The relation between female board representation and the bid-ask spread : case study for Norway

(1)

MSc Finance

_________________________________________________________________________________________________________________________________________________

The relation between female board

representation and the bid-ask spread: Case

study for Norway

_________________________________________________________________________________________________________________________________________________

Name: Lychelle de Lannoy

Date: 1 July 2018

Student number: 10522271

Supervisor: Dr. T. Jochem

Track: Corporate Finance

Faculty: Faculty of Economic

(2)

Statement of Originality

This document is written by Student Lychelle de Lannoy who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

For this paper/ research, I examine the relation between board diversity and information

asymmetry for Norwegian companies. To my knowledge, only 2 papers exist that look into this direct relation and both find that a more diverse board has less asymmetric information. By running a difference-in-difference regression with public Norwegian firms that are listed on the Oslo Stock Exchange and public US firms that are listed on NYSE, we found that there is no significant difference between Norwegian companies that have significantly higher female representations than US companies. The proxy used for information asymmetry is the bid-ask spread and to measure the board diversity, we used 3 different measures. Those are percentage women on the board, BLAU-index and the SHANNON-index.

Additionally, this thesis also zooms in to look at two different channels where female board members might contribute to the information asymmetry concern of a company; these are the monitoring channel and the disclosure channel. In both cases I found insignificant results.

(4)

1. Introduction

One of the important objectives of corporate governance is the concern about the effect of asymmetric information between the company and investors. This means that one party knows more (e.g. private information) about the company than the others. There has been quite some research done over the years about this topic, with regards to corporate disclosure, debt financing and equity financing.

Stock liquidity is directly related to asymmetric information. As can be seen by the various research done about the famous proxy for stock liquidity, which is the bid-ask spread. This spread is the difference between the amount an investor is willing to pay for a stock (bid price) and the price an investor is willing to sell the stock (ask price). This difference exists due to 3 components, namely the order processing cost, inventory holding cost and the asymmetric information cost (Stoll, 2012). The asymmetric information cost is the biggest part of the bid-ask spread and the other 2 components are only a very small portion. Therefore this thesis will use the bid-ask spread as a proxy for information asymmetry (Glosten & Harris, 1988).

Investors’ value high liquidity, which means that they value limited asymmetric information between the company and the investors. High liquidity means that investors can convert their shares in cash easily and vice versa. Consequently, they expect a premium for holding less liquid stocks, because there is a higher risk associated with less liquid stocks (Acharya & Pedersen, 2005). Less liquid stocks translate to a higher degree of market inefficiency. Therefore, stock liquidity is not only important for the investors but also an important aspect for companies to take into account, because by improving stock liquidity they can have a direct effect on the development of the market and the flow of capital (Hoshi et al., 1991). Hence, asymmetric information in the market is an important aspect for both investors and companies.

One of the main players to communicate information to the external parties is the board of directors. The tasks of board of directors vary from establishing the different goals and visions of a company, set a strategy and structure to achieve those goals and communicate to the shareholders (Gabrielsson & Winlund, 2000). However, there is one crucial concern about the board of directors, that has been a very trending topic in the last decade. This is the acknowledgement of the diversity of the board of directors in different countries.

(6)

Over the years, there have been a lot of discussions about the gender equality issue around the world. One of the main focuses of this discussion is the business perspective of gender equality, which includes the board of directors. For this reason, there has been done a lot of research on gender diversity and different measures of firm performance. These discussions and research papers can be divided up into two groups with each having a different opinion about a significant female representation on the board.

There is a group that highly believes that diversity in a board is mandatory and results in positive consequences for the company (Adams & Funk, 2012). However, there is also a group who believes that a board should be formed based on the knowledge and expertise of the board members and not based on gender. Thus forcing women on the board might have negative consequences for the company, according to the last group (Adams & Ferreira, 2009; Eckbo et al., 2016).

Figure 1: evolution of female board representation in Norway

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

(7)

This debate is highly visible when you compare European countries with the United States. In December 2005, Norway was one of the first European countries that introduced a mandatory quota legislation; where is stated that at least 40% of the board members must be female. All public and government companies had to comply with this law by December 2007 (Ahern & Dittmar, 2012). As can be seen in figure 1, the percentage of women on a board increased from 2001 until 2007, where it stabilized after 2007 around the 40%. Other European countries also followed in their footsteps, e.g. Netherlands introduced a gender quota for companies that have more than 250 employees. It is stated that for those companies at least 30% of the board members should be female. However, it is somewhat different from the Norwegian quota, because in Norway the companies get a sanction if they do not comply (mandatory quota) whereas in The Netherlands this is not the case (voluntary quota). Other European countries also followed the footsteps of either Norway or The Netherlands, by introducing either a mandatory gender quota or voluntary gender quotas. This makes clear that European countries firmly believe that such a quota has positive consequences for the companies. On the contrary, the US is still skeptical about such a gender quota. Therefore this thesis will answer the following research question: Does the mandatory quota for Norwegian firm induce a significant relation between percentage of women on the board and the bid-ask spread compared to US firms?

Over the years, various research has been done on the effect of a gender diverse board on different aspects of management, such as the amount spend on R&D (innovation), the amount of risk taken, monitoring, disclosure, performance of a firm etc. These research papers contribute indirectly to the relation between female board representation and the information asymmetry of a firm. For example, Kidwell, Stevens & Bethke (1987) finds that men are more likely to cover up mistakes compared to women. This indicates that women tend to be more honest and therefore there would be better information exchange if women would be on the board. Additionally, Adams and Ferreira (2009) find that women tend to be more on monitoring committees. Based on these findings, this would result in more transparency when having more women on the board.

(8)

Furthermore, Cohen et al. (1998) state that there is also a significant difference in leadership styles between men and women. Women tend to have a more cooperative approach and are less overconfident than men (Gul et al, 2011). Therefore, they value high standards of verification. Moreover, Brown and Hillegeist (2007) found that the relation between a more gender diverse board and the quality of disclosure is positive, which in return results in a decrease of information asymmetry. Hence, based on these findings of previous papers, I expect the relation between female board representation and the bid-ask spread to be negative. Furthermore, this thesis also looks into two specific channels to provide in-depth analyses of the results. The first channel is the transparency channel, which focuses on the relation between female board representation and the quality of earnings management. The second channel is the corporate disclosure channel, which focuses on the effect of women on the board and the relation between EPS and the share price.

The results of the main analysis provide no evidence that there is a significant relation between a gender diverse board and information asymmetry for Norwegian companies. Additionally, by looking into the two channels, I found that indeed there is no significant relation between female board representation and transparency or disclosure for Norwegian firms. Therefore, it confirms the results of the main analysis.

This thesis contributes to the limited research done on the direct relation between women on the board and asymmetric information. To my knowledge, only 2 other studies did such a research, but their focus was on Spain and Australia. Hence this thesis contributes to the literature by doing research for Norwegian companies. Furthermore, this thesis focuses on Norway because it provides us with a natural experiment. Namely, it was the first one in Europe that introduced a mandatory quota, which came as a surprise to the public.

Additionally, this thesis compares two countries with each other that have different viewpoints on the gender diversity issue. To my knowledge, that has not been done before in previous research that tried to investigate the relation between female board representation and information asymmetry. Lastly, this thesis also contributes to the general prior variety of research that focus on having women on the board and its effect on different characteristics of the firm.

(9)

The remaining of this paper will be as follow; In section 2 I will provide an overview of the implementation of the gender quota in Norway, followed by an overview of existing literature on the female board representation and the bid-ask spread. Section 3 will provide the hypothesis and the methodologies and section 4 will elaborate on the data used for the analyses. Further, in section 5 the results will be discussed followed by the conclusion and some limitations in section 6.

(10)

2. Literature Review

In this section I will start of by explaining the process of the implementation of the gender board quota in Norway. Followed by a section that describes various research about the difference between men and women on the board. Thirdly, I will explain in detail what the bid-ask spread is and the components of the bid-bid-ask spread. Furthermore, I will provide an overview of existing literature about the relation between female board representation and the bid-ask spread.

2.1 Board gender quota in Norway

As already mentioned in the introduction, Norway was the first European country that implemented a board gender quota. The process of implementing such a quota started of in the Cabinet. This discussion was not directly focused on the board structure, but rather on the general gender equality problem in Norway, which ranged from a committee level to a board level. Not until 2001, the Cabinet decided to present this amendment to the parliament for a vote. A year later, in 2002 the Cabinet proposed a board gender quota to the parliament. This was submitted to the parliament in June 2003 and in November 2003 the parliament voted in favor of the board gender quota (Eckbo et al., 2016). In this version of the law it was stated that they would implement a voluntary quota for the public and government firms. However, if the firms did not comply with this law by 2005, they would implement the mandatory quota. By 2005, the majority of the firms did not comply with the voluntary quota; therefore the parliament was forced to include the quota in their corporate laws on December 2005, with a monetary sanction if the companies did not comply by December 2007 (Eckbo et al., 2016). For companies that went to the stock exchange between the period of 2003 and 2005, they had to comply with the gender board quota immediately, whereas the other listed companies had the period up to December 2007 to comply with this law (Eckbo et al., 2016).

2001-2003

introduction period - June 2003: Parlement voted in favor for gender quota

2004-2005 voluntary quota period - December 2005: gender quota implemented in corporate law 2006-2007 mandatory quota period (without sanctions) - December 2007: Sanction were introduced 2008-after mandatory quota period (with sanctions)

(11)

2.2 Difference between male and female board members

Various research cover the aspect of the difference between male and female directors. These papers come from different fields, such as psychological studies, medical studies, behavioral studies, but also economic and financial researchers started to look into this difference.

Terjesen, Sealey & Singh (2009) mentioned that there are 3 theories which state that a female board director can contribute to the performance of the board, these are: agency theory, resource dependency theory and gender role theory.

The agency theory addresses the difference between the objectives of the “agent” and the “principal” (Eisenhardt, 1989). A simple illustration of the agency theory is the relationship between CEO and shareholders. A familiar conflict between CEO and shareholders is the amount of risk the company should take. The CEO should act in the best interest of the shareholders, however the benefits for the CEO might not lie directly in line with the interest of the shareholders. Hence a conflict of interest arises. This is a common concern in corporate governance. Terjesen et al. (2009) argues that according to the agency theory, female directors can contribute to the performance of the board by bringing new insides to the table and limit the conflicts between the agent and principal. The second theory that Terjesen et al. (2009) introduces is the gender role theory, which states that female directors bring the feminine aspects to the boardroom. Because the feminine aspects and characteristics are the new “insides” that the female board members bring to the table, these two theories go hand in hand. Within these theories it is clear that there are different aspects that differ between men and women. Research has been done on multiple of these aspects, which indicate that there is indeed a difference between male directors and female directors.

One important category within these theories is based on a psychological aspect of men and women. Men and women have different ways of thinking and behaving. Kidwell, Stevens & Bethke (1987) looked into the difference between ethical decision-making of men and women. This research was based on questionnaires that were handed out to 60 male managers and 60 female managers in the U.S. Their findings were that men and women have similar perspective of ethical values in most of the situations except for one. That is; men are more likely to conceal one’s errors compared to women. This indicates that women are more likely to

(12)

be honest to others compared to men. As the board is one of the key players to make decisions and monitor the management it is obvious that having a women on the board would be beneficial for the information flow of the company.

A second category within these theories is the behavior of board members. Adams & Feirreira (2009) looked into the monitoring behavior of women on the board of U.S companies. They found that female board members attend more meetings than male board members. However, they also found that female attendance to board meetings positively influence male board members attendance. Hence, female directors engage in board monitoring and have a significant impact on the behavior of male directors. Additionally, they find that female directors are more often on the monitoring committee than male directors. Equity holders and bondholders rely on the internal control of the company, which is done by the monitoring committees. Therefore for the information exchange it is important that there are reliable/critical people on these committees, which seem to be the case with having female board members who are seen as independent members (Adams & Feirreira, 2009). Although there are differences between the board structures of different companies, all the boards should have at least 1 monitoring committee. The term “monitoring committees” includes e.g. audit committee, governance committee & compensation committee (Upadhyay, Bhargava & Faircloth, 2014). The audit committee is in charge for the earnings management of a company. There is quite some research done about gender diverse boards and the quality of earnings management. Srinidhi, Gul & Tsui (2011) looked into this relation for U.S companies and found that a board with a higher level of female representation has a higher quality of earnings management. This relationship is tested using a Two-Staged Least Square (2SLS) analysis. The main argument for the positive relation between female representation and earnings quality is that female board members tend to be independent directors. Adams et al. (2012) confirms this argument by stating that because women on boards are not part of the “old-boys club”, they are most likely independent directors.

Another aspect that falls under the behavior differences between men and women are the level of confidence of directors. Barber & Odean (2001) find that female investors are less overconfident than male investors. This finding in confidence level is also applicable on a director level. It indicates that female directors will be more likely to implement higher

(13)

standards of verification, which in return has a significant positive effect on the informativeness of disclosure (Gul et al, 2011).

Literature also finds that the leadership styles differ between men and women. Cohen et al. (1998) state that the leadership style of women is based on cooperation and trust, whereas the leadership style of men is based on command and competition. Implementing a more cooperative leadership style on the board indicates that there should be better communication and more exchange of information within the board (Klenke, 2003). This might trigger more discussion in the boardroom and in consequence different insights are shared. These arguments indicate that a gender diverse board provides a better information-sharing environment (Gul et al, 2011), which might lead to better information exchange.

The last theory that is mentioned by Terjesen, Sealey & Singh (2009) is the resource dependence theory. The theory is based on the fact that the company is exposed to an open system and need resources to survive. Resources are anything that a company can benefit from. Hence, these resources also lie in the skills of the board of directors, which can differ from their network to the knowledge about a specific industry. The resources are linked to the backgrounds of the directors (Siciliano, 1996). There are different channels where a director can contribute to the company according to the resource dependency theory.

One of the channels is the “advice and counsel”-channel (Hillman & Dalziel, 2003).

This channel specifically focuses on the knowledge of the directors. Some directors obviously know more about a particular industry, product or market than other directors. Such an expertise can contribute to the monitoring role of a director.

A more important channel that contributes to the information environment is the

“channel for communication” (Hillman & Dalziel, 2003). This channel focuses on the way information is communicated to outside investors/organizations. Hence, a bigger network contributes to more information exchange. Ibarra (1992) state that women generally have a more diverse network. However, Lalanne & Seabrights (2011) highlights the difference in social behavior between men and women executives and non-executives with regard to their social network. They find that men and women do not differ in the number of social connections they have but rather that men use their social network more wisely than woman, when it comes to remuneration. The argument they mention is that woman invests in a small

(14)

number of connections but they are indicated as “strong”-relations, whereas men have a lot of connections, but those are indicated as “weak”-relations. These differences in connections between men and woman might also have an influence on the information flow and the returns of the stocks.

2.3 Bid-ask spread

The bid-ask spread is the difference between the amount that the company is willing to sell a particular stock (ask price) for and the amount the investor is willing to pay for a particular stock (bid price). It is also seen as a measure for liquidity, because the bid and the ask price are determined by the demand and the supply for a particular stock. Hence for a very liquid stock, there is a high demand and supply that results in a relatively small bid-ask spread. The opposite holds for an illiquid stock (Glosten & Harris, 1988).

The central entity that manages and links the demand and supply for a particular stock is called a market maker. The market maker registers the bid-ask spread and uses that to offset its risk (Ellis et al., 2000). The bid-ask spread is necessary to prevent arbitrage opportunities. For example, an investor cannot buy a stock and sell it immediately for a much higher price. You will always sell it for the same price, due to the inclusion of the bid-ask spread. However, in the long run this can differ due to market movements and other factors that influence the market.

The bid-ask spread consists of 3 different components, namely the asymmetric information cost, inventory holding cost and order processing cost (Madhaven, 2002).

The most important component of the bid-ask spread is the asymmetric information cost. Asymmetric information cost arises when the one party knows more about the firm than another party (Stoll, 2000). An obvious example is informed trading. The reason that a bid-ask spread arises is because the market maker tries to eliminate the losses it makes by trading with informed parties by making sure it gets a higher gain when it trades with the less informed parties (Gregoriou, Ioannidis & Skerratt, 2005).

The second component is the inventory holding cost. A famous paper by Amihud & Mendelson (2006) stated that a less liquid market have higher inventory holding cost compared to a liquid market. This suggests that the market values continuous trading. Because a dealer has inventory it can buy more stocks for a lower price and sell them for a higher price. However, to be compensated for the fact that it took on risk of holding these stocks, it is compensated by the bid-ask spread (Stoll, 2000).

(15)

The last component is the order processing cost. The paper of Demsetz (1968) was the first paper that researched the order processing cost in detail. He states that there are 2 costs a trader always has to pay; which is the order processing cost and the broker fee. However, the order processing cost are relatively small compared to the other components and arises from the fact that there are multiple processes to undergo before you can actually sell or buy a stock (Stoll, 2000). This results in somewhat inefficient market. An example of such a process is a trader has to go to the market maker to get it’s stock on the market, where the market maker is going to find him an appropriate stock thus it should monitor the market.

2.4 Female board representation and bid-ask spread

The direct relation between female board representation and the asymmetric information has not been studied extensively. There were only two papers that I could find that investigates this (direct) relation but with a focus on Spain and Australia.

The paper of Abad et al (2017) is one of the few studies that investigate the direct link between having female board members and the information asymmetry. This research focuses on companies that are listed on the Spanish stock market for the period of 2004-2009. By using the Generalized Method of Moments (GMM), they find a negative relation between female board representation and asymmetric information. They argue that by having female board representation, there is a better quality of public disclosure and more information exchange. This study uses different measures for information asymmetry, which are bid-ask spread, price impact measure and the PIN1 (Abad et al., 2017).

Furthermore, in this paper the authors argue that there are 2 important channels where female board representation contributes to the asymmetric information. The first channel is associated with the agency theory. The positive side states that more women on the board increase the effectiveness of the monitoring of the board and therefore increases the transparency towards the shareholders, which in return reduces the information asymmetry. However, it might also be that if woman are in the minority they are seen as a tokens and do not increase transparency. The second channel is based on the arguments of psychologists and economist sociologists (Abad et al, 2017). According to various research, it is stated that having a diversified board brings different perspectives to the table and therefore increases the firm’s public disclosure.

1_{PIN stands for probability of informed trading. This is an empirical method, that was introduced by Easley et al. (1996), to estimate} the probability of informed trading.

(16)

For example, different research papers state that women have better communication skills and therefore contribute to better connections with stakeholders, which in returns result in better/more information exchange. However, a counter argument might be that a more diversified board brings more critical vision to the board and might create barriers.

The last article I could find that directly tests the relation between having female board representation and the stock liquidity is the paper of Ahmed & Ali (2017). This paper focuses on the listed companies on the Australian Stock Exchange during the period 2008-2013. By conducting a multivariate analysis, they also find a negative relation between female board representation and market liquidity. They also test if there is a reverse causality between having women on the board and the bid-ask spread. But they found no significant results. Hence they prove that having women on the board increases liquidity and therefore decreases the information asymmetry.

Furthermore, the differences between men and women on the board that was mentioned in section 2.2 also contributes to the effect on the information asymmetry.

Bethke (1987) argued that women engage less in conceals of one’s error. This indicates that women are more straightforward and will provide all the correct information towards other parties. This will lead to less information asymmetry, because the parties will be aware of what is going on in the firm. This can also be translated to; having women will result in less corporate fraud. Uzun, Szewczyk & Varma (2004) looked into the effect of different board characteristic on corporate fraud for U.S. companies. Their main findings were that the structure of the board and the structure of the monitoring committees have a significant effect on fraud incidents. Having more independent directors on the board will lead to a reduction of corporate fraud, which in return will lead to less information asymmetry. They also find that an audit committee with more independent directors will also lead to a reduction of corporate fraud, and in return reduce the probability of corporate fraud. In section 2.2 was already mentioned that Srinidhi, Gul & Tsui (2011) argues that women are more likely to be independent board members, because they are not part of the “old boys club”. Furthermore, Adams & Feirreira (2009) also argued that women are more likely to join a monitoring committee. Combining these arguments will state that having more independent women on the board and in the monitoring committees will lead to a reduction of asymmetric information.

(17)

3. Hypothesis & Methodology

In this section I will introduce the hypotheses and explain what methodology will be used to test the hypotheses.

Besides our main research question, I will go a little further by analyzing some of the channels that contributes to the relation between female board representation and asymmetric information. By doing this, I want to find out what drives the findings of the difference-in-difference results. This translates into 3 hypotheses. For each hypothesis, I designed an empirical model to test the hypothesis.

The first hypothesis is based on our research question. Based on the previous literature analysis that is discussed in part 2.4, I came up with the following hypothesis:

H1: A firm with a more gender diverse board has less information asymmetry compared to a

firm with a less gender diverse board

To test this hypothesis I will use a difference-in-difference model. The reason I choose a difference-in-difference method is because the gender quota introduction in Norway supplied me with an experimental research design. It is an exogenous shock that affected all the listed Norwegian companies. Furthermore, by using a difference-in-difference model, I ensure that the effect of a more gender diverse board on the bid-ask spread is solely caused by the implementation of the gender quota.

I will be comparing U.S public firms (control group) that are listed on the NYSE with the Norwegian public firms (treatment group) that are listed on the Oslo stock exchange. In this case, we can assume that Norwegian firms have a more diversified board compared to U.S. firms because of the gender quota they have to comply with. Second, I will analyze the difference between the period before December 2007 and the period after December 2007. As discussed in the previous section, on December 2007 the board gender quota was implemented in corporate law and a sanction would be given if the firms did not comply. Hence, after December 2007 we are sure that the boards are diversified.

(18)

The difference-in-difference method provides me with the opportunity to interpret a causal relation between the dependent and independent variable by using observational data, if the following criteria are met:

1. The treatment group has similar underlying outcome trends as the control group 2. The treatment only affects the treatment group.

Furthermore, the difference-in-difference model removes potential unobserved effects that can cause biased results. E.g If you run a regression on pre-period and post-period, it might be that there are other factors, besides the treatment, that also influence outcome. By assuming that the control group has similar fundamentals as the treatment group, you control for such an unobserved factors.

This will lead us to the following equation:

Spreadit = α + λ1 Nor_firmsi + λ2 postt +

λ 3 Nor_firmsi*postt + λ4 varianceit + (1)

λ5 log(MV)it + λ6 log(price)it + Λ7 boardsize +εit

Where variable Nor_firms is a dummy variable that indicates a 1 if the firms is listed on the Oslo Stock Exchange and the variable post is also a dummy variable that indicates a 1 if the period is after December 2007. The interaction term Nor_firms x period is the variable of interest.

To complete such a regression, I included different control variables. Stoll (2000) and Glosten & Harris (1988) argues that you should include: the volatility of the returns of a particular stock, the market value of the company and the price of the stock to control for the variation of the bid-ask spread on a cross-sectional level. The rationale for the inclusion is mainly to control for the order processing cost and the inventory cost. An increase in market value of a company will lead to a more liquid market for a particular stock and therefore will match the demand a supply faster (Stoll, 2000). This results into less need for inventory cost and therefore reduces the risk of inventory. The volatility of the return of a stock measures the risk that inventory might increase due to unfavorable price changes (Stoll. 2000). The log of price is included to control

(19)

for the overall risk of a stock, because a low price might be more risky than a high price stock (Stoll, 2000). Furthermore I also included the board size to control for corporate governance.

As discussed in section 2.4, there are different channels through which female board representation can effect the asymmetric information. To complete my research, I will look into these 2 channels and see if these had a significant impact in Norway and if they can explain the findings of hypothesis 1.

The first channel, which Abad, Lucas-Pérez, Minguez-Vera & Yagüe (2017) also touches upon, is the effectiveness of monitoring. If women tend to be more independent than men (Srinidhi, Gul & Tsui, 2011) and tend to join monitoring committees more often (Amas & Ferreira, 2009), this would result in a negative relation for information asymmetry. Therefore the second hypothesis is:

H2: Having women on the board (diversified board) increases the transparency (monitoring)

Following the paper of Peni & Vähämaa (2010), I will use a model for earnings management to test this hypothesis. The reasoning behind it is that if female board members monitor better and reduce information asymmetry, this must result in an increase of quality of earnings reporting. In the first step of this methodology we will estimate the discretionary accruals in reported earnings. I will do that using two models, namely the DD model; which is the accrual model that was suggested by Dechow & Dichev (2002) and the “modified DD” model; which was introduced by McNichols (2002). McNichols (2002) introduced the implied DD model, which combines the theory of Dechow & Dichev (2002) and Jones (1991). Jones (1991) tried to distinguish discretionary accruals and non-discretionary accrual, whereas Dechow & Dichev (2002) looked at total accrual. Jones (1991) included some variables in their model that according to him has an impact on the accruals. On the contrary, Dechow et al. (2002) did not include these variable, as he states that accruals are only based on cash flows. Therefore, McNichols (2002) included the variables of Jones (1991) to the regression of Dechow & Dichev (2002), and it resulted in a significant increase of explanatory power.

(20)

The reason I choose to use the two models is because discretionary accruals depend on the model you are using. Hence, it is good to see what the results would be for both models and see what drives the results. Peni & Vähämaa (2010) state “accruals are adjustments that resolve timing problems in the underlying cash flows at the cost of making assumptions and estimates”. Therefore, to test the quality of the earnings reported I will use the estimated error term of the following regressions:

ACC = α0 + β1 CFi,t-1 + β2 CFi,t + β3 CFi,t+1 + εi,t (2)

ACC = α0 + β1 CFi,t-1 + β2 CFi,t + β3 CFi,t+1 + β4 Δsalesi,t + β5 PPEi,t + εi,t (3)

Where ACC is total current accruals for a particular firm in a particular year, which is calculated by change current asset – change current liabilities – change cash + change debt in current liabilities.

Furthermore CF stands for operating cash flows, which is calculated by net income – total accruals. [Total accruals = total current accruals – depreciation and amortization]. In the second regression we also include the change in sales and the gross value of property, plant and equipment (PPE). After regressing the regressions we estimate the error term, which is going to be the discretionary accruals and be used in the second part of the methodology as our dependent variable.

In the second part, the following regression will be used:

DAj,t = α0 + β1 DIVERSITYi,t + β2 levi,t + β3 Lossi,t + β4 Sgrowthi,t + β5 Sizei,t + εj,t (4)

Where DA is equal to the estimated error term from the first part. DIVERSITY is the proxy that measures diversity of the board, Lev is total liabilities divided by total assets, Loss is a dummy variable if the firm made a loss in that year, Sgrowth is sales growth calculated as the percentage change from one year to another and SIZE is the natural logarithm of total assets.

(21)

The second channel, which Abad, Lucas-Pérez, Minguez-Vera & Yagüe (2017) also touches upon, is the corporate disclosure channel. Corporate disclosure is defined as inside parties2 providing information to the outside parties. This information can range from firm performance to governance topics. According to various research, characteristics of women tend to contribute to this channel. Torchia & Calabro (2016) found a positive relation between independent board members and corporate disclosure and a negative relation between board size and corporate disclosure. As mentioned in part 2.4, women are more likely to be independent, than their male colleagues (Adams & Ferreira, 2009). This results in the second hypothesis:

H3: Having women on the board (diversified board) increases firms’ public disclosure of the

firm.

Following the paper of Dimitripoulos & Asteriou (2010), I will use a simple OLS model to test the relation between earnings per share and the price of the share. This indicates how much information is priced in the share price, which I took as a proxy for disclosure. This will indicate how informative the prices of stocks are. By including an interaction term between the board diversity proxy and the earnings per share, we can find the answer to our question. This results in the following equation:

Pi,t = α0 + β1 EPSi,t + β2 BOARDSIZEi,t +

Β3 DIVERSITYi,t + β4 EPSi,t x DIVERSITYi,t + β5 LEVi,t + β6 SIZEi,t + (5)

Β7 GROWTHi,t + εi,t

where EPS stand for earnings per share, BOARDSIZE represents the total number of board members, DIVERSITY stands for the board diversity measures. LEV indicates the amount of leverage, which is calculated as the total liabilities divided by total assets. SIZE is the natural logarithm of the total assets of a firm and GROWTH indicates the sales growth from one year to another, which is calculated as the change in revenues from t-1 to t.

2_{E.g board of directors}

(22)

Abad et al. (2017) uses different measures as proxies for board gender diversity. These measures are the BLAU-index and the SHANNON-index. I also included these measure in the earnings management analysis and the disclosure analysis.

Blau (1977) introduced these indexes that measure the diversity based on two characteristics in a group. First, it includes the “mixture” of a group; this might be based on different characteristics such a religion, ethnicity or gender or a combination. In our case it is just focused on gender. Furthermore, it also includes a “balance” component, which indicates how the proportion of men and women is divided on the board.

The BLAU-index is calculated as follow:

(5)

where p stands for the number of women divided by the total board members and n stand for the number of different categories. In this case, we only have 2 categories, which are men and women. Therefore, the higher the percentage of women on the board, the more diversified the board is, which results in a higher BLAU index. E.g if 40% of the board is female, than the BLAU-index will be 0.483_{. The BLAU-index ranges from 0 to 0.5.}

The second diversity measure is the SHANNON-index, which is calculated as follow:

(7)

where p is also the number of women on the board divided by the total number of board members and n is the number of categories, which is also two in this case. The SHANNON-index differs from the BLAU-SHANNON-index in 2 ways, namely it obtains higher values and it is more sensitive for changes in the board structure due to the natural logarithm term. Both indexes have a higher value when the board is more diverse. The SHANNON-index ranges from 0 to 0.69. All these variables will be included in the methodology for the second hypothesis to provide an overall analysis of the effect of gender diverse boards on earnings management.

3_{Calculation for the BLAU- index: 1 – (0.4}2_+0.62)

(23)

4. Data

In this section I will explain how the variables are constructed and where I found the data. Followed by what potential effect it might have on the dependent variable.

The database I used to gather the data for the majority of the variables is DataStream. This is an add-in function in excel and provides data for different variables of public companies all over the world. DataStream gives you the option to choose whether you want; daily, monthly, quarterly or yearly data. For all the variables obtained from DataStream I chose the yearly option.

To calculate the bid ask spread, I needed the bid price and the ask price of each company, both on the Oslo Stock Exchange and the NYSE. DataStream provided me with the information. The data was obtained for the period of 2000 – 2014.

First I eliminated all the companies that delisted between 2000 en 2005. Because after a company has delisted, it is not publicly traded anymore, therefore you cannot find data on their bid and ask prices. Furthermore, I also deleted the companies that listed after 2005. The reason for that is that we cannot see what the transformation was from before and after the period that the quota was introduced. This elimination of companies was done for both the Norwegian companies and the US companies.

The data the percentage of woman on the board was also retrieved from DataStream. The name of the variable in DataStream is Value – board structure/board diversity. DataStream provided me with limited observations for this variable; hence I complemented the data by using the annual reports of the companies.

Some of the control variables were also found in DataStream, namely the share price, the market value of the firms and the board size.

There were different proxies for share price, but I choose the “adjusted for default” price. This variable provides the official closing price of the companies stocks, which is adjusted for capital actions and adjusted figures. I expect a positive relation between the relation of share price and bid-ask spread. As already mentioned above, price is included to control for different risks. Hence, when the price is higher, it initiates more risk; therefore there will be a higher bid-ask spread set by the market maker.

Although DataStream does have a variable for market value of a company, I choose to multiply the share price with the common shares outstanding. These variables were also retrieved from

(24)

DataStream. The reason for that is, that the data for the market value was limited compared to the calculation of the market value. This is done for the Norwegian firms and for the US firms. I expect the relation between market value of a firm and their bid-ask spread to be negative. The reason for that, as already mentioned above, is that a higher market value will result in a more liquid market. The inventory cost will decrease, because it is easier to sell and buy stock. The variance of the return variable is calculated by using the prices we obtained from DataStream. However, in this case I downloaded the daily prices and calculated the daily returns by using the formula (P1 - sP0)/P0. After that I calculated the variance in excel for each

year for the period 2000-2014 and for each company. The reason for this method is that I couldn’t find return data for Norwegian companies on CRSP. Although I could find the data for U.S companies, I choose to use the same method I used for the Norwegian companies to calculate the variance of the return.

I expect the relation between the volatility of the returns and the bid-ask spread to be positive. Stoll (2000) argued that this has an effect on the inventory cost, whereby a high volatility of returns increases the risk of having a big inventory.

The governance variable, board size, was also retrieved from DataStream. This indicates the total number of directors that are on the board for each year. No adjustments were needed. Unfortunately, DataStream only provided me with limited data points. Therefore I used the annual reports of the Norwegian and U.S companies to complement the data I found from DataStream.

I expect the relation between board size and bid-ask spread to be positive. This is based on the findings of Cai, Keasley & Short (2006) who find that a smaller board has less information asymmetry, which should results in a lower bid-ask spread. The opposite also holds true, a larger board has more information asymmetry and therefore a higher bid-ask spread.

Additionally to our DataStream database, I also used COMPUSTAT for the earnings management analysis and the disclosure analysis.

For the first stage of the analysis, I retrieved all the variables from COMPUSTAT. This included total current assets, total current liabilities, cash, debt in current liabilities, net income and depreciation and amortization. For the second part, the percentage women on the board

(25)

represent the same variable as we also used in the difference-in-difference analysis. Hence it is retrieved from DataStream and complemented by data from annual reports.

The leverage variable is constructed by dividing total liabilities over total assets. Both of the variables were collected from COMPUSTAT. I expect that the level of leverage of a firm will be positively related to the quality of earnings. Becker, DeFond, Jiambalvo & Subramanyam, K. R. (1998) argues that leverage is related to debt covenants. When a firm is close to violating their debt covenant, they are more likely to adjust their discretionary accruals. Schipper & Smith (1986) argue that an increase in leverage of a company signals that the company is doing well and is able to pay off their debt payments. Therefore, investors are willing to pay more for the stock. Therefore, I expect a positive relation between leverage and share price

Furthermore, the size of the company, measured as the natural logarithm of total assets, is also obtained from the COMPUSTAT database. I expect the relation between the quality of earnings management and the size of the company to be negative. Lobo & Zhou (2001) argues that larger firms report and inform more because of the high demand for disclosure. Additionally, the average cost of disclosure lowers when the firm size increases. The reason for that is; larger firms are monitored more intensively by more of investors. Therefore, I expect the relation between firm size and share price to be positive. For obvious reasons, if the company increases it will be worth more, which results in a higher share price.

Another variable included as a control variable is LOSS. This variable is a dummy variable that equals 1 if the firm made a loss in that year. A loss is measures as a negative net income. This data is provided by the COMPUSTAT database. I expect a positive relation between the LOSS dummy variable and quality of earnings management. DeAngelo et al. (1994) argue that a firm that is in financial trouble is more likely to adjust their earnings; hence the quality will go down. Jaggi & Lee (2002) also found that firms that are in financial distress have a higher incentive to adjust their earnings.

The last control variable for this analysis is the sales growth ratio. This variable is constructed by calculating the change of total revenue from t-1 to t. The data for revenue is obtained from COMPUSTAT. As this variable is a proxy for the growth of a firm, I expect that sales growth and discretionary accruals have a positive relation. Peni & Vähämaa (2010) argue that firms that are in their growth phase are less likely to have a transparent approach and are more eager to

(26)

use a more opportunistic approach. Additionally, I also expect that the sales growth ratio and the share price have a positive relation.

For the last analysis, I also needed earnings per share and board size, which I obtained from COMPUSTAT and annual reports.

(27)

5. Results

In this section I will provide an analysis of the different empirical results. Firstly, I will start of with an analysis of the data in general, where I will use the data descriptions. The second part of this section will focus on the main results that test the 3 hypothesis.

5.1 Data analysis

In table 1 you can find the data description for the difference-in-difference analysis.

The data is split up into 4 groups, namely a group with the period before December 2007 and a group after the period of December 2007 and this is done for the companies for each country.

Table 1: Data descriptive for difference-in-difference analysis

2000-2007 2008-2014

obs mean st.dev min max obs mean st.dev min max

a. Norwegian Companies Bid-Ask spread 519 4.95 13.503 0.1 78.33 1006 3.59 10.976 0.100 78.33 % women-to-board 345 0.25 0.162 0 0.571 757 0.38 0.100 0.125 0.75 Board size 346 7.27 2.466 3 17 801 6.91 2.170 3 17 MV(mln) 674 8.16 32.949 0.125 525.765 960 14.72 94.061 0.125 1496.233 Variance 735 0.00 0.002 0 0.021 1091 0.00 0.003 0 0.042 Price 708 22.21 23.906 1.175 77.865 1032 16.91 17.676 1.155 57.895 b. US Companies Bid-Ask Spread 187 0.07219 0.17713 0.0025 1.697 783 0.06 0.18022 0.003 1.697 % women-to-board 313 0.09 0.100982 0 0.375 741 0.10 0.103 0 0.375 Board size 297 9.69 3.7235 4 18 740 9.2 2.544 4 17 MV(mln) 337 216.66 3537.956 74.12 27008.65 768 316.76 4683.89 40.57 27008.65 Variance 431 33.09 188.0375 0.030076 1994.344 784 56.19 244.029 0.03 1994.344 Price 346 26.68 15.45315 7.84 63.22 784 27.86 18.275 7.84 63.22

In table 1a, we can see that the bid-ask spread of Norwegian companies decreased a little from 4.95 to 3.59. Compared to the US companies (table 1b), the bid-ask spread is significantly higher in both periods. Based on this observation we can conclude that the U.S. market is more liquid than the Norwegian market.

Furthermore, the diversity measures (women-to-board) increased from one period to another for the Norwegian firms. For example, the percentage women on the board increased from an average of 25% to an average of 38%. This diversity measure also increased for the US

(28)

companies, but not as drastically as the Norwegian companies4. Furthermore, I do observe that the minimum threshold of the percentage of women on the board for Norwegian firms during the period 2008-2014 is 12.5%. This indicates that even after the sanctions were introduced, there are still some firms that did not manage to comply with the law. The average percentage of women on the board during the period 2008-2014 is 38%, which is lower than what is stated in the law. But as we just discovered, not all firms managed to comply with the law, therefore this does not come as a surprise.

Another interesting observation is that the boards of U.S. companies are on average larger than that of Norwegian companies in both periods. We do, however, see a (small) decreasing trend from the first period to the second period for both countries.

The market values of the companies that are listed in U.S. are significantly larger than that of Norway and that comes with a higher volatility of returns. Furthermore, there is not really a significant difference between the prices of the Norwegian companies and the US companies when translated to US dollars. Hence we can conclude that the market value is driven by the amount of shares outstanding.

In table 2 we can see the data description for the earnings management analysis. The analysis is done only for Norwegian firms for the entire period (2000-2014). As descripted in the previous section, for this analysis I will be using 2 different accruals model. In table 2, you can see that there is not a significant difference between the DD-accruals and the modified DD-accruals model. Furthermore, the average percentage of women on the board is about 34% during the entire period, with a maximum of 60% of women on the board.

4_{This can also be seen in the graph in the appendix}

(29)

Table 2: Data descriptive for the pooled OLS regression for the discretionary accrual analysis

Obs Mean St. Dev Min Max

DA1 1515 0.084 0.234 -0.435 1.343 DA2 151 0.080 0.231 -0.435 1.423 % women-to-board 901 0.341 0.138 0 0.600 BLAU-index 898 0.411 0.133 0 0.500 SHANNON-index 834 0.310 0.124 0 0.368 Leverage 2916 0.558 0.242 0.035 1.416 Size 2916 6.609 2.059 1.600 12.022 Sales Growth 2456 0.631 2.914 -0.997 23.709 Loss 3729 0.286 0.452 0 1.000

In table 3, you can see the data description for the disclosure analysis. This analysis is also done on only Norwegian firms during the entire period. Most of the control variables are the same as the analysis for earnings management.

The share price fluctuates between 1.21 and 77.89, with an average price of 16.15 dollar. Also, the EPS fluctuates in a large range from -40.79 to 94.78. This indicates that some of the

companies are losing money during some periods. However, the average EPS of the sample is positive at 2.62 dollar per share.

Table 3: Data descriptive for the pooled OLS regression for the disclosure analysis Obs Mean St. Dev Min Max

Price 1378 16.147 15.838 1.205 77.865 EPS 2455 2.268 13.891 -40.790 94.778 Board size 897 7.062 2.216 3.000 17.000 % women-to-board 901 0.341 0.138 0 0.600 BLAU-index 898 0.411 0.133 0 0.500 SHANNON-index 834 0.310 0.124 0 0.368 Leverage 2748 0.558 0.244 0.035 1.488 Sales Growth 2302 0.609 32.819 -0.998 23.297 Size 2748 6.621 3.066 1.674 12.022

(30)

5.2 Regression analysis

To test the relation between female board representation and the bid-ask spread, I conducted a difference-in-difference analysis that is shown in equation 1. These regressions include control variable that control for company characteristics and corporate governance characteristics. In all the columns of table 4, the dependent variable is the bid-ask spread. In model 1, the time span runs from 2000-2014. In this model I found a negative insignificant relation between the interaction term and the bid-ask spread. Looking at a more narrow time span in model 2 and model 3, the significance level does not change. The definition of the coefficient of interest indicates that the bid-ask spread of Norwegian firms decreased in the period 2008-2014 compared to US firms in the period 2000-2007. Only in model 2 the coefficient of the interaction term becomes positive. To further look into this relationship, I included year fixed effects and industry fixed effects or both in the next following model in table 4. The year fixed effects are included to account for different trends in the economy of a country, such as economic growth. Furthermore, the firm fixed effects are included to account for trends within a firm, which might influence the bid-ask spread of that company.

In the 4th column of table 4, I included only year fixed effects and in the 5th column, I included only firm fixed effects. Lastly, column 6 includes both year fixed effects and firm fixed effects. In these models the DiD-coefficient stays negative and insignificant. With these insignificant findings I cannot reject nor accept the null hypothesis. However, I can conclude that the positive coefficient in model 2 has to be some kind of error in the data.

There are various reasons why the results are insignificant. The first reason might be because of the treatment date I used, which was the period when everybody had to comply with the gender quota. In this analysis, we take December 2007 as the threshold between pre-period and post-period, because after December 2007 we can assume that all companies have about 40% women on their board. However, before this period there were already 2 other events5, which indicated that a gender quota would be introduced. Therefore the companies could already prepare themselves for the changes. This concludes that, the period I took as the treatment date cannot directly be seen as an exogenous shock. Some elaboration of this research can be done by using different dates as the treatment date, such as the date when the parliament voted in favor of a

(31)

gender quota. Furthermore, the treatment date falls in the period of the Global Financial Crisis (GFC). During the Global Financial Crisis, all the markets became less liquid, meaning that the asymmetric information component of the bid-ask spread would not have been the larger portion of the spread. Because less liquid markets increases the inventory holding cost which also indicates that the risk of holding a stock increases Hence, this might have also been a significant portion of the bid-ask spread. Therefore, we might have not gotten significant results.

The second reason why there might not be a significant difference Norway and the US, is because the effect on the bid-ask spread does not depend on the gender of a board member, but rather on their independence status. Srinidhi, Gul & Tsui (2011) argues that women have an effect on the bid-ask spread because they tend to be independent. However, in the US it is stated in the law that companies have to indicate how many independent directors they have on the board. Goh, Lee, Ng. Ow Yong (2016) found that more independent directors leads to less information asymmetry. Because in the US it is important to have independent directors, it might be that all the women on the board are independent. Whereas, because it is not stated in the law in Norway, it might be that there are more women on the board but not all are independent directors. Therefore there might not be a significant difference between Norway and US.

Another potential reason why the results are insignificant is that U.S. companies are not a good control group. The reason for that is because the trends might not be the same as in Norway. The US has different laws and regulation than Norway. The argument before (independent directors) is a clear example of a difference between the countries. Hence, maybe for future research it might be an idea to compare another country with Norway, who does have similar laws and regulation except for the gender quota.

The control variables included were all significant expect for the board size variable. The coefficients were all in the same directions as expected, which were explained in the previous chapter, except for the variance variable. I expected the volatility of the returns to be positive, because this would results in a higher risk for inventory cost. However, in this regression analysis it is negative.

(32)

Table 4: The effect of % women-to-board on the bid-ask spread of a company.

This table presents the coefficients from the difference-in-difference regression of equation 1 for the Norwegian and US companies and the period before and after December 2007. The dependent variable for all the models is the bid-ask spread. The regression in column one covers the period of 2000-2014, columns 2 covers the period of 2004-2010 and column 3 covers the period of 2006-2008. Furthermore, in model 4, I included Year FE, model 5 includes only Firm FE and in model 6 both Year FE and Firm FE are included. The independent variable “Norwegian firms” is a dummy variable that indicates 1 if it is a Norwegian firms. The “post” variable is also a dummy variable that indicates 1 if the year is in the period of 2008-2014. Additional also some control variables are included to control for some characteristics of the firms. Furthermore, in the last 3 columns, I included fixed effects. The standard errors, which are adjusted for clustering at company level, are reported in parenthesis.

(1) (2) (3) (4) (5) (6) VARIABLES model 1 2000-2014 model 2 2004-2010 model 3 2006-2008 model 4 2000-2014 model 5 2000-2014 model 6 2000-2014 Norwegian firm -0.0144 -0.172 -0.125 0.141 -0.0144 0.141 (0.295) (0.317) (0.382) (0.298) (0.295) (0.298) Post 0.0918*** 0.167*** 0.192*** -0.0981 0.0918*** -0.0981 (0.0285) (0.0371) (0.0364) (0.298) (0.0285) (0.298)

Post x Norwegian firm -0.104 0.0778 -0.136 -0.134 -0.104 -0.134

(0.103) (0.132) (0.151) (0.125) (0.103) (0.125) Board size -0.0156 -0.0139 -0.0245 -0.0162 -0.0156 -0.0162 (0.0145) (0.0185) (0.0150) (0.0150) (0.0145) (0.0150) Variance -0.00159*** -0.00187*** -0.00224*** -0.00211*** -0.00159*** -0.00211*** (0.000409) (0.000498) (0.000541) (0.000519) (0.000409) (0.000519) Ln(Market Value) -0.168*** -0.182*** -0.175*** -0.149*** -0.168*** -0.149*** (0.0423) (0.0487) (0.0543) (0.0424) (0.0423) (0.0424) Ln(Price) 0.403*** 0.443*** 0.554*** 0.441*** 0.403*** 0.441*** (0.0713) (0.0871) (0.117) (0.0725) (0.0713) (0.0725) Constant 0.0998 0.0300 -0.299 -0.129 0.0998 -0.129 (0.213) (0.290) (0.353) (0.343) (0.213) (0.343) Observations 1,636 748 347 1,636 1,636 1,636 Number of ID R-squared 232 0.0316 182 0.0310 166 0.0144 232 0.0695 232 0.0316 232 0.0695

Year FE NO NO NO YES NO YES

Firm FE NO NO NO NO YES YES

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

(33)

In table 5, the results of the regression of equation 4 is illustrated. In model 1,3 and 5, the results of the DD-model are illustrated and in model 2,4,6 the results of the modified DD-model are illustrated. In all the regression in table 5, I included year fixed effects and industry fixed effects. This is to account for potential macroeconomic trends and industry trends. The effect of having women on the board and the discretionary accruals are positive but

insignificant. In this model, the coefficient of the percentage women on the board indicates that having more women on the board increases the discretionary accruals, hence quality of earnings management decreases. However, because this variable is insignificant I can provide a causal relationship and therefore cannot reject nor accept the hypothesis.

The results of the modified DD-model (model2) also provide us with an insignificant positive coefficient. A reasoning for the positive sign, instead of the negative lies might lie in the fact that women in Norway do not necessarily provide better monitoring skills.

Ahern & Dittmar (2012) found that women on the board of Norwegian companies have less CEO experience and are younger than their male colleagues. Additionally, Tacheva & Huse (2006) conclude that female board members in Norway have a negative effect on the tasks of a board, namely the financial control and service tasks. Combining these findings with the

conclusion of Ahern & Dittmar (2012) we can argue that due to the lack of experience of female directors, they do not contribute to the efficient monitoring process and therefore earnings management does not necessarily change when women are added to the board.

I also ran the same regression with two different diversity measures. The results are provided in model 3-6. The BLAU-index provides the same results as the percentage of women on the board. However, surprisingly the SHANNON-index provides us with a negative but insignificant coefficient. A reason for this finding is that the SHANNON-index is more sensitive for changes in the board structures. This is captured by the natural logarithm in the equation. Hence a small change can already make a huge difference.

Furthermore, only 2 out of the 4 control variables are significant, which are the size of the firm and the loss-variable.

Based on the significance level of the size-variable, I can conclude that a bigger company provides better quality of earnings management. As discussed in the previous section, larger companies are obligated to disclose and inform more towards the outsiders, because they are being monitored more intensively by a larger group of investors (Lobo & Zhou, 2001).

(34)

Table 5: The effect of % women-to-board on discretionary accruals of Norwegian companies.

This table shows the coefficients from the pooled OLS regression of equation 4 for the Norwegian. The dependent variable for all the models is the discretionary accruals, which is calculated as the residuals of the DD-model of modified DD-model. The regression in column 1,3 and 5 uses the discretionary accruals from the DD-model and the regression in columns 2, 4 and 6 uses the discretionary accruals from the modified DD-model. Besides the independent variable, % women-to-board, I also included some control variables to control for firm characteristics. I also regressed the regression on 2 other diversity measure, that are BLAU-index and SHANNON-index. The standard errors, which are adjusted for clustering at company level, are reported in parenthesis.

(1) (2) (3) (4) (5) (6)

VARIABLES model 1 model 2 model 3 model 4 model 5 model 6

leverage 0.0933 0.0703 0.0914 0.0686 0.120 0.0987 (0.173) (0.167) (0.172) (0.167) (0.181) (0.174) Size of firm -0.132*** -0.131*** -0.132*** -0.131*** -0.126*** -0.123*** (0.0449) (0.0418) (0.0449) (0.0419) (0.0473) (0.0438) Loss -0.0920*** -0.100*** -0.0917*** -0.100*** -0.0980*** -0.102*** (0.0185) (0.0172) (0.0185) (0.0172) (0.0204) (0.0192) Sales growth -0.00251 -0.00313 -0.00263 -0.00321 -0.00274 -0.00375

% women on the board

(0.00556) 0.0613 (0.0881) (0.00528) 0.0649 (0.0831) (0.00554) (0.00526) (0.00746) (0.00708) BLAU-index 0.0433 0.0582 (0.101) (0.0990) SHANNON-index -0.0500 -0.0300 (0.193) (0.172) Constant 1.008*** 1.006*** 1.007*** 1.004*** 0.973*** 0.948*** (0.296) (0.279) (0.299) (0.283) (0.338) (0.313) Observations 526 512 526 512 501 488 R-squared 0.244 0.275 0.243 0.275 0.219 0.242 Number of company 77 76 77 76 76 75

Year FE YES YES YES YES YES YES

Industry FE YES YES YES YES YES YES

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

The relation between female board representation and the bid-ask spread : case study for Norway

MSc Finance