The effect of mandatory gender representation in the board on firm performance; evidence from Norway

The effect of mandatory gender representation in

the board on firm performance; evidence from

Norway

Stefan Alberda

S2752670

s.alberda@student.rug.nl

University of Groningen, MSc. Accountancy

Dr. S. Mukherjee

University of Groningen, Supervisor

Dr. N. Hussain

University of Groningen, Co-assessor

June 25, 2018

Word count: 7.023

Abstract: In this study, I investigate the effect of the Norwegian implementation of the

voluntary gender quota in 2003 on firm performance. This study introduces a separation of five different backgrounds of female directors in a corporate board and investigates the effect of these different backgrounds on firm performance. A difference-in-difference research design for the period 2000-2008 with Sweden as the control group is executed. Based on 490 firm years of 247 unique companies, this study found a significant negative relationship for female directors with an academic background and a significant positive relationship for female directors with a scientific background on firm performance. This study contributes to the growing literature on the effect of board gender quota by evaluating the effect of the implementation of a board gender quota on firm performance and attributes this effect to particular director backgrounds.

Keywords * Gender Diversity * Gender Quota Legislation * Director Background * Norway

* Tobin’s Q Ratio * Difference-in-Difference *

Acknowledgement: I would like to thank my supervisor dr. S. Mukherjee from the University of Groningen

for his guidance and useful comments. Next to this, I would like to thank C. Schiphorst, T. Meijer, E. Meijer and L. Brouwer for the cooperation on collecting the data for this thesis.

TABLE OF CONTENTS

Contents

INTRODUCTON

In 2003, the Norwegian government was the first government that introduced a voluntary board gender quota, stating that Norwegian companies should have at least a 40 per cent female representation in their board of directors (Ahern & Dittmar, 2012). In 2006, Norway made the quota mandatory and by then companies had to comply (Matsa & Miller, 2013). Since most of the Norwegian companies did not comply with the quota when the law was announced, more female directors had to be recruited. This study introduces five different director background groups and places all female directors, based on their primary profession in one of those background groups. This study investigates the effect of mandatory women representation in corporate boards on firm performance, where a distinction between five different director background groups will be made.

Gender inequalities and gender differences are still major corporate governance issues. O’Brien, Fitzsimmons, Crane and Head (2017) show that there is a minority of women in corporate boards, women are earning less for the same labour and women are in the majority in unpaid labour. These factors are resulting in a greater workspace inequality (O'Brien, Fitzsimmons, Crane, & Head, 2017). Arguments for increased levels of board gender diversity, and therefore implementing a board gender quota, can be divided in two categories: ethical and economic (Campbell & Mínguez-Vera, 2008). From an ethical point of view, male and female are the same, so they should be treated the same when new directors are appointed. To equalize the gender differences found by O’Brien et al. (2017), a board gender quota can be implemented. This results in a more equally distribution of male and females in corporate boards. The economic view, on the other hand, claims that female directors will lead to an increased firm performance, meaning that the board gender quota should be implemented to increase firms’ economic performance. This study researches the proposition of the economic view by investigating the effect of mandatory gender representation on firm performance and dividing this effect to particular background groups.

To investigate the effect of mandatory women representation in corporate boards, this study uses a difference-in-difference research design, where Norway is the treatment group and Sweden is the control group. Norway and Sweden are homogenous in terms of gender and nationality (Randoy, Thomsen, & Oxelheim, 2006), but Sweden does not have a board gender quota and therefore the effect of the board gender quota in Norway can be measured. In a difference-in-difference research design, the variable of interest is the interaction variable,

which includes treatment multiplied by post and background. This interaction variable gives the effect of the particular background in Norway after the implementation of the voluntary quota in 2003 on firm performance. In this study, five interaction variables are created, where each interaction variable represents one of the five background groups. These interaction variables are used to execute the regressions.

The data on the backgrounds of female directors in this study is hand-collected by five Master’s students of the University of Groningen. Background information of the female directors is gathered by the use of publicity available sources. Based on this background information like primary profession, female directors are placed in one of the five background groups. The five background groups identified in this study are academic, social, scientific, service and invisible.

This study found two significant relationships between the backgrounds of women in a corporate board and firm performance. First female directors with an academic background have a significant negative relationship with firm performance after the implementation of a board gender quota. Second, this study found that female directors with a scientific background have a significant positive relationship with firm performance after the implementation of a board gender quota. This study found no significant relationship between the other three backgrounds and firm performance after the implementation of a board gender quota. These results will be elaborated in the results section.

This study contributes to the literature by expanding on the existing board gender diversity literature. This will be done by researching the effect of the backgrounds of female directors on firm performance in a mandatory setting This study places all women directors, based on their primary profession in a background group. There are studies that investigated the expertise of directors and matched this expertise with the companies. Faleye, Hoitash and Hoitash (2018) found that companies with particular characteristics are more likely to appoint industry expert directors. Furthermore, prior research found that outside directors with an unconventional background are the best directors in terms of independence and monitoring (Carter, Simkins, & Simpson, 2003). This study extends on this literature by investigating the effect of female director backgrounds on firm performance in a mandatory setting. Since Norway implemented a voluntary board gender quota in 2003, this study investigates the effect of this quota, and particularly the backgrounds on firm performance. The relationship between the background groups and firm performance in a mandatory setting in Norway is not

researched before in a difference-in-difference research design setting for the period 2000-2008, and therefore this study gives new insights in this relationship.

With this study, countries that are considering implementing a board gender quota can gain more insights in the effects of the implementation of a board gender quota. This study investigates the effect of board gender diversity and the backgrounds of female directors on firm performance. Countries that consider implementing a quota might need this information to make a thought-out decision. Next to this, companies that face the consequences of the implementation of a board gender quota can use this study to make recruitment decisions. The remainder of this study is organized as follows: first, the background will be elaborated and the hypothesis of this study will be developed. Thereafter, the methodology will be explained and the data will be summarized. In the third part, the results will be elaborated and this study will conclude with the conclusions and limitations of this study.

HYPOTHESIS DEVELOPMENT

In this section, the background of board gender quota will be elaborated and the hypotheses of this study will be formed.

In 2003, Norway was the first country in the world that introduced a board gender quota, where companies could comply with on a voluntary basis (Matsa & Miller, 2013). Since the first of January 2006, the quota was enforced as mandatory; all existing public listed companies had two years to comply with the quota. Companies founded after the first of January 2006 had to comply directly (Matsa & Miller, 2013). By January first 2008, all companies had to comply with the quota, where non-compliance firms can face dissolution (Ahern & Dittmar, 2012). The Norwegian quota law (Law of Public Limited Companies) states that the quota applies for both directors elected by shareholders and directors elected by employees. With this distinction, companies are not able to comply with the quota by just adding more employee elected female directors. The Norwegian quota law describes the mandatory number of female directors in a board in relation to the total board size. When a corporate board has more than ten directors, at least 40 per cent of the directors should be female. There specific rules for corporate boards with fewer than ten directors. For nine directors, four women are required, six to eight directors, three female directors are required. For four to five directors, two female directors are required and for a board with two or three directors, each gender should be represented at least ones (Law of Public Limited Companies).

Board gender diversity has been researched many times before. Studies found relations between board gender diversity and governance (Adams & Ferreira, 2009), informativeness of stock prices (Gul, Srinidhi, & Ng, 2011), earnings quality (Srinidhi, Gul, & Tsui, 2011) and firms financial performance (Campbell & Mínguez-Vera, 2008). Adams and Ferreira (2004) investigated the effects of gender diversity in the boardroom. They found a significant and robust negative relation between board gender diversity and risk. This suggests that when there are more women in the corporate board, fewer risks will be taken. Adams and Ferreira (2004) also found a positive relationship between board gender diversity and performance related pay. This shows that boards with more females reduce the fixed salary of board members and increase the relative importance of performance related pay. The reduction of risk-taking results in a more informal and less opportunistic disclosure. The increase in relative importance of performance related pay results in more alignment between shareholders and management.

Therefore, diversity in corporate boards is related to corporate governance and to the agency theory.

The agency theory of Jensen and Meckling (1973) describes the separation between ownership and control of a firm. In this relation, management can be seen as the agent, which gets the authority to make decisions for the firm on behave of the shareholders. The shareholders are the principals in this relation; they give the decision-making authority of the firm to the hired management. The shareholders have to trust the management they hired. In the agency theory there is a basic principle that describes that the interests of the principals and the agents differ (Jensen & Meckling, 1973). Corporate governance is a mechanism that prevents shareholders from agency problems; it aims to align managers and shareholders interest.

Norway and Sweden are both Scandinavian countries where boards are homogenous in terms of gender and nationality (Randoy, Thomsen, & Oxelheim, 2006). However, there is a difference in terms of laws and regulations regarding board gender diversity, because there is a gender quota in Norway. Norway introduced the voluntary quota for board gender diversity in corporate boards (Matsa & Miller, 2013), whereas Sweden does not have such a quota. The t-test in the methodology section shots that despite not having a quota, the board gender diversity in Sweden also increased in the 2000-2008 period. This increase was smaller than the increase in Norway, although still significant. Legitimacy theory explains the increase of board gender diversity by arguing that in the Scandinavian countries, society expects a higher proportion of women in corporate boards. This is the reason that Norway implemented the board gender quota in 2003 (Ahern & Dittmar, 2012). With this board gender quota, companies had to comply or explain and in this way the Norwegian government tried to create a more equal gender distribution in corporate boards. Since 2006, the quota is mandatory and by then it is crucial for the companies to comply. So, according to the legitimacy theory, the board gender quota is introduced because a higher proportion of gender equality in corporate board was expected by society. Because companies have to comply with the board gender quota, board gender diversity increased in Norway.

The relationship between board gender diversity and firm performance is researched many times before. These studies were executed in different settings and the results are mixed. In a setting with US firms, Kochan et al. (2003), Farrell and Hersch (2005) and Randoy et al. (2006) found no significant relationship between the presence of females and firm performance, where Charter et al. (2003) found a positive relation between board gender diversity and firm

performance. The results from a Scandinavian setting are mixed as well. Bohren and Strom (2010) investigated Norwegian firms in the 1989-2002 period. This was the period before the gender quota was announced. Bohren and Strom (2010) found a negative relationship between board gender diversity and shareholder value. Smith, Smith and Verner (2006), on the other hand, made a distinction between female directors elected by employees and female directors elected by shareholders. They found a negative relationship between firm performance and female directors elected by shareholders, and a positive relation between female directors elected by employees and firm performance (Smith, Smith, & Verner, 2006).

Looking at prior research on firm performance and board gender diversity, firm performance is mainly measured in two ways: by accounting measures and by Tobin’s Q. This study follows the papers of Ahern and Dittmar (2012), Coles, Daniel, and Naveen (2008) and Yermack (1996) by using Tobin’s Q as the measure of firm performance. Prior research used the Tobin’s Q as a measure for firm performance for two reasons. First, according to Montgomery and Wernerfelt (1988), Tobin’s Q reflects the market’s expectations of future earnings and therefore shows the competitive advantage of the firm. Second, Ahern and Dittmar (2012) focused on Tobin’s Q rather than other performance measurement indicators, because of the accounting changes in Norway in the period around 2005. Before 2005 most of the firms followed the Norwegian Generally Accepted Accounting Principles (NGAAP). However, since 2005 Norwegian firms have to use International Financial Reporting Standards (IFRS) (Ahern & Dittmar, 2012). When accounting rules change, accounting-based performance measures are less reliable than market value performance measures, such as Tobin’s Q (Oystein, Knivsfla, & Saettem, 2008). This studies timeframe is from 2000-2008, so the accounting change is also included in this study. Following Ahern and Dittmar (2012), I use Tobin’s Q as the measure of firm performance.

Arguments for increased levels of board gender diversity can be split in two categories: economic and ethical (Campbell & Mínguez-Vera, 2008). Based on the agency theory and the paper of Adams and Ferreira (2004), an increase in board gender diversity would mean that there is fewer risk-taking in a corporate board, and board members act more in favour of the share- and stakeholders. This means that according to the agency theory, there is an economic reason for increased levels of board gender diversity, resulting in a positive expected relationship between board gender diversity and firm performance.

Legitimacy theory, on the other hand, explains the increase in board gender diversity as mandatory to comply with the board gender quota. Matsa and Miller (2013) state that policymakers in Norway implemented the board gender diversity quota to promote equality rather than increasing firm performance. This would mean that women are added to the corporate board, because the company has to comply with the board gender quota. In this case, Ahern and Dittmar (2012) state that binding legal constraints result in a lower firm performance.

Backgrounds

Previous studies found relationships between board characteristics and firm performance. Adams and Ferreira (2009) for example on board gender diversity, and Oxelheim and Randoy (2003) on foreigners in the corporate board. However, backgrounds of females in corporate boards are a unique setting. This study identifies five different backgrounds: invisible, academic, service, social and science. This paragraph elaborates these backgrounds further. Invisible directors include all directors where no information was found about their primary profession. Ullmann (1985) summarized previous studies on voluntary disclosure and firms economic performance, where he concluded that there is a small positive relationship between voluntary disclosure and the economic performance. Therefore, companies who do not disclose have a slightly lower firm performance. Non-disclosure, and therefore companies with directors were no information is available of might have the same effect.

There is a considerable gap between researchers and recommendations and actual practices of management (Miller, Greenwood and Hinings, 1997; Rynes, Bartunek and Daft, 2001). This gap can be explained by the inconsistence of the findings and research relevance (Rynes, Bartunek, & Daft, 2001). The science-practice gap is not only in organizational sciences, but also in nearly all businesses where researchers and practitioners are (Rynes, Bartunek, & Daft, 2001). Due to this science-practice gap, adding academics to the corporate board might affect firm performance in a negative way.

Previous studies made no distinction in the backgrounds of directors when researching board characteristics. Therefore, empirical evidence based on the backgrounds of women in a board is not available. For that reason, service, social and science as backgrounds do not have clear definitions. Appendix C summarizes the characteristics of women with all the different backgrounds.

When interpreting the different identified backgrounds, there can be concluded that the expected relationship between backgrounds and firm performance are not available or mixed. Invisible, technical, service, social and science backgrounds might either have a positive, negative or no relationship with firm performance. Therefore the hypotheses in this study is:

Hypothesis: Difference in backgrounds of women in corporate board influence (+ / -) firm

METHODOLOGY

This section describes how I conduct the research, by describing the research method, research plan, data collection methods and the measures that I use. The first section elaborates on the sample and the data, the second section describes the variables, the third section lists the empirical model and the last section describes the summary statistics.

Sample and data

This study examines the influence of the identified backgrounds on Tobin’s Q when implementing a quota. To examine these relationships, this study makes use of time-series and cross-sectional data by looking into the differences between public listed companies in Norway and Sweden in the period of 2000-2008.

I include Norway and Sweden, because this study examines the influence of implementing a gender quota. Norway and Sweden are both Scandinavian countries where boards are homogenous in terms of gender and nationality (Randoy, Thomsen, & Oxelheim, 2006). A difference between both countries in regards to board gender diversity is that Norway implemented a board gender quota in 2003 (Ahern & Dittmar, 2012). Sweden, on the other hand, does not have such a law or regulation. I investigate the period until 2008 in this study, because Norway made the board gender quota in 2008 a hard quota (Matsa & Miller, 2013). This led to a significant change in circumstances, because since 2008 Norway serves penalties for non-compliance.

The data collected for this study comes mainly from three sources. The corporate board data used in this study comes from the BoardEx database. The market and accounting data comes from the Thomson Reuters Worldscope database. Data regarding the backgrounds of the directors in the corporate boards is hand-collected. Five Master’s students of the University of Groningen (Appendix B) gathered the hand-collected data. The data is collected from the sources listed in Appendix D.

In line with Matsa and Miller (2013), this study excludes all banking, insurance and financial firms from the sample, due to the different laws and regulations regarding ownership. Table 1 gives an overview of the data used.

Number of

observations NORWAY SWEDEN

GRAND TOTAL

COMPANYID INDIVIDUAL COMPANYID INDIVIDUAL COMPANYID INDIVIDUAL

2000 41 299 65 604 106 903 2001 52 385 78 738 130 1123 2002 58 432 83 755 141 1187 2003 61 456 94 857 155 1313 2004 65 470 97 873 162 1343 2005 67 474 97 869 164 1343 2006 67 492 98 850 165 1342 2007 68 494 100 848 168 1342 2008 68 499 99 824 167 1323 TOTAL 547 4001 811 7218 1358 11219

Table 1: sample distribution

Variables

This section gives a description of the variables used in this study.

Dependent variable

To compute the firm value, this study follows prior research on firm value and computes firm values as the yearly Tobin’s Q rate (Yermack, 1996, Coles et al., 2008, Ahern and Dittmar, 2012). The Tobin’s Q ratio (TOBINSQ_W) is measured by the total price of the firms market value divided by the total assets of the firm (Ahern & Dittmar, 2012). This Tobin’s Q was winsorized at one percent to reduce the effect of outliers.

𝑇𝑜𝑏𝑖𝑛′_{𝑠 𝑄 =} 𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠

When the Tobin’s Q ratio is above one, the Tobin’s Q ratio indicates that the company’s internal organization is very good in the eyes of the market (Hermalin & Weisbach, 1991). The stock price of an organization (Total assets + market equity) are more expensive than the replacement costs of the assets (Total assets). When the Tobin’s Q value is below one, the stock value is lower than the replacements costs of the assets (Hermalin & Weisbach, 1991).

Independent variables

Women’s background (PER_INVISIBLE, PER_SERVICE, PER_SCIENCE, PER_SOCIAL and

PER_ACADEMIC) is the independent variable in this study. This variable is measured as a

percentage of the total number of board members. So the percentage of women with a service background (PER_SERVICE) is measured by the number of women with a service background divided by the total number of board members.

𝑃𝐸𝑅_𝑆𝐸𝑅𝑉𝐼𝐶𝐸 =𝑊𝑜𝑚𝑒𝑛 𝑤𝑖𝑡ℎ 𝑎 𝑠𝑒𝑟𝑣𝑖𝑐𝑒 𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑜𝑎𝑟𝑑 𝑚𝑒𝑚𝑏𝑒𝑟𝑠

The five different backgrounds were linked to individual board members by evaluating their primary profession. An overview of the distribution of women in a particular background group can be found in appendix C.

Control variables

According to Lee (2009) revenues (LN_REVENUES_W) has a positive relation with firm performance, since firms with more revenues have higher efficiency or increased levels of market power. The revenues variable is lagged for one year, to measure the effect of a change in revenues on Tobin’s Q one year later. The revenues variable is also logged in order to create smaller coefficients.

Return On Assets (ROA_W) is often used as a proxy for firm performance in other studies regarding board diversity (Shrader, Blackburn, & Iles, 1997). Erhardt, Werbel and Shrader (2003) found a positive relationship between board gender diversity and ROA in a US setting. ROA is used as a control variable in this study, because a previous study found a positive relationship between ROA and the Tobin’s Q ratio (Kananoppadol & Pariwatnanont, 2012). According to Adams, Hermalin and Weisbach (2010), board size (LN_BOARDSIZE) may affect performance directly. This can be caused by the risk of free riding and the increased symbolic status of the board (Hermalin & Weisbach, 2003). Coles et al. (2008), on the other hand, found a U-shaped relation between board size and Tobin’s Q. This was caused by firms’ complexity, where complex firms had an increase Tobin’s Q ratio when the board size increases.

Board independence (BOARD_INDEPENDENCE) is used as a board specific control variable in this study. Prior literature found a relationship between board independence and firm performance, where a higher proportion of board independence resulted in increased levels of firm performance (Yuetang & Xiaoyan, 2006)

In studies with a difference-in-difference research design, it is standard to control for industry effects. Fama and French (1997) introduced forty-eight industries. In this study, these industries (_IFF48_2 - _IFF48_48) are matched with the firms in the sample to control for industry specific influences. Industry effects might influence the results in this study, because particular

industries are male orientated and other industries are female orientated. The nursing industry for example is an industry were women are overrepresented (Gardiner & Tiggemann, 1999).

Empirical model

This study makes use of a difference-in-difference (DiD) research design, wherein there is a comparison between the differences in the Tobin’s Q rate in Norway and Sweden before and after the implementation of the voluntary quota in Norway in 2003. Within this study, the model takes the following form:

𝑇𝑂𝐵𝐼𝑁𝑆𝑄_𝑖𝑗𝑡 = 𝛽₀+ 𝛽₁𝑇𝑅𝐸𝐴𝑇𝑀𝐸𝑁𝑇_𝑗+ 𝛽₂𝑃𝑂𝑆𝑇_𝑄𝑈𝑂𝑇𝐴_𝑖𝑡+ 𝛽₃(𝑇𝑅𝐸𝐴𝑇𝑀𝐸𝑁𝑇 ∗ 𝑃𝑂𝑆𝑇)𝐷𝐼𝐹𝐹𝑖𝑛𝐷𝐼𝐹𝐹𝑖𝑗𝑡+ 𝐵𝑥CONTROLS𝑖𝑗𝑡+ 𝛼𝑖 + 𝜆𝑡+ 𝑢𝑖𝑡

In this study, each observation of TOBINSQ represents the performance rate of company i in country j at time t. In a difference-in-difference research design, the variable of interest is the

INTERACTION variable. INTERACTION has been calculated as TREATMENT * POST_QUOTA * BACKGROUND. TREATMENT is an indicator that makes a distinction

between Norway with a board gender quota (labelled 1) and Sweden without a board gender quota (labelled 0). POST_QUOTA is an indicator which makes a distinction between before and after the implementation of the voluntary quota in Norway in 2003, therefore before 2003 is labelled ‘0’ and after 2003 is labelled as ‘1’. BACKGROUND indicates the number of women with a particular background as a percentage of the total board. INTERACTION captures the percentage of women with a particular background after the implementation of the quota in 2003 and in Norway.

Due to the existence of other variables that may affect the Tobin’s Q ratio, this study controls for firm characteristics (LN_REVENUES_W and ROA_W), board characteristics (LN_BOARDSIZE and BOARD_INDEPENDENCE), industries and years. LN_REVENUES_W has been calculated as the natural logarithm of the revenues or net sales of a company. ROA_W has been calculated as the ebit divided by the total assets of a company.

LN_BOARDSIZE is the natural logarithm of the total number of directors in the board of

directors. BOARD_INDEPENDENCE has been calculated as the percentage of non-executive directors of the total number of directors in the board of directors. INDUSTRY is a dummy for all forty-eight industries identified by Fama and French (1997). YEAR is a dummy for the years 2000-2008.

This study will be done in a fixed effects setting. In the empirical model, alpha I is the error term for the firm fixed effect, delta t is the error term for time fixed effects and Uit stands for the idiosyncratic error.

This study makes use of time series and cross sectional data, therefore heteroscedasticity is expected. To deal with this heteroscedasticity, robustness of standard errors is included in the models for this study. Next to this, a robustness check will be executed to examine the robustness of the results in a different setting. The robustness check will be done by winsorizing all variables at a five per cent level and compare the results with the original results.

Summary statistics

The descriptive statistics of the variables used in this study are summarized in table 2. This table shows the number of observations, average, standard deviation, minimum and maximum per variable used in this study. This table gives information about the data used in this study.

2000-2008 (1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Dependent TOBINSQ_W 1,297 1.047 1.236 0.0159 7.448 Independent PER_ACADEMIC 1,358 0.00983 0.0319 0 0.250 PER_INVISIBLE 1,358 0.0275 0.0577 0 0.333 PER_SERVICE 1,358 0.1272 0.171 0 0.429 PER_SCIENCE 1,358 0.0313 0.0636 0 0.400 PER_SOCIAL 1,358 0.0108 0.0382 0 0.333 Control LN_BOARDSIZE 1,189 2.059 0.363 0.693 3.045 BOARD_INDEPENDENCE 1,358 0.310 0.302 0 1 ROA_W 1,329 0.0349 0.199 -0.937 0.409 LN_REVENUES_W 1,336 14.32 2.343 5.684 18.71 BRD_BGD_W 1,358 0.166 0.141 0 0.429 Number of companyid 247 247 247 247 247

Table 2: Descriptive statistics

Notable numbers in this table are the average percentage of board gender diversity in Norway and Sweden (16.6 per cent). The descriptive statistics are split up in the period before and after 2003 in appendix E. This appendix shows that the board gender diversity in Norway and Sweden was only 7.8 per cent on average in the period of 2000-2002 and almost twenty per cent on average in the period of 2008. Next to this, it is notable that in the period of

2003-2008 there is still a board with a board gender diversity of zero per cent. In 2003-2008 there are even two Norwegian firms with a board gender diversity percentage of zero per cent; Subsea 7 INC (De-listed 01/2011) and Solvang ASA.

Another notable number is the average of the Tobin’s Q ratio (1.047). This indicates, according to Hermalin and Weisbach (1991), that on average, the internal organization is good in the eyes of the market.

T-test

To compare the means of the different variables used in this study, an independent sample t-test is executed. This is done twice to compare the means of Norway and Sweden before the implementation of the voluntary quota in 2003 and after the implementation of the voluntary quota in 2003. Based on the study of Randoy et al. (2006), it is expected that there is not much difference in terms of gender before the implementation of the gender quota. Table 3 shows the results of the independent sample t-test before the implementation and table 4 shows the results of the independent sample t-test after the implementation of the voluntary quota in 2003.

Group 2000-2002 Mean Norway Mean Sweden Std. Dev. Norway Std. Dev. Sweden Difference T-stat TOBINSQ_W 0.912 1.044 1.237 1.384 -0.133 -0.9257 PER_ACADEMIC 0.007 0.009 0.033 0.029 -0.002 -0.503 PER_INVISIBLE 0.024 0.014 0.064 0.038 0.010** 1.980 PER_SERVICE 0.034 0.061 0.086 0.108 -0.027** -2.5798 PER_SOCIAL 0.003 0.009 0.015 0.029 -0.006** -2.3129 PER_SCIENCE 0.016 0.011 0.051 0.032 0.005 1.084 LN_BOARDSIZE 1.949 2.205 0.388 0.347 -0.256*** -5.416 BOARD_INDEPENDENCE 0.120 0.178 0.210 0.211 -0.058** -2.639 ROA_W -0.084 0.009 0.498 0.296 -0.093** -2.262 LN_REVENUES_W 13.628 14.604 2.092 2.448 -0.976*** -3.972 BRD_BGD_W 0.0697 0.083 0.114 0.097 -0.014 -1.238 * p < 0.05, ** p < 0.01, *** p < 0.001 Table 3: T-test 2000-2002

The independent sample t-test of the period before the implementation of the voluntary gender quota did not find a significant difference in the means of board gender diversity between Norway and Sweden. This is in line with the results from the study of Randoy et al. (2006). Next to this, there is no significant difference in means in the Tobin’s Q ratio. When looking at the control variables of this study, the independent sample t-test in table 3 shows a significant lower mean in Norway for all four control variables. This means that board size, board

independence, return on assets and revenues are significantly lower in Norway in comparison to Sweden in the 2000-2002 period.

Group 2003-2008 Mean Norway Mean Sweden Std. Dev. Norway Std. Dev. Sweden Difference T-stat TOBINSQ_W 0.925 1.158 0.970 1.316 -0.233*** -2.919 PER_ACADEMIC 0.010 0.011 0.035 0.031 -0.001 -0.302 PER_INVISIBLE 0.043 0.023 0.072 0.049 0.019*** 5.041 PER_SERVICE 0.216 0.117 0.213 0.147 0.100*** 8.650 PER_SOCIAL 0.009 0.015 0.037 0.045 -0.006** -1.989 PER_SCIENCE 0.039 0.038 0.071 0.068 0.001 0.287 LN_BOARDSIZE 1.932 2.123 0.338 0.349 -0.190*** -8.283 BOARD_INDEPENDENCE 0.293 0.423 0.328 0.287 -0.123*** -6.567 ROA_W 0.047 0.052 0.202 0.176 -0.004 -0.358 LN_REVENUES_W 14.054 14.743 1.907 2.301 -0.689*** -4.865 BRD_BGD_W 0.240 0.172 0.155 0.120 0.068*** 7.708 * p < 0.05, ** p < 0.01, *** p < 0.001 Table 4: T-test 2003-2008 When looking at the independent sample t-test of the period 2003-2008, a significant difference in means of the Tobin’s Q ratio between Norway and Sweden was found. The average Tobin’s Q ratio is significantly lower (-0.233) in Norway than in Sweden after the implementation of the voluntary quota. In the results section, this study investigates whether the independent variables of this study are an explanation for this decline in the Tobin’s Q ratio. There is also a significant difference in means of the board gender diversity in Norway and Sweden. The board gender diversity in Norway is on average 6.8 per cent higher than in Sweden. Since there was no significant difference in means of the board gender diversity in the period before the implementation of the board gender quota, this implies that the implementation of the board gender quota in Norway might have caused the difference in means of board gender diversity in Norway and Sweden.

To conclude, in the independent sample t-test in the period after the implementation of the voluntary quota in Norway, three out of the four control variables of this study still have a significant difference in means. For board size and revenues, the difference in means between Norway and Sweden has declined when comparing the difference in means before and after the implementation of the voluntary quota in 2003. The difference in means for return on assets is no longer significant and the difference in means of board independence has grown when comparing the difference in means before and after the implementation of the voluntary quota.

RESULTS

In the previous section, I analyzed the descriptive statistics. To follow up on these descriptive statistics, in this section multicollinearity will be elaborated. Thereafter, the results of the hypotheses will be discussed. This section concludes with an overview of the results in a regression table.

Multicollinearity

Table 5 shows the correlation between the different variables used in this study. To compute these correlations a Pearson correlation test is executed. Multicollinearity gives the degree in which the variables of this study correlate with each other. Multicollinearity influences the results of the regressions executed in the next paragraph, by reducing the reliability of the results (Field, 2013). When a correlation is higher than 0.7, multicollinearity might be an issue (Farrar & Glauber, 1967). In the correlationmatrix below, shows some significant correlation between the variables used in this study. However, these correlations are not higher than 0.7 and therefore it is assumed that multicollinearity is not an issue in this study. (An & Davey, 2011)

The correlation between PER_SERVICE and BRD_BGD_W is 0.697, which is very close to the limit of 0.7 stated by An and Davey (2011). This correlation can be explained by the distribution of the women in the different background groups. Most of the women in the sample used in this study do have a background in the service industry. Therefore, the number of board gender diversity and the percentage of women with a service background are significantly correlated with a coefficient of 0.697.

Correlation matrix Correlationmatrix 1 2 3 4 5 6 7 8 9 10 11 (1) TOBINSQ_W 1 (2) BRD_BGD_W 0.0501 1 (3) BOARD_INDEPENDENCE 0.0293 0.188*** 1 (4) PER_INVISIBLE -0.112*** 0.388*** 0.0544 1 (5) PER_SERVICE 0.0189 0.697*** 0.0968** 0.0288 1 (6) PER_SCIENCE 0.121*** 0.340*** 0.151*** -0.0630* -0.0913** 1 (7) PER_SOCIAL -0.0242 0.233*** -0.0326 -0.00918 -0.0225 -0.0574 1 (8) PER_ACADEMIC -0.0550 0.217*** 0.00783 0.0185 -0.000459 0.00166 -0.00905 1 (9) LN_BOARDSIZE -0.203*** 0.104*** -0.0397 0.110*** 0.00634 0.0157 0.0456 0.207*** 1 (10) ROA_W 0.0431 0.111*** 0.109*** 0.0301 0.119*** -0.0485 -0.0251 0.0570 0.159*** 1 (11) LN_REVENUES_W -0.373*** 0.191*** 0.135*** 0.149*** 0.131*** -0.0418 0.0252 0.185*** 0.601*** 0.401*** 1 * p < 0.05, ** p < 0.01, *** p < 0.001

Results backgrounds

This section elaborates the results of the different backgrounds on the Tobin’s Q rate. Each background will be discussed separately.

Invisible

Model 1 in table 6 shows the interaction coefficient (0.717) of directors with an invisible background on the Tobin’s Q rate. This coefficient implies a positive relationship between companies who have board members where they do not report about on publicity available sources on Tobin’s Q after the implementation of a board gender quota. However, the relationship between directors with an invisible background and the Tobin’s Q rate is not significant. This means that the hypotheses that expects a relationship between directors with an invisible background and the Tobin’s Q rate, based on the science-practice gap cannot be accepted.

Service

The interaction coefficient of directors with a service background (-0.767) is not significant as well. This coefficient implies a positive relationship between companies who have directors with a service background in their board and firm performance after the implementation of a board gender quota. Due to the insignificance, the hypotheses regarding the relation between directors with a service background and firm performance cannot be accepted. The interaction coefficient of directors with a service background and Tobin’s Q can be found in Model 2 in table 6.

Academic

Model 3 in table 6 shows the interaction coefficient (-9.772) between directors with an academic background and the Tobin’s Q ratio. This coefficient is significant at a five per cent level, so this study found a significant relation between directors with an academic background and firm performance. The coefficient of -9.772 implies that adding a women director with an academic background after the implementation of a board gender quota, leads to a decrease of -9.772 in the Tobin’s Q ratio. With this significant coefficient, the hypotheses regarding the relation between directors with an academic background and firm performance can be accepted.

Social

In this study, a coefficient of -2.019 was found for the interaction between female directors with a social background and Tobin’s Q. This coefficient implies a negative relationship between female directors with a social background on firm performance, after the implementation of a board gender quota. However, this coefficient is not significant and therefore the hypotheses regarding the relation between female directors with a social background and firm performance cannot be accepted. Model 4 in table 6 provides an overview of the regression executed.

Science

Model 5 in table 6 gives an overview of the regression regarding female directors with a science background and Tobin’s Q. The interaction this coefficient (3.099) implies a positive relation between female directors with a scientific background and the Tobin’s Q ratio. This coefficient is significant at a ten per cent level and therefore the hypotheses regarding female directors with a scientific background can be accepted. The positive coefficient of 3.099 implies that adding a women director with a scientific background to the corporate board would increase the Tobin’s Q ratio by 3.099.

Regressions (DiD)

Difference-in-difference (1) (2) (3) (4) (5) VARIABLES Invisible Service Academic Social Science NORWAY_DUMMY 1.280*** 1.318*** 1.267*** 1.257*** 1.278*** (0.422) (0.412) (0.423) (0.424) (0.417) >=2003_DUMMY -0.039 -0.063 -0.010 -0.037 -0.038 (0.210) (0.220) (0.207) (0.208) (0.207) INTERACTION INVISIBLE 0.717 (2.144) INTERACTION_SERVICE -0.767 (1.364) INTERACTION_ACADEMIC -9.772** (4.331) INTERACTION_SOCIAL -2.019 (3.062) INTERACTION_SCIENCE 3.099* (2.407) ROA_W 0.670 0.664 0.738 0.636 0.632 (0.437) (0.447) (0.457) (0.453) (0.428) LN_REVENUES_W = L, -0.248 -0.270 -0.273 -0.252 -0.240 (0.194) (0.192) (0.196) (0.193) (0.193) BOARD_INDEPENDENCE = L, -0.417 -0.434 -0.474 -0.412 -0.398 (0.561) (0.546) (0.593) (0.562) (0.564) LN_BOARDSIZE = L 1.264* 1.164 1.259* 1.228* 1.235* (0.657) (0.718) (0.641) (0.674) (0.652) PER_INVISIBLE -1.445 -0.958 -1.195 -0.844 -0.908 (1.618) (1.372) (1.241) (1.356) (1.305) PER_SERVICE -0.022 0.536 -0.006 0.017 -0.046 (0.602) (1.224) (0.592) (0.609) (0.583) PER_SCIENCE 0.043 0.124 0.180 0.054 -1.144 (1.650) (1.618) (1.596) (1.647) (1.865) PER_SOCIAL 0.652 0.860 0.890 1.749 0.675 (2.105) (2.111) (2.082) (3.083) (2.161) PER_ACADEMIC 1.465 1.271 7.457* 1.492 1.316 (3.620) (3.687) (4.014) (3.628) (3.556) Constant 0.555 0.718 0.734 0.624 0.540 (1.739) (1.764) (1.728) (1.758) (1.728) Observations 490 490 490 490 490 R-squared 0.909 0.909 0.911 0.909 0.909

Country fixed-effects Yes Yes Yes Yes Yes

Year fixed-effects Yes Yes Yes Yes Yes

Industry fixed-effects Yes Yes Yes Yes Yes

Firm fixed-effects Yes Yes Yes Yes Yes

Robust standard errors in parentheses Table 6: Difference-in-difference model *** p<0.01, ** p<0.05, * p<0.1

Robustness

To verify the results found in this study, a robustness check is executed. This robustness check did the same regression, but winsorized all variables at a five per cent level. The results of this regression can be found in appendix F.

When comparing the results of the robustness check with the original results, there can be concluded that the coefficients of the interaction terms still have the same positive or negative

direction. Some coefficients have slightly changed, but there are, except for the interaction social coefficient, no big changes. The interaction coefficient for women with a social background changed from -2.019 to 0.297. However, none of the interaction coefficients for women with a social background is significant and therefore, no conclusions can be made. The significant interaction coefficients of female directors with an academic and a science background are still significant in the robustness check. The interaction coefficient for female directors with a science background only slightly changed (3.099 to 3.271). The interaction coefficient for female directors with an academic background, on the other hand, changed from -9.772 to -5.202. Due to the similarities in the original results and the robustness check executed, there can be concluded that the results of this study can be seen as robust.

CONCLUSION

This study investigated the effect of the backgrounds of women in a board. Here, a distinction was made between five different backgrounds; invisible, service, academic, social and science. With the difference-in-difference research design, this study found two significant coefficients. First, women with an academic background (-9.772), this coefficient shows that women with an academic background in a board are resulting in a lower firm performance after the implementation of a board gender quota. Second, women with a science background (3.099). This means that having women with a science background in a board lead to increased levels of firm performance. Therefore, this study concludes that having women with an academic or science background in the board after the implementation of a board gender quota leads to a significant change in firm performance.

Previous studies did not research these backgrounds and therefore no empirical evidence regarding these backgrounds is available. However, this Ullmann (1985) found evidence of a relationship between non-disclosure and firm performance, where non-disclosure leads to a lower firm performance. Since directors with an invisible background are defined as directors where no information is available, this study hypothesized a negative relation between directors with an invisible background and firm performance. This study did not find a significant relationship between women with an invisible background a firm performance, meaning that this study does not provide any evidence that supports the study of Ullmann (1985)

Limitations and implication for further research

This study has some limitations that might affect the results of this study. Five different students gathered the data regarding the backgrounds of the women in the boards, therefore a risk of inconsistency exist. Outside directors with an unconventional background are the best directors in terms of independence and monitoring (Carter, Simkins, & Simpson, 2003). Further research can use the backgrounds identified in this study to match those backgrounds with the directors. With these matches, other research can investigate whether the unconventional backgrounds also lead to increased levels of firm performance in Norway and Sweden.

REFERENCES

Adams, R., & Ferreira, D. (2004). Gender diversity in the boardroom. European Corporate

Governance Institute, Finance Working paper, 57(30).

Adams, R., & Ferreira, D. (2009). Women in the boardroom and their impact on governance and performance. Journal of Financial Economics, 94(2), 291-309.

Adams, R., Hermalin, B., & Weisbach, M. (2010). "The Role of Boards of Directors in Corporate Governance: A Concepual Framework and Survey.". Journal of Economic

Literature, 48(1), 58-107.

Ahern, K., & Dittmar, A. (2012). "The changing of the boards: The impact on firm valuation of mandated female board representation.". Quaterly Journal of Economics, 127(1), 137-197.

An, Y., & Davey, H. (2011). Towards a comprehensive theoretical framework for voluntary IC disclosure. Journal of Intellectual Capital, 571-585.

Bohren, O., & Strom, O. (2010). Governance and Politics: Regulation Independence and Diversity in the Board Room. Journal of Business Finance and Accounting, 37(9-10), 1281-1308.

Campbell, K., & Mínguez-Vera, A. (2008). Gender Diversity in the Boardroom and Frim Financial Performance. Journal of Business Ethics, 83(3), 435-451.

Carter, D., Simkins, B., & Simpson, W. (2003). Corporate governance, board, diversity, and firm value. Financial Review, 38(1), 33-53.

Coles, J., Daniel, N., & Naveen, L. (2008). "Boards: Does One Size Fit All?". Journal of

Financial Economics, 87, 329-356.

Erhardt, N., Werbel, J., & Shrader, C. (2003). Board of Director Diversity and Firm Financial Performance. Corporate Governance, 11(2), 102-111.

Faleye, O., Hoitash, R., & Hoitash, U. (2018). Industry expertise on corporate boards. Review

Fama, E., & French, K. (1997). 'Industry costs of equity.'. Journal of Financial Economics,

43(2), 153-193.

Farrar, D., & Glauber, R. (1967). Multicollinearity in regression analysis: the problem revisited. The Review of Economic and Statistics, 92-107.

Farrell, K., & Hersch, P. (2005). Additions to corporate boards: The effect of gender. Journal

of Corporate Finance, 11(1), 85-106.

Field, A. (2013). Discovering statistics using IBM SPSS statistics. Sage.

Gardiner, M., & Tiggemann, M. (1999). Gender differences in leadership style, job stress and mental health in male- and female-dominated industries. Journal of Occupational ad

Organizational Psychology, 72, 301-315.

Gul, F., Srinidhi, B., & Ng, A. (2011). Does board gender diversity improve the

informativeness of stock prices? Journal of Accounting and Economics, 51(3), 314-338.

Hermalin, B., & Weisbach, M. (1991). The effects of board composition and direc incentives on firm performance. Financial Management, 20(4), 101-112.

Hermalin, B., & Weisbach, M. (2003). Boards of Directors as an Endogenously Determined Institution: A Survey of the Economic Literature. Economic Policy Review, 7-26. Jensen, M., & Meckling, W. (1973). Theory of the firm: Managerial behavior, agency cost

and capital structur. Journal of Financial Economics, 3, 305-360.

Kananoppadol, J., & Pariwatnanont, C. (2012). Original Ownership Structure, Capital Structure, and Corporate Performance of Construction Material Segment in the Stock Exchange of Thailand. Journal of Science and Technology, 31, 673-680.

Kochan, T., Bezrukova, K., Ely, R., Jackson, S., Joshi, A., Jehn, K., . . . Thomas, D. (2003). The effects of diversity on business performance: Report of the diversity research network. Human Resource Management, 42(1), 3-21.

Lee, J. (2009). Does Size Matter in Firm Performance? Evidence from US Public Firms.

Matsa, D., & Miller, A. (2013). 'A Female Style in Corporate Leadership? Evidence from Quotas.'. American Economic Journal: Applied Economics, 5(3), 136-169.

Miller, D., Greenwood, R., & Hinings, B. (1997). Creative chaos versus munificent momentum? The schism between normative and academic views of organizational change. Journal of Management Inquiry, 6, 71-78.

O'Brien, K., Fitzsimmons, T., Crane, M., & Head, B. (2017). Workplace gender inequality as a Wicked Problem: Implications for research and practice. Academy of Management

Proceedings, 2017(1), 14717.

Oxelheim, L., & Randoy, T. (2003). The impact of foreign board membership on firm value.

Journal of Banking and Finance, 27(12), 2369-2392.

Oystein, G., Knivsfla, K., & Saettem, F. (2008). "The Value-Relevance of Adopting IFRS: Evidence from 145 NGAAP Restatements,". Journal of International Accounting and

Taxation, 17, 92-112.

Randoy, T., Thomsen, S., & Oxelheim, L. (2006). A Nordic Perspective on Corporate Board

Diversity. Norway: Nordic Innovation Centre.

Rynes, S., Bartunek, J., & Daft, R. (2001). Across the great divide: knowledge creation and transfer between practitioners and academics. Academy of Management Journal,

44(2), 340-355.

Shrader, C., Blackburn, V., & Iles, P. (1997). Women in Management and Firm Financial Performance: An Exploratory Study. Journal of Managerial Issues(9), 355-372. Smith, N., Smith, V., & Verner, M. (2006). Do women in top management affect firm

performance? A panal study of 2,500 Danish firms. International Journal of

Productivity and Performance Management, 55(7), 569-593.

Srinidhi, B., Gul, F., & Tsui, J. (2011). Female directors and earnings quality. Contemporary

Accounting Research, 28(5), 1610-1644.

Ullmann, A. (1985). Data in search of a theory: A critical examination of the relationship among social performance, social disclosure, and economic performance of US firms.

Wernerfelt, B., & Montgomery, C. (1988). Tobin's q and the importance of focus in firm performance. The American Economic Review, 246-250.

Yermack, D. (1996). "Higher Market Valuation of Companies with a Small Board of Directors,". Journal of Financial Economics, 185-211.

Yuetang, W., & Xiaoyan, W. (2006). Does Independence of the Board Affect Firm Performance? Economic Research Journal, 5, 62-73.

APPENDICES

Appendix A – Variables

Variable Definition Source

TOBINSQ_W Market

Capitalization/(Total Assets + Total Liabilities)

Worldscope

PER_INVISIBLE – PER_ACCOUNTING

Women with Background X / Number of Board Members

Hand-collected

REVENUES Revenues Worldscope

ROA Earnings before Interest

and Taxes/Total Assets

Worldscope

LN_BOARDSIZE Number of Board

Members (Directors)

BoardEx BOARD_INDEPENDENCE Number of Non-Executive

(Independent)

Directors/Board Size

BoardEx

_iyear_2000 - _iyear_2008 Year BoardEx / Worldscope _iff48_2 - _iff48_48 Fama and French (1997)

industry

Worldscope

This table shows the variables used in this study.

_________________________________________________________________________________

Appendix B – Data gathering distribution

This table shows the distribution of the data collection process.

_________________________________________________________________________________

Students Data collected Percentage

C.W. Schiphorst 621 12.51 E. Meijer 1164 23.45 L. Brouwer 1051 21.18 S. Alberda 1059 21.34 T. Meijer 1068 21.52 Total 4963 100.00

Appendix C – Background grouping

Backgrounds (5) Backgrounds (13) Primary proffesion

Invisible Invisible N/A

Service Accounting, Business,

Financial, Lawyer and Marketing. Accountant, Advisor, Banker, Business, Consultancy, Controller, Finance, Founder, Insurance, Investment, Lawyer, Logistics, Management, Marketing, Operations, Sales,

Strategy, Tax and Treasurer.

Science Scientist and Technical. Applications, Biology, Chemist, Engineer, Electrician, IT,

Laboratory, Mechanical, Medical, Pharmacist, Scientist and Technical.

Social Politician, Social, HR and

Military.

Diplomat, Editor, Ethics, Government, HRM, Journalist, Media, Military, Politician and Union Workers.

Academic Academic Professor, Researcher and

Economist. This table presents the background merging from thirteen to five background groups.

___________________________________________________________________________ Appendix D – Used sources for hand-collected data

Data source Used by

Annual reports E. Meijer, T. Meijer, S. Alberda

Management Scope E. Meijer, S. Alberda

Bloomberg (website) E. Meijer, T. Meijer, S. Alberda, C.W. Schiphorst

Linkedin E. Meijer, T. Meijer, S. Alberda, C.W.

Schiphorst

Parlement.com E. Meijer

Wikipedia E. Meijer, T. Meijer, S. Alberda, C.W.

Schiphorst

Proff.no T. Meijer

Companies websites T. Meijer, S. Alberda, C.W. Schiphorst

Newsweb.no T. Meijer

4-Traders S. Alberda

News articles S. Alberda

Business.dk C.W. Schiphorst

This table shows the main sources used to collect the data. L. Brouwer did not provide his sources. __________________________________________________________________________________

Appendix E – Descriptive statistics

2000-2008 (1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Dependent TOBINSQ_W 1,297 1.047 1.236 0.0159 7.448 Independent PER_ACADEMIC 1,358 0.00983 0.0319 0 0.250 PER_INVISIBLE 1,358 0.0275 0.0577 0 0.333 PER_SERVICE 1,358 0.127 0.171 0 0.429 PER_SCIENCE 1,358 0.0313 0.0636 0 0.400 PER_SOCIAL 1,358 0.0108 0.0382 0 0.333 Control LN_BOARDSIZE 1,189 2.059 0.363 0.693 3.045 BOARD_INDEPENDENCE 1,358 0.310 0.302 0 1 ROA_W 1,329 0.0349 0.199 -0.937 0.409 LN_REVENUES_W 1,336 14.32 2.343 5.684 18.71 BRD_BGD_W 1,358 0.166 0.141 0 0.429 Number of companyid 247 247 247 247 247 2000-2002 (1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Dependent TOBINSQ_W 361 0.993 1.329 0.0159 7.448 Independent PER_INVISIBLE 377 0.0182 0.0501 0 0.333 PER_SERVICE 377 0.0503 0.1001 0 0.429 PER_SCIENCE 377 0.0128 0.0406 0 0.333 PER_SOCIAL 377 0.00620 0.0244 0 0.143 PER_ACADEMIC 377 0.00814 0.0306 0 0.250 Control LN_BOARDSIZE 251 2.108 0.383 0.693 3.045 BOARD_INDEPENDENCE 377 0.155 0.212 0 1 ROA_W 372 -0.00639 0.252 -0.937 0.409 LN_REVENUES_W 374 14.18 2.410 5.684 18.71 BRD_BGD_W 377 0.0778 0.104 0 0.429 Number of companyid 247 247 247 247 247 2003-2008 (1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Dependent TOBINSQ_W 936 1.067 1.198 0.0233 7.448 Independent PER_INVISIBLE 981 0.0311 0.0600 0 0.300 PER_SERVICE 981 0.1567 0.1835 0 0.429 PER_SCIENCE 981 0.0384 0.0692 0 0.400 PER_SOCIAL 981 0.0126 0.0422 0 0.333 PER_ACADEMIC 981 0.0105 0.0324 0 0.200 Control LN_BOARDSIZE 938 2.046 0.357 0.693 3.045 BOARD_INDEPENDENCE 981 0.370 0.311 0 1 ROA_W 957 0.0510 0.171 -0.937 0.409 LN_REVENUES_W 962 14.38 2.314 5.684 18.71 BRD_BGD_W 981 0.1995 0.139 0 0.429 Number of companyid 247 247 247 247 247

Appendix F – Robustness test

DIFFERENCE-IN-DIFFERENCE (1) (2) (3) (4) (5)

VARIABLES Invisible Service Academic Social Science

NORWAY_DUMMY 1.207*** 1.217*** 1.200*** 1.208*** 1.205*** (0.224) (0.213) (0.222) (0.225) (0.220) >=2003_dummy 0.072 0.070 0.090 0.079 0.072 (0.146) (0.146) (0.147) (0.148) (0.146) INTERACTION_INVISIBLE 0.684 (2.090) INTERACTION_SERVICE -0.215 (0.581) INTERACTION_ACADEMIC -5.202** (2.189) INTERACTION_SOCIAL 0.297 (2.884) INTERACTION_SCIENCE 3.271* (1.910) ROA_W 0.848** 0.843** 0.882** 0.848** 0.809** (0.339) (0.335) (0.344) (0.336) (0.329) LN_REVENUES_W = L -0.137 -0.142 -0.150 -0.135 -0.130 (0.097) (0.101) (0.098) (0.097) (0.095) BOARD_INDEPENDENCE = L, -0.329 -0.335 -0.361 -0.332 -0.310 (0.340) (0.336) (0.343) (0.340) (0.334) LN_BOARDSIZE = L 1.509*** 1.475*** 1.502*** 1.505*** 1.478*** (0.465) (0.471) (0.457) (0.463) (0.461) PER_INVISIBLE -1.791 -1.344 -1.463 -1.373 -1.269 (1.547) (1.055) (1.022) (1.102) (0.994) PER_SERVICE -0.266 -0.113 -0.260 -0.279 -0.291 (0.379) (0.433) (0.370) (0.379) (0.338) PER_SCIENCE -0.406 -0.410 -0.350 -0.451 -1.655 (1.415) (1.398) (1.373) (1.382) (1.553) PER_SOCIAL -0.124 -0.015 0.034 -0.202 -0.108 (1.759) (1.737) (1.713) (3.104) (1.747) PER_ACADEMIC 0.483 0.373 3.638* 0.387 0.334 (1.805) (1.826) (1.902) (1.815) (1.738) Constant -0.528 -0.488 -0.437 -0.548 -0.543 (0.896) (0.919) (0.889) (0.899) (0.887) Observations 490 490 490 490 490 R-squared 0.937 0.937 0.938 0.937 0.938

Country fixed-effects Yes Yes Yes Yes Yes

Year fixed-effects Yes Yes Yes Yes Yes

Industry fixed-effects Yes Yes Yes Yes Yes

Firm fixed-effects Yes Yes Yes Yes Yes

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1