.92) (0.0256/1.92) (0.0282/1.87) (0.0192/1.98) (0.0269/1.98)
board size − 0.0014*** − 0.0008*** − 0.0003** − 0.0015*** − 0.0008***
(0.0043/2.08) (0.0056/2.12) (0.0047/2.03) (0.0039/2.10) (0.0049/2.13)
dual regime 0.0104 0.0133* 0.0155*** 0.0111 0.0127**
(0.2330/3.48) (0.2013/3.49) (0.1519/3.53) (0.1639/3.60) (0.1491/3.60)
OFDI extra EU − 0.0027 − 0.0088 − 0.0070 0.0035 − 0.0048
(0.2446/1.21) (0.2296/1.21) (0.2287/1.18) (0.2492/1.19) (0.2787/1.19)

country-controls Yes Yes Yes Yes Yes

year-controls Yes Yes Yes Yes Yes
industry-controls Yes Yes Yes Yes Yes

observations 3098 3098 2739 2732 2732

correct classifications 98.48% 98.48% 98.54% 98.54% 98.54%
Wald test of exogeneity 6.74*** 11.94*** − 8.37*** 12.66***
log-likelihood − 209.02 − 208.13 − 182.32 − 179.66 − 179.72

Notes: This table reports probit regression results for the effects of diversity (column 1), and gender and nationality diversity (column 2) on the probability of initiating
outward foreign direct investment, controlling for OFDI region of destination, as estimated in Equation (1); for the effects of performance on the probability of
initiating outward foreign direct investment, controlling for OFDI region of destination, as estimated from Equation (4) (column 3); for the effects of diversity (column
4) and gender and nationality diversity (column 5) on the probability of initiating outward foreign direct investment, controlling for performance and OFDI region of
destination, as estimated in Equation (5). Average marginal effects are displayed. Robust, firm-clustered standard errors/VIF coefficients are in round brackets. Su­
perscript *, **, and *** corresponds to statistical significance at one-, five- and ten-percent level, respectively.

On the other hand, attitudes toward risk has been invoked to explain next section, we investigate the role of knowledge and experience in
the behaviour of gender-diverse boards engaging in M&A activities (Levi international markets to explain the fewer OFDI decisions by
et al., 2014; Bernile et al., 2018).31 As long as operating abroad increases nationality-diverse boards.35
variability in firm profits, gender-diverse boards would be reluctant to
initiate OFDI. 5. Robustness checks
Our evidence supports the view that gender-diverse boards reluc­
tance to approve OFDI is a matter of imposing managerial discipline. In Section 4, we establish a case for the negative effect of board di­
Indeed, controlling for performance, we show that the negative effect of versity on OFDI and firm performance as modifier and mediator of this
gender diversity on OFDI is the stronger, the more productive firms effect. This section explores the robustness along several dimensions of
are.32 Because of the fixed costs it entails, OFDI profitability is increasing the results, on which our case rests.36
in firm productivity. This places managers of highly productive firms in Our first extension addresses the issue that firms initiating OFDI
a better position to pursue ‘empire building through OFDI’.33 To curb it, between 2011 and 2015 may retain some OFDI experience from foreign
gender-diverse boards take a more severe stance toward initiating OFDI subsidiaries that have been opened and closed before our time win­
in highly productive firms.34 dow.37 Thus, we explore whether OFDI experience plays a role in our
To the contrary, risk-averse boards would take a more severe stance model by including a control variable to capture such an experience. We
toward initiating OFDI in less productive, and thus less profitable firms define a new binary variable, OFDI memory, which takes the value of one
or the same stance in all firms, depending on whether constant relative if firm i had foreign subsidiaries from 2007 to 2010 and zero other­
or absolute risk aversion prevails among board members. wise.38 Overall, 34 firms in our sample initiated OFDI from 2011 to
Considering the negative direct impact of nationality-diverse boards 2015, and did not have any foreign subsidiaries from 2007 to 2010.39
on OFDI decisions, evidence (Hahn and Lasfer, 2016; Frijns et al., 2016)
reveals that such boards face coordination problems and fail to impose
managerial discipline. However, having directors who are knowledge­ 35
Lack of detailed data on board meetings prevents us from investigating
able about foreign markets increases the likelihood of cross-border ac­
further alternative mechanisms determining board diversity’s effect on OFDI.
quisitions (Masulis et al., 2012; Stroup, 2017; Dou et al., 2019). In the More diverse boards tend to engage in lengthier deliberations (Levi et al., 2014)
and exhibit greater persistence in adhering to adopted policies (Bernile et al.,
2018). Cultural and language barriers among board members (Frijns et al.,
31
However, evidence shows that gender-driven differences in risk preference 2016; Piekkari et al., 2015) increase the costs of deliberation, and thus, may
observed in laboratory experiments disappear in professional populations explain the reluctance of diverse boards to initiate OFDI.
36
(Croson and Gneezy, 2009). We thank the associate editor and an anonymous reviewer for their valuable
32
See the results reported in Table 7. suggestions.
33 37
They can pledge larger profits to raise capital from investors and to By definition, OFDI starters have no foreign subsidiaries in the two previous
convince board members that no better alternative for firm’s funds is available. years; Section 4 provides further information.
34 38
In line with our interpretation, gender-diverse boards are more likely to Note that our panel does not cover the entire period spanning 2007 to 2015
impose high dividend payouts (Pucheta-Martinez and Bel-Oms, 2016; Chen because board data were only available from 2011 to 2015.
39
et al., 2017) and undertake share buybacks (Evgeniou and Vermaelen, 2017) to We emphasise that restricting our sample to 34 firms would decrease the
reduce the agency costs of free cash and curb ‘empire building’ by the number of relevant observations drastically, and thus, impact our results’
management. reliability.

12
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 13 Table 14
The effect of board diversity on the probability of initiating outward foreign The effect of diversity on performance, with performance measured by TFP.
direct investment with performance measured by TFP as a modifier. log(TFP) (1) (2)
StartOFDI (1) (2)
gender diversity − 5.5408***
gender diversity − − 0.1541*** (− ) (1.4873)
(− ) (0.0584/1.85) nationality diversity − 1.9535**
nationality diversity − − 0.0406 (− ) (0.8990)
(− ) (0.0492/2.46) diversity 4.3289*** −
diversity − 0.1751** − (1.0789) (− )
(0.0757/2.05) (− ) log(tangible fixed assets) 0.0485* 0.0630***
log(TFP) 0.0009** 0.0009** (0.0269) (0.0198)
(0.0005/1.48) (0.0004/1.38) board size 0.0138*** 0.0137***
gender diversity*high_TFP − − 0.0623 (0.0028) (0.0033)
(− ) (0.0427/1.96) dual regime 0.0280 0.0293
nationality diversity* high_TFP − − 0.0385 (0.0658) (0.0736)
(− ) (0.0667/2.20)
country-controls yes Yes
diversity*high_TFP − 0.0495** −
year-controls yes Yes
(0.0205/1.74) (− )
industry-controls yes Yes
gender diversity*low_TFP − − 0.0743
(− ) (0.0975/1.62) observations 5085 5076
nationality diversity* low_TFP − − 0.0439 R2 0.1278 0.1059
(− ) (0.0808/1.93) root MSE 0.7482 0.7127
diversity*low_TFP − 0.0425 − Robust score test for endogeneity 4.5141** 11.392***
(0.0322/1.51) (− )
log(tangible fixed assets) 0.0013 − 0.0001
Notes: This table reports linear regression results for the effect of diversity
(0.0011/2.06) (0.0004/2.08) (column 1), and gender and nationality diversity (column 2) on firm performance
board size − 0.0005*** − 0.0005** measured by TFP. All results are estimated from Equation (3). Regression co­
(0.0002/2.07) (0.0002/2.10) efficients are displayed. Robust, firm-clustered standard errors/VIF coefficients
dual regime 0.0111** 0.0115** are in round brackets. Superscript *, **, and *** corresponds to statistical sig­
(0.0057/3.60) (0.0050/3.71) nificance at one-, five- and ten-percent level, respectively.
country-controls Yes Yes
year-controls Yes Yes memory, which takes the value of one if firm i had some foreign share­
industry-controls Yes Yes
holders from 2007 to 2010 and zero otherwise.40 Only five firms initi­
observations 4050 4050 ating OFDI within our timeframe did not have any foreign shareholders
root MSE 0.1014 0.1001
from 2007 to 2010, which suggests that almost all firms in our sample
test of exogeneity 5.57*** 7.34***
have some IFDI experience. Table 11 reports the probit estimates of
Notes: This table reports linear regression results for the effect of diversity Equation (1) (columns 1 and 2), (4) (column 3), and (5) (columns 4 and
(column 1), and gender and nationality diversity (column 2) on the probability of 5) with the IFDI memory included among the controls. Our findings
initiating outward foreign direct investment, controlling for performance indicate that the additional IFDI memory control has no direct impact on
measured by TFP. The interaction terms between diversity (column 1), gender and
the probability of opening a foreign subsidiary from 2011 to 2015, and
nationality diversity (column 2) and the dummies high_TFP and low_TFP are the
our main results discussed in the context of Tables 6 and 9 are robust for
explanatory variables of interest. All results are estimated from Equation (2).
Regression coefficients are displayed. Robust, firm-clustered standard errors/ the inclusion of IFDI memory.
VIF coefficients are in round brackets. Superscript *, **, and *** corresponds to As discussed in Section 3, the firms in our sample have a balanced
statistical significance at one-, five- and ten-percent level, respectively. composition of foreign subsidiaries inside and outside the European
Union. However, we argue that investing outside the European Union is
Table 10 reports the probit estimates of Equation (1) (columns 1 and 2), more complex than investing internally, as legislation in the latter is
(4) (column 3), and (5) (columns 4 and 5) with the OFDI memory harmonised and business practices are familiar. Thus, we determine
included among the controls. whether the OFDI destination affects our previous findings regarding
We acknowledge that having foreign subsidiaries has a negative ef­ board diversity’s effect on the probability of opening a foreign subsidi­
fect on the probability of initiating OFDI. This is as anticipated, as most ary by defining an additional control, OFDI Extra_EU, which takes the
firms that had past foreign subsidiaries simply retained them in the value of one if firm i has at least one subsidiary outside the European
present; therefore, they are less likely to initiate OFDI within our Union and zero otherwise.41 Table 12 reports the probit estimates of
timeframe than firms with no past foreign subsidiaries. Importantly, Equation (1) (columns 1 and 2), (4) (column 3), and (5) (columns 4 and
comparing the estimates displayed in Tables 6 and 9, including OFDI 5) with OFDI Extra_EU included among the controls. Results indicate that
memory does not affect our overall results concerning board diversity’s OFDI Extra_EU is never significant, revealing that the OFDI destination is
impacts on OFDI. Additional information can be obtained when irrelevant to the probability of initiating OFDI. The comparison of Ta­
considering gender and nationality diversity. OFDI experience does not bles 6, 9 and 12 shows that the impacts of diversity, gender diversity,
affect the sign, relative magnitude, and significance of gender diversity, nationality diversity, and labour productivity are almost unchanged.
while nationality diversity remains insignificant. In summary, controlling The last robustness check that we consider differs in spirit and ad­
for OFDI experience does not modify the effect of gender diversity but dresses an alternative measure of performance. Our results in
reduces the effect of nationality diversity on the probability of opening a Tables 7–12 have been obtained using labour productivity as a measure
foreign subsidiary. This suggests that OFDI experience absorbs the na­ of performance. In Tables 13–15, we replicate our econometric analysis
tionality diversity effect, which is consistent with the results of Estelyi with total factor productivity (TFP) used as a measure of performance.
and Nisar (2016), Masulis et al. (2012) and Pisani et al. (2019). Specifically, we assume the Cobb-Douglas production function of firm i
After exploring our results’ robustness for the OFDI experience, we at time t as
replicate our econometric analysis by controlling for the past presence of
foreign shareholders. The latter might influence the propensity for
40
future investments abroad through an accumulation of knowledge about Note that IFDI stands for inward foreign direct investment.
41
international markets. To this end, we create a new binary variable, IFDI Due to our panel’s limited number of observations, we could not define a
StartOFDI variable by destination.

13
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table 15
The effect of diversity and performance on the probability of initiating outward foreign direct investment, with performance measured by TFP.
StartOFDI (1) (2) (3) (4) (5)

gender diversity − − 0.2221*** − − − 0.2630***

(− ) (1.7736) (− ) (− ) (1.2795)
nationality diversity − − 0.1282** − − − 0.1017
(− ) (1.3505) (− ) (− ) (1.9233)
diversity − 0.3606*** − − − 0.4588*** −
(1.4291) (− ) (− ) (1.2251) (− )
log(TFP) − − 0.0002** 0.0003 0.0002**
(− ) (− ) (0.0030/1.21) (0.0039) (0.0040)
log(tangible fixed assets) 0.0048** 0.0009 − 0.0019*** 0.0068*** 0.0011
(0.0373) (0.0244) (0.0207/1.82) (0.0195) (0.0205)
board size − 0.0014*** − 0.0009*** − 0.0004** − 0.0017*** − 0.0009***
(0.0052) (0.0064) (0.0054/2.00) (0.0051) (0.0060)
dual regime 0.0104 0.0129* 0.0151*** 0.0108 0.0122**
(0.1927) (0.1916) (0.1487/3.53) (0.1596) (0.1461)

country-controls yes yes Yes Yes Yes

year-controls yes yes Yes Yes Yes
industry-controls yes yes Yes Yes Yes

observations 3098 3098 2733 2726 2726

correct classifications 98.52% 98.52% 98.50% 98.50% 98.50%
Wald test of exogeneity 5.96** 14.80** − 7.42*** 14.85***
log-likelihood − 211.62 − 210.83 − 183.17 − 182.15 − 180.43

Notes: This table reports probit regression results for the effects of diversity (column 1), and gender and nationality diversity (column 2) on the probability of initiating
outward foreign direct investment as estimated in Equation (1); for the effects of performance measured by TFP on the probability of initiating outward foreign direct
investment as estimated from Equation (4) (column 3); for the effects of diversity (column 4) and gender and nationality diversity (column 5) on the probability of
initiating outward foreign direct investment, controlling for performance measured by TFP, as estimated in Equation (5). Average marginal effects are displayed.
Robust, firm-clustered standard errors/VIF coefficients are in round brackets. Superscript *, **, and *** corresponds to statistical significance at one-, five- and ten-
percent level, respectively.

Yi = δ + wL Li + wK Ki + wM Mi + wi + εi (6) performance is measured by TFP, the mediated effect of diversity is

smaller in percentage terms compared to its counterpart when perfor­
where Yi is the logarithm of firm i’s output measured by the value added; mance is measured by labour productivity.
Li and Mi are the logarithms of the freely variable labour and interme­
diate inputs approximated by the number of employees and the cost of 6. Conclusions
raw materials, respectively; and Ki denotes the logarithm of the state
variable capital, which is directly available in our dataset. The TFP has This study examines a panel of firms in European countries to discuss
been obtained according to the Levinsohn and Petrin (2003) method­ the effect of board diversity on OFDI. We adopt a nonlinear specification
ology, by implementing the standard ‘levpet’ routine available in STATA to model the probability of opening a foreign subsidiary and use IV/CF
for our balanced panel of 1283 firms.42 methods to address endogeneity in board diversity.
In Table 13, we replicate the results reported in Table 7 using TFP as We find that board diversity affects OFDI directly and through the
a measure of performance. Again, we define high_TFP and low_TFP as modifying and mediating roles of firm performance. Although board
binary variables taking the value of one if the firm is collocated in the diversity positively affects OFDI through firm performance, firms with
fourth and first quantiles of the spectrum of TFP values, respectively, more female directors and foreign directors are less likely to open
and zero otherwise. The results in the first column of Table 13 are very foreign subsidiaries. The negative effect of board diversity on OFDI
similar to those of our main specification in Table 7; and in the second decisions is stronger for more productive firms and firms with gender-
column, we notice that the interaction variables preserve the same sign diverse boards. These results are robust for several alternative specifi­
as the corresponding values in Table 7 but are no longer significant. cations and econometric models.
Table 14 presents the IV/2SLS estimates for Equation (3). The gen­ In our sample, boards that are more diverse lead to better firm per­
eral pattern of the results agrees with that shown in Table 8. formance and fewer OFDI decisions. A tougher and more informed
Finally, Table 15 reports the probit estimates of Equation (1) (col­ monitoring of managerial activities can explain these effects. Our results
umns 1 and 2), (4) (column 3), and (5) (columns 4 and 5) using TFP as a are in line with the literature that suggests that gender-diverse boards
measure of performance. The results are largely in line with those dis­ are better monitors, whereas foreign directors contribute local knowl­
played in Tables 6 and 9 Similar to Section 4, when performance is edge and expertise to foreign market operations.
measured by TFP, we can calculate that variations in performance To the best of our knowledge, this is the first study to explore firms’
explain (in absolute value) about 0.5% of the total board diversity effect self-selection into OFDI based on board diversity while disentangling the
on OFDI, where direct and mediated effects have opposite signs and the complex nexus between board composition, firm performance, and
direct effect exceeds in absolute value the mediated one, pointing to a internationalisation. On the one hand, our results explain the de­
negative overall effect of board diversity on OFDI. We also find that terminants of OFDI. On the other hand, they reveal the channels through
variation in performance explains about 0.6% and 0.5% of the effect of which board diversity affects OFDI, thus contributing to the literature on
board diversity on OFDI due to gender and nationality, respectively, OFDI and firm performance and to the literature on board diversity, firm
again with direct and mediated effects of opposite signs. Thus, when performance, and strategic decisions.
However, we acknowledge the limitations of our study. Due to data
constraints, our panel extends over a relatively short time span. This
42 prevents us from investigating the medium and long-term effects of the
Intermediate estimates of labour and capital from the levpet routine amount
to 0.6861 (0.0289) and 0.1870 (0.0377), respectively, where the standard er­
OFDI decision for firms with more diverse boards. In addition, our
rors are reported in brackets. period of observation, from 2011 to 2015, coincides with years of great

14
V. Gattai et al. Economic Modelling 120 (2023) 106156

economic uncertainty in Europe due to the sovereign debt crisis suffered Credit author statement
by Union members. The observation period may account for the rela­
tively small number of firms initiating OFDI in our sample. Furthermore, Valeria Gattai: Conceptualization, Methodology, Formal analysis,
we are unable to observe the effect of board diversity on OFDI in less Data curation, Writing – original draft, Writing – review & editing.
volatile environments and, thus, cannot assess any interaction between Piergiovanna Natale: Conceptualization, Methodology, Writing – orig­
board diversity and economic volatility in shaping a firm’s OFDI strat­ inal draft, Writing – review & editing. Francesca Rossi: Conceptualiza­
egy.43 Detailed information about board members such as education, tion, Methodology, Formal analysis, Data curation, Writing – original
past professional experience and independence, in addition to infor­ draft, Writing – review & editing.
mation about committee participation, would allow a better under­
standing of the effect of board composition on OFDI decisions. Declaration of competing interest
Unfortunately, such data would be very difficult to compile for a large
sample of firms, headquartered in different countries. Last, data con­ The authors declare that they have no known competing financial
straints prevent us from observing firm exporting activities or input interests or personal relationships that could have appeared to influence
sourcing on international markets. It follows that we are unable to the work reported in this paper.
establish whether board diversity affects differently alternative inter­
nationalisation strategies. Data availability
Despite these limitations, we believe that our results provide inter­
esting policy implications. A large fraction of the remarkable changes in The authors do not have permission to share data.
board diversity observed across Europe is the outcome of legislative
interventions aimed at promoting gender equality (Ferreira and Kirch­ Acknowledgements
maier, 2013). However, in recent years a lively debate has developed
around the notion of a ‘business case’ for board diversity. This case rests The authors are indebted to the editor, associate editor, and re­
on the positive impact of board diversity on a firm’s performance. Our viewers for their valuable comments and suggestions. The authors are
study demonstrates that board diversity improves firm performance; grateful to Emilio Colombo and Lars Oxheleim for their helpful com­
however, it also proves that board diversity directly impacts inter­ ments. This paper benefitted from comments by participants at the 2021
nationalisation choices and has effects not mediated by performance. World Finance Conference, 2021 SIE Annual Conference, 2020 SIEPI
This suggests that caution should be exercised in adopting measures that Annual Conference, and 2019 NESPUTT Conference. We thank Michele
promote board diversity solely based on their impact on firm perfor­ Rocca for research assistance.Valeria Gattai and Piergiovanna Natale
mance. Policymakers and shareholders must also consider the impact of gratefully acknowledge the financial support from the Università degli
board diversity on the multifaceted strategy selection process, especially Studi di Milano-Bicocca (2017-ATE-0362 and 2018-ATE-0340).
when strategy selection has long-term economic consequences as in the
case of internationalisation choices.

Appendix

This appendix describes our variables (Table A1) and presents pairwise correlations (Table A2) for the independent variables used in our
econometric analysis.

Table A1
Description of firm-level variables.

Variable Description

StartOFDI Dummy variable; 1 if firm i had no subsidiary abroad in the years t − 2 and t − 1 and has at least one in year t; 0 if the firm has no subsidiary abroad in the years
t − 2, t − 1 and t.
dual regime Dummy variable; 1 if firm i is governed by a dual system, with a supervisory board and a management board; 0 if firm i is governed by a unitary system, with a
board of directors.
board size Number of members of the board of directors for firms in unitary systems; sum of the numbers of members of supervisory and management boards for firms in
dual systems.
diversity Proportion of female and foreign directors in the board of firm i in year t.
gender diversity Proportion of female directors in the board of firm i in year t.
nationality diversity Proportion of foreign directors in the board of firm i in year t.
labour productivity Firm i’s value added divided by employees.
high_labour Dummy variable; 1 if firm i belongs to fourth quartile of the spectrum of labour productivity; 0 otherwise
productivity
low_labour Dummy variable; 1 if firm i belongs to first quartile of the spectrum of labour productivity; 0 otherwise
productivity
tangible fixed assets Firm i’s tangible fixed assets.
TFP Total factor productivity, Levinsohn and Petrin (2003) estimates.
high_TFP Dummy variable; 1 if firm i belongs to fourth quartile of the spectrum of TFP; 0 otherwise
low_TFP Dummy variable; 1 if firm i belongs to first quartile of the spectrum of TFP; 0 otherwise
OFDI memory Dummy variable; 1 if firm i had at least one foreign subsidiary from 2007 to 2010;
0 otherwise.
IFDI memory Dummy variable; 1 if firm i had at least one foreign shareholder from 2007 to 2010;
0 otherwise.
OFDI Extra EU Dummy variable; 1 if firm i has at least one subsidiary outside the European Union;
0 otherwise.

43
Data limitations prevent us from investigating board diversity’s impact on the intensive margin of OFDI.

15
V. Gattai et al. Economic Modelling 120 (2023) 106156

Table A2
Pairwise correlations of the independent variables.

labour board tangible gender nationality diversity log(TFP) dual OFDI IFDI OFDI
productivity size fixed assets diversity diversity regime memory memory Extra EU

labour 1.0000
productivity
board size 0.0171 1.0000
tangible fixed 0.0098 0.2636 1.0000
assets
gender 0.0102 − 0.2032 0.0283 1.0000
diversity
nationality 0.0145 − 0.0471 0.0840 0.1529 1.0000
diversity
diversity 0.0164 − 0.1557 0.0840 0.7105 0.8040 1.0000
log(TFP) 0.9268 0.0967 0.1002 0.0148 0.0371 0.0353 1.0000
dual regime − 0.0249 − 0.2939 − 0.0042 0.0252 0.0976 0.0843 − 0.0549 1.0000
OFDI memory − 0.0274 0.1742 0.0788 0.0017 0.0796 0.0577 − 0.0040 0.0245 1.0000
IFDI memory − 0.0337 0.1489 0.0418 − 0.0175 0.0534 0.0276 − 0.0139 − 0.1290 0.0791 1.0000
OFDI Extra EU − 0.0188 0.1604 0.0917 0.0337 0.1043 0.0945 0.0098 0.0121 0.5301 0.1198 1.0000

References Christiansen, Lone Engbo, Lin, Huidan, Pereira, Joana, Topalova, Petia, Turk, Rima,
2016. Gender diversity in senior positions and firm performance; evidence from
Europe. IMF Working Papers 16/–50.
Adams, Renee B., Ferreira, Daniel, 2007. A theory of friendly boards. J. Finance 62 (1),
Ciavarella, Angela, 2017. Board Diversity and Firm Performance across Europe. Consob
217–250, 02.
Quaderni di Finanza, 85.
Adams, Renee B., Ferreira, Daniel, 2009. Women in the boardroom and their impact on
Colombo, Emilio, Rotondi, Valentina, Stanca, Luca, 2018. Macroeconomic conditions
governance and performance. J. Financ. Econ. 94 (2), 291–309, 11.
and well-being: do social interactions matter? Appl. Econ. 50 (28), 3029–3038.
Adams, Renee B., Hermalin, Benjamin E., Weisbach, Michael S., 2010. The role of boards
Comi, Simona, Grasseni, Mara, Pagani, Laura, Origo, Federica, 2020. Where women
of directors in corporate governance: a conceptual framework and survey. J. Econ.
make a difference. Gender quotas and firms’ performance in three European
Lit. 48 (1), 58–107, 03.
countries. Ind. Labor Relat. Rev. 73 (3), 768–793.
Ahern, Kenneth R., Dittmar, Amy K., 2012. The changing of the boards: the impact on
Croson, Rachel, Gneezy, Uri, 2009. Gender differences in preferences. J. Econ. Lit. 47 (2),
firm valuation of mandated female board representation. Q. J. Econ. 127 (1),
448–474.
137–197, 02.
Dale-Olsen, Harald, Pal, Schone, Verner, Mette, 2013. Diversity among Norwegian
Amore, Mario Daniele, Garofalo, Orsola, Minichilli, Alessandro, 2014. Gender
boards of directors: does a quota for women improve firm performance? Fem. Econ.
interactions within the family firm. Manag. Sci. 60 (5), 1083–1097, 05.
19 (4), 110–135, 10.
Arnold, Jens M., Hussinger, Katrin, 2010. Exports versus FDI in German manufacturing:
Damijan, Joze P., Polanec, Saso, Prasnikar, Janez, 2007. Outward FDI and productivity:
firm performance and participation in international markets. Rev. Int. Econ. 18 (4),
micro-evidence from Slovenia. World Econ. 30 (1), 135–155, 01.
595–606, 09.
Dang, Rey, Houanti, L’Hocine, Ammari, Aymen, Le Nhu, Tuyen, 2018. Is there a
Aw, Bee Yan, Lee, Yi, 2008. Firm heterogeneity and location choice of Taiwanese
’business case’ for board gender diversity within French listed SMEs. Appl. Econ.
multinationals. J. Int. Econ. 75 (1), 167–179, 05.
Lett. 25 (14), 980–983, 08.
Baiardi, Donatella, Gattai, Valeria, Natale, Piergiovanna, 2021. Estimating the ex-ante
Dang, Rey, Houanti, L’Hocine, Reddy, Krishna, Simioni, Michel, 2020. Does board
and the ex-post effects of Chinese outward. FDI. World Econ. 44 (12), 3641–3673.
gender diversity influence firm profitability? A control function approach. Econ.
Baldwin, Richard, Wyplosz, Charles, 2019. The Economics of European Integration.
Modell. 90, 168–181.
McGraw-Hill.
Dou, Junsheng, Jacoby, Gady, Li, Jialong, Su, Youyi, Wu, Zhenyu, 2019. Family
Barba Navaretti, Giorgio, Castellani, Davide, 2004. Investments Abroad and Performance
involvement and family firm internationalization: the moderating effects of board
at Home: Evidence from Italian Multinationals. CEPR Discussion Papers, 4284.
experience and geographical distance. J. Int. Financ. Mark. Inst. Money 59, 250–261.
Barba Navaretti, Giorgio, Castellani, Davide, Disdier, Anne-Celia, 2010. How does
Eckbo, Espen B., Nygaard, Knut, Thorburn, Karin S., 2016. Does gender-balancing the
investing in cheap labour countries affect performance at home? Firm-level evidence
board reduce firm value? CEPR Discussion Papers,, 11176.
from France and Italy. Oxf. Econ. Pap. 62 (2), 234–260, 04.
Engel, Dirk, Procher, Vivien, 2012. Export, FDI and firm productivity. Appl. Econ. 44
Baron, Reuben M., Kenny, David A., 1986. The moderator–mediator variable distinction
(13–15), 1931–1940, 05.
in social psychological research: conceptual, strategic, and statistical considerations.
Estelyi, Kristina Saghy, Nisar, Tahir M., 2016. Diverse boards: why do firms get foreign
J. Pers. Soc. Psychol. 51 (6), 1173–1182.
nationals on their boards? J. Corp. Finance 39, 174–192, 08.
Bernard, Andrew B., Jensen, J. Bradford, 1995. Exporters, jobs, and wages in U.S.
Evgeniou, Theodoros, Vermaelen, Theo, 2017. Share buybacks and gender diversity.
Manufacturing: 1976-1987. Brookings Pap. Econ. Activ. 67–112, 0.
J. Corp. Finance 45, 669–686, 08.
Bernile, Gennaro, Bhagwat, Vineet, Scott, Yonker, 2018. Board diversity, firm risk, and
Federico, Stefano, 2010. Outsourcing versus integration at home or abroad and firm
corporate policies. J. Financ. Econ. 127 (3), 588–612, 03.
heterogeneity. Empirica 37 (1), 47–63, 0.
Borin, Alessandro, Mancini, Michele, 2016. Foreign direct investment and firm
Fedorets, Alexandra, Gibert, Anna, Burow, Norma, 2019. Gender Quotas in the
performance: an empirical analysis of Italian firms. Rev. World Econ. -
Boardroom: New Evidence from Germany. DIW Berlin Discussion Papers, 1810.
Weltwirtschaftliches Archiv 152 (4), 705–732, 11.
Ferreira, Daniel, 2010. Board diversity. In: Kent Baker, H., Anderson, Ronald (Eds.),
Bose, Udichibarna, Mallick, Sushanta, Tsoukas, Serafeim, 2020. Does easing access to
Corporate Governance: A Synthesis of Theory, Research, and Practice. Wiley,
foreign financing matter for firm performance? J. Corp. Finance 64, 101639.
Hoboken, N.J., pp. 225–242
Bose, Udichibarna, Filomeni, Stefano, Mallick, Sushanta, 2021. Does bankruptcy law
Ferreira, Daniel, Kirchmaier, Tom, 2013. Corporate boards in Europe: size,
improve the fate of distressed firms? The role of credit channels. J. Corp. Finance 68,
independence, and gender diversity. In: Belcredi, Massimo, Ferrarini, Guido (Eds.),
101836.
Boards and Shareholders in European Listed Companies. Cambridge University
Bruno, Giovanni, Ciavarella, Angela, Linciano, Nadia, 2018. Boardroom Gender Diversity
Press, Cambridge, pp. 191–224.
and Performance of Listed Companies in Italy. Consob Quaderni di Finanza, 87.
Frijns, Bart, Dodd, Olga, Cimerova, Helena, 2016. The impact of cultural diversity in
Campbell, Kevin, Minguez-Vera, Antonio, 2010. Female board appointments and firm
corporate boards on firm performance. J. Corp. Finance 41 (12), 521–541.
valuation: short and long-term effects. J. Manag. Govern. 14 (1), 37–59, 02.
Garcia-Meca, Emma, Garcia-Sanchez, Isabel, Martinez-Ferrero, Jennifer, 2015. Board
Casaburi, Lorenzo, Gattai, Valeria, Minerva, Gaetano Alfredo, 2007. Firms’ international
diversity and its effects on bank performance: an international analysis. J. Bank.
status and heterogeneity in performance: evidence from Italy. Riv. Politic. Econ. 97
Finance 53, 202–214, 04.
(5–6), 151–187, 05.
Gattai, Valeria, Trovato, Valentina, 2016. Estimating sourcing premia using Italian
Castellani, Davide, Giovannetti, Giorgia, 2010. Productivity and the international firm:
regional data. B. E. J. Econ. Anal. Pol. 16 (2), 1029–1067.
dissecting heterogeneity. Journal of Economic Policy Reform 13 (1), 25–42, 03.
Girma, Sourafel, 2005. Technology transfer from acquisition FDI and the absorptive
Castellani, Davide, Zanfei, Antonello, 2007. Internationalisation, innovation and
capacity of domestic firms: an empirical investigation. Open Econ. Rev. 16 (2),
productivity: how do firms differ in Italy? World Econ. 30 (1), 156–176, 01.
175–187, 04.
Castellani, Davide, Mariotti, Ilaria, Piscitello, Lucia, 2008. The impact of outward
Green, Colin P., Swarnodeep, Homroy, 2018. Female directors, board committees and
investments on parent company’s employment and skill composition: evidence from
firm performance. Eur. Econ. Rev. 102, 19–38, 02.
the Italian case. Struct. Change Econ. Dynam. 19 (1), 81–94, 03.
Greenaway, David, Kneller, Richard, 2007. Firm heterogeneity, exporting and foreign
Chen, Guoli, Crossland, Craig, Huang, Sterling Zehnrui, 2016. Female board
direct investment. Econ. J. 117 (517), F134–F161, 02.
representation and corporate acquisition intensity. Strat. Manag. J. 37 (2), 303–313.
Gregory-Smith, Ian, Main, Brian G.M., O’Reilly III., Charles A., 2014. Appointments, pay
Chen, Jie, Leung, Woon Sau, Goergen, Marc, 2017. The impact of board gender
and performance in UK boardrooms by gender. Econ. J. 124 (574), F109–F128, 02.
composition on dividend payouts. J. Corp. Finance 43, 86–105 04.

16
V. Gattai et al. Economic Modelling 120 (2023) 106156

Grossman, Gene M., Helpman, Elhanan, Szeidl, Adam, 2006. Optimal integration Masulis, Ronald W., Wang, Cong, Xie, Fei, 2012. Globalizing the boardroom — the
strategies for the multinational firm. J. Int. Econ. 70 (1), 216–238 09. effects of foreign directors on corporate governance and firm performance.
Hahn, Peter D., Lasfer, Meziane, 2016. Impact of foreign directors on board meeting J. Account. Econ. 53 (3), 527–554, 06.
frequency. Int. Rev. Financ. Anal. 46, 295–308, 07. Matsa, David A., Miller, Amalia R., 2013. A female style in corporate leadership?
Hannan, Timothy H., Mavinga, Ferdinand, 1980. Expense preference and managerial Evidence from quotas. Am. Econ. J. Appl. Econ. 5 (3), 136–169, 07.
control: the case of the banking firm. Bell J. Econ. 11, 671–682. Melitz, Marc J., 2003. The impact of trade on intra-industry reallocations and aggregate
Hayakawa, Kazunobu, Kimura, Fukunari, Machikita, Tomohiro, 2012. Globalization and industry productivity. Econometrica 71 (6), 1695–1725, 11.
productivity: a survey of firm-level analysis. J. Econ. Surv. 26 (2), 332–350, 04. Moriconi, Simone, Peri, Giovanni, Pozzoli, Dario, 2020. The role of institutions and
Head, Keith, Ries, John, 2003. Heterogeneity and the FDI versus export decision of immigrant networks in firms’ offshoring decisions. Can. J. Econ. 53 (4), 1745–1792.
Japanese manufacturers. J. Jpn. Int. Econ. 17 (4), 448–467, 12. Parrotta, Pierpaolo, Pozzoli, Dario, Sala, Davide, 2016. Ethnic diversity and firms’ export
Helpman, Elhanan, Melitz, Marc J., Yeaple, Stephen R., 2004. Export versus FDI with behavior. Eur. Econ. Rev. 89 (10), 248–263.
heterogeneous firms. Am. Econ. Rev. 94 (1), 300–316, 03. Piekkari, Rebecca, Oxelheim, Lars, Randöy, Trond, 2015. The Silent Board: how
Huang, Jiekun, Kisgen, Darren J., 2013. Gender and corporate finance: are male language diversity may influence the work processes of corporate boards. Corp.
executives overconfident relative to female executives? J. Financ. Econ. 108 (3), Govern. Int. Rev. 23 (1), 25–41.
822–839, 06. Pisani, Niccolò, Muller, Alan, Paula, Bogăţan, 2019. Top management team
Huang, Youxing, Zhang, Yan, 2017. How does outward foreign direct investment internationalization and firm-level internationalization: the moderating effects of
enhance firm productivity? A heterogeneous empirical analysis from Chinese home-region institutional diversity and firm global focus. J. Int. Manag. 24 (3),
manufacturing. China Econ. Rev. 44 (7), 1–15. 239–256.
IMF, 1993. Balance of Payments Manual, fifth ed. IMF, Washington, DC. Preacher, Kristopher J., Hayes, Andrew F., 2008. Asymptotic and resampling strategies
Joecks, Jasmin, Pull, Kerstin, Vetter, Karin, 2013. Gender diversity in the boardroom and for assessing and comparing indirect effects in multiple mediator models. Behav.
firm performance: what exactly constitutes a "critical mass". J. Bus. Ethics 118 (1), Res. Methods 40 (3), 879–891.
61–72. Pucheta-Martinez, Maria, Bel-Oms, Inmaculada, 2016. The board of directors and
Judd, Charles M., Kenny, David A., 1981a. Estimating the Effects of Social Interventions. dividend policy: the effect of gender diversity. Ind. Corp. Change 25 (3), 523–547,
Cambridge University Press, Cambridge, England. 06.
Judd, Charles M., Kenny, David A., 1981b. Process analysis: estimating mediation in Raff, Horst, Ryan, Michael, Stahler, Frank, 2012. Firm productivity and the foreign-
treatment evaluations. Eval. Rev. 5 (5), 602–619. market entry decision. J. Econ. Manag. Strat. 21 (3), 849–871, 10.
Kalemli-Ozcan, Sebnem, Sorensen, Bent, Villegas-Sanchez, Carolina, Rhode, Deborah L., Packel, Amanda K., 2014. Diversity on corporate boards: how much
Volosovych, Vadym, Yesiltas, Sevcan, 2015. How to Construct Nationally difference does difference make? Del. J. Corp. Law 39 (2), 377–425, 0.
Representative Firm Level Data from the ORBIS Global Database. NBER Working Ribeiro, Samuel Pinto, Menghinello, Stefano, De Backer, Koen, 2010. The OECD ORBIS
Papers, 21558. database: responding to the need for firm-level micro-data in the OECD. OECD
Kanter, Rosabeth Moss, 1997. Men and Women of the Corporation. Basic Books, New Statistics Working Papers, 2010/1.
York. Rivas, Jose Luis, 2012. Diversity & internationalization: the case of boards and TMT’s.
Kim, Daehyun, Starks, Laura T., 2016. Gender diversity on corporate boards: do women Int. Bus. Rev. 21 (1), 1–12.
contribute unique skills? Am. Econ. Rev. 106 (5), 267–271, 05. Rivers, Douglas, Vuong, Quang H., 1988. Limited information estimators and exogeneity
Kimura, Fukunari, Kiyota, Kozo, 2006. Exports, FDI, and productivity: dynamic evidence tests for simultaneous probit models. J. Econom. 39 (3), 347–366, 11.
from Japanese firms. Rev. World Econ. - Weltwirtschaftliches Archiv 142 (4), Sabatier, Mareva, 2015. A women’s boom in the boardroom: effects on performance?
695–719, 0. Appl. Econ. 47 (25–27), 2717–2727, 05.
Kirsch, Anja, 2018. The gender composition of corporate boards: a review and research Schmid, Thomas, Urban, Daniel, 2016. The Economic Consequences of a ’Glass-Ceiling’:
agenda. Leader. Q. 29 (2), 346–364, 06. Women on Corporate Boards and Firm Value. AFA 2016 San Francisco Meetings
Kohler, Wilhelm K., Smolka, Marcel, 2011. Sourcing premia with incomplete contracts: Paper SSRN-id2344786-1.
theory and evidence. B E J. Econ. Anal. Pol.: Contrib. Econ. Anal. Pol. 11 (1), 1–39. Sherman, Hugh D., Kashlak, Roger J., Joshi, Maheshkumar P., 1998. The effect of the
Kohler, Wilhelm K., Smolka, Marcel, 2012. Global sourcing decisions and firm board and executive committee characteristics on the degree of internationalization.
productivity: evidence from Spain. In: Stern, Robert M. (Ed.), Quantitative Analysis J. Int. Manag. 4 (4), 311–335.
of Newly Evolving Patterns of International Trade: Fragmentation; Offshoring of Singh, Tarlok, 2010. Does international trade cause economic growth? A survey. World
Activities; and Vertical Intra-Industry Trade. World Scientific Publishing, Singapore, Econ. 33 (11), 1517–1564.
pp. 139–189. Stroup, Caleb, 2017. International deal experience and cross-border acquisitions. Econ.
Levi, Maurice, Li, Kai, Zhang, Feng, 2014. Director gender and mergers and acquisitions. Inq. 55 (1), 73–97.
J. Corp. Finance 28 (10), 185–200. Tomiura, Eiichi, 2007. Global Sourcing, Technology, and Factor Intensity: Firm-Level
Levinsohn, James, Petrin, Amin, 2003. Estimating production functions using inputs to Relationships.. RIETI Discussion Paper Series, 24.
control for unobservables. Rev. Econ. Stud. 70 (2), 317–342. Torchia, Maria Teresa, Calabrò, Andrea, Huse, Morten, 2011. Women directors on
Levy, Orly, Beechler, Schon, Taylor, Sully, Boyacigiller, Nakiye, 2007. What we talk corporate boards: from tokenism to critical mass. J. Bus. Ethics 102 (2), 299–317.
about when we talk about ‘global mindset’: managerial cognition in multinational UNCTAD, 2020. World Investment Report 2019. International Production beyond the
corporations. Journal of International Businsess Studies 38, 231–258. Pandemic. United Nations, Geneva.
Lewellen, Wilbur, Loderer, Claudio, Rosenfeld, Ahron, 1989. Mergers, executive risk Wagner, Joachim, 2006. Exports, foreign direct investment, and productivity: evidence
reduction, and stockholder wealth. J. Financ. Quant. Anal. 2, 459–472. from German firm level data. Appl. Econ. Lett. 13 (6), 347–349, 05.
Li, Huijuan, Cai, Weihong, Li, Wenxiu, 2021. Does global value chains participation Wagner, Joachim, 2007. Exports and productivity: a survey of the evidence from firm-
improve skill premium? Mediating role of skill-biased technological change. Econ. level data. World Econ. 30 (1), 60–82, 01.
Modell. 99, 105489. Wagner, Joachim, 2012. International trade and firm performance: a survey of empirical
Lopez, Ricardo A., 2005. Trade and growth: reconciling the macroeconomic and studies since 2006. Rev. World Econ. - Weltwirtschaftliches Archiv 148 (2),
microeconomic evidence. J. Econ. Surv. 19 (4), 623–648, 09. 235–267, 06.
MacKinnon, David P., Dwyer, James H., 1993. Estimating mediated effects in prevention Wagner, Joachim, 2016. A survey of empirical studies using transaction level data on
studies. Eval. Rev. 17 (2), 144–158. exports and imports. Rev. World Econ. - Weltwirtschaftliches Archiv 152 (1),
Marques, Helena, 2015. Does the gender of top managers and owners matter for firm 215–225, 02.
exports? Fem. Econ. 21 (4), 89–117, 10.

17