The interplay between institution-based trust, inter-personal trust
within the board of directors, and firm performance
Master's thesis International Economics & Business
Author: Dirk Romeijnders s2177137
d.romeijnders@student.rug.nl
Supervisor: prof. dr. H. van Ees
2
Abstract
I examine the mediating effect of inter-personal trust within the board of directors on the relationship between institution-based trust and firm performance. I construct a measure of trust within a board, by calculating the average trust between directors using aggregated survey data on generalized trust perceptions of nationalities. By using a sample of 1,494 firms that are located in 32 different countries including 7,255 directors in the year 2015, I find that institution-based trust positively affects inter-personal trust and firm performance, but I find no proof that inter-personal trust is related to firm performance in a nonlinear inverted U-shaped way.
3
Table of Contents
1. Introduction ...4
2. Theory and hypothesis development ...5
2.1. Inter-personal trust ...5
2.2. Inter-personal trust, board performance, and firm performance ...7
2.3. Institution-based trust ...10
2.4. The complete interplay ...11
2.5. Control variables ...12
3. Data and Methods ...13
3.1. Sample ...13 3.2. Data collection ...13 3.3. Measures ...16 3.3.1. Dependent variables ...16 3.3.2. Independent variables ...16 3.3.3. Control variables ...17 3.4. Method ...19 3.4.1. Statistical assumptions ...20 3.4.2. Robustness tests ...21
4. Results and analyses ...22
4.1. Summary statistics ...22
4.2. Assumption tests ...22
4.3. Baron and Kenny's (1986) mediation test ...24
4.4. Generalized structural equation modeling ...24
4.5. Instrumental variables regression ...27
4.6. Robustness tests ...29
4.6.1. Robustness of board national diversity ...29
4.6.2. Robustness of board trust ...31
4.6.3. Sample bias test ...35
4
1. Introduction
Numerous studies discuss the role of trust in relation to firm performance. In the academic field of corporate governance, this mostly relates to the performance of the board of directors of a firm. Trust influences the interaction between directors and the management of a firm, and this affects the performance of the board of directors. Understanding how trust functions within this context, and by what it is determined, is therefore important for the performance of a firm. However, trust is an abstract concept and therefore difficult to perceive and measure. Perhaps unsurprisingly, scholars have found inter-personal trust within the board of directors to both positively and negatively affect firm performance (Frijns, Dodd, & Cimerova, 2016); (Fulmer & Gelfand, 2012).
Besides inter-personal trust, scholars have debated institution-based trust (Fulmer & Gelfand, 2012); (Child & Mollering, 2003); (Sydow, 2006). This type of trust can help facilitating trusting behavior between people due to the quality of certain institutions in a particular environment. For instance, the presence of effective repressive state apparatuses causes people to trust others in (business) transactions more easily due to the expectations that the other behaves fairly in the face of certain repercussions (Lane & Bachmann, 1996). Besides that, scholars have discussed the effects of institution-based trust on firm performance, where the consensus is that higher quality institutions lower firms' transaction costs (Gaviria, 2002); (Elango & Lahiri, 2014). Institution-based trust thus facilitates both inter-personal trust and firm performance.
The existing literature on corporate governance has not yet explored the complete interplay between institution-based trust, inter-personal trust, board behavior, and firm performance. Since scholars argue that institution-based trust affects both inter-personal trust within corporate boards and firm performance, inter-personal trust arguably mediates this effect as inter-personal trust is also related to firm performance. This mediating effect is currently ambiguous, as scholars do not fully agree on the effect of inter-personal trust on firm performance. This interplay leads to new insights into the role of inter-personal trust within corporate boards. Firms cannot directly control the level of institution-based trust, but they can control inter-personal trust. Understanding the interplay between institution-based trust, inter-personal trust, and firm performance can help firms to optimize their performance by adapting or influencing the level of inter-personal trust within the board of directors.
In that regard, it is important to understand the drivers of inter-personal trust. Scholars point out that homogeneity generates trust, as people tend to trust others who look, think, and act in a similar way (Guiso, Sapienza, & Zingales, 2009). Heterogeneity then causes distrust due to various distances between persons. Some scholars consider inter-personal trust to be a latent variable of distance between directors (Doney, Cannon, & Mullen, 1998); (DeBruine, 2002). This simplifies modeling inter-personal trust, as differences between directors can be measured more easily than the trust behavior of directors. Many scholars therefore use nationality differences to proxy for distance between board members (Frijns et al. 2016).
5 performance? I use a sample of 1,494 firms that are located in 32 different countries with directors that have one of 15 European nationalities in the year 2015. By using generalized structural equation modeling, I find that institution-based trust positively affects inter-personal trust and firm performance, but I find no proof that inter-personal trust is significantly related to firm performance in a nonlinear inverted U-shaped way. The effect of inter-personal trust on firm performance, and its mediating effect, therefore remains ambiguous.
This paper makes two contributions to the literature. First, I put the relationship between institution-based trust and inter-personal trust in the context of the board of directors. To my knowledge, scholars have not yet explored this particular relationship. This adds to the growing literature on the factors that influence board behavior and performance. This paper’s findings imply that research on inter-personal trust in the board of directors should take into account the role of institution-based trust. Second, the literature has not yet explored the mediating effect of inter-personal trust within the board of directors on the relationship between institution-based trust and firm performance. Although this mediating effect remains inconclusive, this paper uncovers its theoretical foundations and contributes to the advancement of future research on the interplay between institution-based trust, inter-personal trust within the board of directors, board behavior, and firm performance.
This paper proceeds in the following way. Section 2 includes the theoretical background and development of hypotheses on inter-personal trust, board behavior, firm performance, institution-based trust, and their interrelatedness. Section 3 discusses the data, the construction of the variables, and the methodology this paper utilizes. Section 4 presents and analyzes the results and passes judgment on the hypotheses. Lastly, section 5 provides a conclusion along with the limitations of this paper and suggestions for future research.
2. Theory and hypothesis development 2.1. Inter-personal trust
Even though the definition of trust has widely been debated in the academic discourse, many scholars (Davis et al., 2000); (Colquitt, Scott, & LePine, 2007); (Vanneste, 2016) tend to agree on the definition by Mayer et al. (1995), which states that trust is "the willingness of a party to be vulnerable to the actions of another party based on the expectation that the other will perform a particular action important to the trustor, irrespective of the ability to monitor or control that other party." This concept of trust is used in a variety of academic disciplines, including (international) business and, more specific, corporate governance.
6 different ways of perception and interpretation" (Frijns et al., 2016) has increased over the last decades.
Scholars use nationality as a proxy of the multifaceted determinants of inter-personal trust. The idea and assumption is that a person's nationality determines certain factors which in turn may affect inter-personal trust. Some of these factors are for instance culture, religion, language, and appearance. Nationality provides a very generalizing notion of these factors, as people with the same nationality often do not share these factors in a similar fashion. However, scholars use these generalized factors often in the academic literature. For instance, studies that examine the influence of cultural differences on corporate practices (Bryan, Nash, & Patel, 2015); (El Ghoul & Zheng, 2016); (Zheng, El Ghoul, Guedhami, & Kwok, 2012) use Hofstede's criteria of cultural dimensions (Hofstede 2001). Hofstede's national culture indicators are constructed for individual countries. This implies that the dimensions are the product of a country's nationality. Frijns et al. (2016) used these dimensions to research whether cultural distance between members of the board of directors affects firm performance, where national diversity causes cultural distance between directors. Scholars and institutions have used nationality to conceptualize distances between persons and groups for various variables, including inter-personal trust (Guiso et al., 2009); (Eurobarometer 2008). Although less precise, it is then easier to model inter-personal trust using these generalized measures than conducting surveys or interviews.
National distance between people has a negative impact on inter-personal trust. For example, coordination and communication in cultural diverse groups is more difficult (Anderson, Reeb, Upadhyay, & Zhao, 2011); (Ferreira, 2010) and Bjørnskov (2008) argues that cultural diversity in groups can therefore cause lower levels of trust. "People have a given trust radius; they trust people who are inside the trust radius and distrust people who are outside this radius. In more diverse groups, many people fall outside the trust radius and, therefore, there is less trust in a more heterogeneous group" (Frijns et al., 2016). This not only applies to cultural distance between people, but also to religious distance, geographical distance, and distance in terms of appearance (Guiso et al., 2009). National diversity has therefore a negative effect on inter-personal trust, as the distances that are determined by national distance function as latent variables that affect inter-personal trust.
7 However, national diversity is not the only aspect that affects inter-personal trust among board members. Directors' gender and age also have an impact on the dynamic of a corporate board (Anderson et al., 2011) because they reduce inter-personal trust among directors in corporate management, as "social similarity breeds trust" (Adams & Ferreira, 2004). When taking this into account, I formulate the following hypothesis:
Hypothesis 1. National distance between board members negatively affects inter-personal trust within the board of directors
2.2. Inter-personal trust, board performance, and firm performance
The academic consensus is that inter-personal trust within corporate boards has a positive effect on firm performance. The idea is that inter-personal trust improves the performance of the board, which in turn increases firm performance. Theoretically, this is a big step, as many factors influence board and firm performance. There are also many latent variables in this assumed causal relationship. However, scholars generally argue that trust reduces conflict within groups (Beccerra & Gupta, 1999), "reduces the need for formal contracts ... and hierarchical controls" (Davis et al., 2000), and therefore reduces transaction costs (Hosmer, 1995). These mechanisms enhance board members' task effectiveness and the board's cohesiveness (Roberts, McNulty, & Stiles, 2005). In this way, inter-personal trust can be a source of competitive advantage (Barney & Hansen, 1994); (Hancke, Rhodes, & Thatcher, 2008) and has therefore a positive effect on firm performance (Goergen et al., 2013); (de Wit, Greer, & Jehn, 2012); (Beccerra & Gupta, 1999); (Audi, Loughran, & McDonald, 2016).
Board members contribute to the performance of a firm by performing mainly a monitoring and strategic role. The monitoring role refers "to the responsibility of directors to monitor managers on behalf of shareholders" (Hillman & Dalziel, 2011). Garg (2013) explains that "the board and its individual directors ... track the significant behaviors of executives, the outcomes of their actions, and the performance" of the firm where they take "corrective action" if necessary. The strategic role refers to the provision of resources which includes activities as "providing a firm legitimacy, expertise, and counsel,” but also connecting the firm to external stakeholders, “facilitating the firm's access to resources, and aiding the firm in the formulation of strategy" (van Ees, van der Laan, & Postma, 2008). Effective board performance of both roles tends to have a positive effect on firm performance.
8 The strategic role of board members in relation to firm performance rests in the resource-based view. "The unique combination of the expertise and wider experience of the board (board's human and relational capital) and the quality of top management will positively contribute to the strategic decision-making and ultimately to successful performance of the firm" (van Ees et al., 2008). Thus if board members perform their strategic role effectively, firm performance increases as resources are more effectively utilized.
Forbes and Milliken (1999) point out three board processes that are positively related to board members' task performance: effort norms, cognitive conflict, and use of knowledge. First, effort norms refer to "the group's shared beliefs regarding the level of effort each individual is expected to put toward a task" (Forbes & Milliken, 1999). Kanfer (1992) points out that effort derives from motivation and improves individual behavior towards the group's task. Norms strongly influence the behavior of members (Feldman, 1984), which means that effort norms enhances "the effort of individual group members," (Forbes & Milliken, 1999) and consequently the task performance of the board. Second, cognitive conflict refers to "task-oriented differences in judgment among group members" (Forbes & Milliken, 1999). Van Ees et al. (2008) point out that "cognitive conflict in boards that involve the use of critical and investigative processes improves board role performance." Third, use of knowledge refers to "the board's ability to tap the knowledge and skills available to it and then apply them to its tasks (Forbes & Milliken, 1999). Van Ees et al. (2008) argue that "use of knowledge is positively related to the monitoring and strategy role performance" of the board.
Inter-personal trust within a corporate board positively affects these three board processes which in turn positively affect board performance. "The prevalence of trust may make board members and managers more confident to share information and knowledge and make them more active in board decision-making" (van Ees et al., 2008). Van Ees et al. (2008) argue that "trust plays a facilitating role" in the positive relationship between the three discussed "board processes and board monitoring." They argue that "in a high trust-context, the three board processes may enhance monitoring performance, because of positive expectations between board members vis-à-vis top management. In this situation, it is likely that more input of effort norms, use of knowledge and cognitive conflict lead to higher monitoring performance." In a similar fashion, inter-personal trust also enhances the performance of the board's strategic role. Inter-personal trust thus positively affects board performance where the board processes function as latent variables. Since board performance is positively related to firm performance, inter-personal trust also positively affects firm performance.
inter-9 personal trust therefore reduces the quality of board performance. Consequently, firm performance decreases as well.
I therefore expect a nonlinear relationship between inter-personal trust within the board of directors and firm performance in the shape of an inverted U, where both board processes and performance function as first and second stage latent mediators of personal trust. This relationship functions in the following way. An underinvestment in inter-personal trust reduces firm performance. When inter-inter-personal trust improves, the negative effect on firm performance diminishes until it turns into a positive effect. This positive effect then increases, but the increase diminishes when the level of inter-personal trust becomes too high. When there is an overinvestment in inter-personal trust, the positive effect on firm performance starts to decline until it turns negative again. Figure 1 provides a graphical depiction of this relationship.
Figure 1
Graphical depiction of the relationship between inter-personal trust and firm performance
D
B C
A C
This figure depicts the nonlinear inverted U-shaped relationship between inter-personal trust and firm performance. In area A, there is an underinvestment in inter-personal trust, which has a negative effect on firm performance. Area B shows the middle ground where inter-personal trust has a positive effect on firm performance. The increase of this positive effect diminishes every time inter-personal trust increases. In area C, there is an overinvestment in trust. Every additional increase in inter-personal trust sharply declines its positive effect on firm performance until it turns negative. Point D points towards the inflection point where an overinvestment in inter-personal trust occurs. This is also the sweet spot where boards should aim for to optimize their effect on firm performance.
Inter-personal trust
Firm
performan
10 Inter-personal trust is therefore a double-edged blade. The nonlinear relationship implies a particular middle ground in terms of the level of inter-personal trust where it positively affects firm performance. There is a sweet spot where the positive effect on firm performance reaches an inflection point due to an overinvestment in inter-personal trust. Boards should aim for that particular level of inter-personal trust to optimize their effect on firm performance. I formulate the following hypothesis:
Hypothesis 2. The relationship between inter-personal trust within the board of directors and firm performance is positive but nonlinear in the shape of an inverted U.
2.3. Institution-based trust
Scholars have debated the relevance of institution-based trust in various academic disciplines (Fulmer & Gelfand, 2012); (Child & Mollering, 2003); (Sydow, 2006). Broadly taken, institution-based trust refers to all environmental and institutional structures in which people and firms operate that "make an environment feel trustworthy" (McKnight, Choudhury, 2002). Higher quality structures improve this feeling of trustworthiness, which increases institution-based trust. One relevant perspective on this is "structural assurance" (McKnight & Chervany, 2001). This "means that one believes that protective structures—guarantees, contracts, regulations, promises, legal recourse, processes, or procedures—are in place that are conducive to situational success" (McKnight & Chervany, 2001), where situational success refers to a positive outcome after having trusted someone. Institution-based trust is predominantly measured by countries' legal system, rule of law (Bachmann & Inkpen, 2011); (Yu, Beugelsdijk, & de Haan, 2015), regulatory quality (Zhang, Liu, Sayogo, Picazo-Vela, & Luna-Reyes, 2016), government effectiveness (La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1997) and control of corruption (Hakhverdian & Mayne, 2012), where institution-based trust tends to be lower when these measures are weak.
Institution-based trust is positively related to inter-personal trust. Institution-based trust facilitates trusting behavior "because of the secure feeling structural assurance engenders" (McKnight & Chervany, 2001). This means, for instance, if agents operate in an environment with an effective legal system, inter-personal trust will be higher as an agent will expect another agent to behave according to certain legal conditions (Lane & Bachmann, 1996). Institutions therefore function as "a personal third party guarantor" that can significantly reduce "the risk that a trustee will behave untrustworthily" (Bachmann & Inkpen, 2011). Backmann and Inkpen (2011) argue that "institutions may find access into potential trustors behavior in that they lend meaning to the circumstances in which the actors are embedded before any relationship is built." Institution-based trust is therefore inherent in the creation of inter-personal trust. This is an iterative process in which an increase in the quality of institutions positively affects inter-personal trust.
11 2011). However, this does not nullify the effect of national institutions on inter-personal trust within the board. Wang & Gordon (2011) argue that "the very foundation of these contextual effects is concerned with macro-level institutions." Both environments are therefore relevant for inter-personal trust in the board of directors and so are arguably a lot more. National institutions are on top of the hierarchy of systems and environments that can influence personal trust. I therefore expect institution-based trust to have a positive effect on inter-personal trust and thus expect inter-inter-personal trust within the corporate board to be higher when the firm operates in a country with high institution-based trust. This leads to the following hypothesis:
Hypothesis 3. Institution-based trust is positively related to inter-personal trust within the board of directors.
Besides inter-personal trust, I also expect institution-based trust to have a positive effect on firm performance. Scholars agree on the positive effects of a country's control of corruption (Faruq, Webb, & Yi, 2013); (Gaviria, 2002); (Sahakyan & Stiegert, 2012), rule of law (Elango & Lahiri, 2014), government effectiveness (Choi, Jiang, & Shenkar, 2015); (Elango & Lahiri, 2014), and regulatory quality (Hallward-driemeier, Wallsten, & Xu, 2006) on firm performance. Quality institutions improve firm performance by lowering transaction hazard among firms. Since institutions generate institution-based trust, I expect a direct positive effect of institution-based trust on firm performance. I therefore formulate the following hypothesis:
Hypothesis 4. Institution-based trust is positively related to firm performance
2.4. The complete interplay
12 trust on firm performance is therefore contingent on the level of inter-personal trust within the board. Its positive effect on a firm can be lower if it causes an overinvestment in inter-personal trust within the board, and higher if it remains or moves into the middle ground between under-and overinvestment. This discussion leads to my final hypothesis:
Hypothesis 5. Inter-personal trust within the board of directors partially mediates the relationship between institution-based trust and firm performance
2.5. Control variables
Besides inter-personal and institution-based trust, there are various other variables that have an impact on firm performance. Based on prior studies, with respect to firm performance I have to take into account firm characteristics like firm size, age, sales growth, and leverage (Frijns et al., 2016); (Short & Keasey, 1999); (Veprauskaite & Adams, 2013), and board characteristics such as board size, age, gender, and national diversity (Adams & Ferreira, 2009); Anderson et al., 2011); (Masulis, Wang, & Xie, 2012). I also include industry and country effects. By including these control variables I can isolate the effect of inter-personal and institution-based trust on firm performance. Figure 2 depicts the theoretical model.
Figure 2
Depiction of the theoretical model
+ + +/- +/-
This figure depicts the theoretical model. Please note that board performance is a latent variable of inter-personal trust within the board of directors. Institution-based trust has both a direct and indirect effect on firm performance.
The board characteristics gender, age, and national diversity have both a direct effect on firm performance and inter-personal trust. They affect inter-personal trust by increasing board heterogeneity, and affect firm performance by influencing board performance. For
Mediating variable
Inter-personal trust within corporate boards
Independent variable Institution-based trust Dependent variable Firm performance Board performance Control variables
13 instance, "women are not part of the 'old boys' network, which allows them to be more independent" (Kang, Cheng, & Gray, 2007). Moreover, "they may have a better understanding of consumer behavior, the needs of customers, and opportunities for companies in meeting those needs" (Kang et al., 2007). This can impact board performance. In terms of the age of directors, Kang et al. (2007) explain that while the older group of directors "can provide experience, wisdom, and usually the economic resources, the middle group carries the major positions of active responsibilities in corporates and in society, whereas the younger group has the energy and drive to succeed and plan ahead for the future." These different dynamics can also influence board performance. In relation to national diversity, "the appointment of a foreign national may sharpen differences between domestic and overseas directors" (Manzoni, Strebel, & Barsoux, 2009). Boards can then "become polarized or split into factions determined by how the different members perceive the new director and his or her contributions" (Manzoni et al., 2009). For instance, "behaviors like interrupting or excitability may have been the norm in a director's previous surroundings, but they can raise eyebrows where they are not generally accepted" (Manzoni et al., 2009). Board gender, age and national diversity can therefore function as control variables for both inter-personal trust and firm performance.
3. Data and Methods
3.1. Sample
The sample I use consists of 1494 publicly listed and privately-owned firms that are located in 32 different countries. I include all 19 Bureau van Dijk's major sectors except for the financial sectors (banks and insurance companies) and the public administration & defense sector. I exclude the latter due to a lack of firm observations. Due to data limitations, the sample only includes firms from 2015 with directors that have one of 15 European nationalities. The sample contains 7,255 directors in total. Table 1 and 2 provide comprehensive overviews of the firm sample and table 3 of the sample of directors.
Tables 1 and 2 show that some countries and sectors feature more often than others, which may point towards sample bias. This means that the results can be unrepresentative of the population of firms to which it applies. Firms from France and the United Kingdom are arguably overrepresented in the sample (21.35% and 21.02%), as well as firms in the other services sector (33.33%). Table 3 reports that Italian (17.82%), French (19.88%), and British (25.57%) directors are the largest group of directors in the sample. This may also point to sample bias. However, since directors do not necessarily cluster together at the sector and country level, the distortion effect of the statistical analysis arguably diminishes.
3.2. Data collection
14 the firm from the sample. Also, when there was unclear information such as directors with multiple nationalities I decided to omit all nationalities except the first listed. In terms of firm characteristics, I obtain data on firm performance, firm size, sales growth, leverage, and the age of the firm. Besides that, I gather data from Guiso et al. (2009) on inter-personal trust and various national distance measures including data from Hofstede (2001) on cultural distance.
Table 1
Number of firms located in a particular country
Country Number of firms % Total
Austria 3 0.20 Belgium 159 3.55 Bulgaria 2 0.13 Canada 1 0.07 Croatia 1 0.07 Côte d'Ivoire 1 0.07 Denmark 26 1.74 Finland 22 1.47 France 319 21.35 Germany 29 1.94 Greece 85 5.69 Hong Kong 1 0.07 India 10 0.67 Ireland 12 0.80 Israel 1 0.07 Italy 187 12.52 Lithuania 1 0.07 Luxembourg 8 0.54 Malta 1 0.07 Morocco 1 0.07 Netherlands 15 1.00 Norway 127 8.50 Poland 2 0.13 Portugal 9 0.60 Romania 1 0.13 Senegal 1 0.13 Spain 35 2.34 Sweden 216 14.46 Turkey 2 0.13 Ukraine 1 0.07 United Kingdom 314 21.02
United States of America 5 0.33
Total 1,494 100.00
15
Table 2
Number of firm observations per sector
Sector Number of firms % Total
Chemicals, rubber, plastics, non-metallic products 128 8.57
Construction 41 2.74
Education, Health 26 1.74
Food, beverages, tobacco 56 3.75
Gas, Water, Electricity 38 2.54
Hotels & restaurants 29 1.94
Machinery, equipment, furniture, recycling 222 14.86
Metals & metal products 48 3.21
Other services 498 33.33
Post & telecommunications 37 2.48
Primary sector 46 3.08
Publishing, printing 62 4.15
Textiles, wearing apparel, leather 43 2.88
Transport 57 3.82
Wholesale & retail trade 135 9.04
Wood, cork, paper 28 1.87
Total 1,494 100.00
The sectors are classified according to Bureau van Dijk's major sectors. Banks, insurance companies, and public administration & defense are omitted. The second column shows the number of firms in each sector. The third column reports the percentage of firms in each sector.
Table 3
Number of directors from each country
Country Unique directors % Total
Austria 29 0.40 Belgium 203 2.79 Denmark 120 1.65 Finland 155 2.14 France 1442 19.88 Germany 224 3.09 Greece 303 4.12 Ireland 85 1.17 Italy 1293 17.82 Netherlands 98 1.35 Norway 628 8.66 Portugal 31 0.43 Spain 133 1.83 Sweden 656 9.04 United Kingdom 1855 25.57 Total 7,255 100.00
16
3.3. Measures
The following paragraphs elaborate on the variables I utilize in my empirical analysis. Table 6 reports detailed summary statistics of all the variables.
3.3.1. Dependent variables
For firm performance, I follow Frijns et al. (2016) who use Tobin's Q and return on assets (ROA) as measures. Some scholars also use return on equity (ROE) as measure (Zabri, Ahmad, & Wah, 2016); (Hatem, 2014); (Veprauskaite & Adams, 2013), but this is highly correlated with ROA and therefore redundant. "Tobin's Q is calculated as the book value of total assets minus the book value of equity plus the market value of equity, all divided by the book value of total assets" (Frijns et al., 2016). A Tobin's Q ratio between 0 and 1 means that the book value of a firm is lower than its market value, while a ratio higher than 1 implies that the firm's market value is higher than its book value. If a firm's Tobin's Q is increasing, it means the firm is performing better. "ROA is calculated as operating income divided by the year-end book value of total assets" (Frijns et al., 2016) expressed in percentages. It provides an indicator of how efficient a firm is using its assets in generating profits. A higher ROA also implies better firm performance. I manually remove several extreme outliers before I winsorize both variables at 1% on each side of the distribution and transform Tobin's Q into its natural logarithm.
3.3.2. Independent variables
17 To obtain a single indicator of inter-personal trust within the board of each firm, I have to calculate how much each director trusts another director in the board. In other words, I need to calculate the average trust distances for all director pairs within a board. Since the data on trust is asymmetric, I use the following equation adapted from Frijns et al. (2016):
= ∑ ( ) (1)
where Board trusti is the measure of inter-personal trust within the corporate board of firm i,
and m is the number of directors. Trustlj is the bilateral trust between each two directors l and
j. The measure is then normalized for the size of the board.
To construct a measure for institution-based trust, I use the quality of governance dataset constructed by the Worldwide Governance Indicators (Kaufmann, Kraay, & Mastruzzi 2010). They provide aggregate indicators of six broad dimensions of countries' governance from 1996 until 2015. The WGI has gathered this data "from a number of survey institutes, think tanks, non-governmental organizations, international organizations, and private sector firms" (Kaufmann et al., 2015). The dataset includes data for all 32 countries in my sample. The relevant measures as proxies of institution-based trust are government effectiveness, regulatory quality, rule of law, and control of corruption. The measures are indexed from -2.5 to 2.5, where -2.5 implies a very weak judgment and 2.5 a very strong judgment of the quality of the institution. As a proxy, the intuition is that a higher value means more institution-based trust in the country where a firm is located. Table 4 reveals that these four measures are highly correlated. Factor analysis reports that institution-based trust is mostly defined by a country’s rule of law. I therefore omit the other three measures and use rule of law as the variable for institution-based trust.
Table 4
Correlation matrix of institution-based trust
Variable Government
effectiveness
Rule of law Regulatory quality Control of corruption Government effectiveness 1 Rule of law 0.993 1 Regulatory quality 0.933 0.933 1 Control of corruption 0.974 0.9883 0.924 1
This table reports the correlation statistics of the proxies of the institution-based trust variable.
3.3.2. Control variables
I control for four director-level characteristics: board age, board gender, board size, and board national diversity. Board agei measures the average age of directors, Board genderi the
percentage of female directors, and Board sizei the number of directors within the corporate
18 within the corporate board of firm i, which means that a high index implies low national diversity.
Besides data on trust, Guiso et al. (2009) have constructed several distance measures between ordinary citizens of the 15 European countries. I use this data on geographical, genetic, somatic, religious, historical, and cultural distance between directors to test the robustness of the national diversity and inter-personal trust variables. Geographical distance is measured as the log of the distance in km between the capitals of two countries. Genetic distance is based on "the difference between the frequencies of alleles in two populations" (Guiso et al., 2009). Somatic distance is based on the average height, hair color, and facial features (cephalic index) of nationalities. Religious similarity is based on "the empirical probability that two randomly chosen individuals in two countries will share the same religion" (Guiso et al., 2009). Contrary to the other measures, religious similarity is a reversed distance index, as two directors with the same nationality receive a value of 1 instead of 0 (as in zero distance). Historical distance refers to "the number of years a country pair has been engaged in a war between 1000 and 1970" (Guiso et al., 2009). Lastly, cultural distance is based on four cultural dimensions related to a nationality: power distance, individualism, masculinity and uncertainty avoidance. These dimensions are expressed in values between 0 and 100. A high value implies that a director contains a strong presence of the cultural dimension.
Similarly as formula (1) for Board trusti, I construct a bilateral distance measure which
captures the distance between each pair of directors for each variable. The data is symmetric, which implies that there is no distance between directors with the same nationality. To determine for instance the genetic distance within the board of a firm, I use the following equation:
= ∑ ( ) (2)
where Genetici is the average genetic distance measure of the board of firm i, and m the
number of directors. Genlj refers to the genetic distances between each two directors l and j.
The measure is then normalized for the size of the board. By using this formula, I obtain the average distance measure of all nine variables of the board of firm i. The intuition as proxy of inter-personal trust is that more distance implies lower trust. A board with no national diversity will have zero average distance for all six variables, except for religious similarity, which will yield 1.
Besides director characteristics, I control for the following firm characteristics: firm size, firm age, leverage, and sales growth. I measure firm size by taking a firm's total assets and net sales, Firm age as the age of the firm since the date of incorporation, Leverage as "the proportion of long-term debt to total assets," (Veprauskaite & Adams, 2013) and Sales growth by taking the yearly change of net sales. I winsorize Sales growthi at 3% on each side of the
19 scholars using similar variables (Chen, Leung, & Goergen, 2017); (Frijns et al., 2016); (Veprauskaite & Adams, 2013).
Lastly, I control for the data’s clustering effect by adding industry and country dummies. Table 6 reports detailed summary statistics of all variables.
3.4. Method
Since the theoretical model is characterized by a mediation effect, I use generalized structural equation modeling (GSEM) as method. GSEM is an extension of structural equation modeling (SEM) that allows the inclusion of factor variables in the regressions. "SEM is a powerful technique that can combine complex path models with latent variables (factors)" (Hox & Bechger, 2008). It is used "to estimate a system of equations" (Huber, 2014). When applied to my theoretical model, GSEM can account for the effect of institution-based trust on inter-personal trust before estimating the direct effects of both variables on firm performance. It essentially connects two regression equations where the last equation is affected by the first. It also allows me to isolate the direct and indirect effect of institution-based trust on firm performance. In this way, I can investigate whether inter-personal trust partially mediates the relationship between institution-based trust and firm performance. In that sense, I utilize GSEM as a convenient alternative to Baron and Kenny's (1986) method of estimating mediation. Their estimation method incorporates three separate regression equations which GSEM performs simultaneously. For the sake of clarity, I follow their steps of establishing mediation. When applied to my theoretical model, it functions in the following way. The first step is to regress institution-based trust on firm performance using regression equation (3):
= + - + ℎ +
ℎ + + + (3)
where subscript i denotes ith firm. lnTobin’s Qi and ROAi are the firm performance measures
and Institution-based trusti is the measure of institution-based trust. Board characteristicsi
include Board sizei, Board genderi, Board agei, and Board nat. diversityi as control variables,
and firm characteristicsi include Firm agei, lnAssetsi, lnSalesi, Sales growthi, and Leveragei as
control variables. Industryeffectsi and Countryeffectsi are dummies to control for clustering
effects, and εi is the error term. Institution-based trust should be significantly related to firm
performance.
The second step is to regress institution-based trust on inter-personal trust using regression equation (4):
= + - + + +
. + (4)
where Board trusti is the measure of inter-personal trust within the board of the ith firm.
20 board characteristics that are included in the regression equation have both an effect on inter-personal trust and firm performance as explained in section 2.5. This may cause multi-collinearity issues, which I will address in section 3.4.1. For the second step, institution-based trust should be significantly related to inter-personal trust.
The last step is to regress both institution-based trust and inter-personal trust on firm performance using regression equation (5):
= + - + +
ℎ + ℎ + +
+ (5)
where both institution-based trust and inter-personal trust are predicting firm performance. Assuming that the previous two steps were successful, inter-personal trust completely mediates based trust if the latter turns insignificant. It partially mediates based trust if both variables remain significant and the regression coefficient of institution-based trust differs from regression equation (3). If any of the three steps are not met, then Baron and Kenny's (1986) mediation test has failed, which means the data is inconsistent with the hypothesis that inter-personal trust within the board of directors partially mediates the relationship between institution-based trust and firm performance (H5).
Since I hypothesized the relationship between inter-personal trust and firm performance to be nonlinear in the shape of an inverted U, I also have to include the variable Board trust2i, which is Board trusti in its squared form. This variable captures how much the
slope of Board trusti changes every time it changes one unit. The final regression equation that
captures the complete interplay between institution-based trust, inter-personal trust within the board of directors, and firm performance is therefore the following:
= 0+ - + + +
ℎ + ℎ + + +
(6)
I will perform GSEM on regression equation (6), which essentially includes all steps of Baron and Kenny's (1986) method of establishing mediation. By doing so, I will be able to pass judgment on all five hypotheses.
3.4.1. Statistical assumptions
21 conducting a Durbin-Wu Hausman test following Davidson and MacKinnon (1993). This test requires instruments that need to be significantly related to inter-personal trust, but not to firm performance. Scholars argue that religious and somatic distance (Brown, Gray, McHardy, & Taylor, 2015); (Guiso et al., 2009); (Knack & Keefer, 1997) are appropriate instruments. Differences in religion can cause distrust between persons (Putnam, 1993) and there is "no direct association between religion and firm performance" (Brown et al., 2015). Besides that, DeBruine (2002) argues that people tend to distrust others whose appearance is different than their own. Guiso et al. (2007) confirm this argument when using somatic distance to explain trust differences, and there is to my knowledge no observed significant relationship between somatic distance and firm performance. I will therefore use somatic and religious distance as instruments.
3.4.2. Robustness tests
This paper performs three robustness checks to test the robustness of the results of the previously discussed method. First, I test the robustness of the relation between national distance and inter-personal trust by substituting the additional distance measures variables for the national distance variable. These distance measures are essentially caused by national diversity, and therefore function, similar to board national diversity, as proxies of inter-personal trust. I therefore conduct nine separate regressions. This is depicted by regression equation (7):
= + - + +
+ + + (7)
where Additional distance measuresi includes the distance variables: lnGeographici, Genetici,
Somatici, Historici, and Relgiousi, and Cultural distancei the distance variables: Poweri,
Individualismi, Masculinityi, and Uncertainty avoidancei. Please note that these distance
measures are not pooled together, but separately included in the regressions.
22
= 0+ - + +
+ ℎ + ℎ +
+ + (8)
where additional distance measures2i captures the nonlinear effect of the distance measures.
Similarly to regression equation (7), the additional distance measures are not pooled together, but separately included in the regressions.
Lastly, I address the potential sample bias by conducting additional regressions using regression equation (6) without firm observations from France and the United Kingdom. The sample is representative of the whole population if the significance of the variables does not change when firms from one of these countries are excluded. If a sample bias exists, the results may be inaccurate.
4. Results and analyses
In section 4, I first report and analyze the results of the constructed measures, assumption tests, proposed method, and robustness tests. Afterwards, I relate the analysis of the findings to the hypotheses.
4.1. Summary statistics
Table 6 reports the detailed summary statistics of all the variables. The average Tobin’s Q of a firm is 1.090 (-0.514 when expressed in its natural logarithm) and the average ROA is 1.811. This means that the average firm in my sample is performing well, as its market value is slightly higher than its book value and the ROA is positive (1.8%). The average Board trust is 3.229 and Institution-based trust 1.429. These values indicate that both inter-personal and institution-based trust are rather high in my sample. The average Board nat. diversity is 0.907, which implies that national diversity within boards is very low. The median value is 1, which means that there is no national diversity at all. Most boards are therefore characterized by directors that have the same nationality. Table A2 in the appendix depicts the correlation matrix including all variables. The reported values do not instigate any concerns.
4.2. Assumption tests
This section analyzes the results of the tests that address various statistical assumptions. First, the Breusch-Pagan test and White’s general test for heteroskedasticity report that my empirical model exhibits heteroskedasticity. To control for this, I use robust standard errors in all regression equations. Second, the Variance Inflation Factor (VIF) test reports that equation (6) exhibits high multicollinearity due to Board trusti and Board trust2i. By mean centering
Board trust2, the VIF test no longer shows high multicollinearity. Third, the Skewness and Kurtosis test reports that both dependent variables, lnTobin’s Qi and ROAi, are not normally
distributed. Transforming a variable into its natural logarithm may help solve this issue. However, lnTobin’s Qi is already expressed in its natural logarithm and ROAi has negative
23 when transforming ROAi plus constant into its natural logarithm, the Skewness and Kurtosis
test still reports a non-normal distribution. Despite that, this is not necessarily an issue, as it does not violate the Gauss-Markov theorem. Fourth, the Durbin-Wu Hausman test reports the relationship between inter-personal trust within the corporate board and firm performance to be characterized by endogeneity only when lnTobin’s Qi is the dependent variable. I therefore
conduct a two stage least squares regression for regression equation (6) using somatic and religious distance as instruments for inter-personal trust when lnTobin’s Qi is the dependent
variable. Tables A3-A6 depict the results of all the assumption tests.
Table 6
Detailed summary statistics of all variables
Obs. Mean Median Min Max Variance Skewness Kurtosis
Firm performance lnTobin's Q 1,267 -0.514 -0.503 -4.12 1.994 1.370 -0.412 3.516 Tobin's Q 1,267 1.090 0.605 0.015 7.346 1.759 2.543 10.261 ROA 1,407 1.811 3.432 -41.47 38.94 140.91 -0.699 5.194 Trust Board trust 1,494 3.229 3.277 2.39 3.69 0.071 -0.405 2.582 Institution-based trust 1,494 1.429 1.782 -0.801 2.073 0.432 -1.006 2.627 Additional distance measures lnGeographic 1,494 0.828 0 0 7.566 2.861 1.961 5.741 Genetic 1,494 0 0 0 0.014 0 4.677 33.06 Somatic 1,494 0.289 0 0 5 0.48 3.23 15.593 Historic 1,494 0.006 0 0 0.198 0 5.659 44.71 Religious 1,494 0.923 1 0 1 0.029 -2.597 9.752 Cultural distance Power 1,494 1.805 0 0 40 26.596 3.926 20.57 Individualism 1,494 1.524 0 0 38 17.532 4.355 27.341 Masculinity 1,494 2.885 0 0 61 60.167 3.503 16.774 Uncertainty avoidance 1,494 3.13 0 0 65 66.473 3.476 17.03 Board characteristics Board age 1,494 56.755 56.667 33.75 80.14 39.075 0.0923 3.34 Board gender 1,494 0.135 0 0 1 0.029 1.136 3.98 Board size 1,494 4.786 4 2 20 7.13 1.469 6.192
Board nat. diversity 1,494 0.907 1 0.28 1 0.033 -1.68 4.339
Firm characteristics lnAssets 1,492 11.555 11.502 3.477 19.85 4.761 0.101 3.144 lnSales 1,368 11.119 11.233 0.365 19.26 6.021 -0.479 4.001 Sales growth 1,312 0.002 -0.034 -0.61 0.971 0.086 1.192 6.127 Leverage 1,484 0.524 0.533 0 2.176 0.087 0.725 5.536 Firm age 1,482 35.949 27 0 278 1142.84 2.191 9.669
24
4.3. Baron and Kenny's (1986) mediation test
Table 7 reports the results of regression equation (3), (4), and (5). The first two columns depict the results of the first step of Baron and Kenny's (1986) method of establishing mediation. The results show that Institution-based trusti is significantly and positively related
to lnTobin's Qi, but insignificantly to ROAi. When Institution-based trusti changes by one
unit, Tobin's Qi increases by 334.9%. The first step of establishing mediation is therefore
successful.
The third column reports the results of the second step of the mediation test. The results show that Institution-based trusti is significantly and positively related to Board trusti,
where the Board trust score increases by 0.300 when the former changes one unit. This confirms the second step of the mediation test.
The last two columns report the last step of the mediation test. The results show that Board trusti is insignificantly related to both firm performance variables. This means that
Baron and Kenny's (1986) mediation test is unsuccessful. Institution-based trusti is again
positively and significantly related to lnTobin's Qi, but not to ROAi. However, its coefficient
is slightly lower. The three steps show that inter-personal trust does not fully or partially mediate the relationship between institution-based trust and firm performance. Before I draw any conclusions, I will first report and analyze the results of the complete model.
4.4. Generalized structural equation modeling
Table 8 reports the results of regression equation (6). This model includes the Board trust2i
variable that captures the nonlinear effect of inter-personal trust on firm performance. The first two columns show that Institution-based trusti is significantly related to both Board trusti
and Board trust2i, where it has a positive effect on Board trusti and a negative effect on Board
trust2i. This could possibly capture the discussed effect institution-based trust has on
inter-personal trust, where inter-inter-personal trust moves up on the curve which characterizes its relationship with firm performance.
The columns also show that Board nat. diversityi is significantly related to both Board
trust variables. It increases Board trusti by 0.518 and reduces Board trust2i by 0.120. This
means that national diversity has a negative effect on inter-personal trust. This also means that a higher concentration of nationality in the board has a similar effect on inter-personal trust as institution-based trust. Board genderi is only significantly related to Board trust2i, while Board
agei is insignificantly related to both Board trust variables. The effect of gender diversity on
inter-personal trust is therefore ambiguous.
The last two columns show that the GSEM including Board trust2i does not cause any
significant changes compared to the results of regression equation (5). The variable is insignificantly related to both firm performance variables. The coefficients of Institution-based trusti and Board trusti slightly change, but their significance level remains the same. It
confirms the findings of the mediation test, where inter-personal trust does not mediate the relationship between institution-based trust and firm performance.
25 sizei is only significantly related to ROAi on which it has a negative effect, while Leveragei
and lnAssetsi are negatively related to both firm performance measures. Sales growthi is only
significantly related to lnTobin's Qi on which it has a positive effect. Board agei, Board nat. Table 7
Result of regression equations (3), (4), and (5)
lnTobin's Q (3) ROA (3) Board trust (4) lnTobin’s Q (5) ROA (5)
Variables (1) (2) (3) (4) (5) Board trust 0.074 3.240 (0.354) (4.163) Institution-based trust 3.349** -1.635 0.300*** 3.307* -3.427 (1.653) (9.924) (0.009) (1.691) (10.295) Board characteristics Board age -0.005 0.033 -0.001 -0.004 0.035 (0.005) (0.057) (0.001) (0.005) (0.057) Board gender 0.306* 5.711** -0.033 0.305* 5.685*** (0.170) 1.981 (0.027) (0.170) (1.980)
Board nat. diversity -0.143 4.196** 0.518*** -0.185 2.364 (0.158) (1.985) (0.033) (0.244) (3.341) Board size 0.018 -0.346** 0.018 -0.353** Firm characteristics (0.012) (0.146) (0.012) (0.147) Leverage -1.376*** -10.693** -1.376*** -10.707*** (0.161) (2.000) (0.161) (1.997) lnAssets -0.139*** -1.515*** -0.140*** -1.523*** (0.028) (0.381) (0.028) (0.380) lnSales 0.061** 3.386*** 0.061** 3.389*** (0.027) (0.353) (0.027) (0.352) Sales growth 0.302** 0.680 0.061** 0.689 (0.115) (1.492) (0.027) (1.493) Firm age 0.000 0.003 0.000 0.003 (0.001) (0.009) (0.001) (0.009)
Industry effects YES YES NO YES YES
Country effects YES YES NO YES YES
Constant -5.476* -19.429 2.397*** -5.612* -25.433
(3.026) (18.730) (0.057) (3.129) (20.492)
Observations 1,180 1,242 1,494 1,180 1,242
R-squared 0.433 0.223 0.599 0.433 0.223
26
Table 8
Results of regression equation (6) using GSEM
Board trust Board trust2 lnTobin’s Q ROA
Variables (1) (2) (3) (4) Board trust 0.383 7.362 (0.511) (5.809) Board trust2 0.615 8.589 (0.658) (7.302) Institution-based trust 0.300*** -0.031*** 3.365* -2.609 (0.009) (0.004) (1.838) (11.777) Board characteristics Board age -0.001 -0.000 -0.005 0.036 (0.001) (0.000) (0.005) (0.057) Board gender -0.033 0.051*** 0.310* 5.708*** (0.027) (0.014) (0.166) (1.982)
Board nat. diversity 0.518*** -0.120*** -0.277 1.103
(0.033) (0.018) (0.262) (3.559) Board size 0.019 -0.345** Firm characteristics (0.012) (0.147) Leverage -1.374*** -10.616*** (0.157) (1.990) lnAssets -0.139*** -1.508*** (0.027) (0.377) lnSales 0.059** 3.358*** (0.027) (0.379) Sales growth 0.301*** 0.657 (0.112) (1.491) Firm age 0.000 0.004 (0.001) (0.009)
Industry effects NO NO YES YES
Country effects NO NO YES YES
Constant 2.397*** 0.116** -6.689* -39.792
(0.057) (0.023) (3.578) (26.615)
Observations 1,494 1,494 1,180 1,242
R-squared 0.599 0.111 0.434 0.224
27 diversityi, and Firm agei are all insignificantly related to firm performance. The opposite signs
of lnAssetsi and lnSalesi are noteworthy, as both variables represent the size of a firm. The
effect of firm size on firm performance is therefore ambiguous.
Table 9 reports the indirect and total effect of institution-based trust on firm performance. Although inter-personal trust has no significant mediating effect, the results show that the coefficient of Institution-based trusti is actually affected by both Board trust
variables. The total effect of Institution-based trusti on Tobin's Qi that includes the mediation
effect of Board trusti is 348%. The inclusion of the mediation effect of Board trust2i causes
Institution-based trusti's effect to be 334.6%. The mediation effect of Board trusti is 11.5%
and the mediation effect of Board trust2i is -1.9%. Although the indirect effects are
insignificant, they do show the resemblance of an inverted U-shaped effect, where the mediating effect of inter-personal trust can both increase and decrease the effect of institution-based trust on firm performance.
Table 9
Indirect and total effect of institution-based on firm performance
lnTobin’s Q ROA
Variables (1) (2)
Institution-based trust:
Indirect effect via
Board trust 0.115 2.210
(0.153) (1.745)
Board trust2 -0.019 -0.265
(0.020) (0.227)
Total effect via
Board trust 3.480* -0.399
(1.832) (11.714)
Board trust2 3.346* -2.874
(1.838) (11.769)
Notes: ***p<0.01, **p<0.05, *p<0.1. Robust standard errors in parentheses. Column (1) includes the GSEM of equation (6) with lnTobin's Q as dependent variable, and column (2) with ROA as dependent variable.
4.5. Instrumental variables regression
Since regression equation (6) exhibits endogeneity with lnTobin’s Qi as dependent variable, I
perform a two-stage least squares regression using somatic and religious distance as instruments for inter-personal trust. Table 10 reports the results of this instrumental variable regression.
The first stage regression only shows significant results for the instruments Religiousi
and Somatici, Board nat. diversityi, and the firm size variables. Institution-based trusti is no
longer significantly related to Board trusti. This seems to contradict previous findings.
28
Table 10
Instrumental variable regression of equation (6) with lnTobin’s Q as dependent variable 1st stage Board trust 1st stage Board trust2 2nd stage lnTobin’s Q Variables (1) (2) (3) Board trust -2.378 (3.466) Board trust2 -4.476 (5.401) Institution-based trust 0.216 -0.132 2.999*** (0.413) (0.196) (1.143) Board characteristics Board age -0.000 -0.000 -0.005 (0.000) (0.000) (0.005) Board gender -0.006 -0.001 0.274 (0.030) (0.021) (0.172)
Board nat. diversity 0.005 0.224*** -0.598
(0.061) (0.060) (1.262) Board size -0.001 -0.000 0.012 Firm characteristics (0.001) (0.001) (0.013) Leverage 0.003 -0.005 -1.391*** (0.007) (0.005) (0.162) lnAssets 0.002 -0.003** -0.148*** (0.002) (0.002) (0.029) lnSales -0.001 0.003** 0.074** (0.002) (0.001) (0.031) Sales growth -0.001 0.002 0.305*** (0.008) (0.006) (0.116) Firm age -0.000 -0.000 0.000 (0.000) (0.000) (0.001) Instruments Religious 0.571*** -0.290*** (0.074) (0.077) Somatic -0.027* 0.037** (0.016) (0.018)
Industry effects YES YES YES
Country effects YES YES YES
Constant 2.490*** 0.302 2.608
(0.732) (0.366) (11.026)
Observations 1,180 1,180 1,180
R-squared 0.944 0.667 0.395
29 The second stage regression reports that both Board trust variables remain insignificantly related to lnTobin’s Qi. Thus when controlling for endogeneity, inter-personal
trust is still not related to firm performance. The coefficient of Institution-based trusti slightly
decreases but the variable is now significant at the 1% level. The signs of the control variables remain the same. Only Board genderi is no longer significantly related to lnTobin’s Qi. The
effect of gender diversity on firm performance is therefore equivocal. Contrarily, the effect of Leveragei on firm performance is consistently negative, and the effect of Sales growthi to a
lesser extent consistently positive. The effect of firm size is again ambiguous.
4.6. Robustness tests
4.6.1. Robustness of board national diversity
Table 11 reports the results of regression equation (7), which tests the robustness of the relationship between national distance and inter-personal trust. The results show that all variables that substitute Board nat. diversityi are significantly and negatively related to Board
trusti. Only Religiousi is positively related to Board trust, but since this is a similarity
measure, it also causes Board trusti to decrease with more distance. Institution-based trusti is
consistently significantly and positively related to Board trusti. The results depict Board nat.
diversityi in relation to Board trusti as very robust.
Board genderi is eight out of ten times significantly and negatively related to Board
trusti. This contradicts previous findings where the effect of Board genderi seemed to be
ambiguous. Board agei is seven out of ten times significantly and negatively related to Board
trusti. This also contradicts previous findings that reported Board agei to be unrelated to Board
trusti. The robustness test therefore confirms the theoretical notion that gender and age
30
Table 11
Result of regression equation (7) with Board trust as dependent variable Board trust
Variables (1) (2) (3) (4) (5) (6)
Board nat. diversity 0.518*** (0.033) Additional distance measures lnGeographic -0.056*** (0.004) Genetic -51.379*** (5.801) Somatic -0.149*** (0.008) Historic -4.639*** (0.331) Religious 0.675*** (0.031) Institution-based trust 0.300*** 0.298*** 0.285*** 0.294*** 0.281*** 0.292*** (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) Board gender -0.033 -0.043 -0.058** -0.068*** -0.053** -0.066*** (0.027) (0.027) (0.028) (0.025) (0.026) (0.025) Board age -0.001 -0.001* -0.002** -0.001** -0.001 -0.001* (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Constant 2.397*** 2.922*** 2.941*** 2.941*** 2.925*** 2.261*** (0.057) (0.044) (0.047) (0.044) (0.047) (0.053) Observations 1,494 1,494 1,494 1,494 1,494 1,494 R-squared 0.600 0.604 0.549 0.626 0.590 0.664
31
Table 11 continued
Result of regression equation (7) with Board trust as dependent variable
Variables (7) (8) (9) (10) Cultural distance Power -0.020*** (0.001) Individualism -0.023*** (0.002) Masculinity -0.010*** (0.001) Uncertainty avoidance -0.011*** (0.001) Institution-based trust 0.279*** 0.283*** 0.291*** 0.288*** (0.009) (0.009) (0.009) (0.009) Board gender -0.062** -0.067** -0.053* -0.070*** (0.025) (0.026) (0.027) (0.027) Board age -0.001 -0.002** -0.001* -0.002** (0.001) (0.001) (0.001) (0.001) Constant 2.931*** 2.959*** 2.931*** 2.952*** (0.045) (0.046) (0.047) (0.046) Observations 1,494 1,494 1,494 1,494 R-squared 0.628 0.613 0.554 0.590
Notes: ***p<0.01, **p<0.05, *p<0.1. Robust standard errors in parentheses. All columns show the result of regression equation (7) with Board trust as dependent variable. However, each column has one different independent variable.
4.6.2. Robustness of board trust
Tables 12 and 13 report the results of regression equation (8), which tests the robustness of Board trusti and Board trust2i in relation to firm performance. Since institution-based trust is
theoretically only related to inter-personal trust, where the latter is a latent variable of distance between directors, I am not interested in any mediating effects.
With lnTobin’s Qi as dependent variable, Institution-based trusti is consistently
positively and significantly related to firm performance, while all distance variables are consistently insignificantly related to firm performance. With ROAi as dependent variable,
Institution-based trusti remains insignificantly related. All distance variables are insignificant
except for historic distance. The results show that Historici increases ROAi by 77.933 and
Historic2i decreases the slope of Historici by 587.19 every time it changes one unit. This is in
32 redundant. Regardless of that, the robustness test confirms the insignificant relationship between inter-personal trust and firm performance.
Table 12
Result of regression equation (8) with lnTobin's Q as dependent variable lnTobin’s Q Variables (1) (2) (3) (4) (5) Institution-based trust 3.346* 3.180* 3.340** 3.021* 3.092* (1.760) (1.705) (1.655) (1.686) (1.705) Distance measures lnGeographic 0.094 (0.160) lnGeographic2 -0.012 (0.017) Genetic -76.164 (79.000) Genetic2 8,722.705 (6,646.436) Historic 4.119 (3.421) Historic2 -29.162 (19.674) Somatic -0.162 (0.164) Somatic2 0.018 (0.034) Religious -0.534 (0.894) Religious2 -1.037 (0.942)
Control variables YES YES YES YES YES
Industry effects YES YES YES YES YES
Country effects YES YES YES YES YES
Constant -5.961* -4.914 -5.587* -4.468 -4.586
(3.522) (3.222) (3.027) (3.162) (3.171)
Observations 1,180 1,180 1,180 1,180 1,180
R-squared 0.433 0.434 0.434 0.434 0.434
33
Table 12 continued
Result of regression equation (8) with lnTobin's Q as dependent variable Variables (6) (7) (8) (9) Institution-based trust 3.346* 3.180* 3.340** 3.021* (1.760) (1.705) (1.655) (1.686) Distance measures Power 0.021 (0.018) Power2 -0.001 (0.001) Individualism -0.002 (0.024) Individualism2 0.000 (0.001) Masculinity 0.009 (0.015) Masculinity2 -0.000 (0.000) Uncertainty avoidance -0.003 (0.013) Uncertainty avoidance2 0.000 (0.000)
Control variables YES YES YES YES
Industry effects YES YES YES YES
Country effects YES YES YES YES
Constant -5.666* -5.837* -5.569* -5.428*
(3.103) (3.327) (3.093) (3.160)
Observations 1,180 1,180 1,180 1,180
R-squared 0.434 0.433 0.434 0.433
34
Table 13
Result of regression equation (8) with ROA as dependent variable ROA Variables (1) (2) (3) (4) (5) Institution-based trust -0.361 -7.370 -2.091 -8.127 -4.351 (11.194) (10.573) (9.990) (9.627) (9.475) Distance measures lnGeographic 0.186 (2.013) lnGeographic2 0.013 (0.212) Genetic -1,107.934 (1,072.157) Genetic2 62,449.334 (78,662.987) Historic 77.933* (41.165) Historic2 -587.19** (232.720) Somatic -1.985 (2.096) Somatic2 -0.067 (0.478) Religious 1.764 (11.552) Religious2 -3.923 (12.578)
Control variables YES YES YES YES YES
Industry effects YES YES YES YES YES
Country effects YES YES YES YES YES
Constant -23.880 -3.840 -20.745 -1.037 -13.258
(27.985) (21.787) (18.964) (19.921) (19.511)
Observations 1,242 1,242 1,242 1,242 1,242
R-squared 0.223 0.224 0.226 0.227 0.223
