1

Regulation and Corporate Board Composition

PhD Dissertation

Siv Staubo

August 30, 2013

Department of Financial Economics

Norwegian Business School, BI

2

Acknowledgements

First of all, I would like to thank Øyvind Bøhren, my supervisor, for his guidance, advice,

patience throughout my time as a PhD student. His input has been invaluable for the

completion of this dissertation. Next, I am grateful to the former and present Head of

Department of Financial Economics, Dag Michalsen and Richard Priestley for making it

possible to combine my work in the department with PhD studies. I thank Øystein Strøm and

Charlotte Østergaard for valuable comments and suggestion on my pre-doc defense. I

appreciate valuable comments from my colleagues at the Norwegian Business School, in

particular I thank, Janis Berzins, Bogdan Stacescu, and Øyvind Norli. Encouraging and

enjoyable chats with Siri Valseth, Kjell Jørgensen, and Andreea Mitrache have been of great

importance throughout the years working with this dissertation. Finally, I want to thank my

wonderful family and my really good friends for their support.

3

Contents Page:

1 Introduction 7 1.1 Does mandatory gender balance work? Changing organizational form to avoid board upheaval. 8 1.2 Female directors and board independence. Evidence from boards with mandatory gender balance. 9 1.3 Determinants of board independence in a free contracting environment. 10

2 Does mandatory gender balance work? Changing organizational form to avoid board upheaval. 11 2.1 Introduction 12

2.2 Predictions 19

2.2.1 Compliance costs 20 2.2.2 Compliance benefits 22 2.2.3 Benefits regardless of the GBL 22

2.3 Data and descriptive statistics 23

2.4 Statistical tests 26

2.4.1 The base case 28 2.4.2 Robustness 29 2.4.3 Entry 31

2.5 Summary and conclusions 33

3 Female directors and board independence. Evidence from boards with mandatory gender balance. 49 3.1 Introduction 50

3.2 Predictions 55

3.2.1 Board independence 56 3.2.2 Regulatory determinants 58 3.2.3 Non-regulatory determinants 59

4

3.3 Data and summary statistics 61 3.4 Empirical methodology and base-case results 65 3.5 Robustness 67

3.5.1 Econometric techniques 67 3.5.2 Board independence 68 3.5.3 Non-regulatory determinants of board independence 69

3.6 Conclusions 71

4 Determinants of board independence in a free contracting environment. 89 4.1 Introduction 90 4.2 Theory and predictions 94 4.2.1 Measures of board independence 96 4.2.2 Determinants of board independence 97 4.3 Data, sample, and summary statistics 100 4.4 Research design, methodology, and estimation results 102 4.5 Robustness 105 4.5.1 Alternative econometric techniques 105 4.5.2 Alternative proxy for board independence 108 4.5.3 Non-linear determinants of board independence 108 4.6 Conclusion 110

5 Summary 132

5

Figures, tables, and appendices

1 Does mandatory gender balance work? Changing organizational form to avoid board upheaval. 37 1.1 Tables

1.1.1 Responding to the gender balance law by choosing organizational form 37 1.1.2 Sample size by listing status, exit behavior, and entry behavior 38 1.1.3 Characteristics of exit firms and non-exit firms 39 1.1.4 Exit propensity in Norway and in neighboring countries 40 1.1.5 The base-case estimates 41 1.1.6 Alternative estimation methods 42 1.1.7 Alternative definitions of family control 43 1.1.8 Definition of exit status 44 1.1.9 The entry decision 45

1.2 Appendices 1.2.1 Regulatory differences between limited liability firms

with alternative organizational forms 46 1.2.2 Regulation of gender balance in corporate boards across the world 47 1.2.3 The empirical variables 48

2. Female directors and board independence. Evidence from boards with mandatory gender balance. 76 2.1 Figures

2.1.1 The fraction of female directors in Norwegian firms exposed to the gender balance law 76

2.2 Tables 2.2.1 The empirical proxies 77 2.2.2 Distributional properties of key variables 78 2.2.3 Characteristics of listed and non-listed firms 79 2.2.4 Director types 80 2.2.5 Board size 81 2.2.6 Multiple directorships 82 2.2.7 Bivariate correlation coefficients between the determinants

of board independence 83

6

2.2.8 Estimates of the base-case model 84 2.2.9 Alternative econometric techniques 85 2.2.10 Alternative proxies for board independence 86 2.2.11 Alternative non-regulatory determinants of board independence 87

2.3 Appendices 2.3.1 Classifying directors as inside, grey or outside: Examples 88

3. Determinants of board independence in a free contracting environment. 117 3.1 Tables

3.1.1 The empirical variables 117 3.1.2 Distributional properties of the variables 118 3.1.3 Firm characteristics by ownership 119 3.1.4 Bivariate correlation coefficients between the determinants

of board independence 121 3.1.5 Estimates of the base-case model – Agency problem 1 122 3.1.6 Estimates of the base-case model – Agency problem 2 123 3.1.7 Estimates of the base-case model in subsamples 124 3.1.8 Alternative estimation methods – Agency problem 1 126 3.1.9 Alternative estimation methods – Agency problem 2 127 3.1.10 Alternative proxy for board independence – Agency problem 1 128 3.1.11 Estimates of the base-case model with non-linear determinants

– Agency problem 1 129

3.2 Appendices 3.2.1 Characteristics of non-listed firms and listed firms 130 3.2.2 Properties of the instrumental variable (IV) 131

7

1. Introduction

This thesis consists of three essays investigating the effects of regulations on board composition. Two regulations affecting the selection of board members have been put into effect during the last decade:

Regulation 1: The Gender Balance Law (GBL). In 2003, the Norwegian government passed a law requiring at least 40% of each gender in the board of directors of ASA firms.

Regulation 2: The Independence Code (IC). In 2006, boards of listed ASA firms were recommended by a corporate governance code from the Oslo Stock Exchange to appoint at least 50% independent directors to their boards.

The GBL is unique to Norway. Although ‘women on boards’ is a hot topic in countries across the world, Norway is the first country to establish a 40% gender quota by law. Firms that do not comply with the law will be liquidated.

The IC is a recommendation, following the principle of comply-or-explain. This recommendation is one of the codes in ‘The Norwegian code of practice for Corporate Governance’. Similar codes exist in most countries across the world. An independent director is neither a member of the firm’s management team nor family-related to members of the management team. A more detailed definition of independent directors is given in the second essay.

This thesis is in the field of corporate governance. The corporate governance structure involves laws, rules, and regulations on the distribution of rights and responsibilities among the different stakeholders in the firm. Due to several financial crises in the late 1990s and early 2000s, there has been an increased interest in the regulation of corporate governance. In particular, the composition of the corporate board has achieved extensive attention.

The first essay in the thesis investigates stockholders’ reactions to the GBL.

The second essay mainly addresses a supposedly unintended consequence of the GBL. Furthermore, this essay explores the link between the GBL and the IC.

Finally, using a sample of firms that are neither exposed to the GBL nor the IC, the third essay explores which firms will benefit and which firms will suffer if they had to comply with the IC.

The motivation for writing three essays on the regulation of board composition is that such regulations come with both costs and benefits. Therefore, regulation might be costly to some firms and beneficial to others. This possibility is analyzed in detail in the three essays. If we assume that stockholders are able to compose an optimal board for their firm, restrictions to board composition might result in non-optimal boards for some firms. This happens if a regulation makes it costly for stockholders to compose a board with the same qualities as the board had before the firm was exposed to the regulation.

8

To better understand why the GBL and the IC might have both costs and benefits, we explain the background of the two regulations.

The GBL, was proposed by a politician who believed that more value for stockholders can be created by increasing diversity in top management and on corporate boards. Other politicians argue that it makes a fairer society when firms include more women in their board room. For example, Prime Minister David Cameron recently stated that: “There is clear evidence that ending Britain’s male-dominated business culture would improve performance, and that Britain’s economic recovery is being held back by a lack of women in the boardroom” (The Guardian 2012).

The IC was first proposed in the United States, and later included in the Sarbane-Oxeley Act (SOX). Prior to SOX, boards in United States, as well as in most other countries, were dominated by insiders who were members of the management team. During the last decade, most countries have developed corporate governance codes that recommend stockholders to appoint a majority of independent directors. Following several financial scandals, policy makers suggested that enhancing the boards’ monitoring role would help prevent further scandals. Independent directors are assumed to be better at monitoring management than dependent directors. To illustrate, the European Commission’s Recommendation from October 6, 2006 states the following: ‘The role of independent non-executive directors features prominently in corporate governance codes. The presence of independent representatives on the board, capable of challenging the decisions of management, is widely considered a means of protecting the interests of stockholders and, where appropriate, other stakeholders.’

We believe that the corporate governance rationale for these opinions need to be further investigated. As far as we can judge, existing research provides no robust support for these opinions. That is, there is neither convincing theory nor convincing empirical evidence showing that more board independence unconditionally improves firm value. The reason is simply that more board independence has both benefits and costs and that the costs may outweigh the benefits. Moreover, this relationship between costs and benefits may vary from firm to firm.

1.1 Does mandatory gender balance work? Changing organizational form to avoid board upheaval

In the first essay, we study stockholders’ reactions to the gender balance law (GBL). We find that, after the GBL ruled that the firm will be liquidated unless at least 40% of each gender is present on the board, half the firms exited to an organizational form that is not exposed to the law. In Norway, as in many other countries, there are two organizational forms for limited liability firms. All firms operating in the most advanced organizational form (ASA) had to change their boards by 2008. The only way to avoid the GBL was to exit to a less advanced organizational form (AS), where gender diversity in the board room is not regulated. It is

9

reasonable to infer that the new regulation is costly to many firms, since the stockholders of half the firms decided to exit the exposed organizational form.

We also show that the costs are firm-specific. Exit is more common when the firm is non-listed, successful, small, young, has powerful owners, no dominating family owner, and few female directors. It is important to notice that listed firms have to delist if they change organizational form. That is, all listed firms have to operate in the most advanced organizational form. Consequently, if the benefits of being listed are greater than the costs of changing the board, the GBL is still costly even though the firm does not exit. Correspondingly, certain unexposed firms may hesitate to become exposed because the expected benefits of operating in the most advanced organizational form are lower than the cost of changing the board.

Overall, we find that mandatory gender balance may produce firms with either inefficient boards or inefficient organizational forms.

1.2 Female directors and board independence. Evidence from boards with mandatory gender balance.

The second essay explores whether gender quotas have other effects on the composition of corporate boards than the implied upward shift in gender diversity. We analyze the impact on board independence of the GBL that requires at least 40 percent of a firm’s directors to be of each gender. We find that the average fraction of independent directors grows by 20 percentage points after the passage of the law. This upwards shift occurs because 84 percent of the female directors are independent, while only 50 percent of the men are.

This large increase in board independence may be costly to some firms because the demand for monitoring by independent directors is firm-specific. That is, optimal board independence requires a trade-off between monitoring by independent directors and advice by dependent directors. This conflict between monitoring and advice suggests that board quality will suffer if forced gender balance pushes the board’s independence level above its optimal level.

We find that demand for an independent board is lowest in small, young, profitable, non-listed firms with few female directors and powerful stockholders. Such firms need monitoring by independent directors the least and advice by dependent directors the most. These firms are hit hardest by excessive board independence, which may be an unintended side effect of mandatory gender balance.

One may wonder whether increased board independence is driven not by the GBL, but rather by the IC, which was introduced in the middle of our sample period. This code is soft law based on the principle of comply-or-explain, recommending that half the firm’s directors be independent. However, the code applies to the listed (public) firms, but not to the non-listed (private). Hence, whereas the GBL imposes the same indirect restriction on board independence regardless of listing status, the IC restricts board independence only in listed

10

firms. We exploit this difference to separate the effects on independence stemming from the two regulations. Roughly, half the firms in the population are listed and implicitly exposed to both the GBL and the code. The other half consists of non-listed firms and is exposed only to the GBL. Therefore, the smaller the difference in growth of board independence between listed and non-listed firms, the higher the likelihood that the regulatory effect on independence is due to the GBL rather than the IC.

Our evidence shows that the impact does not come from the IC, but rather from the GBL. The GBL produces the same upward shift in board independence regardless of the firm’s listing status. That is, because the entire pool of female director talent has so few dependent candidates, one cannot select both many women and many dependent women simultaneously. Therefore, choosing a female director very often means having to choose an independent director, even though that was not the intention. 1.3 Determinants of board independence in a free contracting environment

The essay is the first to explore the demand for monitoring and advice on the board by the owners of firms that are not required by regulation to appoint independent directors. The sample of firms in this study is regulated neither by the GBL nor by the IC. Our focus is on the potential conflict between monitoring and advice and on the idea that the relative value of these two board functions varies across firms.

We explore the board’s monitoring role not just relative to the CEO, but also relative to potential conflicts between large and small stockholders. The first of these two monitoring functions is the only focus in the existing literature. This function of the board is to reduce the principal-agent problem that arises when managers exploit their control rights at the stockholders’ expense. This situation is called the first agency problem in the literature, and directors who are independent of management are supposed to be better at reducing this problem.

The board’s second monitoring function is to oversee the conflict between majority and minority stockholders, which has been called the second agency problem. Directors who are independent of influential stockholders are supposed to be better at protecting the rights of minority stockholders. As far as we are aware, we are the first to study the second monitoring function in a board independence setting.

Our evidence shows that well established, small, and profitable firms with concentrated ownership need advice from dependent directors more than monitoring of their management by independent directors. The analysis shows similar results when we investigate the demand for board independence driven by potential conflicts between large and small stockholders. Unlike earlier research, we find that female directors are just as likely to be advisors as monitors when the firm operates in a free contracting environment regarding gender balance and independence. Our results support the idea that optimal board independence is firm-specific.

11

2.

Does mandatory gender balance work? Changing organizational form to avoid board upheaval

by

Øyvind Bøhrena,b

Siv Stauboa

March 20, 2013

Abstract Norway is the first, and so far only, country to mandate a minimum fraction of female and

male directors on corporate boards. We find that after a new gender balance law surprisingly

stipulated that the firm must be liquidated unless at least 40% of its directors are of each

gender, half the firms exit to an organizational form not exposed to the law. This response

suggests that forced gender balance is costly. These costs are also firm-specific, because exit

is more common when the firm is non-listed, successful, small, young, has powerful owners,

no dominating family owner, and few female directors. These characteristics reflect high costs

of involuntary board restructuring and low costs of abandoning the exposed organizational

form. Correspondingly, certain unexposed firms hesitate to become exposed. Overall, we find

that mandatory gender balance may produce firms with either inefficient organizational forms

or inefficient boards.

Keywords: Corporate governance. Organizational form. Regulation. Boards. Gender quota JEL classification codes: G30. G38

a BI Norwegian Business School, Nydalsveien 37, N0442 Oslo, Norway. b Corresponding author. Telephone: +4746410503. Email address: oyvind.bohren@bi.no.

12

1. Introduction

The choice of organizational form determines regulatory constraints on firms’ governance

system, such as stockholders’ ability to design the board, to separate cash flow rights from

voting rights, and to choose the principles for financial reporting. Therefore, a regulatory shift

may change the optimal way to organize the firm (Hansmann 1996, p. 151). This paper

analyzes how a large and unexpected shift in corporate law, with a liquidation penalty for

non-compliers, influences the firm’s choice of organizational form. In particular, we are the

first to study how a new law for mandatory gender balance in the boardroom induces firms to

exit from or not enter into the organizational form that suddenly becomes exposed to the

stricter regulation.

We find that half the initially exposed firms choose to exit, and that exit propensity is driven

by firm characteristics. This result suggests that the regulation is costly for firms in general,

more costly for some firms than for others, and that even non-exiting firms may end up with

suboptimal boards because the benefit of keeping their exposed organizational form exceeds

the inherent cost of forced gender balance. Correspondingly, our findings for the entry

decision indicate that firms choosing not to enter may keep their optimal board composition,

but fail to obtain their best organizational form. Thus, the observed changes in exit and entry

propensities do not reflect the full corporate costs of mandatory gender balance.

The Norwegian Parliament passed a regulation in 2003 requiring that at least 40% of the

firm’s directors be of each gender. Ahern and Dittmar (2012) argue that this gender balance

law (GBL) represents a massive, surprising shock to the stockholders’ ability to optimally

design their firm’s board. The authors also notice that the GBL represents a natural

experiment that allows the researcher to study the choice of corporate governance

mechanisms with less worry than usual about endogeneity problems (Adams, Hermalin, and

Weisbach 2010). Ahern and Dittmar document the magnitude of the shock by observing that

the average proportion of female directors in listed firms was about 10% when the GBL was

passed. During the next five years until the end of the transition period in 2008, firms

complying with the 40% quota replaced about one third of their male directors by females.

The number of female directorships increased by 260% (from 165 to 592 seats), while the

number of male directorships dropped by 38% (from 1,516 to 938 seats).

13

Our paper identifies firm characteristics that separate firms that chose to comply with the

GBL by making this large board restructuring, from the firms that avoided it by exiting their

current organizational form. We also consider the flip side of the exit decision by analyzing

how the GBL’s passage influenced the tendency of unexposed firms to enter the exposed form.

Table 1 shows how firms in our sample can respond to the GBL by changing or not changing

their current organizational form.

Table 1

Norwegian firms with limited liability can choose between two organizational forms called

ASA and AS, respectively.1 This dual system is dominant worldwide, although exceptions

exist in Canada, the United States, and a few other countries (Lutter 1992). The Norwegian

ASA and AS forms resemble, respectively, A/S and ApS in Denmark, the S.A. and S.A.R.L.

in France, the AG and GmbH in Germany, the AB (publ.) and AB in Sweden, and the Plc. and

Ltd. in the United Kingdom.

The GBL applies to every ASA, but to no any AS. Hence, an ASA may respond to the GBL

by keeping its current organizational form. If it does, the 40% gender quota must be filled.2

This choice response corresponds to the first row of table 1 (Stay). Alternatively, the ASA

may convert into an AS (Exit), which is the response shown in the second row. Unlike for

Stay, the Exit option allows the firm to continue having a board with the preferred gender mix.

AS firms in rows 3 and 4 are not exposed to the GBL. If the AS chooses to become an ASA

(Enter), it must meet the gender quota. Alternatively, the firm remains an AS (Do not enter)

and chooses whatever gender balance owners prefer.

Existing research has focused on firms in the first row of table 1, which are the ASA firms

that choose to remain ASA and hence comply with the GBL. The findings suggest that the

large, forced upwards shift in the demand for female directors by ASA firms made it difficult

to design post-GBL boards with pre-GBL qualities. For instance, 69% of the retained male

directors had CEO experience, compared to 31% of the entering females. The new female

1 ASA (AS) is short for allmennaksjeselskap (aksjeselskap). The dual system was introduced in 1996 to align Norwegian corporate law with legislation in the European Union. 2 The 40% quota applies only to boards with more than nine members. For smaller boards the quota is specified as a minimum number of directors per gender. There must be at least one director of each gender if the board has two or three members, at least two of each if there are four or five members, at least three of each if there are six to eight members, and at least four of each gender if the board has nine members. These thresholds imply that the quota may vary between 33% and 50% in the cross-section of compliers.

14

directors had less board experience and were on average eight years younger than their male

co-directors (Ahern and Dittmar 2012). Thus, most female directors in ASAs post-GBL differ

from their male colleagues in terms of less experience and younger age.

This difference means that although the GBL regulates only gender balance per se, the law

may effectively restrict stockholders’ ability to choose a board with desired qualities. The

reason is that such director qualities may correlate with gender. In particular, the two pools of

potential male and female directors differ considerably along dimensions that may matter for

the board’s ability to create firm value, such as work experience in general and leadership

experience in particular.

This impression of restricted board competence in ASAs after the GBL is supported by Ahern

and Dittmar. They estimate an average announcement return of –3.5% for listed firms with no

female directors when the Minister of Trade and Industry announced his plans to mandate

gender balance. The remaining firms experienced no abnormal announcement return, but

firms with no women on the board represented about three quarters of all listed firms at that

time. This result is consistent with evidence from a period the period 1989–2002, which is

before the GBL was announced. Listed firms would most likely lose value in that period if

they had voluntarily improved their boards’ gender balance (Bøhren and Strøm 2010). The

subsequent value drop when the regulatory intent was announced in 2002 shows that

stockholders did indeed expect a prospective GBL to be likely and costly. Moreover, this

value drop does not appear to be a temporary overreaction. Typically, firms that had to change

their boards the most experienced an abnormal 15% drop in their market-to-book ratio during

the five subsequent years.

The four alternative responses to the GBL, shown in table 1, suggest that the cost of gender

balance may differ across firms. First, the compliance costs may vary among ASA firms that

choose to keep their organizational form (Stay). For instance, the reported announcement

returns support the notion that boards with more female directors must sacrifice less board

competence to reach the 40% quota. Second, firms converting from ASA to AS (Exit) may

experience different exit costs depending on the firm’s listing status. This is because the GBL

applies to all ASAs regardless of whether they are listed (public) or non-listed (private).

However, only listed firms are required to be an ASA. Therefore, exit to avoid the quota

automatically triggers delisting for a listed ASA, but obviously not for a non-listed firm. Third,

15

because the GBL changes the benefit of having the exposed organizational form, the law may

not influence just the exit decision, but also may influence when an unexposed firm chooses

to become exposed (Enter or Do not enter).

To improve the understanding of how this one-size-fits-all regulation of gender balance has

heterogeneous effect across firms, we study how the GBL affects the choice of organizational

form of all exposed (ASA) and unexposed (AS) firms during nine years. This approach

provides new insight for four reasons. First, following the firms’ behavior during an extended

time turns out to be important. For instance, we find that among the ASAs that existed when

the GBL was passed in 2003 and that did not subsequently merge or go bankrupt, 51% had

chosen to exit into AS by the time the law became binding five years later. Second, including

non-listed firms is essential not just for a priori reasons, but also because the exit propensity

turns out to be much higher for non-listed firms than for listed firms.

Third, we find that the tendency to enter the ASA form does not constitute a mere mirror

image of the tendency to exit the ASA form. Thus, studying both exit and entry deepens

insight into the regulatory effect. Finally, the two existing studies on valuation effects of the

GBL report conflicting results. Ahern and Dittmar (2012) find negative valuation effects,

while Nygaard (2011) finds positive effects using a different event date and a different sample.

We avoid these ambiguities by analyzing how firms respond to the regulatory shift by

changing organizational form. This direct evidence on altered firm behavior may improve the

understanding of what forced gender balance does to different firms and why announcement

returns may vary across firms. The key is to identify how certain characteristics enable the

firm to influence the cost of the regulatory shock by either keeping or changing its

organizational form.

The GBL was implemented on January 1, 2006 with a two-year grace period.3 Among the 309

ASAs in 2002 that did not subsequently merge, fail, or exit for other reasons unrelated to the

GBL, we find that 151 firms existed in 2008. This exit behavior represents a drop of 51%.

These exiting firms chose the unexposed AS form. As shown by Appendix 1, the financial

reporting and the corporate governance mechanisms are less tightly regulated for AS than for 3 The law as passed in 2003 would have been withdrawn if the firms had voluntarily filled the gender quota by July 1, 2005. Because that did not happen, the GBL became mandatory in 2008. All firms had complied by April 2008, including the 72 firms that violated the January 2008 deadline (Nygaard 2011). The regulation states that the firm will be liquidated three months after non-compliance, although the government may abstain from liquidation if the firm is considered particularly important for society. No firm has been liquidated for non-compliance so far. A likely reason is that the alternative to fill the quota as an exposed ASA is exit into the unexposed AS.

16

ASA. For instance, an ASA must have ten times larger minimum share capital, produce

financial reports containing greater detail, and provide compensation data containing greater

specificity about its officers and directors. Unlike for most AS firms, CEO-chair duality is

illegal in ASAs, and not more than half their share capital can be non-voting.

In all, 42% of the ASAs were non-listed by year-end in 2009. Appendix 1 shows that there is

less discretion in the design of corporate governance mechanisms when the ASA is listed. For

instance, only listed ASAs are subject to comply-or-explain governance codes, flagging

requirements, and tender offer rules.

We find that unlike before the GBL and unlike in neighboring countries where firms were not

exposed to gender balance regulation, exit is much more common than entry. However, this

exit from ASA to AS primarily happens among the non-listed firms. For instance, the number

of listed ASAs in our sample increases by 11% from 2002 to 2008, while the number of non-

listed ASAs decreases by 49%. Thus, listed firms, which cannot remain listed unless they

keep the ASA form, exit much less often. Also, and unlike before the GBL, the propensity for

an AS to enter the ASA form and hence become exposed to the GBL is higher if the firm

immediately goes public upon ASA entry rather than staying private. These findings support

the argument that the GBL more often induces a change of organizational form when the

firm’s listing benefits are low.

This evidence shows that a study of exits by listed firms only would miss most of the

interesting cases. Moreover, because the change in the number of ASAs is the net of exits and

entries, both exit from ASA to AS and entry from AS to ASA must be addressed.

Regardless of listing status, we find that most firms converting are those that perform well

and have powerful owners. This finding supports the idea that independently of the GBL,

profitable firms with low agency costs benefit the least from the strictest regulatory standards

for transparency and governance. Exit is also more common among non-family firms. This

may indicate that family owners are better able than others to radically change the board’s

gender balance. Moreover, most firms that exit have few female directors, suggesting that

regulatory costs are higher the more the board must be restructured. Finally, most firms that

convert are small or young firms. This result may reflect that the compliance cost is fixed

relative to firm size, and that the cost of changing organizational form grows as the firm

matures.

17

Most of these relationships are supported by the evidence for firms converting from AS to

ASA. The exception is that unlike for exiting firms, the fraction of female directors is not a

significant predictor of conversion from AS to ASA. This result may be driven by the fact that

whereas an ASA must either meet the mandatory gender quota or exit to AS, an AS faces no

such pressure. An AS enters only if it expects the benefits will exceed the compliance costs.

Radically changing the gender mix is apparently not an important compliance-cost driver for

AS firms that voluntarily choose to enter. A possible reason is that they have had sufficient

time to ensure easy access to the pool of qualified female directors.

Our results are robust to alternative econometric techniques, to the definition of family control,

and to how we measure performance. The definition of exit matters, however. The fraction of

female directors is a strong determinant of exit if the firm is classified as an exit firm also in

the years before it actually exits. The relationship is considerably weaker if the firm is

classified as an exit firm only in the actual exit year. This result may reflect the empirical fact

that after the GBL was passed, gender balance gradually increased also in firms that

ultimately exited. Thus, ignoring the years before the ASA actually exits misses the general

trend towards more gender balance in all ASAs before 2008. In particular, the approach

misses the cost this increasing trend imposes on firms that gradually improve their gender

balance, but ultimately decide to exit.

These findings do not imply that the GBL is more costly for firms that exit than for those that

stay. The reason is that the non-exiting firms may find the cost of changing organizational

form to be even higher than the cost of complying with the GBL. Thus, abandoning the more

strongly regulated ASA form may be more burdensome than being forced to radically change

the board’s gender balance. This happens particularly often to the listed firms in our sample,

because exit implies losing the listing benefit. Correspondingly, AS firms that choose not to

enter the ASA form may still incur a GBL-related cost. This is because these firms will not

enter whenever the cost of complying with the GBL exceeds the ASA benefits that are

independent of the GBL. Examples of such benefits are easier access to financing and

stronger legal protection of minority stockholder rights.

Our paper is related to the empirical literature on the economics of corporate governance

regulation. Bushee and Leuz (2004) study the effect of stricter SEC disclosure requirements

for firms trading on the OTC Bulletin Board. They find that almost 75% of the firms either go

18

private or exit to the pink sheets market, which is not exposed to the new regulation. The exit

propensity is strongest for small, profitable firms with low leverage. Engel, Hayes, and Wang

(2004) analyze corresponding effects of the 2002 Sarbanes-Oxley Act, finding a slightly

increased tendency to go private. Small firms with high ownership concentration go private

more often than others. A study of 17 European countries shows that firms go private more

often when corporate governance codes are introduced and when minority protection is

increased. Exit is more common among small and profitable firms (Thomsen and Vinten

2007). These results are generally consistent with ours.

Ahern and Dittmar (2012) briefly address exits after the GBL in their valuation study, but

consider only listed firms. Because ownership characteristics are not included in their study,

Ahern and Dittmar ignore agency costs as a determinant of exit. This bias also applies to the

valuation study of Nygaard (2011), who makes a robustness test of whether the firm’s listing

status influences the relationship between exit and the fraction of female directors. Moreover,

both approaches are biased towards finding excessive exit because they include financial

firms. Such firms were allowed to convert from ASA to AS one year before the gender quota

became mandatory. Finally, the entry decision is not addressed.

Although Norway was the first country to regulate gender balance in the private sector,

mandating gender quotas for corporate boards is currently a hot political topic internationally.

The obvious reason is that corporate boards are strongly dominated by men. The highest

fraction of female directors in listed firms outside Norway is 27% (Sweden), and 70% of the

43 countries included in a recent survey have fewer than 10% of their board positions filled

by women (www.catalyst.org). France, Iceland, Netherlands, and Spain will implement quotas

in 2013–2016. Proposals along the same lines have recently been made in Australia, Belgium,

Canada, Italy, and the EU Commission. Gender balance rules for state-owned firms have

already been launched in Ireland, South Africa, and Switzerland.

Some of these countries consider whether to use mandatory law like in Norway or the softer

comply-or-explain system, which is the common standard in national corporate governance

codes among more than 100 countries worldwide (www.ecgi.org). Appendix 2 shows the

details. So far, however, only Norway has experience with gender quotas in the private sector,

and no other country has chosen a mandatory system with liquidation penalty. Hence, our

findings from the first country to adopt a radically new regulation on board diversity may

19

contribute to a more informed choice elsewhere. In particular, we can document effects from

a regulatory regime that is mandatory rather than voluntary, dictates the same gender balance

in all boards rather than allows for firm-specific discretion, ensures full compliance by a

liquidation penalty, and applies to listed firms as well as to some non-listed firms rather than

to all firms or just to listed firms.

The rest of the paper is organized as follows: Section 2 specifies our predictions, and section 3

presents the data and summary statistics. We explain the methodology and test the predictions

in section 4, and we summarize and conclude in section 5.

2. Predictions

The firm should transform itself from the ASA (exposed) to the AS (unexposed)

organizational form when the firm benefits from doing so. This benefit, B(Exit), has three

components:4

(1) B (Exit) = Compliance costs – Compliance benefits – Benefits regardless of the GBL

Exit is optimal if B(Exit) is positive. If negative, the best choice is to continue being an ASA

in compliance with the GBL. Section 4 deals with the entry decision, where the benefit of

entry is the negative of (1). Either way, changing organizational form requires a two-thirds

majority vote at the stockholder meeting.

If there are no market imperfections such as irrational owners or conflicts of interest between

owners and managers, the compliance benefits in the second term of (1) are zero. The only

effect of the GBL in such a case is to add a new constraint to the owners’ value maximization

problem by ruling out in an ASA any board design involving fewer than 40% of the positions

going to each gender. This added restriction will at best leave the owners’ opportunity set

unchanged.

Consequently, the GBL must be rationalized economically by its ability to reduce negative

effects on stockholder wealth of market imperfections. In the absence of such benefits, the

new regulation produces only compliance costs for owners as reflected in the first term of (1).

Consistent with this view, Ahern and Dittmar (2012) argue that lack of leadership experience 4 Similar logic is used by Engel, Hayes, and Wang (2004).

20

among female directors may be a major driver of compliance costs and the resulting loss of

firm value.

The third term in (1) explains why an ASA with low compliance benefits and high

compliance costs may still decide not to exit. This happens when the net compliance cost of

the GBL (the first two terms) is smaller than the benefit of being an ASA for reasons

unrelated to the GBL (the third term). These latter reasons are independent of board

composition and were also valid before the GBL. Examples are the benefit of having a liquid

stock if the ASA is listed and of having the option to go public without first changing

organizational form if the ASA is non-listed. Regardless of listing status, any ASA may also

benefit from regulation ensuring more transparency and stronger protection of minority

stockholders.

The composite nature of the exit benefit in (1) has two immediate implications. First, cost

measures based on the exiting firms alone will underestimate the full cost of the GBL. This

happens because regulatory costs based on just exiting firms ignore the costs for firms that

stay. The latter costs are particularly relevant for the listed firms in our sample, because they

cannot exit without simultaneously delisting. Second, firms more likely to exit are not just

those with high compliance costs and low compliance benefits. Exit to AS is also optimal for

firms with low benefits from being an ASA in the first place. For such firms, the third

component in (1) is too small to overcome even a moderate cost of the GBL.

We next hypothesize how firm characteristics will influence the three components of B(Exit)

in (1) and hence the likelihood of switching from the ASA to the AS organizational form after

the GBL.

2.1 Compliance costs

The costs of complying with the GBL consist of search costs for new directors, increased

compensation costs for these directors once hired, and reduced private benefits for owners,

who lose control because the board is restructured.

If the owners have chosen the optimal board composition before the GBL, forced board

changes, and hence the compliance costs, will be higher the fewer women the board has. This

logic is supported by Ahern and Dittmar (2012), who find that on average, only firms with no

21

female directors lose market value at the GBL announcement. We predict that the lower the

fraction of female directors, the higher the propensity to exit (H1).

If earlier top management experience matters for director quality, the findings by Ahern and

Dittmar (2012) that women have less such experience than men do imply that new qualified

directors must be drawn from a smaller pool than earlier. Thus, the GBL will increase both

search costs and compensation costs. Because these increased costs seem rather independent

of firm size, however, compliance costs will more often produce a positive exit benefit when

the firm is small. We expect that the smaller the firm, the higher the propensity to exit (H2).

Family firms often have members of the controlling family in board and CEO positions

(Anderson and Reeb 2003). To illustrate, family-controlled firms in our sample have a median

ownership concentration of 50%, a family chairperson in 38% of these cases, and a family

CEO in 30%. In contrast, non-family firms have a median ownership concentration of 26%,

and the largest owner is chairperson or CEO in 17% of the cases. The GBL may therefore

threaten the family’s ability to extract private benefits in the ASA whenever the gender mix

among the family’s director candidates does not match the mandated gender quota. Such

concerns for family-internal recruiting to the board suggest that ASAs controlled by families

convert more often to the AS form than other firms after the GBL.

Nevertheless, this concern for family-internal recruiting may not tell the full story about the

family firm’s compliance costs. The high ownership concentration and the family’s

governance involvement during extended periods suggest that family firms often have

particularly powerful and committed owners. The long and deep experience with the firm and

its environment may have enabled the family to establish a rich network with resourceful

individuals outside the firm. Therefore, the controlling owner may know the outside pool of

potential female directors particularly well. This argument suggests that unlike for the

family’s ability to recruit female directors from inside the family, it may be relatively easy to

fill the gender quota by recruiting from outside the family. Hence, compared to other firms,

family-controlled firms may exit the exposed organizational form less often rather than more

often. We define a family-controlled firm as one where ultimate owners by blood or marriage

hold more than half the equity. The two conflicting arguments imply that the expected

relationship between family control and exit propensity is unspecified (H3).

22

2.2 Compliance benefits

The hypotheses discussed so far assume that owners always know their best interest,

including the ability to establish an optimally designed board before the GBL. Allowing for

imperfections in terms of gradual learning, however, firms may need time to locate the pool of

director candidates and pick the best team. Such a limited ability to choose the optimal board

may be particularly relevant for gender mix, because boards and recruiting committees were

strongly dominated by men before the GBL (Rosener 2011). Hence, older firms with a long

learning history pre-GBL may have been closer to their value-maximizing gender balance

than were younger firms with a shorter history.

This logic suggests that older firms will be hurt by a rule mandating the same gender mix for

every firm, while younger firms may benefit from being forced to establish a more gender-

balanced board. On the other hand, older firms may find it harder to change organizational

form because they are more complex and rigid (Boone et al. 2007). This argument suggests

older firms are less rather than more prone to exit. Thus, the expected relationship between

firm age and exit propensity is unspecified (H4).

2.3 Benefits regardless of the GBL

The ASA firms in our sample are subject to tighter reporting requirements than the AS firms

are. This higher transparency of ASAs reduces the asymmetric information between old and

new stockholders, between majority and minority stockholders, and between borrowers and

lenders. Thus, being organized under the most demanding organizational form may reduce the

cost of raising outside finance. This option is more valuable the more financially constrained

the firm (Myers and Majluf 1984). Using leverage to proxy for financial constraints, we

predict that the weaker the financial constraint, the higher the propensity to exit (H5).

For similar reasons, profitable firms may suffer less after exit because they can more easily

finance growth internally. Measuring profitability as operating returns to assets after taxes

(ROA), we expect that the more profitable the firm, the higher the propensity to exit (H6).

The higher transparency of ASAs than ASs because of regulatory differences may induce less

costly and more intense monitoring by financiers, analysts, and the media. The resulting lower

information asymmetry in ASA firms is more valuable the higher the potential agency costs,

that is, the weaker the owners’ incentives and power to monitor management (Morck, Shleifer,

and Vishny 1989). Hence, an ASA with low potential agency costs has fewer governance

23

benefits from being an ASA. Moreover, these low agency costs increase the likelihood that

the firm will rationally choose organizational form according to the value-maximizing exit

criterion in (1). We relate agency costs to ownership concentration, which we measure as the

fraction of outstanding equity held by the firm’s largest ultimate owner. We hypothesize that

the higher the ownership concentration, the higher the propensity to exit (H7).

Unlike a listed ASA, a non-listed ASA does not change listing status when exiting to AS.

Thus, owners of listed firms have more to lose by not having their stock traded in a liquid

market (Bahrat and Dittmar 2006). Listed firms also have a much wider stockholder base,

which makes them more vulnerable to free-rider and coordination problems when concerted

action would benefit all stockholders as a group (Shleifer and Vishny 1986). For instance,

Norwegian listed and non-listed ASAs of similar size have on average roughly 4,000 and 10

stockholders, respectively (Bøhren 2011). We expect that non-listed firms will exit more often

than listed firms do (H8).

Summarizing predictions H1–H8, we hypothesize that a firm with the ASA organizational

form, which is exposed to the GBL, will exit more often to the unexposed AS form when the

firm has low leverage, high profitability, high ownership concentration, small size, few

female directors, and when the firm is non-listed. The expected effects on exit of firm age and

family control are left unspecified.

3. Data and descriptive statistics

The official initiatives to regulate gender balance in corporate boards were made in 1999 and

once more in 2001 through public hearings about possible overhaul of the Equal

Opportunities Act from 1978. The first public announcement of the planned regulation was

made in February 2002. The regulation was passed as corporate law by Parliament in

December 2003 and once more in June 2005, with the added provision of a liquidation

penalty for non-compliers. The transition period from the old to the new regime ended at

year-end 2007, although 77 firms were allowed to postpone compliance until the end of

February 2008.

To allow for approximately two non-event years at the beginning and end of these regulatory

events, our sample period is 2000–2009. The sample for the analysis of exits is based on the

24

population of ASA firms by year-end.5 We ignore firms that exit due to merger or bankruptcy.

Financial firms are also excluded because they had to choose the exposed form until a new

law lifted this requirement in 2007.6 Because both merging firms and financial firms may also

have left the ASA form partly because of the GBL, this sample restriction biases our tests

towards accepting the null hypothesis that the GBL has no effect on the choice of

organizational form.

Table 2 shows the number of sample firms by year-end during the sample period. The total

number of sample firms in panel A (All) represents 264 observations per year on average,

which is 53% of the population.7 This large difference between their population and the

sample suggests that our filters are important for eliminating firms that have probably not

exited because of the GBL.

The number of ASAs is largest in 2001, monotonically decreasing thereafter to a minimum in

2009. Although not shown in the table, the peak in 2001 becomes more obvious if we also

include every year from when the dual system of ASA and AS was established in 1996. The

number of ASA firms starts at 177 in 1996 and grows every year until 2001.

Table 2

Panel A also documents that the decline after 2001 only happens in the sub-sample of non-

listed firms. The number of non-listed firms drops by 56% from 2001 to 2009, while the

number of listed firms grows by 6%. This large difference suggests that if the underlying exit

and entry decisions are partially driven by the GBL’s introduction, the benefit of changing

organizational form to avoid the GBL is considerably larger for non-listed firms.

The change in the number of firms from one year to the next in panel A reflects the difference

between entering firms (from AS to ASA) and exiting firms (from ASA to AS) during the

year. Panel B shows the exits, entries, and net exits. As already documented by panel A, net

exit (exit minus entry) is generally positive and increasing. There were altogether 217 exit 5 Our data set is organized by the Centre for Corporate Governance Research (www.bi.edu/ccgr). The data on family relationships are delivered by the tax authorities (www.skatteetaten.no), while Experian (www.experian.no) has delivered the accounting data and the corporate governance data. 6 Financials are also regulated differently than other firms regarding capital structure and corporate governance. For instance, the risk-adjusted leverage of banks cannot exceed 92% according to the Basel regulation, and Norwegian banking law stipulates that no investor can own more than 10% of a bank’s equity without the government’s permission. 7 The population of ASA firms averages 482 firms per year. Excluding financial ASAs reduces this number to 340, dropping further to 264 when we also exclude ASAs that go bankrupt or become AS because of a merger.

25

firms and 146 entry firms, producing a net exit of 71 firms during the sample period. Panel C

confirms that this tendency to exit is much stronger among non-listed firms. For instance,

while panel A shows that the number of firm years is rather independent of listing status, non-

listed firms account for over four times more of the exits (175 vs. 42, respectively). In all,

12.1% of the non-listed firms exit yearly on average, while only 3.3% of the listed firms do.

Panel D shows that when an AS decides to become ASA and hence becomes exposed to the

GBL, the firm more often enters as non-listed (no IPO) than listed (IPO), where the latter

means going public in the entry year (92 firms vs. 54 firms, respectively). However, this

tendency is reversed after the passage of the GBL, when it becomes much more common to

go public directly. Whereas 21% of the entry firms go public directly up to 2003, 54% do so

subsequently. This shift in IPO propensity by entry firms suggests that although compliance

with the GBL may produce similar costs and benefits regardless of listing status, listed firms

earn more benefits that are unrelated to the GBL. Therefore, firms considering to become

exposed after the GBL increasingly find that entry does not pay off unless the firm quickly

reaps the listing benefits as well. We focus on the exits in the following, leaving discussion of

the entry decision until section 4.

The empirical variables are defined in Appendix 3. Unreported summary statistics for these

variables show that non-listed firms account for 55% of our observations, the largest

stockholder owns on average 43% of the equity, the average board has 17% of its positions

filled by female directors and 5.6 members, while 20% of the firms are majority controlled by

a family.

A firm is classified as exiting if it transforms from ASA to AS during the sample period. An

exiting firm leaves the sample the year it actually exits. The firm is called non-exiting if it

never abandons the ASA form. Table 3 shows that compared to ASA firms that stay exposed

to the GBL, the ASAs that convert to AS and hence become unexposed are different

according to most of the hypothesized determinants. For instance, exiting firms have

significantly fewer women on the board (10% vs. 20%), are younger (20 years vs. 29 years),

have higher ownership concentration (53% vs. 38%), and are more often non-listed (78% vs.

42%). Compared to our hypotheses from section 2, these univariate relationships are

consistent with H1, H4, H7, and H8, respectively.

Table 3

26

Unreported tests show that when we split the sample based on listing status rather than exit

status, a similar pattern emerges as for exit vs. non-exit firms. This similarity suggests that

whether or not the firm is listed correlates both with the exit/non-exit choice and with other

determinants of exit beyond listing status. Thus, not controlling for listing status may create a

serious omitted variable bias in a regression where exit is the dependent variable. That does

not imply, however, that any other hypothesized determinant than listing status is redundant in

an exit model. This argument is supported by the bivariate correlation coefficients, which

show that listing status does not correlate alarmingly with any other determinant.8

4. Statistical tests Table 2 shows that the number of firms exposed to the GBL has been dropping every year

since 2002, when the intention to regulate gender balance in corporate boards was announced.

Although table 2 ignores all ASA firms that become AS because of merger, bankruptcy, or

regulatory change for financials, our filtering criteria might still have failed to exclude other

exogenous exit determinants that are unrelated to the GBL. To account for this possibility, we

compare the exit propensity for ASA firms in Norway with the exit propensity from the

similar organizational form in the neighboring countries Denmark and Sweden. These two

countries do not mandate gender-balanced boards, but they have the same system of dual

organizational forms as Norway has.

We use a difference-in-difference approach to test whether the tendency to change

organizational form by Norwegian firms differs from what it is in the neighboring countries.

The event is the passage of the GBL in 2003, the event group is the Norwegian ASA firms,

and the alternative non-event groups are the firms in Denmark, Sweden, or both, with similar

organizational form as the firms in the event group. The post-event period is 2003–2009,

while the pre-event period is 1996–2002. Hence, all firms in our sample have the option to

change organizational form any time during the sample period. Only Norwegian firms may

consider exiting to avoid the GBL, however. Moreover, this can be a valid reason only in the

post-event period.

8 The listed/non-listed dummy correlates the strongest with ownership concentration. The Pearson correlation coefficient is -0.42, which is nevertheless well below the rule-of-thumb critical limit of +/- 0.8 (Greene 2007).

27

The statistic of interest is the difference-in-difference statistic D ≡ ΔNorway – ΔForeign, where

ΔNorway is the difference between the number of Norwegian firms in the post-event period and

the pre-event period, respectively. Correspondingly, ΔForeign is the difference between the

number of firms in the two periods in the foreign country (Denmark, Sweden, or both). We

estimate D by the model

(2) 0 1 2 3 ,= + + + ⋅ +it i t i t ity EG PE EG PEβ β β β ε

where yit is the number of firms in group i at time t. EGi is a dummy variable which is 1 if the

firm is in the event group and 0 if the firm is in the non-event group. Similarly, PEt is 1 if t is

in the post-event period and 0 if t is in the pre-event period.

The estimator of D is the OLS estimate of β3 in (2). This coefficient reflects the effect on the

number of firms if the observation is from Norway (rather than a neighboring country) in the

event period (rather than in the non-event period).

The number of firms in the three countries is shown in panel A of table 4. Two major factors

explain why the number of firms in Denmark is much higher than in the two other countries.

First, Norway and Sweden wrote their laws for ASA type of firms around 1995 based

primarily on an EU directive. Denmark, however, wrote its ASA law twenty years earlier

primarily based on its existing AS law. Second, both Danish laws are less restrictive than their

Norwegian and Swedish counterparts are (Gomard and Schaumburg-Müller 2011). For these

reasons, we will estimate (2) using alternatively Denmark, Sweden, and both as the non-event

group.

Table 4

Panel B shows the estimate of β3, which is negative and significant in every case. This result

reflects that the drop in the number of firms with organizational form exposed to the GBL as

observed in tables 1 and 2 is a unique Norwegian phenomenon.

The findings in table 4 strengthen the impression of an inverse relationship between the

introduction of the GBL and the choice of ASA as organizational form. We analyze this link

more closely in the following by relating the exit and entry behavior to firm characteristics as

specified in section 2. We first report the findings from the base case, followed by a series of

robustness tests. Finally, we present estimates of an entry model.

28

4.1 The base case Our base-case model is the following:

(3) 1 2 3 4

5 6 7 8

it it it it it

it it it it it

Exit Female directors Firm size Family control Firm ageFinancial constraints Performance Ownership concentration Listed u

α β β β ββ β β β

= + + + +

+ + + + +

Exitit is a dummy variable that equals 1 if firm i leaves the exposed form (ASA) during the

sample period and 0 otherwise. We estimate (3) as a logit model, using the GLM and a sample

from 2000–2009.

Table 5 shows the estimates. Consistent with our prediction based on the compliance cost

component of B(Exit) in (1), the table documents that firms with few female directors exit

more often from the exposed to the unexposed organizational form (H1). This inverse

relationship suggests that the costs of complying with the GBL are higher the more the board

must be restructured in general, and the more that male directors must be replaced by females

in particular. The finding is also in line with Ahern and Dittmar (2012), who report that firms

with no female directors lost value when plans for a GBL were announced.

Table 5

Smaller firms exit more often than other firms do. This finding supports the economies of

scale argument that compliance costs are fixed relative to firm size, such as the cost of

searching for new female directors and having to pay them higher compensation because of

short supply (H2). Moreover, firms controlled by families exit less often than other firms do.

This result is inconsistent with the logic that the family’s ability to extract private benefits is

threatened by a GBL that mandates a gender-based board composition the family cannot

match. Rather, the finding supports the argument that family owners have better access than

others to female directors who can protect the owners’ interests (H3).

Turning next to potential compliance benefits, the estimates show that younger firms exit

more often than older firms do. This result is inconsistent with the ignorant-owner argument,

but supports the notion that mature firms find it more costly to change organizational form

(H4).

The third component of B(Exit) in (1) is benefits of having the exposed organizational form

that are independent of the GBL. Table 5 shows that unlike what we predicted, the exit

decision is not significantly related to financial constraints as measured by leverage (H5). A

29

more profitable firm is more willing to leave the exposed form, however, possibly because it

can more easily self-finance investments by high earnings and can afford the higher financing

costs as the firm becomes less transparent (H6). Exit is also more common when ownership

concentration is high. This is evidence that strong owners can be a substitute for the

disciplining effect of a stricter regulatory regime (H7).

Finally, non-listed firms are more prone to exit. This result supports with the notion that listed

firms have more to lose by exiting for reasons unrelated to the GBL, such as better stock

liquidity, continuous pricing of their stock, and closer following by financial analysts and the

media (H8).

4.2 Robustness Table 6 estimates the base-case model (1) with five alternative econometric techniques. The

table documents that the results are insensitive to whether we use logit (the base case), probit,

a standard panel method with random effects, a logit panel method with random effects,

pooled OLS, or pooled OLS with standard errors adjusted for clustering at the firm level.

Thus, the choice of econometric technique is not driving the base-case results.

Table 6

Family control may be operationalized in several ways. Table 7 shows what happens when we

measure family control by other proxies than the one used in table 5, which is whether the

family holds a majority ownership stake (family firm). We alternatively measure family

control by the fraction of board seats held by the family (family board), by whether the CEO

is recruited from the family (family CEO), by the number of family members owning stock in

the firm (family size), and by whether a family member heads the board (family chair), and by

the family’s equity in the firm (family ownership). The first column of results copies the base-

case result from table 5, where we measure family control by whether the largest family by

ownership has a majority stake.

The table shows that the relationship between the exit decision and all other variables than

ownership concentration is insensitive to how we measure family control. As in the base case,

higher ownership concentration always increases the expected exit propensity, but the

statistical significance is weaker when we use family control variables that do not directly

reflect the family’s ownership rights. This result suggests that formal power at the stockholder

30

meeting is the key ownership determinant of the exit decision. Overall, table 7 reflects that the

estimates of (3) are robust to how family control is measured.

Table 7

Unreported tests use alternative empirical measures for financial constraints and for

performance. Instead of measuring financial constraints by leverage, we use annual real sales

growth during either the current, the last two, or the last three years. The estimates are

equivalent to those in table 5. We find the same result if we measure annual ROA during the

current or the last two years.

We have so far used the convention that if the firm exits at time t, it is classified as an exit

firm also before t. Table 8 repeats the base-case result from table 5 in model I, while model II

classifies the firm as exiting only in the year it actually switches from ASA to AS. The

estimates show that female directors and firm performance become insignificant determinants

in model II, and that firm size becomes positive and significant at the 4% level. The roles are

unaltered for ownership concentration, family control, listing status, and firm age.

The insignificant relationship between exit and female directors may turn up because all firms

in our sample tend to increase their use of female directors over time, regardless of whether

the firm ultimately exits. For instance, firms that exit increase the average fraction of females

on their boards before exit from 8% in 2002 to 30% in 2008. This increasing use of female

directors regardless of exit behavior means that when a firm is classified as exiting only in the

year it actually exits, it is an exit firm in our test only when its female director fraction is the

highest and hence the closest to the female fraction for non-exiting firms. This is probably

why this firm characteristic is unable to separate exiting firms from the non-exiting.

Table 8

Summarizing, we have shown that the base-case results are independent of whatever

econometric technique we use, how we define family control, how we measure return on

assets, and whether we use leverage or growth to measure financial constraints. The definition

31

of an exit firm matters, because the fraction of female directors is not a significant predictor of

exit behavior when the firm is classified as exiting only in the year it actually exits.9

4.3 Entry The entry decision is expected to be driven by almost the same firm characteristics as for the

exit decision. We specify the following model:

(4) 1 2 3 4

5 6 7 8

it it it it it

it it it it it

Entry Female directors Firm size Family control Firm ageFinancial constraints Performance Ownership concentration IPO u

α β β β ββ β β β

= + + + ++ + + + +

Entry is a dummy variable that is 1 for firms that enter the exposed organizational form and 0

otherwise. Based on the theoretical arguments for the exit decision in section 2 as specified in

hypotheses H1–H8, we predict that entry is more likely if the firm has many female directors

(β1 > 0), large size (β2 > 0), binding financial constraints (β5 > 0), low performance (β6 < 0),

and low ownership concentration (β7 < 0). Like we did for exit, we leave unspecified the

expected effects on entry of family control (β3) and firm age (β4).

Because an AS firm considering entry must necessarily be non-listed, listing status is an

irrelevant determinant. However, and as suggested by table 2, we expect the GBL will

increase the tendency of entering firms to become listed (make an IPO; go public) directly

upon entry rather than to stay non-listed in their new ASA form. The reason is that listed

firms enjoy more of the benefits of being an ASA that are independent of the GBL. Hence, the

GBL makes it relatively more attractive to be listed than non-listed once the firm is already an

ASA. The dummy variable IPO in (4) is 1 if the entering firm chooses to become listed in the

entry year and 0 otherwise. We expect a positive relationship between the propensity to enter

and the tendency to become listed upon entry (β8 > 0).

The sample is AS firms during the period 2000–2009 and ASA firms in their entry year. To

qualify as an entry candidate, the AS must have at least three board members and sales not

less than the lowest sales observed among the ASA firms that year.

The estimates of the relationship in (4) are shown in table 9, which reports the results from

two models. In model A, an AS that converts to ASA during the sample period is classified as 9 Boards in firms with more than 200 employees have one third of their directors elected by and from the employees. We find no evidence that the exit decision depends on whether the board has employee directors. Hence, even if employee directors may be more positive to the GBL than stockholders are, stockholders do not seem so be influenced by this view when making the exit decision.

32

entering every year until the year it enters, which is the firm’s last sample year. Model B uses

the alternative definition, where the firm is classified as entering only in the year it actually

enters. Unlike in the exit case, where every exit candidate already is an ASA, it seems more

appropriate for the entry case to use the definition in model B. This definition is more

appropriate because an entry candidate is not exposed to the GBL until it voluntarily chooses

to enter. Moreover, many non-entering firms that qualify for entry may not even consider

entry a relevant option. Finally, firms choosing to enter know what new regulation they must

comply with. Hence, model B seems to reflect the more reasonable definition of an entering

firm.

Table 9

The estimates are consistent with our predictions for the effect of firm size, family control,

firm age, ownership concentration, and the decision to go public directly. The negative

coefficient for financial constraints is inconsistent with the hypothesis, strengthening the

impression gathered from the exit model that financial constraints do reliably predict changes

in organizational form. More surprisingly, the fraction of female directors is not a significant

determinant of entry in model B and is even significantly negative in model A.10

This finding suggests that non-entering firms are not held back more than entering firms are

by the GBL’s requirement to replace a large portion of the male directors. This result is

apparently puzzling, given our earlier finding in table 5 that the existing gender mix in the

board is a strong determinant of exit. However, and as already discussed, exiting and entering

firms are fundamentally different in a GBL compliance sense. An ASA firm is already

exposed to the GBL and has no choice in the sense that unless it exits to AS, it must comply

with the 40% quota or accept to be liquidated. An AS, however, chooses to enter and hence

comply only if it wants to. Therefore, AS firms that voluntarily decide to become ASA do so

because their owners think the cost of filling the gender quota is small relative to the benefit

of being an ASA. Apparently, the owners also consider this cost to be independent of the

board’s current gender mix.

Finally, tables 5 and 9 jointly document that family-controlled firms are more hesitant than

are other firms both to exit from and to enter into the ASA form. This finding suggests that 10 Unreported regressions show that the fraction of female directors continues to be insignificant when we re-estimate the two models on the subsample of firms that enter after December 2005, when Parliament decided to punish non-compliers with liquidation.

33

when new regulation changes the benefit of the status quo, family firms are inclined to keep

their organizational form, whatever that is. One possible reason is that family-controlled firms

have transaction costs of organizational change that are not well accounted for by the other

independent variables in our entry and exit models.

Overall, we have shown that the radical board restructuring mandated by the GBL has strong

effects on the choice of organizational form. This evidence suggests the regulation is costly,

and estimates made by others may indicate the magnitude of this cost. First, Ahern and

Dittmar (2012) find that listed firms with no female directors experienced an abnormal price

drop of 3.5% when the intention to regulate was announced. This estimate is an upper limit

for the average GBL cost to ASAs, because it reflects the listed ASAs only, including the

three quarters of them that must restructure their boards the most if they decide not to exit.

Second, Fjesme (2012) finds an average first-day IPO return of 2.7% at the Oslo Stock

Exchange during 2000–2008. This evidence of IPO underpricing may represent the minimum

average benefit for an ASA firm of being listed instead of non-listed. Hence, 3% may be a

rough estimate of the average cost of the GBL for owners of listed firms. The average cost for

non-listed ASAs is below 3%, because they lose less when value when exiting than listed

firms do. Importantly, however, all these estimates are averages and do not necessarily apply

to an individual firm in the sample. This limitation follows from our finding that the cost and

benefit of exiting or entering the organizational form exposed to the GBL depends on a series

of firm characteristics.

5. Summary and conclusions

The findings of this paper support the idea that firms may respond to more restrictive

regulation by changing their organizational form. Such change occurs when the added cost of

the new regulatory constraint makes the firm’s current organizational form less attractive than

the best alternative. Strikingly, we find that when a new law mandates at least 40% of men

and women in Norwegian boardrooms, half the firms choose to exit into an organizational

form that is not exposed to the law. This tendency to avoid costly regulation by changing

organizational form varies systematically with firm characteristics. We find that exit is

significantly more likely when the firm is profitable, small, young, and non-listed. Exiting

firms also tend to have powerful owners, no controlling family, and few female directors.

34

Most of these characteristics also influence the decision to enter the exposed organizational

form by firms that are not exposed to the gender balance law. Even though far fewer listed

firms exit than do non-listed firms, listed firms that do not exit may nevertheless have to bear

the highest cost of the new regulation.

This evidence is consistent with theoretical predictions and existing empirics on how firms

respond to regulatory shocks. The results are also in line with earlier findings that board

composition matters for firm value, and that compulsory gender balance in the boardroom

shrinks the pool of competent directors and reduces stockholder wealth. Our evidence

supports the notion that optimal board composition and the best response to regulatory shocks

varies from firm to firm. Moreover, gender balance regulation may be less disruptive if firms

have the option to exit into organizational forms where the law does not apply.

Recent political signals indicate that the exit option we analyze in this paper may soon

disappear. In particular, gender balance in corporate boards may be made mandatory for more

than just one organizational form.11 If that happens, Norway will not just be special for being

the first and only country to mandate a massive, rapid shift in the composition of corporate

boards and to punish non-compliers with liquidation. The regulators may also decide to

eliminate the possibility firms currently have to mitigate the costs of regulatory shocks by

transforming into organizational forms that are not exposed to the law. Every other country

considering gender balance regulation seems to favor the comply-or-explain system or

considerably milder sanctions than liquidation. Such regulatory regimes would leave the

gender balance choice to the firm’s discretion and hence allow for firm heterogeneity in board

design. Our findings suggest that compared to this more flexible alternative, the mandatory

approach, and particularly one without exit options, is a costly way to the regulate gender

balance of corporate boards.

11 http://e24.no/jobb/naa-vil-regjeringen-ha-kvinner-i-alle-styrer/20060520.

35

Acknowledgements We are grateful for valuable feedback from Janis Berzins, Tore Bråthen, Ilan Cooper, Daniel Ferreira, Miguel Garcia-Cestona, Tom Kirchmaier, Espen Moen, Øyvind Norli, Charlotte Ostergaard, Luc Renneboog (the editor), Amir Sasson, R. Øystein Strøm, Danielle Zang, and from participants at BI’s Brownbag Seminar in Economics, the CBS Conference on Board Diversity and Economic Performance, London School of Economics, and the 13th Workshop on Corporate Governance and Investment at Cardiff Business School.

36

References Adams, Renée B., Benjamin E. Hermalin, and Michael S. Weisbach. 2010. "The Role of

Boards of Directors in Corporate Governance: A Conceptual Framework and Survey." Journal of Economic Literature no. 48 (1):58–107.

Ahern, Kenneth R., and Amy Dittmar. 2012. "The Changing of the Boards: The Impact on Firm Valuation of Mandated Female Board Representation." Quaterly Journal of Economics no. 127 (1):137–197.

Anderson, Ronald C., and David M. Reeb. 2003. "Founding-Family Ownership and Firm Performance: Evidence from the S&P 500." Journal of Finance no. 58 (3):1301–1328.

Bahrat, Sreedhar T, and Amy K. Dittmar. 2006. "To Be or Not to Be (Public)." SSRN eLibrary.

Boone, Audra L., Laura C. Field, Jonathan M. Karpoff, and Charu G. Raheja. 2007. "The Determinants of Corporate Board Size and Composition: An Empirical Analysis." Journal of Financial Economics no. 85 (1):66–101.

Bushee, Brian J., and Christian Leuz. 2004. "Economic Consequences of SEC Disclosure Regulation: Evidence from the OTC Bulletin Board." SSRN eLibrary.

Bøhren, Øyvind. 2011. "Eierne, Styret og Ledelsen: Corporate Governance i Norge." Bergen: Fagbokforlaget.

Bøhren, Øyvind, and R. Øystein Strøm. 2010. "Governance and Politics: Regulating Independence and Diversity in the Board Room." Journal of Business Finance & Accounting no. 37 (9–10):1281–1308.

Engel, Ellen, Rachel M. Hayes, and Xue Wang. 2004. "The Sarbanes-Oxley Act and Firms' Going-Private Decisions." SSRN eLibrary.

Fjesme, Sturla F. 2012. "Laddering in Initial Public Offering Allocations." AFA 2012 Chicago Meetings Paper. SSRN eLibrary.

Gomard, Bernhard, and Peer Schaumburg-Müller. 2011. "Kapitalselskaber." Copenhagen: Jurist- og Økonomforbundets Forlag.

Greene, William. H. 2007. "Econometric Analysis." Upper Saddle River, NJ: Prentice Hall. Hansmann, Henry. 1996, page 151. "The Ownership of Enterprise." Cambridge, MA: Harvard

University Press. Lutter, Marcus. 1992. "Die Entwicklund der GmbH in Europa und in der Welt.” In Marcus

Lutter, Peter Ulmer and Wolfgang Zollner (eds), “Festschrift 100 Jahre GmbH-Gesetz." Cologne: Verlag Dr. Otto Schmidt KG.

Morck, Randall, Andrei Shleifer, and Robert W. Vishny. 1989. "Alternative Mechanisms for Corporate Control." American Economic Review no. 79 (4):842–852.

Myers, Stewart C., and Nicholas S. Majluf. 1984. "Corporate Financing and Investment Decisions when Firms have Information that Investors Do Not Have." Journal of Financial Economics no. 13 (2):187–221.

Nygaard, Knut. 2011. "Forced Board Changes: Evidence from Norway." SSRN eLibrary. Rosener, Judy B. 2011. "The 'Terrible Truth' About Women On Corporate Boards."

ForbesWoman no. 06072011. Shleifer, Andrei, and Robert W. Vishny. 1986. "Large Shareholders and Corporate Control."

The Journal of Political Economy no. 94 (No. 3, Part 1):461–488. Thomsen, Steen, and Frederik Vinten. 2007. "Delisting in Europe and the Cost of

Governance." SSRN eLibrary.

37

Table 1: Responding to the gender balance law by choosing organizational form

Response Before gender balance law After gender balance law

Stay ASA ASAExit ASA ASEnter AS ASADo not enter AS AS

Organizational form

This table shows the relationship between alternative responses to the gender balancelaw (Stay, Exit, Enter, and Do not enter) and the inherent choice of alternativeorganizational forms (ASA and AS). The gender balance law applies to every ASA, butto no AS.

38

Table 2: Sample size by listing status, exit behavor, and entry behavior

Year All Non-listed Listed Exit Entry Net exit All Non-listed Listed All IPO No IPO

2000 313 187 126 8 33 -25 8 6 2 33 29 42001 317 190 127 14 18 -4 14 13 1 18 15 32002 309 187 122 17 9 8 17 13 4 9 6 32003 292 175 117 23 6 17 23 18 5 6 2 42004 283 166 117 19 10 9 19 13 6 10 5 52005 270 146 124 25 12 13 25 23 2 12 3 92006 263 130 133 28 21 7 28 24 4 21 11 102007 248 110 138 46 31 15 46 36 10 31 17 142008 231 95 136 22 5 17 22 16 6 5 3 22009 217 83 134 15 1 14 15 13 2 1 1 0Average 274 147 127 22 15 7 22 18 4 15 9 5Total 2,743 1,469 1,274 217 146 71 217 175 42 146 92 54

This table shows the number of sample firms by alternative classification criteria. Panel A shows the total number of ASA firms (All)and the number of such firms by listing status (Non-listed and Listed), while Panel B shows the number of ASA firms that exit to AS(Exit), the number of AS firms that enter into ASA (Enter), and the difference between the two (Net exit). Panel C splits the number ofexits from panel B into non-listed and listed firms, while panel D splits the sample of entry firms from panel B according to whether thefirm makes an initial public offering (IPO) in the entry year. ASA firms are exposed to the gender balance law, while AS firms are not.Listed firms are quoted on the Oslo Stock Exchange. The sample is all Norwegian AS firms entering the ASA form and all NorwegianASA firms that are not financials or have not exited the ASA form because of takeover or bankruptcy during the sample period.

A: Listing status B: Entry and Exit C: Exit by listing status D: Entry by IPO propensity

39

Table 3: Characteristics of exit firms and non-exit firms

Exit less Exit Non-exit Non-exit t-value (p-value)

General firm characteristics

Listed 0.221 0.582 -0.361 -

20.982 (0.000) Financial constraints 0.542 0.519 0.023 2.108 (0.035) Growth 1.659 1.847 -0.188 -1.112 (0.268) Performance 6.864 6.813 0.051 0.129 (0.893) Firm age 19.800 29.428 -9.628 -8.288 (0.000) Firm size 658.930 3,225.282 -2,566.352 -4.299 (0.000)

Ownership characteristics Ownership concentration 52.768 37.699 15.069 11.783 (0.000) Family ownership 35.132 30.754 4.378 4.078 (0.000) Inside ownership 13.934 12.042 1.892 2.144 (0.032)

Board characteristics

Female directors 0.100 0.201 -0.101 -

15.911 (0.000)

Board size 5.129 5.926 -0.797 -

10.545 (0.000)

Family characteristics Family firm 0.222 0.189 0.033 1.874 (0.062) Family size 1.847 2.109 -0.262 -5.072 (0.000) Family chair 0.200 0.221 -0.021 -1.244 (0.214) Family CEO 0.184 0.214 -0.030 -1.923 (0.055) Family board 0.134 0.124 0.010 1.639 (0.101)

N 1,100 1,900

This table compares exit firms to non-exit firms in terms of their mean value for general firm, ownership, board, and family characteristics. The difference between the two mean values, the t-value, and the p-value (in parentheses) of this difference are reported in the three right-most columns. Performance is censored at the 0.5% tail and then winzorized at the 1% and 99% tails. Financial constraints and growth are winzorized at the 1% and 99% tails. Appendix 3 defines the variables, and the sample is all Norwegian ASA firms in 2000–2009 that are not financials or have not exited to the AS form because of takeover or bankruptcy during the sample period. Unlike AS firms, ASA firms must comply with the gender balance law.

40

Table 4: Exit propensity in Norway and in neighboring countries

A: Number of firms

Year Norway Denmark Sweden Denmark and Sweden

1996 177 22,572 286 22,858 1997 226 22,898 448 23,346 1998 253 23,589 420 24,009 1999 288 24,855 460 25,315 2000 313 25,978 470 26,448 2001 317 29,164 452 29,616 2002 309 30,540 430 30,970 2003 292 31,894 420 32,314 2004 283 32,519 447 32,966 2005 270 38,021 487 38,508 2006 263 39,525 533 40,058 2007 248 41,087 595 41,682 2008 231 41,778 592 42,370 2009 217 41,280 575 41,855 Average 263 31,836 473 32,308

B: Difference-in-difference regressions

Non-event group Estimate (p-value)

Adjusted R2 N

Denmark -0.426 (0.000) 0.593 27 Sweden -0.271 (0-027) 0.494 27 Denmark and Sweden -0.424 (0.000) 0.592 27

This table compares the number of ASA firms in Norway with the number of firms in Denmark and Sweden that have a similar organizational form, but that are not subject to gender balance regulation. Panel A reports the number of firms per year, while Panel B shows estimates of the difference-in-difference equation defined in model (2) of the main text. The estimate in the first column of results reflects the difference between the number of Norwegian firms (the event group) in the event period and the non-event period relative to the corresponding difference in Denmark, Sweden, or both (the non-event group). The sample period is 1996–2009, where 1996–2002 is the pre-event period and 2003–2009 is the post-event period. We exclude Norwegian ASAs that are financials or firms that have exited the ASA form because of takeover or bankruptcy during the sample period. The Danish, Swedish, and the combined Danish and Swedish samples consist of all firms with the organizational form that is similar to the Norwegian ASA form. Sources for the Danish and Swedish data are www.dst.dk and Finbas, respectively.

41

Independent variable Prediction Estimate

Female directors (-) -3.064(0.000)

Firm size (-) -0.104(0.000)

Family control (-/+) -0.701(0.000)

Firm age (-/+) -0.011(0.000)

Financial constraints (-) 0.349(0.222)

Performance (+) 0.025(0.000)

Ownership concentration (+) 0.010(0.000)

Listed (-) -1.182(0.000)

Constant 1.868(0.000)

N 1,560LR chi2(8) 377.470Prob > chi2 (0.000)Pseudo R2 0.182

Table 5: The base-case estimates

This table shows the estimated coefficients for a logit regression of ownership, board, family, andgeneral firm characteristics on the decision to exit or not exit the organizational form exposed to thegender balance law. The relationship is specified in model (3) of the main text. The predicted signs ofthe coefficients are shown in the second column, and the p-values of the estimated coefficientsreported in the third column are stated in parentheses underneath. The dependent variable is 1 if thefirm is an exit firm and 0 otherwise. Female directors is the proportion of shareholder-elected boardmembers who are women. Firm size is the log of sales in constant 2009 millions of NOK. Family controlis a dummy variable which equals 1 if the largest family owns more than 50% of the equity and 0otherwise. Firm age is the number of years since the firm was founded. Financial constraints is totaldebt divided by total assets. Performance is the average real return on assets per year from year t-3 tot. Ownership concentration is the fraction of equity held by the largest stockholder. Listed is a dummyvariable which is 1 if the firm is quoted on the Oslo Stock Exchange and 0 otherwise. Performance iscensored at the 0.5% tail and then winzorized at the 1% and 99% tails. Financial constraints iswinzorized at the 1% and 99% tails. The sample is all Norwegian ASA firms in 2000–2009 that are notfinancials or have not exited to the AS form because of takeover or bankruptcy during the sampleperiod. Unlike AS firms, ASA firms must comply with the gender balance law.

42

Independent variable Logit Probit Standard panel Logit panel Pooled OLS Clustered OLS

Female directors -3.064 -1.809 -0.897 -9.547 -0.561 -0.561(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Firm size -0.104 -0.061 -0.026 -0.65 -0.019 -0.019(0.001) (0.001) (0.032) (0.003) (0.001) (0.039)

Family control -0.701 -0.421 -0.158 -4.387 -0.139 -0.139(0.000) (0.000) (0.028) (0.000) (0.000) (0.005)

Firm age -0.011 -0.007 -0.001 -0.493 -0.002 -0.002(0.000) (0.000) (0.223) (0.001) (0.000) (0.007)

Financial constraints 0.349 0.215 -0.022 1.054 0.058 0.058(0.222) (0.209) (0.859) (0.603) (0.273) (0.499)

Performance 0.025 0.015 0.008 0.186 0.005 0.005(0.000) (0.000) (0.009) (0.000) (0.000) (0.018)

Ownership concentration 0.01 0.006 0.003 0.351 0.002 0.002(0.000) (0.000) (0.002) (0.062) (0.000) (0.033)

Listed -1.182 -0.706 -0.162 -3.945 -0.248 -0.248(0.000) (0.000) (0.009) (0.000) (0.000) (0.000)

Constant 1.868 1.105 0.581 9.093 0.856 0.856(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

N 1,560 1,560 1,560 1,246 1,560 1,560LR chi2(8)/Wald chi2(8) 377.47 377.99 407.48Prob > chi2 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)F(t,n) 53.97 24.28Prob > F(t,n) (0.000) (0.000)Pseudo R2 0.182 0.183 0.213 0.214 0.212Random firm effects no no yes yes no no

Table 6: Alternative estimation methods

This table shows the effect of estimating the base case model (3) in the main text with alternative econometric techniques. The dependent variable is 1 if the firm is anexit firm and 0 otherwise. Logit is the base case from table 5, Probit is a probit model, Standard panel is a random effects model with exit considered a continuousvariable, Pooled OLS uses no panel method and treats exit as a continuous variable, while clustered OLS treats exit as a continuous variable and uses standard errorsadjusted for dependence between observations at the firm level. The p-values are stated in parentheses. Female directors is the proportion of shareholder-elected boardmembers who are women. Firm size is the log of sales in constant 2009 millions of NOK. Family control is a dummy variable which equals 1 if the largest family ownsmore than 50% of the equity and 0 otherwise. Firm age is the number of years since the firm was founded. Financial constraints is total debt divided by total assets.Performance is the average real return on assets per year from year t-3 to t. Ownership concentration is the fraction of equity held by the largest stockholder. Listed is adummy variable which is 1 if the firm is quoted on the Oslo Stock Exchange and 0 otherwise. Performance is censored at the 0.5% tail and then winzorized at the 1% and99% tails. Financial constraints is winzorized at the 1% and 99% tails. The sample is all Norwegian ASA firms in 2000–2009 that are not financials or have not exited tothe AS form because of takeover or bankruptcy during the sample period. Unlike AS firms, ASA firms must comply with the gender balance law.

Method

43

Family Family Family Family Family FamilyIndependent variable firm board CEO size chair ownership

Female directors -3.064 -2.956 -2.876 -2.830 -2.790 -2.841(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Firm size -0.104 -0.129 -0.125 -0.113 -0.107 -0.125(0.001) (0.000) (0.001) (0.002) (0.003) (0.001)

Family control -0.701 -1.575 -0.807 -0.195 -0.174 -0.0090.000 0.001 0.000 0.264 0.003 0.002

Firm age -0.011 -0.013 -0.012 -0.013 -0.011 -0.014(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Financial constraints 0.349 0.219 0.167 0.227 0.106 0.272(0.222) (0.514) (0.621) (0.497) (0.751) (0.417)

Performance 0.025 0.027 0.029 0.026 0.025 0.029(0.000) (0.002) (0.001) (0.002) (0.004) (0.001)

Ownership concentration 0.010 0.004 0.003 0.004 0.004 0.008(0.000) (0.112) (0.294) (0.200) (0.150) (0.015)

Listed -1.182 -1.156 -1.208 -1.097 -1.114 -1.171(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Constant 1.868 2.120 2.051 2.231 1.955 2.235(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

N 1,560 1,145 1,145 1,145 1,145 1,145LR chi2(8) 377.470 269.910 279.890 259.940 267.920 268.820Prob > chi2 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)Pseudo R2 0.182 0.179 0.186 0.173 0.178 0.179

Table 7: Alternative definitions of family control

Family control definition

This table shows estimates of model (3) in the main text under six alternative proxies for family control.The first column of results reproduces the base-case findings from table 5. The dependent variable is 1if the firm is an exit firm and 0 otherwise.The p-values are stated in parentheses. Female directors isthe proportion of shareholder-elected board members who are women. Firm size is the log of sales inconstant 2009 millions of NOK. Performance is the average real return on assets per year from year t-3 to t. Ownership concentration is the fraction of equity held by the largest stockholder. Listed is adummy variable which is 1 if the firm is quoted on the Oslo Stock Exchange and 0 otherwise.Performance is censored at the 0.5% tail and then winzorized at the 1% and 99% tails. Financialconstraints is winzorized at the 1% and 99% tails. The sample is all Norwegian ASA firms in2000–2009 that are not financials or have not exited to the AS form because of takeover or bankruptcyduring the sample period. Unlike AS firms, ASA firms must comply with the gender balance law.

44

Independent variable I II

Female directors -3.064 -0.541(0.000) (0.503)

Firm size -0.104 0.124(0.001) (0.040)

Family control -0.701 -0.651(0.000) (0.022)

Firm age -0.011 -0.014(0.000) (0.010)

Financial constraints 0.349 -0.377(0.222) (0.480)

Performance 0.025 -0.002(0.000) (0.878)

Ownership concentration 0.010 0.017(0.000) (0.000)

Listed -1.182 -1.881(0.000) (0.000)

Constant 1.868 -4.194(0.000) (0.000)

N 1,560 1,060LR chi2(8) 377.47 109.77Prob > chi2 (0.000) (0.000)Pseudo R2 0.182 0.162

Table 8: Defining exit status

This table shows the effect of defining exit in two alternative ways. Model I is thebase case from table 5, where the firm is classified as an exit firm every year until itleaves the ASA oganizational form and enters the AS form at some point during thesample period. Model II assigns exit status to the firm only in the year it becomes anAS. The dependent variable is 1 if the firm is an exit firm and 0 otherwise.The p-values are stated in parentheses. Female directors is the proportion of shareholder-elected board members who are women. Firm size is the log of sales in constant 2009millions of NOK. Family control is a dummy variable which equals 1 if the largestfamily owns more than 50% of the equity and 0 otherwise. Firm age is the number ofyears since the firm was founded. Financial constraints is total debt divided by totalassets. Performance is the average real return on assets per year from year t-3 to t.Ownership concentration is the fraction of equity held by the largest stockholder.Listed is a dummy variable which is 1 if the firm is quoted on the Oslo Stock Exchangeand 0 otherwise. Performance is censored at the 0.5% tail and then winzorized at the1% and 99% tails. Financial constraints is winzorized at the 1% and 99% tails. Thesample is all Norwegian ASA firms in 2000–2009 that are not financials or have notexited to the AS form because of takeover or bankruptcy during the sample period.Unlike AS firms, ASA firms must comply with the gender balance law.

Model

45

Independent variable Prediction A B

Female directors (+) -1.381 -0.336(0.001) (0.546)

Firm size (+) 0.226 0.178(0.000) (0.000)

Family control (-/+) -1.455 -1.705(0.000) (0.000)

Firm age (-/+) -0.130 -0.149(0.000) (0.000)

Financial constraints (+) -1.640 -2.115(0.000) (0.000)

Performance (-) -0.009 -0.003(0.206) (0.775)

Ownership concentration (-) -0.013 -0.012(0.000) (0.002)

IPO (+) 1.062 2.019(0.000) (0.000)

Constant -5.277 -5.160(0.000) (0.000)

N 126,152 126,152LR chi2(8) 813.858 484.060Prob > chi2 (0.000) (0.000)Pseudo R2 0.232 0.280

Table 9: The entry decision

Model

This table shows the estimated coefficients from a logit regression of ownership, family, and general firm characteristics on the decision of an AS firm to become an ASA and thereby become exposedto the gender balance law. The predicted signs of the coefficients are shown in the second column.Model A classifies the AS firm as an entry firm every year before entry if it becomes ASA duringthe sample period. Model B assigns entry status to the firm only the year it actually enters. Thedependent variable is 1 if the firm is an entry firm and 0 otherwise. Female directors is theproportion of shareholder-elected board members who are women. Firm size is the log of sales inconstant 2009 millions of NOK. Family control is a dummy variable which equals 1 if the largestfamily owns more than 50% of the equity and 0 otherwise. Firm age is the number of years sincethe firm was founded. Financial constraints is total debt divided by total assets. Performance is theaverage real return on assets per year from year t-3 to t. Ownership concentration is the fraction ofequity held by the largest stockholder. IPO is a dummy variable which is 1 if the firm becomeslisted in the entry year and 0 otherwise. We censor financial constraints and performance at the +/-2% tails. The p-value of the estimated coefficient's t-statistic is shown in parentheses. The sampleis AS firms that are entry candidates and ASA firms in their entry year during the period2000–2009. To qualify as an entry candidate, the AS firm must have at least three board membersand sales not smaller than the lowest sales among the ASA firms that year.

46

Appendix 1: Regulatory differences between limited liability firms with alternative organizational forms

Exposed listed (Public ASA) Exposed non-listed (Private ASA) Unexposed (AS)

1 Minimum share capital 1 million Norwegian kroner 1 million Norwegian kroner 0.1 million Norwegian kroner2 Corporate governance code

(comply-or-explain)The annual report must specify, item by item, whetherthe firm complies with the corporate governance codeof the OSE.

No corporate governance code No corporate governance code

3 CEO-chair duality Illegal Illegal Legal if share capital is below 3 millionNOK.

4 Gender balance on board At least 40% of each gender* At least 40% of each gender* No gender balance requirement

5 Non-voting shares Up to 50% of the shares can be non-voting. Up to 50% of the shares can be non-voting. No restriction on non-voting shares

6 Mandatory flagging An investor passing up and down through thethresholds of 5%, 10%, 20%, 1/3, 50%, 2/3, and 90% ofthe outstanding cash flow rights or voting rights mustnotify both the firm and the OSE no later than the nextmorning.

No flagging rule No flagging rule

7 Mandatory tender offer An investor passing the 1/3 ownership thresholdmust offer to buy all the remaining stock in the firm.

No tender offer requirement No tender offer requirement

8 Reporting of trades by corporate insiders

Insiders (the firm's officers and directors) must reporttheir trades to the OSE no later than the next morning.

Insiders must report to the board. Theinformation is not public.

No insider reporting required

9 Ownership recording The firm must report every transaction in itsoutstanding equity securities to the VPS. Thenotification must specify the identity of the buyer andseller, the exact time of the transaction, the number ofsecurities traded, and the security price.

The firm must report every transaction in itsoutstanding equity securities to the VPS. Thenotification must specify the identity of thebuyer and seller, the exact time of thetransaction, the number of securities traded,and the security price.

The firm must have a register that keeps track of every trade in the firm's stock.The information is not public.

10 Accounting rules The firm must report in detail on the executives' anddirectors' compensation, and on the stockholdings ofthe officers, directors, and their close family. The firmmust report the interest adjustment dates of theirbonds.

Less detailed reporting requirements than forlisted firms on compensation, ownership, anddebt

Less detailed reporting requirementsthan for listed firms on compensation,ownership, and debt

Regulation

This table shows regulatory differences between Norwegian limited liability firms with alternative organizational forms. An exposed firm is subject to the gender balance law, while anunexposed firm is not. OSE is the Oslo Stock Exchange. VPS is Verdipapirsentralen (The Securities Registry).

Organizational form

* The minimum fraction of each gender varies between 33% and 50%, depending on board size. The minimum is 40% when the board has at least ten members.

47

Percentage of Gender Comply or explain(ce)/Country female directors quota Regulatory activity Mandatory(m)Australia 8.4 no Firms should adopt and publicly explain a diversity policy. ce

Belgium* 7.7 yes A draft for a law was filed in December 2009 requiring boards of listed firms and certain non-listed firms to chooseat least one third of their directors from each gender. No sanction

pending

Canada 10.3 no The Canadian Board of Diversity was launched in November 2009 with a goal of improving diversity on boards,including gender diversity.

pending

France* 12.7 yes Parliament proposed a law in January 2011. The law requires a quota whereby 25% of the directors are femalewithin 3 years and 40% within 6 years for firms employing at least 500 and with revenues over 50 million EUR. Nosanction

ce

Germany* 11.2 no The German Corporate Governance Code, which applies to listed firms, has recommendations aimed at promotingmore women on boards.

ce

Iceland 3.8 yes Law passed in March 2010; quota of 40%. No sanction pending

Italy* 3.7 yes A law was passed in December 2010 requiring one third of each gender on boards of listed firms and state-ownedfirms. The law needs Senate approval and will will apply to new board nominations six months after suchapproval. No sanction

pending

Netherlands* 14.0 yes A law on gender quotas for the executive and the supervisory boards received govermental approval inDecember 2009. The law proposes at least 30% of each gender for the board of listed firms and for the boards ofnon-listed firms that meet certain financial and employment criteria. No sanction

ce

New Zealand 7.5 no The New Zealand Shareholders' Association will make board diversity one of three priorities. pendingNorway 40.1 yes A law passed in 2003 and implemented in 2008 mandates at least 40% of each gender on the board of listed firms

and certain non-listed firms.** Non-complyers are liquidated.m

Spain* 9.3 yes Parliament passed the Law of Equality in 2007, which requires listed firms to appoint females to 40%–60% of theboard positions. Firms are allowed until 2015 to comply. No penalty ***

ce

United Kingdom* 15.0 no The Corporate Governance Code recommends gender diversity on boards. The Conservative Party hasannounced that it will require that females constitute at least 50% of the candidates on the long list ofdirectorship candidates.

pending

United States 16.1 no The SEC approved a rule in December 2009 requiring disclosure of whether and how board nominationcommittees consider diversity when identifying director candidates. If the committee or the board has a diversitypolicy, the SEC rule requires disclosure of how this policy is implemented and how the nomination committee orthe board assesses the policy's effectiveness. The rule was implemented in February 2010.

ce

Appendix 2: Regulation of gender balance in corporate boards across the world

This table shows the regulatory status on gender quotas and the actual fraction of females in boards across the world as of 2012. In addition to the countries specified above, Ireland, SouthAfrica, and Switzerland have gender quotas for state-owned firms. Sources: Ahern and Dittmar (2012), Catalyst (2012), www.corpgov.deloitte.com, and www.nho.no.

* A EU draft from November 2012 proposes a 40% target for each gender by 2020 on boards of listed firms with at least 250 employees. Each members state is supposed to decide whether the quota should be mandatory and what sanctions should be used for non-compliers.** Boards with less than ten members have the quota stated as a minimum number of members per gender.*** No formal penalty will apply to non-compliers, but the government will take compliance into account when awarding state contracts to private firms.

48

Variable DefinitionExit Dummy variable which equals 1 if the firm leaves the organizational form

exposed to the gender balance law during the sample period and 0 otherwiseEntry Dummy variable which equals 1 if the firm enters the organizational form

exposed to the gender balance law during the sample period and 0 otherwise

General firm characteristicsListed Dummy variable which is 1 if the firm is listed on the Oslo Stock Exchange

and 0 otherwiseFinancial constraints Total debt divided by total assets (leverage)Growth The average percentage increase in real sales per year from year t-3 to tPerformance The average real return on assets (ROA) per year from year t-3 to tFirm age The number of years since the firm was foundedFirm size Sales in constant 2009 millions of NOK. Log transformed in regressionsIPO Dummy variable which is 1 if the firm becomes listed the same year it enters

the exposed organizational form and 0 otherwise

Ownership characteristicsOwnership concentration The fraction of equity held by the largest stockholderFamily ownership The fraction of equity held by the family with the largest equity stakeInside ownership The fraction of equity held by the officers and directors

Board characteristicsFemale directors The proportion of shareholder-elected board members who are womenBoard size The number of shareholder-elected board members

Family characteristicsFamily firm Dummy variable which equals 1 if the largest family owns more than 50% of

the equity and 0 otherwiseFamily size The number of owners in the largest family by ownershipFamily chair Dummy variable which equals 1 if the chair belongs to the largest family by

ownership and 0 otherwiseFamily CEO Dummy variable which equals 1 if the CEO belongs to the largest family by

ownership and 0 otherwiseFamily board The fraction of directors coming from the largest family by ownership

Appendix 3: The empirical variables

This table defines the variables used in the empirical analysis. The ownership characteristics are basedon ultimate (direct plus indirect) equity holdings.

49

3.

Female directors and board independence:

Evidence from boards with mandatory gender balance

by*

Øyvind Bøhren Siv Staubo

April 8, 2013

Abstract

This paper explores if gender quotas have other effects on the composition of corporate

boards than the implied upwards shift in gender diversity. We analyze the impact on board

independence of an unexpected and radical law in Norway requiring that at least 40 percent of

a firm’s directors be of each gender. Our evidence shows that this regulatory shock has strong

and firm-specific effects. The average fraction of independent directors grows by 20

percentage points. This upwards shift occurs because 84 percent of the female directors are

independent, while only 50 percent of the men are. We find that demand for an independent

board is lowest in small, young, profitable, non-listed firms with few female directors and

powerful stockholders. Such firms need monitoring by independent directors the least and

advice by dependent directors the most. These firms are hit hardest by excessive board

independence, which may be an unintended side effect of mandatory gender balance.

Keywords: corporate governance; regulation; board independence; gender balance

JEL classifications: G30; G38.

* Department of Financial Economics, BI Norwegian Business School, N-0442 Oslo, Norway. Our email addresses are oyvind.bohren@bi.no and siv.staubo@bi.no. We are grateful for valuable comments to an earlier draft from Renée Adams, Rosemarie Koch, Øyvind Norli, Charlotte Østergaard, and R. Øystein Strøm.

50

1. Introduction The low proportion of females on corporate boards has attracted widespread attention in many

countries from practitioners, regulators, researchers, and the media (Farrell and Hersch 2005;

McKinsey & Company 2007; Terjesen, Sealy, and Singh 2009; Langli 2011; Adams and

Kirchmaier 2013). Norway was the first country to act politically on this issue by mandating

gender balance in the boardroom, and other countries have followed suit. France, Iceland,

Netherlands, and Spain will implement gender quotas in 2013–2016, and proposals along

similar lines have recently been made in Australia, Belgium, Canada, the EU Commission,

and Italy. 12 Prime Minister David Cameron recently stated, “There is clear evidence that

ending Britain’s male-dominated business culture would improve performance, and that

Britain’s economic recovery is being held back by a lack of women in the boardroom” (The

Guardian 2012).

The Norwegian gender balance law was announced to the surprise of many in 2002, was

passed by Parliament in 2003, and became mandatory from 2008. Figure 1 shows that the

average fraction of directorships filled by females increased monotonically from 11 percent

when the law was passed to 42 percent five years later, when the 40 percent quota became

mandatory, including a liquidation penalty after three months for non-compliers.

Figure 1

This paper explores whether such a massive, involuntary shift in gender balance changes

other board characteristics than just the gender balance. In particular, we analyze the impact

of the gender balance law on board independence, which regulators consider the key

characteristic of a board with high monitoring skills (Bhagat and Black 1998; Adams,

Hermalin, and Weisbach 2010). However, the idea that board independence is always

beneficial has no support in the research literature. As has been shown both theoretically

(Adams and Ferreira 2007) and empirically (Linck, Netter, and Yang 2008; Duchin,

Matsusaka, and Ozbas 2010), optimal board independence requires a tradeoff between the

value of monitoring provided by independent (outside) directors and the value of advice

provided by dependent (inside) directors. This inherent conflict between monitoring and

advice suggests that board quality will suffer if forced gender balance pushes the board’s

independence above its optimal level.

12 Gender balance rules for the boards of state-owned firms have been implemented in Ireland, South Africa, and Switzerland.

51

We show that the Norwegian gender balance law (GBL) increases board independence quite

dramatically. Whereas the average fraction of independent directors was 46 percent when the

GBL was passed, the fraction rose to 67 percent when the gender quota became mandatory

with a non-compliance penalty five years later. Even though all firms exposed to the GBL had

to eventually fill the 40 percent quota, we document that the effect of the law on board

independence varies in the cross-section.13 The firms affected the most are those that need

independence the least, because they have low demand for the monitoring provided by

independent directors and high demand for the advice provided by dependent directors. We

find that those firms that ultimately have the most excessive board independence tend to be

small, young, private, profitable, owned by powerful stockholders, and to have had few

female directors before the quota became mandatory.

This finding means that although the GBL regulates just one board characteristic per se

(gender balance), the law affects another characteristic (independence) as well. This shift

from dependence towards independence also shifts the balance of skills away from advice

towards monitoring. This effect occurs because independence is a much more widespread

characteristic among female director candidates than among males. That is, the pool of male

director candidates and the pool of female director candidates are not equal regarding the two

fundamental director skills, which are monitoring and advice.

One may wonder whether increased board independence is driven not by the GBL, but rather

by the corporate governance code, which was introduced in the middle of our sample period.

This code is soft law based on the principle of comply-or-explain, recommending that half the

firm’s directors be independent.14 However, the code applies to the listed (public) firms, but

not to the non-listed (private). Hence, whereas the GBL imposes the same indirect restriction

on board independence regardless of listing status, the governance code restricts board

independence only in listed firms.

13 The exact 40 percent quota applies only to boards with more than nine members. The quota for smaller boards is specified as a minimum number of directors per gender. There must be at least one director of each gender if the board has two or three members, at least two if there are four or five members, at least three if there are six to eight members, and at least four of each gender if the board has nine members. These thresholds imply that the minimum fraction of each gender may vary between 33 percent and 50 percent in a cross-section of compliers. 14 The code was implemented in 2006 and states, “The majority of the stockholder-elected members of the board should be independent of the company’s executive personnel and material business contacts.” Moreover, at least two stockholder-elected directors should be independent of the main stockholder (www.nues.no). According to corporate law, one third of the board must be elected by and from the employees in firms with more than 200 employees. Employee-elected directors are probably dependent by nature.

52

We exploit this difference to separate the effects on independence stemming from the two

regulatory sources. Roughly half the firms in the population are listed and hence exposed to

both the GBL and the code. The other half consists of non-listed firms and hence is exposed

only to the GBL. Therefore, the less the growth in board independence differs between listed

and non-listed firms, the higher the likelihood that the regulatory effect on independence is

due to the GBL rather than to the code. Our evidence shows that the impact does not come

from the code, but rather from the GBL. The GBL produces the same upward shift in board

independence regardless of the firm’s listing status. Specifically, the average independence

level rises from 52 percent to 72 percent in listed firms and from 41 percent to 59 percent in

non-listed firms.

This finding also suggests that unlike what has been argued, it is not the independence

requirement in governance codes that causes the high fraction of independent female directors

in countries that have governance codes but no gender quotas. The idea behind this argument

is that if independence is easier to find among female director candidates than among males,

stockholders will rationally choose females rather than males in order to meet the code’s

independence requirement (Beecher-Monas 2007). However, we show that female directors

are as often independent in firms that are not exposed to the independence code, but only

exposed to the GBL. Hence, the high independence among female directors in countries other

than Norway may not be the outcome of stockholders who look for female directors to fill the

independence quota. Rather, the entire pool of female director talent has so few dependent

(inside) candidates that one cannot select both many women and many dependent women

simultaneously. Therefore, choosing a female director very often means having to choose an

independent director, even though that was not the intention. This problem becomes clear for

firms that try to fill the gender quota.

Stockholders could have complied with the GBL in at least two ways that would have

increased board independence less dramatically. We find that these options were not widely

used. First, just adding female directors to the existing board increases independence less than

when independent females replace dependent males. While extensive use of this option would

have increased board size by about 50 percent in our sample, this response is not what we

observe. The average board, excluding employee directors, has 5.6 members of whom about

10 percent are women at the start of the sample period. There are 5.9 members of whom about

40 percent are women at the period’s end. Nevertheless, increasing board size to mitigate the

53

independence effect may be rational. The reason is that studies in several countries have

found an inverse relationship between board size and firm performance regardless of firm size

(Yermack 1996; Eisenberg, Sundgren, and Wells 1998; Bøhren and Strøm 2010).15

A second way to mitigate excessive board independence is by recruiting females with

multiple directorships, who have been shown to be less independent (Fields and Keys 2003).

However, the only trend we find is that men hold multiple seats considerably less often after

the GBL than before. There is no clear evidence in our sample that women’s holding of

multiple directorships becomes more common.

Ahern and Dittmar (2012) find that Norwegian firms without female directors, which

represent three quarters of their sample, lost on average 3.5 percent in market value when the

plan for the GBL was announced. The authors argue that this value drop is not a temporary

overreaction. The Tobin’s Q ratio of the firms with no female directors at the GBL

announcement typically fell by 15 percent when the law was passed. These firms continued

underperforming relative to the other firms until the law was implemented five years later.

The number of female directorships increased by 260 percent (from 165 to 592 seats) during

this period, while the number of directorships held by males dropped by 38 percent (from

1,516 to 938 seats). Ahern and Dittmar also find that the incoming female directors had

considerably less leadership experience than the exiting males. While 69 percent of the

retained male directors had CEO experience, this was the case for only 31 percent of the new

female directors, who were on average also eight years younger.

Our evidence suggests that this loss of firm value partly happens because the GBL produces

excessive board independence. Most incoming female directors are more independent than

most exiting males are. This greater independence is because the incoming females generally

have less prior experience with the firm and its environment. Hence, the value drop may not

be driven by the shift in gender balance per se. Rather, the drop happens because skills that

are relevant for board work correlate with gender. According to Adams and Kirchmaier

(2013), a key reason for this correlation is that men and women with equal formal education

choose different subsequent careers. This choice matters for the ability to build director skills.

15 Coles, Daniel, and Naveen (2008) have recently challenged this result by finding a positive empirical relationship between Tobin’s Q and board size for firms with high leverage or many business segments, both of which are used to measure firm complexity.

54

Empirical tests of the Adams and Ferreira (2007) model find that, as predicted, the demand

for independent directors with monitoring skills varies from firm to firm (Linck, Netter, and

Yang 2008; Duchin, Matsusaka, and Ozbas 2010). We extend this literature by showing that a

regulatory floor on gender diversity may produce a side effect in terms of excessive board

independence relative to dependence and hence excessive monitoring relative to advice. This

distortion of optimal board design is particularly strong in firms that need monitoring the least

and advice the most. Our evidence shows that firms with low demand for monitoring have

low agency costs related to the manager-stockholder relationship. Those firms have few and

strong stockholders (private firm and high ownership concentration) who can easily stay well

informed (small and young firm). Hence, they have little need for monitoring by non-owner

directors. However, these firms may have high need for advice, because their low age and

small size suggest they lack the internal resources to successfully make strategic decisions

and to establish key external networks without the help from experts on their boards.

Earlier research on the role of women on corporate boards has studied only settings where the

proportion of female directors is both low and unregulated (Farrell and Hersch 2005; Adams

and Ferreira 2009b). Using the same independence measure as we do, Adams and Ferreira

classify 84 percent of the female directors in US firms as independent, as opposed to 40

percent of the males. 16 These figures make the authors warn against mandating gender

diversity, fearing that a larger proportion of female directors may unduly increase board

independence.

This argument assumes that the pool of potential female directors is biased towards high

independence. However, unlike what we do us, existing research studies only firms where the

governance code recommends that half the firm’s directors be independent, and where about

10 percent of the directors are females. This limited empirical context means that the

unobserved, complete pool of female director candidates may very well be balanced regarding

dependence vs. independence. Nevertheless, firms may still want to recruit more heavily from

the subpool of independent females in order to comply with the independence code.

We show, however, that even firms facing no such independence requirement increase their

board independence by almost 45 percent when one third of their male directors must be

replaced by women. Thus, stockholders find it difficult to avoid excessive board 16 Their sample is an unbalanced panel of S&P 500, S&P MidCap, and S&P SmallCap firms during the period 1996–2003. The data are collected by the Investor Responsibility Research Center.

55

independence when gender quotas are mandated. Therefore, it seems that the complete pool of

female director candidates is indeed heavily biased towards independence. This bias means

that mandatory quotas will be a more serious problem for optimal board design the more

gender balance the regulator enforces.

We present our base-case model and the predictions in Section 2. The data and summary

statistics are discussed in Section 3, while Section 4 explains the methodology and presents

the base-case results. Section 5 analyzes alternative board independence proxies and

alternative independence determinants. We summarize and conclude in Section 6.

2. Predictions

Board independence is affected by regulatory and non-regulatory determinants. Our paper

focuses on the former, which are exogenous restrictions on board composition. Specifically,

we analyze how board independence is influenced by mandatory law in terms of the GBL and

by soft law in terms of the governance code. Nevertheless, our basic framework comes from

the non-regulatory determinants, which are the firm-specific characteristics we will use to

explain why the response to the same regulatory constraint differs across firms. The literature

has so far addressed only non-regulatory determinants, such the firm’s ownership structure,

profitability, complexity, and risk.

Our base-case model is the following

1 2 3

4 5 6

7 8 9

*

A Reg B Non-regit

it it it it

it it it

it it

(1) Board independence X XFemale directors Listed Female directors Listed

Inside ownership Outside ownership PerformanceLeverage Risk

α β βα β β ββ β ββ β β

= + +

= + + +

+ + +

+ + + 10it it itFirm size Firm age u β+ +

The regulatory determinants in the vector XReg are specified in the second row of the model,

while the non-regulatory determinants in XNon-reg are spelled out in the two bottom rows. The

starting point is the tradeoff theory, in which optimal board independence reflects the value-

maximizing combination of monitoring and advice. We outline the tradeoff theory and

operationalize the board independence concept in section 2.1. We predict how board

independence relates to regulatory determinants in section 2.2, and to non-regulatory

determinants in section 2.3. Table 1 summarizes the empirical proxies.

56

Table 1

2.1 Board independence

Independent directors may create firm value because there is a potential conflict of interest

between the manager who runs the firm and the owners who delegate control rights to the

manager (Jensen and Meckling 1976). The more serious the resulting separation between

ownership and control, the more value the directors may create by monitoring management

more efficiently inside the boardroom than the non-director owners can do from the outside.

Different director types have different incentives to fill this monitoring function. Compared to

directors with professional ties to the firm or personal ties to the manager, directors without

such ties have less to lose by challenging and criticizing the manager. The former director

type is called dependent or inside, while the latter is called independent or outside (Bhagat

and Black 1998; Hermalin and Weisbach 1998; Adams, Hermalin, and Weisbach 2010). The

independent director’s monitoring incentives on a specific board are strengthened further if

the value of his or her human capital primarily depends on a reputation for offering

monitoring services on any board.

The board’s second function is to advise the firm’s management. High advisory skills require

deep insight into the firm, its customers, suppliers, competitors, and industry. Dependent

directors have such skills because of their closeness to the firm, while independent directors

lack them because of their arms-length distance from the firm (Bhagat and Black 1998, 2002).

Because dependent directors may lose reputation by monitoring (control) and build reputation

by advising (support), they have stronger incentives to advise than independent directors do,

and lower incentives to monitor.

This setting implies that the value of the board’s monitoring function stems from reduced

agency costs in the relationship between owners and managers (“monitoring prevents bad

projects”). In contrast, the value of advice stems from the board’s ability to generate new

ideas for strategy and operations that management can develop further and implement

(“advice creates good projects”).

Finally, one would expect that managers operating in this environment would dislike being

intensively monitored. The reason is that monitoring reduces managements’ discretion

regarding the firm’s resources as represented by the free cash flow (Jensen 1986). In contrast,

57

managers like advice, because more advice may increase the free cash flow. Therefore,

management may respond to more monitoring by providing the board with less information.

Adams and Ferreira (2007) formalize this setting and show that optimal board independence

involves a tradeoff between the value of monitoring provided by independent directors and

the value of advice provided by dependent directors. Two of their results are particularly

important for our context. First, over-optimal (too much) independence reflects a board where

the advisory skills are too weak relative to the monitoring skills. There is too much control

and too little support. Firm value is lost because the board generates too few new ideas and

too strongly restricts management.

The second important result is that too much independence pushes not only the board’s

advisory skill below its optimal level. Excessive independence may also reduce the board’s

ability to monitor. Such a loss of both advisory value and monitoring value may occur when

managers respond to being monitored by reducing information flow to the board. This

response creates a problem because an uninformed board cannot properly fulfill its two roles.

Therefore, greater independence caused by new regulation may reduce board quality because

the board gets too low advisory skills and also too little information for advice as well as for

monitoring. This imbalance means the radical change in board composition mandated by the

GBL in our setting may produce uninformed boards that have underoptimal focus on advice

and overoptimal focus on monitoring.

We measure board independence empirically in line with earlier research from the United

States, which classifies a director as either inside, grey, or outside (Baysinger and Butler 1985;

Weisbach 1988; MacAvoy and Millstein 1999; Adams and Ferreira 2009b). Inside directors

are defined as the firm’s full-time employees, former employees, or employees of closely

related firms. Grey (affiliated) directors have professional relationships with management, or

are likely to have business relationships with the firm, such as investment bankers and

lawyers. Outside directors are neither inside nor grey.

We measure board independence in the base case as the fraction of outside directors.

Although this is the most common measure in the literature, our robustness tests will use

several alternatives. Specifically, we alternatively define independent directors as the fraction

of outside directors minus the fraction of inside directors (Bhagat and Black 2002) and as the

fraction of outside plus grey directors (Linck, Netter, and Yang 2008; Ahern and Dittmar

58

2012). Finally, we use an independence measure based on whether a firm’s CEO has a seat on

the board (Carter and Lorsch 2004).

These independence measures may all be criticized for overestimating true independence. The

reason is that the measures ignore social (i.e., non-business) relationships between the

directors and the CEO (Hwang and Kim 2009; Cohen, Frazzini, and Malloy 2012). We think

this is a minor problem in our sample, where men fill 96 percent of the CEO positions. This

fact suggests that if anything, measures that ignore social relationships will overestimate

independence more for men than for women. Consequently, the true difference in

independence between men and women is even larger than what we measure. This bias

strengthens the power of our test, where the null hypothesis is that independence and gender

are unrelated.

2.2 Regulatory determinants

We expect that board independence relates positively to the GBL, which mandates the same

minimum fraction (40 percent) of female directors in every firm. This prediction is supported

by Farrell and Hersch (2005), who examine the characteristics of female directors in 300

firms on the Fortune 1000 list from 1990 to 1999. The use of female directors increased by 7

percentage units during these ten years, and more than 90 percent of the incoming females

were classified as outside directors. The authors argue that because women have limited

experience as managers and stockholders, one would expect board independence to increase

when female directors replace males. Adams and Ferreira (2009b) support this logic by

arguing that because women seldom belong to the so-called old boys’ network, women are

also closer to the theoretical notion of independent directors being arm’s-length monitors. The

authors estimate that while large firms in the United States had about 10 percent females on

the boards in 2004, 84 percent of them were classified as outside directors.17

17 Along similar lines, it has been argued that independence is easier to achieve by ensuring diversity in ethnicity and gender (Fields and Keys 2003; Beecher-Monas 2007). It may be a general belief that diverse boards are more independent (Adams and Ferreira 2009b). Several behavioral differences between men and women reported in the literature support this idea. Female directors may reduce earnings management (Gul, Scrinidhi, and Tsui 2007), women seem to adopt a more democratic, transformational, and trust-building leadership style (Cohen, Pant, and Sharp 2001; Klenke 2003; Trinidad and Normore 2005), females may exhibit higher ethical standards in their decision making (Betz, O'Connell, and Shepard 1989; Mason and Mudrack 1996; Clikeman, Geiger, and O'Connell 2001), and women may be more risk averse (Riley and Chow 1992; Powell and Ansic

59

We expect a positive relationship between having independent directors and being listed

because every sample firm is exposed to the GBL, but only the listed are subject to the

independence code. Nevertheless, listed firms may comply with both the GBL and the code in

one move by appointing female directors who are also independent. To account for this

possibility, we relate the GBL and the code by means of an interaction term for the fraction of

female directors and the firm’s listing status. This interaction term allows us to determine

whether the GBL drives independence differently in firms that must also comply with the

code. If listed firms choose female directors to comply with both the code and the GBL, we

expect a positive coefficient for the interaction term. Conversely, the coefficient will be

negative if female directors are more often independent in non-listed firms than in the listed.

This latter case would occur if dependent female director candidates prefer to sit on listed

firms’ boards, possibly because of higher visibility and pay. Such preferences would tend to

produce a higher fraction of independent female candidates in the talent pool for non-listed

firms than for the listed.

2.3 Non-regulatory determinants We group the non-regulatory determinants of board independence into board, ownership, and

general firm characteristics. Because the inclusion of partly overlapping characteristics in the

same model may cause multicollinearity problems, our base-case model uses only a subset of

the non-regulatory determinants in Table 1 that are common in the literature.

Empirical research has shown that board independence declines as inside ownership increases

(Bhagat and Black 2002; Linck, Netter, and Yang 2008). This finding can be rationalized by

the theoretical argument that there is less need for monitoring when the board and the

stockholders are aligned (Raheja 2005). We measure inside ownership by the aggregate

equity fraction held by the firm’s officers and directors. The equity fraction is measured as

ultimate ownership, which is the investor’s direct equity holding in the firm plus any indirect

holdings through intermediaries such as holding companies. We predict that board

independence and inside ownership are negatively correlated.18

1997; Sundén and Surette 1998). Adams and Funk (2009) have recently challenged the latter point by finding that female directors in Sweden are slightly less risk averse than males are. 18 The term “inside owner” is very different from the term “inside director.” An inside owner is a stockholder in the firm who is also on the board or on the management team. An inside (dependent) director is a board member with close relationships to the firm that are unrelated to ownership. The same relationship goes for outside

60

The principal’s power and incentives to monitor the agent increase with outside ownership

concentration (Shleifer and Vishny 1997). This means powerful owners outside the

boardroom have strong incentives to ensure that their directors monitor management properly

on the owners’ behalf. Hence, board independence may relate positively to outside ownership

concentration. On the other hand, large outside owners may monitor management directly

rather than indirectly through the board. Such monitoring can happen both in the stockholder

meeting and through informal contact with management. The use of such channels reduces

the demand for monitoring directors. Hence, we do not specify the expected relationship

between board independence and outside ownership concentration, which we measure by the

Herfindahl index based on every ultimate ownership stake in the firm.

Research shows both theoretically and empirically that firms may respond to poor

performance by appointing a higher fraction of independent directors (Hermalin and

Weisbach 1991; Bhagat and Black 2002). The rationale is that a CEO who dislikes

monitoring is supposed to have less influence over board composition the weaker the firm’s

performance. We measure performance as the past three years’ average return on assets,

predicting an inverse relationship between board independence and firm performance.

Complex firms are thought to have a high need for monitoring and have also been found to

have more independent boards (Coles, Daniel, and Naveen 2008; Linck, Netter, and Yang

2008). We measure firm complexity by firm size and by firm age, expecting higher values of

both characteristics to correlate positively with board independence. Managers of firms with

high debt have less free cash flow to waste on value-destroying projects. Therefore, high debt,

and the resulting strong monitoring by creditors, is a substitute for the monitoring carried out

by an independent board (Jensen 1986). We predict an inverse relationship between board

independence and financial leverage.

Finally, the optimal fraction of independent directors decreases as the cost of monitoring

increases (Adams and Ferreira 2007). Such monitoring costs are particularly high when firms

with strong information asymmetry are monitored by independent directors (Maug 1997).

Moreover, empirical research has found that the information asymmetry is higher the more

volatile the firm’s stock returns (Fama and Jensen 1983). Because we have stock return data

for only the listed firms in our sample, we use the standard deviation of the book return on owners vs. outside (independent) directors, where an outside owner is not on the board or on the management team.

61

assets to proxy for information asymmetry and hence for monitoring costs. We expect board

independence to correlate negatively with risk in a volatility (total risk) sense.

Summarizing section 2, we predict that the board will have a higher fraction of independent

directors when the firm is listed, has many female directors, low inside ownership

concentration, low performance, low leverage, low risk, and when the firm is large and old.

1 2 3

4 5 6

7 8 9

*

A Reg B Non-regit

it it it it

it it it

it it

(1) Board independence X XFemale directors Listed Female directors Listed

Inside ownership Outside ownership PerformanceLeverage Risk

α β βα β β ββ β ββ β β

= + += + + ++ + ++ + + 10it it itFirm size Firm age u β+ +

3. Data and summary statistics

Our sample is all firms exposed to the GBL, and the data set is an unbalanced panel from

2003 to 2008. Except for the data on director independence, the source is the CCGR database

(www.bi.edu/ccgr). 19 Norwegian firms with limited liability are legally obliged to publish

accounting statements every year. The firm must also report the identity and the equity

holdings of its owners, directors, and CEO. Failure to submit this information within 17

months after fiscal year-end triggers automatic liquidation by the court. The law also

mandates a standardized set of accounting statements certified by a public auditor, regardless

of the firm’s listing status, size, and industry.

The data used to manually classify directors as inside, grey, and outside are from

Brønnøysundregistrene (www.brreg.no) and Proff (www.proff.no). We obtain supplementary

information on director characteristics by manually searching the annual reports. Appendix 1

illustrates how we use this data set to classify the directors.

Table 2 shows distributional properties of the variables. Because the governance code applies

only to listed firms, Table 3 splits the sample by listing status to uncover whether board

composition depends on organizational form.

Table 2

Table 3

According to Table 2, board independence as measured by the fraction of outside directors is

59 percent on average during the sample period. Table 3 documents that this fraction is 64 19 The database includes every limited liability firm registered in Norway from 1994 to the present.

62

percent in listed firms and 50 percent in the non-listed, respectively. Because non-listed firms

are not subject to the governance code on independence, this lower degree of independence in

non-listed firms is as expected from a regulatory perspective. The difference may also reflect

that optimal independence is lower in non-listed firms because the value of monitoring is

lesser. This rationale is supported by the table, which suggests that potential agency problems

between owners and managers are generally smaller in non-listed firms. Ownership is

concentrated in fewer hands, which implies fewer free-riding problems and more incentives

and power to discipline management. For instance, the largest stockholder in non-listed firms

holds on average 63 percent of the equity, compared to 29 percent in the listed. Boards of

non-listed firms also have more owner presence. Officers and directors hold on average 16

percent of the equity in non-listed firms and 8 percent in the listed.

These characteristics mean that demand for independent directors who monitor management

on the owners’ behalf is lower in non-listed firms. Risk as measured by asset return volatility

is also considerably higher. These characteristics increase demand for advice relative to

monitoring, which also follows from the fact that non-listed firms in our sample are generally

smaller and younger.

Table 4 shows the prevalence of outside, grey, and inside directors for the sample as a whole

in panel A. The fraction of outside directors increases every year, growing from 46 percent in

2003 to 67 percent in 2008. An opposite monotone decline occurs for the fraction of inside

directors, which drops from 44 to 27 percent. Finally, the fraction of grey directors drops from

10 to 6 percent. Hence, the large growth in outside directors primarily happens at the expense

of inside directors rather than grey directors.

This pattern means that the fraction of outside directors, which is our base-case measure of

board independence, may capture a fundamental shift in director skills towards more

monitoring at the expense of less advice. Such a shift towards monitoring would have been

more questionable if most of the decline had happened among the grey directors, who are

more independent than the inside directors.

Table 4

Panel B shows that the fraction of outside directors grows from 52 percent to 72 percent in the

listed firms and from 41 to 59 percent in the non-listed. Hence, the difference remains

63

constant, documenting that the increased independence is not due to the governance code,

which was introduced in the middle of the sample period. Linck, Netter, and Yang (2008)

follow a large sample of listed firms in the United States over time. They estimate that the

average fraction of inside directors is around 30 percent in 2003, which is their next-to-last

sample year, and also after the 2002 Sarbanes-Oxley Act was implemented. The

corresponding fraction in our sample that year is 45 percent. Remarkably, however, it takes

only another five years in our sample until the fraction of inside directors has dropped to 28

percent.

Panel C documents the relationship between director type and gender. Females are on average

outside directors in 84 percent of the cases, while the corresponding figure for men is 50

percent. Hence, the average female director is more than 60 percent more likely than a male

director is to be in the outside category. Similarly, whereas on average 13 percent of the

women are inside directors, men are inside directors in 43 percent of the cases. These

relationships remain practically constant over time. Although not reported in the table, this

pattern is also the same regardless of listing status.

This difference in independence between men and women is very close to the difference

estimated by Adams and Ferreira (2009b) and Farrell and Hersch (2005) in the United States,

using the same definition of outside directors as we do. They report that 84 percent of the

females are in this category, as opposed to 40 percent of the males. Our figures are 84 percent

and 50 percent, respectively. This evidence strengthens the impression that owners of

Norwegian firms could not have mitigated excessive independence after the GBL by

recruiting more dependent (inside) female directors than earlier. It seems there is no such

untapped pool of talent better qualified for advice than for monitoring.

Nevertheless, stockholders could have dampened the GBL’s impact on board independence

by merely adding female directors to the existing, male-dominated board. According to Table

5, however, this is not a widespread strategy, because board size does not increase noticeably

during the sample period. The average board has 5.60 members in 2003 and 5.86 members

five years later. In fact, panel B shows that board size even decreases somewhat in non-listed

firms, where the board also tends to be smaller than in the listed (4.97 vs. 6.13 members on

average, respectively).

Table 5

64

The second way to reduce excessive board independence is by appointing female directors

who already hold board seats elsewhere. Directors with multiple seats are likely to be less

independent than those with just one seat because of the former’s links to board members and

top management in other firms. Table 6 documents the holding of directorships by men and

women in 2003, 2006, and 2008, respectively.

Table 6

Panel A documents, as expected, a trend towards a higher total number of multiple seats held

by women and a lower number of such seats held by men. In particular, multiple seats are

held by as many men as women at the end of the sample period (66 vs. 68, respectively),

while men dominate very strongly five years earlier (320 vs. 18, respectively). This pattern

follows almost by implication when 30 percent of the men are to be replaced by women in

short supply during a brief period. A similar impression is given by panel B, where the

number of seats per director increases over time for females and decreases for males. For

instance, the mean number of seats per director increases from 1.16 to 1.22 for women, while

decreasing from 1.25 to 1.10 for men.

Nevertheless, panel C shows that when we consider only directors with multiple seats, there is

no clear tendency that the fraction of women who hold multiple seats increases over time.

Rather, the striking feature is that men have multiple seats considerably less often than earlier.

For instance, the average number of seats held by men decreases from 1.25 to 1.10 during the

sample period (panel B), while the fraction of men who hold multiple seats decreases from 22

percent to 8 percent (panel C).

Overall, Table 6 shows that there is a certain tendency for some women to hold multiple

directorships more often. Nevertheless, the only clear trend is that men as a group hold

multiple seats less often than earlier. Although unreported tests we have made show that

women with multiple directorships are generally more dependent than are women holding just

one seat, it seems that recruiting female directors with multiple seats is not a widespread

strategy to reduce board independence. One possible reason is if stockholders share the view

that busy directors may easily become overstretched and therefore have less value.

65

Alternatively, the pool of females who can potentially hold multiple directorships is so small

that this recruiting source does not matter much.20

Overall, it seems the two alternative strategies for recruiting female directors are not widely

used to dampen the growth in board independence caused by the GBL. Instead, stockholders

comply with the new law by recruiting from the talent pool of female directors with relatively

little leadership experience. These female directors replace males, thereby leaving board size

unchanged.

Table 7 shows bivariate correlation coefficients between the independent variables in the

sample we will use to estimate model (1). Multicollinearity should not be a serious concern,

because no correlation coefficient exceeds the critical limit of 0.8 (Studenmund 2000). Also,

the variance inflation factor (VIF) equals 4.88, which is below the limit of 5 considered a

sufficient reason for not suspecting multicollinearity problems.

Table 7

4. Empirical methodology and base-case results Our data set contains multiple observations of the same firm over time. Therefore, we will use

panel data techniques to reduce potential endogeneity problems caused by unobservable

determinants of board independence (Hsiao 2003).21 We use the fixed-effects approach to

account for firm-specific unobservables.22 Moreover, we control for the possibility that the

20 The effect of women on board independence may also be dampened by appointing experienced women from other countries. Nygaard (2011) shows, however, that the proportion of foreign female directors was 11.7 percent in 2003, while the corresponding number for males was 13.3 percent. Five years later, 12.8 percent of the female directors and 15.7 percent of the males were foreigners. Thus, there is no clear shift from national to international recruitment of female directors. One possible reason is that even though such board members are experienced, they may nevertheless be outside directors in a monitoring sense because their network is from a different country. 21 Endogeneity caused by reverse causation seems to be a minor concern in our study. First, the large increase in female directors is caused by an exogenous shock in terms of an unexpected new law. Second, the board’s independence in listed firms is partially driven by an exogenous corporate governance code that recommends the same minimum level of independence in every firm. Third, firms may certainly reduce inside ownership to ensure that a majority of the directors are independent. Although the correlation between board independence and insider ownership is negative and significant in our sample, the correlation between inside ownership and time is insignificant. Hence, increased inside ownership is not likely to be caused by increased independence over time. Finally, firms exposed to the GBL can avoid the law by exiting to an unexposed organizational form. However, such exit is not an option for a listed firm unless it also chooses to delist. Bøhren and Staubo (2013) show that although half the firms exposed to the GBL in 2003 had left this organizational form by 2008, roughly three quarters of the exiting firms were non-listed. 22 We use fixed firm effects even though the base-case model contains a dummy variable for a firm’s listing status. Because some firms change listing status during the sample period, however, this dummy variable is not a constant over time for all firms. We prefer the fixed-effects approach because it allows the unobservable firm

66

GBL’s effect on board independence in all firms may change as 2008 approaches, when the

law became mandatory. Table 8 shows the estimates of (1), using OLS estimation with fixed

firm effects and fixed time effects.

Table 8

The presence of female directors is associated positively with board independence and is

highly statistically significant. Thus, as predicted, mandating a large change in gender balance

towards more females does not alter gender balance alone. The regulation increases the

board’s independence, too. This result is consistent with that shown in Table 4, which shows

that the sample’s average for monitoring skills in any year is much higher for female directors

than for males. Conversely, women have a comparative disadvantage as advisors. The

stockholders’ problem is, however, that the tradeoff between monitoring and advice after the

GBL must be made within a severely restricted opportunity set regarding the female director

candidates.

Listed firms have boards with significantly greater independence than non-listed firms do.

This result is as predicted, given the regulatory fact that the independence rule in the

governance code applies only to listed firms.23 Moreover, the negative coefficient for the joint

effect of female directors and listing status on independence suggests that listed firms do not

fill the gender quota and the independence quota in one go. That is, a listed firm does not

systematically recruit a female who is independent more often than a non-listed firm does.

Rather, it is even more common for a female director to be independent in non-listed firms,

despite the fact that such firms are subject only to the GBL rather than to both the GBL and

the independence code. The reason is that independence and gender are so closely related in

the talent pool and possibly also because dependent female director candidates prefer listed

firms.

Turning next to the non-regulatory determinants, the table shows that while outside ownership

concentration is not a significant determinant, inside ownership concentration relates

negatively to board independence. This finding is in line with our prediction that the demand

effect to be correlated with the error term. The robustness tests will also use the random-effects approach, which assumes the unobservable firm effect is a random variable that is uncorrelated with the error term. 23 We get the same result both for female directors and for listing status if we exclude all observations after 2005, that is, after Parliament decided not to withdraw the law, but to instead add a liquidation penalty for non-compliers.

67

for directors who primarily monitor is weaker when the CEO or these directors are owners as

well.

As expected, firm performance and board independence are negatively correlated, suggesting

that monitoring becomes stricter when the firm does poorly. Independence decreases as

leverage grows, which is consistent with substitutability between a monitoring board and a

monitoring creditor. However, board independence is not systematically related to asset risk.

This finding suggests that if monitoring costs are reflected in the volatility of asset returns,

such costs are not an important concern in the tradeoff between monitoring and advice.

Finally, and as we predicted, firm complexity as measured by the firm’s size and age

correlates positively with board independence. Hence, the smaller and younger the firm, the

lesser the demand for an independent board.

Summarizing, we have tested the multivariate model in (1) using panel data techniques to

account for unobservable firm and time effects on board independence. Most relationships are

statistically significant, and every significant relationship is consistent with our predictions of

how the demand for an independent vs. a dependent board depends on regulatory and non-

regulatory characteristics. We find that the demand for board independence tends to be lower

when the firm has few female directors, strong owners, high profits, and when the firm is

small, young, and non-listed. Such firms trade off monitoring and advice in a way that reflects

a low value of being monitored and a high value of being advised.

5. Robustness

We first re-estimate the base-case model in (1) using alternative econometric techniques.

Subsequently, we analyze what happens when we use alternative measures of board

independence. Finally, we test the sensitivity to using different empirical proxies than those

used in (1) for some of the non-regulatory determinants of board independence.

5.1 Econometric techniques

The base-case regression in Table 8 uses fixed-effects estimation to control for the influence

of unobservable firm characteristics that remain constant over time. Table 9 repeats these

estimates as the point of reference under technique I. Technique II is random-effects

68

regressions with ML (maximum likelihood) methodology, and III controls for fixed firm

effects by using sample period averages for each variable in every firm.24

Table 9

The estimates reported in II and III are mostly consistent with those of the base case in I. First,

the significant relationships between board independence and female directors, listing status,

inside ownership concentration, and firm size are robust to the econometric technique used.

Second, none of the remaining variables is significant with opposite signs under alternative

techniques. These results show that our major findings in the base-case model are not driven

by the chosen estimation method. Therefore, we continue using the fixed firm effect and the

fixed time effect approaches in the following.

5.2 Board independence

The independence measure in the base case comes from partially hand-collected data that we

use to classify each director as either outside, grey, or inside according to the definitions from

section 2. The data collection and the classification system are both subject to potential bias

because they rely on our judgment. Moreover, independence is an elusive concept because it

is hard to define precisely in terms of a specific empirical proxy. For these reasons, we test

three alternative measures of independence.

The first alternative is whether the CEO is a member of the firm’s board.25 The rationale is

that the party being monitored should not have a say in the monitoring body (Carter and

Lorsch 2004). The CEO is not a board member in 88 percent of our sample firms. This

happens in 92 percent of the listed firms and in 80 percent of the non-listed, respectively. We

measure board independence by a dummy variable that is 1 if the CEO is not a director and 0

otherwise.

24 Technique II reduces standard errors by accounting for cross-sectional variance in the between estimator. Technique III reduces the standard errors by accounting for cross-sectional variance in the between and within estimators. Technique IV is appropriate when the dependent variable can take on values only in a restricted interval, such as in our case, where board independence varies between zero and one. 25 Of course, the CEO is always present in board meetings except when a case cannot be openly discussed unless he or she is absent, as when the board evaluates the CEO’s performance and decides compensation. A law introduced two years after the end of our sample period makes it illegal for the CEO to be a board member.

69

The sample as well as the he determinants of independence correspond to what we used in the

base case, and we estimate the relationship as a logit model with random effects.26 Table 10

reports the estimates under model A. The table shows that model A and the base-case model

from Table 8 have the same set of significant determinants at the 5 percent level or lower

(female directors, listing status, performance, leverage, firm age, firm size), whereas two

determinants are insignificant in both models (outside ownership concentration and risk). Two

determinants are significant only at the 10 percent level in the base case (the interaction term

and inside ownership concentration).

Table 10

Model B follows Bhagat and Black (2002) by defining independent directors as the fraction of

outside directors minus the fraction of inside directors. Model C follows Ahern and Dittmar

(2012) and Linck, Netter, and Yang (2008) by classyfying both outside and grey directors as

independent.27 Again, the deviation from the base case is minor in both models. Reassuringly,

therefore, our findings are not sensitive to alternative ways of measuring board independence.

5.3 Non-regulatory determinants of board independence

The set of potential determinants of board independence is large when moving from

theoretical constructs to empirical proxies. Given the purpose of this paper, however, our

major focus is on determinants that directly or indirectly reflect regulatory effects on board

independence. These determinants are listing status as a direct determinant through the

governance code and female directors as an indirect determinant through the GBL. The

remaining variables in (1) are potential non-regulatory determinants. To analyze whether the

choice of such non-regulatory determinants matters for our findings for the regulatory effects,

we specify a model of board independence where the non-regulatory determinants are as close

as possible to those used by two recent studies of board composition in the United States

(Adams and Ferreira 2009a; Linck, Netter, and Yang 2008):

26 Fixed-effects estimation is not feasible in a logit model (Hsiao 2003). 27 The definition of independent directors used in model C comes closest to the definition used in the Norwegian governance code. This code recommends that half the stockholder-elected directors be unrelated to the executive team or to key business partners.

70

1 2 3

4 5 6

7 8 9

(3) *

it it it it it

it it it

it it

Board independence Female directors Listed Female directors ListedBoard size CEO ownership CEO ownership squaredCEO tenure CEO tenure squared Firm age

α β β ββ β ββ β β

= + + ++ + ++ + +

10 it

it itFirm age squared uβ+ +

The first line of independent variables in (3) specifies the regulatory determinants, which are

identical to those used in (1). Lines 2–4 hold the non-regulatory determinants, which differ

from their counterparts in (1).

Research shows that larger boards are more independent (Boone et al. 2007). Moreover,

larger firms are usually more complex and have larger and more diverse boards (Eisenberg,

Sundgren, and Wells 1998; Linck, Netter, and Yang 2008; Yermack 1996). Accordingly, we

use board size instead of firm size to proxy for firm complexity in (3), expecting board

independence to grow with growing board size.

Board independence has been found to decline with increasing CEO ownership and CEO

tenure (Boone et al. 2007; Coles, Daniel, and Naveen 2008; Linck, Netter, and Yang 2008;

Shivdasani and Yermack 1999; Weisbach 1988). However, Adams and Ferreira (2009a) show

theoretically that one should reconsider the evidence that board independence relates

inversely to CEO ownership and CEO tenure. Adams and Ferreira also find empirically that

board independence does indeed correlate non-monotonically with CEO ownership and with

tenure. Specifically, board independence first decreases and then increases as CEO ownership

and tenure grow. This evidence of a v-shaped relationship is interesting because it suggests

that a change in board independence comes with benefits and costs that vary with the pre-

change level of independence (Adams and Ferreira 2007).

We include CEO ownership, CEO tenure, and their squared values in (3). We expect the

coefficient for both proxies to be negative for the linear term and positive for the quadratic

term. Finally, firm complexity may not increase linearly with firm age, and complex firms

may need boards with greater independence (Boone et al. 2007). We account for this possible

non-linearity by including firm age squared, which we expect to have a negative coefficient.

The estimates are shown in Table 11. The two new measures of CEO characteristics and the

new measure of board diversity are all significant at the 5 percent level with the predicted

signs. Also, the role of female directors corresponds to the base case, while listing status is no

longer significant. Although the results as a whole are generally weaker than in the base case,

71

the relationship between gender balance regulation and board independence is insensitive to

the non-regulatory determinants of independence we account for.

Table 11

6. Conclusions

This paper is the first to analyze the empirical relationship between mandatory gender quotas

and director independence in corporate boards. We show that forcing every firm to radically

and quickly change the board’s gender mix produces a strong upwards shift in board

independence. This happens because director independence is a much more common

characteristic among females than among males. We also find that board independence varies

considerably from firm to firm before the gender quota became mandatory. This result

supports the idea of a firm-specific demand for monitoring skills vs. advisory skills in the

boardroom. As predicted, the evidence shows that firms needing monitoring the least and

advice the most are small, young, private, profitable firms with strong owners and low gender

diversity on the board. Such firms are likely to be hurt the most by mandatory gender balance.

A major political argument for mandating that female directors hold at least 40 percent of the

positions in Norwegian boards was that it would improve corporate governance and economic

performance. The idea was to force supposedly irrational owners to choose human talent from

a wider pool, that is, from both female and male candidates. It was also argued that gender

quotas ensure a more inclusive and fair business society that better reflects basic values in

modern society (Langli 2011). Up from 11 percent when the gender balance law was passed

in 2003, females currently hold about 40 percent of the board seats in firms that must fill the

mandatory gender quota.

Our findings suggest that because the fraction of dependent directors has dropped so sharply,

the gender balance law may have weakened boards’ ability to fill their advisory role. Also,

and despite the fact that independent directors have stronger monitoring incentives than

dependent directors do, the law may even have weakened a board’s ability to monitor. Such

regulatory costs of excessive independence are consistent with theoretical predictions and

earlier findings. Overall, the case of mandatory gender balance on corporate boards illustrates

72

how a large regulatory shift can have strong and unintended side effects that were perhaps not

even considered when the law was written.

73

References

Adams, Renée B., and Daniel Ferreira. 2007. "A Theory of Friendly Boards." Journal of Finance no. 62 (1):217–250.

Adams, Renée B., and Daniel Ferreira. 2009a. "Strong Managers, Weak Boards?" CESifo Economic Studies no. 55 (3–4):482–514.

Adams, Renée B., and Daniel Ferreira. 2009b. "Women in the Boardroom and Their Impact on Governance and Performance." Journal of Financial Economics no. 94 (2):291–309.

Adams, Renée B., and Patricia Funk. 2009. "Beyond the Glass Ceiling; Does Gender Matter?" ECGI Finance Working Paper no. No. 273/2010.

Adams, Renée B., Benjamin E. Hermalin, and Michael S. Weisbach. 2010. "The Role of Boards of Directors in Corporate Governance: A Conceptual Framework and Survey." Journal of Economic Literature no. 48 (1):58–107.

Adams, Renée B., and Tom Kirchmaier. 2013. "From Female Labor Force Participation to Board Gender Diversity." ECGI-Finance Working Paper No. 347/2013.

Ahern, Kenneth R., and Amy Dittmar. 2012. "The Changing of the Boards: The Impact on Firm Valuation of Mandated Female Board Representation." Quaterly Journal of Economics no. 127 (1):137–197.

Baysinger, Barry D., and Henry N. Butler. 1985. "Corporate Governance and the Board of Directors: Performance Effects of Changes in Board Composition." Journal of Law, Economics, and Organization no. 1 (1):101–124.

Beecher-Monas, Erica. 2007. "Marrying Diversity and Independence in the Boardroom: Just How Far Have You Come, Baby?" SSRN eLibrary.

Betz, Michael, Lenahan O'Connell, and Jon M. Shepard. 1989. "Gender Differences in Proclivity for Unethical Behavior." Journal of Business Ethics no. 8 (5):321–324.

Bhagat, Sanjai, and Bernard S. Black. 1998. "Independent Directors." The New Palgrave Dictionary of Economics and the Law. no. 2:283–287.

Bhagat, Sanjai, and Bernard S. Black. 2002. "The Non-Correlation Between Board Indpendence and Long-Term Firm Performance." Journal of Corporation Law no. 27 (2):231–273.

Boone, Audra L., Laura C. Field, Jonathan M. Karpoff, and Charu G. Raheja. 2007. "The Determinants of Corporate Board Size and Composition: An Empirical Analysis." Journal of Financial Economics no. 85 (1):66–101.

Bøhren, Øyvind, and R. Øystein Strøm. 2010. "Governance and Politics: Regulating Independence and Diversity in the Board Room." Journal of Business Finance and Accounting no. 37 (9–10):1281–1308.

Bøhren, Øyvind, and Siv Staubo. 2013. "Does Mandatory Gender Balance Work? Changing Organizational Form to Avoid Board Upheaval." Working Paper, Norwegian Business School.

Carter, Colin B., and Jay W. Lorsch. 2004. "Back to the Drawing Board: Designing Corporate Boards for a Complex World." Boston, MA: Harvard Business School Press.

Clikeman, Paul M., Marshall A. Geiger, and Brendan T. O'Connell. 2001. "Student Perceptions of Earnings Management: The Effects of National Origin and Gender." Teaching Business Ethics no. 5 (4):389–410.

Cohen, Jeffrey R., Laurie W. Pant, and David J. Sharp. 2001. "An Examination of Differences in Ethical Decision-Making Between Canadian Business Students and Accounting Professionals." Journal of Business Ethics no. 30 (4):319–336.

74

Cohen, Lauren, Andrea Frazzini, and Christopher Malloy. 2012. "Hiring Cheerleaders: Board Appointments of "Independent" Directors." Management Science no. 58 (6):1039–1058.

Coles, Jeffrey L., Naveen D. Daniel, and Lalitha Naveen. 2008. "Boards: Does One Size Fit All?" Journal of Financial Economics no. 87:329–356.

Duchin, Ran, John G. Matsusaka, and Oguzhan Ozbas. 2010. "When are Outside Directors Effective?" Journal of Financial Economics no. 96 (96):215–237.

Eisenberg, Theodore, Stefan Sundgren, and Martin T. Wells. 1998. "Larger Board Size and Decreasing Firm Value in Small Firms." Journal of Financial Economics no. 48 (1):35–54.

Fama, Eugene F., and Michael C. Jensen. 1983. "Separation of Ownership and Control." Journal of Law & Economics no. 26 (2):301–326.

Farrell, Kathleen A., and Philip L. Hersch. 2005. "Additions to Corporate Boards: The Effect of Gender." Journal of Corporate Finance no. 11 (1/2):85–106.

Fields, M. Andrew, and Phyllis Y. Keys. 2003. "The Emergence of Corporate Governance from Wall St. to Main St.: Outside Directors, Board Diversity, Earnings Management, and Managerial Incentives to Bear Risk." Financial Review no. 38 (1):1–24.

Gul, Ferdinand A., Bin Scrinidhi, and Judy Tsui. 2007. "Do Female Directors Enhance Corporate Board Monitoring? Some Evidence from Earnings Quality." The Hong Kong Polytechnic University.

Hermalin, Benjamin E., and Michael S. Weisbach. 1991. "The Effects of Board Composition and Direct Incentives on Firm Performance." Financial Management no. 20 (4):101–112.

Hermalin, Benjamin E., and Michael S. Weisbach. 1998. "Endogenously Chosen Boards of Directors and their Monitoring of the CEO." American Economic Review no. 88 (1): 96–118.

Hsiao, Cheng. 2003. "Analysis of Panel Data." Cambridge, UK: Cambridge University Press. Hwang, Byong-Hyoun, and Seoyoung Kim. 2009. "It Pays to Have Friends." Journal of

Financial Economics no. 93 (1):138–158. Jensen, Michael C. 1986. "Agency Costs of Free Cash Flow, Corporate Finance, and

Takeovers." American Economic Review no. 76 (2):323–329. Jensen, Michael C., and William H. Meckling. 1976. "Theory of the Firm: managerial

Behavior, Agency Costs and Ownership Structure." Journal of Financial Economics no. 3 (4): 305–360.

Klenke, Karin. 2003. "Gender Influences in Decision-Making Processes in Top Mangement Teams." Management Decision no. 41 (10):1024–1034.

Langli, John C. 2011. "Kjønnskvoterte Styrer." Praktisk Økonomi & Finans (4):67–92. Linck, James S., Jeffry M. Netter, and Tina Yang. 2008. "The Determinants of Board

Structure." Journal of Financial Economics no. 87 (2):308–328. MacAvoy, Paul W., and Ira M. Millstein. 1999. "The Active Board of Directors and its Effect

on Performance of the Large Publicly Traded Corporation." Journal of Applied Corporate Finance no. 11 (4):8–20.

Mason, E. Sharon, and Peter E. Mudrack. 1996. "Gender and Ethical Orientation: A Test of Gender and Occupational Socialization Theories." Journal of Business Ethics no. 15 (6):599–604.

Maug, Ernst. 1997. "Boards of Directors and Capital Structure: Alternative Forms of Capital Restructuring." Journal of Corporate Finance no. 3 (2):113–139.

McKinsey & Company. 2007. "Women Matter: Gender Diversity, A Corporate Performance Driver." McKinsey & Company, Inc.

75

Nygaard, Knut. 2011. "Forced Board Changes: Evidence from Norway." SSRN eLibrary. Powell, Melanie, and David Ansic. 1997. "Gender Differences in Risk Behaviour in Financial

Decision-Making: An Experimental Analysis." Journal of Economic Psychology no. 18 (6):605–628.

Raheja, Charu G. 2005. "Determinants of Board Size and Composition: A Theory of Corporate Boards." Journal of Financial & Quantitative Analysis no. 40 (2):283–306.

Riley, William B., and K. Viktor Chow. 1992. "Asset Allocation and Individual Risk Aversion." Financial Analysts Journal no. 48 (6):32–37.

Shivdasani, Anil, and David Yermack. 1999. "CEO Involvement in the Selection of New Board Members: An Empirical Analysis." Journal of Finance no. 54 (5):1829–1853.

Shleifer, Andrei, and Robert W. Vishny. 1997. "A Survey of Corporate Governance." Journal of Finance no. 52 (2):737–783.

Studenmund, A. H. 2000. Using Econometrics: A Practical Guide. Vol. 4. Boston, MA: Addison Wesley Longman, Inc.

Sundén, Annika E., and Brian J. Surette. 1998. "Gender Differences in the Allocation of Assets in Retirement Savings Plans." American Economic Review no. 88 (2):207–211.

Terjesen, Siri, Ruth Sealy, and Val Singh. 2009. "Women Directors on Corporate Boards: A Review and Research Agenda." Corporate Governace: An International Review no. 17 (3):320–337.

The Guardian. 2012. "Britain's Boardrooms Need More Women, Cameron says." Guardian News and Media Limited (guardian.co.uk).

Trinidad, Cristina, and Anthony H. Normore. 2005. "Leadership and Gender: A Dangerous Liaison?" Leadership and Organization Development Journal no. 26 (7):574–590.

Weisbach, Michael S. 1988. "Outside Directors and CEO Turnover." Journal of Financial Economics no. 20:431–460.

Yermack, David. 1996. "Higher Market Valuation of Companies with a Small Board of Directors." Journal of Financial Economics no. 40 (2):185–211.

76

Figure 1: The fraction of female directors in Norwegian firms exposed to thegender balance law

This figure shows the average ratio of stockholder-elected female directors to all stockholder-elected directors at year end in firms subject to the Norwegian gender balance law. The lawwas passed in December 2003 and was mandatory from January 2008.

0.0

0.1

0.2

0.3

0.4

0.5

2003 2004 2005 2006 2007 2008

Rat

io o

f fem

ale

dire

ctor

s to

all d

irec

tors

Year

77

Table 1: The empirical proxies

Theoretical variable Empirical proxy

Board characteristicsBoard independence The number of outside board members divided by the number of

stockholder-elected board membersOutside director 0/1 dummy variable which is 1 if and only if the board member is either

not a full-time employee in the firm, a former employee, employed by aclosely related firm, related to a member of management, or has nobusiness relationship with the firm

Board size The number of stockholder-elected board membersFemale director number The number of stockholder-elected directors who are womenFemale directors The proportion of stockholder-elected board members who are womenFemale age The average number of years since the female directors were bornMale age The average number of years since the male directors were born

Ownership characteristicsInside concentration The ultimate fraction of equity owned by the firm's officers and directorsOutside concentration The sum of squared ultimate equity fractions in the firm (Herfindahl Largest owner The ultimate equity fraction held by the firm's largest stockholder

General firm characteristicsListed 0/1 dummy variable which is 1 if the firm is public and 0 otherwiseCEO age The number of years since the CEO was bornCEO tenure The number of years since the CEO took officePerformance The average real book return on assets per year over the last three yearsRisk The standard deviation of performance during the last three yearsLeverage Total debt divided by total assetsFirm age The number of years since the firm was foundedFirm size Sales in constant 2009 millions of NOK. Log-transformed in regressions

This table defines the variables used in the empirical analysis.

78

Table 2: Distributional properties of key variables

Variable Mean Std. 0 5 25 50 75 95 100 N

Board characteristicsBoard independence 0.59 0.30 0.00 0.00 0.40 0.60 0.80 1.00 1.00 750Board size 5.69 1.67 1.00 3.00 5.00 5.00 6.00 9.00 12.00 750Female director number 1.51 1.05 0.00 1.00 1.00 2.00 2.00 3.00 5.00 748Female directors 0.24 1.16 0.00 0.00 0.13 0.25 0.38 0.45 0.67 748Female age 45.72 6.20 30.50 35.00 42.00 45.58 49.67 56.00 72.00 630Male age 51.37 5.86 30.50 41.58 47.50 51.50 55.33 61.60 67.50 747Ownership characteristicsInside concentration 0.12 0.21 0.00 0.00 0.00 0.00 0.16 0.59 1.00 750Outside concentration 0.19 0.24 0.00 0.01 0.04 0.09 0.25 0.91 1.00 719Largest owner 0.42 0.31 0.05 0.08 0.15 0.32 0.56 1.00 1.00 620

General firm characteristicsListed 0.64 0.48 0.00 0.00 0.00 1.00 1.00 1.00 1.00 750CEO age 47.09 7.03 30.00 36.00 42.00 47.00 52.00 59.00 72.00 745CEO tenure 5.13 3.87 0.00 1.00 2.00 3.00 6.00 12.00 12.00 511Performance 0.06 0.16 -4.48 -0.22 0.01 0.05 0.12 0.32 0.69 612Risk 0.18 0.47 0.00 0.01 0.03 0.06 0.15 0.39 0.95 630Leverage 0.43 0.30 0.00 0.03 0.22 0.40 0.63 0.91 3.93 630Firm age 25.13 33.44 0.00 1.00 4.00 11.00 29.00 107.80 161.00 743Firm size 2,513 22,256 0 2 33 155 515 4,298 452,370 743Average 696

This table shows distributional properties of the variables used to measure board, ownership, and generalfirm characteristics. Table 1 defines the variables. Performance, Risk, and Leverage are censored at the 1%and 99% tails. The sample consists of all firms exposed to the Norwegian gender balance law from 2003 to2008.

Percentile, %

79

Table 3: Characteristics of listed and non-listed firms

Variable All Listed Non-listed Difference t-value (p-value)

Board characteristicsBoard independence 0.59 0.64 0.50 0.14 5.63 (0.00)Outside director 3.63 3.78 3.36 0.42 9.21 (0.00)Board size 5.69 5.91 4.95 0.96 7.49 (0.00)Female director number 1.51 1.68 1.21 0.47 5.85 (0.00)Female directors 0.24 0.28 0.23 0.05 3.80 (0.00)Female age 45.72 46.22 44.59 1.63 2.87 (0.00)Male age 51.37 52.00 50.25 1.75 3.87 (0.00)

Ownership characteristicsInside concentration 0.12 0.08 0.16 -0.08 -7.99 (0.00)Outside concentration 0.19 0.13 0.32 -0.19 -8.70 (0.00)Largest owner 0.42 0.29 0.63 -0.34 -11.40 (0.00)

General firm characteristicsCEO age 47.09 47.35 46.62 0.73 1.35 (0.18)CEO tenure 5.13 5.26 4.06 1.20 4.18 (0.00)Performance 0.06 0.06 0.07 -0.01 -0.07 (0.95)Risk 0.18 0.11 0.29 -0.18 -3.74 (0.00)Leverage 0.43 0.41 0.46 -0.05 -1.93 (0.06)Firm age 25.13 31.97 12.88 19.09 9.02 (0.00)Firm size 2,513 2,915 2,262 658 3.07 (0.00)Average N 696 418 278 140

This table compares the mean values of board, ownership, and general firm characteristics forlisted and non-listed firms. The differences between mean values across the two groups, their t-values and p-values (in parentheses) are reported in the three right-most columns. Table 1 definesthe variables. Performance, Risk, and Leverage are censored at the 1% and 99% tails. Thesample consists of all firms exposed to the Norwegian gender balance law during the period2003–2008.

80

Table 4: Director types

AverageDirector type 2003 2004 2005 2006 2007 2008 Mean Diff. z p(z) NOutside 0.46 0.52 0.54 0.58 0.64 0.67 0.57 0.21 11.98 (0.00) 427Grey 0.10 0.09 0.09 0.08 0.07 0.06 0.08 -0.04 -3.13 (0.00) 60Inside 0.44 0.40 0.37 0.34 0.29 0.27 0.35 -0.17 -5.16 (0.00) 263N 706 702 757 825 863 645 750 750

Outside Grey InsideYear Listed Non-listed Listed Listed Non-listed2003 0.52 0.41 0.03 0.10 0.45 0.492004 0.56 0.47 0.03 0.09 0.41 0.442005 0.64 0.50 0.02 0.09 0.34 0.412006 0.64 0.52 0.01 0.09 0.33 0.392007 0.71 0.57 0.01 0.08 0.28 0.352008 0.72 0.59 0.00 0.07 0.28 0.34Average 0.63 0.51 0.02 0.09 0.35 0.40Difference 0.12 -0.07 -0.05z-value 15.36 -27.11 -8.12p(z) (0.00) (0.00) (0.00)2008 less 2003 0.20 0.18 -0.03 -0.03 -0.17 -0.15z-value 5.97 4.13 -3.63 -1.23 -5.09 -3.49p(z) (0.00) (0.00) (0.00) (0.22) (0.00) (0.00)

Outside Grey InsideYear Female Male Female Male Female Male2003 0.84 0.49 0.02 0.07 0.14 0.442004 0.83 0.50 0.03 0.08 0.14 0.422005 0.85 0.50 0.02 0.06 0.13 0.442006 0.83 0.48 0.03 0.09 0.14 0.432007 0.83 0.50 0.04 0.08 0.13 0.422008 0.83 0.51 0.04 0.04 0.13 0.43Average 0.84 0.50 0.03 0.07 0.13 0.43Difference 0.34 -0.04 -0.30z-value 30.70 -25.50 -38.40p(z) (0.00) (0.00) (0.00)2008 less 2003 -0.01 0.02 0.02 -0.03 -0.01 -0.01z-value -0.16 0.51 0.63 -0.84 0.00 -0.18p(z) (0.87) (0.61) (0.55) (.040) (0.99) (0.85)

This table reports the average fraction of outside, grey, and inside directors by listing status and gender.Inside directors are the firm’s full-time employees, former employees, or employees of closely related firms.Grey (affiliated) directors are related to a member of management, or are likely to have businessrelationships with the firm. Outside directors are neither inside nor grey. Listed firms are quoted on the OsloStock Exchange. The sample is all firms exposed to the Norwegian gender balance law.

A. All2008 less 2003

Non-listed

B. By listing status

C. By gender

81

Year Mean Median Mean Median Mean Median2003 5.60 5 0.56 0 5.04 52004 5.65 5 0.89 1 4.76 42005 5.77 6 0.88 1 4.89 52006 5.67 5 1.12 1 4.55 42007 5.56 5 1.60 2 3.96 32008 5.86 5 2.25 2 3.61 3Average 5.69 5 1.22 1 4.47 42008 less 2003 0.26 0.00 1.69 2.00 -1.43 -2.00t-value 23.63 0.00 19.59 18.86 -18.06 -15.89(p-value) (0.00) (1.00) (0.00) (0.00) (0.00) (0.00)

Year All Females Males All Females Males2003 5.91 0.61 5.30 5.12 0.51 4.612004 6.11 0.90 5.21 5.11 0.63 4.482005 6.19 1.21 4.98 5.05 0.66 4.392006 6.22 1.53 4.69 4.93 1.04 3.892007 6.12 1.95 4.17 4.70 1.50 3.202008 6.23 2.52 3.71 4.88 1.95 2.93Average 6.13 1.45 4.68 4.97 1.05 3.922008 less 2003 0.32 1.91 -1.59 -0.24 1.44 -1.68t-value 5.83 31.83 -27.33 -4.90 29.39 -34.29(p-value) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Listed less non-listedt-value(p-value) (0.00)

Table 5: Board size

1.17

This table shows board size as measured by the number of shareholder-elected directors. Thesample is all firms exposed to the Norwegian gender balance law. Listed firms are quoted on theOslo Stock Exchange.

A. All firms

Non-listed firms

B. Mean by listing status

MalesFemalesAll

Listed firms

11.99

82

Table 6: Multiple directorships

Directorships 2003 2006 2008 Average 2003 2006 2008 Average1 124 341 417 294 1,134 1,164 788 1,0292 15 33 47 32 290 93 53 1453 1 16 13 10 24 20 10 184 2 5 3 3.3 4 6 1 3.75 2 2 1.3 2 6 2 3.36 2 0.77 1 0.38 1 0.3 1 0.3All 142 397 485 341 1,454 1,291 854 1,199

Directorships 2003 2006 2008 Average 2003 2006 2008 AverageMean 1.16 1.22 1.22 1.20 1.25 1.15 1.10 1.17Median 1 1 1 1 1 1 1 1Maximum 4 5 8 5.7 5 8 5 62008 less 2003 0.06 -0.15t-value 15.50 -20.50(p-value) (0.00) (0.00)N 165 485 592 414 1,812 1,479 938 1,410

Directorships 2003 2006 2008 Average 2003 2006 2008 AverageNumber 18 23 68 36 320 127 66 171Fraction 0.13 0.06 0.14 0.11 0.22 0.10 0.08 0.142008 less 2003 0.01 -0.14z-value 0.41 -8.75(p-value) (0.68) (0.00)N 142 397 485 341 1,454 1,291 854 1,199

MalesFemalesA. Totals

B. Seats per directorFemales Males

This table shows the total number of directorships and the number per director heldby females and males in 2003, 2006, and 2008. Panel A shows the total number ofdirectors with 1 to 8 seats. Panel B shows the number of seats per director, whilepanel C shows the number and fraction of directors who held more than one seat.The sample is all firms exposed to the Norwegian gender balance law during theperiod 2003–2008.

C. Directors with multiple seats

Females Males

83

Table 7: Bivariate correlation coefficients between the determinants of board independence

Female Female directors Inside Outsidedirectors Listed *Listed concentration concentration Performance Leverage Risk Firm age

Listed 0.141Female directors*Listed 0.612 0.733Inside concentration -0.073 -0.156 -0.147Outside concentration -0.099 -0.052 -0.277 0.203Performance 0.143 0.109 0.035 0.032 0.014Leverage -0.006 -0.085 -0.027 0.035 0.054 0.200Risk -0.233 -0.170 -0.168 -0.014 -0.030 -0.077 0.024Firm age 0.086 0.274 0.237 -0.090 -0.122 0.064 0.143 -0.133Firm size 0.122 0.133 0.149 -0.160 0.077 0.260 0.301 -0.171 0.276

This table shows pairwise Pearson correlation coefficients between the hypothesized determinants of board independence as specified in model (1) ofthe main text. Listed is a dummy variable which is 1 if the firm is public and 0 otherwise. Female directors is the proportion of stockholder-electedboard members who are women. Inside concentration is the ultimate fraction of equity owned by the firm's officers and directors. Outsideconcentration is the sum of squared ultimate equity fractions in the firm (Herfindahl index). Performance at time t is the average real book return onassets per year over the last three years. Leverage is total debt divided by total assets. Risk at time t is the standard deviation of performance overthe last three years. Firm age is the number of years since the firm was founded. Firm size is sales in constant 2009 millions of NOK. Performance,Risk, and Leverage are censored at the 1% and 99% tails. The sample is all Norwegian firms that are exposed to the gender balance law during theperiod 2003–2008.

84

Table 8: Estimates of the base-case model

Determinant Prediction Coefficient (p-value)

Female directors (+) 0.308 *** (0.003)

Listed (+) 0.137 *** (0.003)

Female directors*Listed (+/-) -0.185 * (0.077)

Inside concentration (-) -0.002 * (0.056)

Outside concentration (+/-) -0.076 (0.357)

Performance (-) -0.002 *** (0.007)

Leverage (-) -0.137 ** (0.015)

Risk (-) 0.001 (0.196)

Firm age (+) 0.028 *** (0.000)

Firm size (+) 0.026 ** (0.020)Firm fixed effects YesYear fixed effects Yesp-value (F) (0.000)R2 0.300N 429

This table shows the base-case estimates of model (1) in the main text, which specifies boardindependence as a function of the determinants in column 1. The predicted signs of thecoefficients are shown in column 2, column 3 reports the coefficient estimates, and thecorresponding p-values (in parentheses) are in column 4. Female directors is the proportion ofstockholder-elected board members who are women. Listed is a dummy variable which is 1 if thefirm is public and 0 otherwise. Inside concentration is the ultimate fraction of equity owned by thefirm's officers and directors. Outside concentration is the sum of squared ultimate equity fractionsin the firm (Herfindahl index). Performance at time t is the average real book return on assets per year during the last three years. Leverage is total debt divided by total assets. Risk at time t is thestandard deviation of performance during the last three years. Firm age is the number of yearssince the firm was founded. Firm size is the log of sales in constant 2009 millions of NOK.Performance, Risk, and Leverage are censored at the 1% and 99% tails. Statistical significance at the 1%, 5%, and 10% levels is labeled ***, **, and *, respectively. The sample is all firmsexposed to the Norwegian gender law during the period 2003–2008.

85

Table 9: Alternative econometric techniques

Determinant I II III

Female directors 0.308 *** 0.410 *** 0.410 ***

Listed 0.137 *** 0.094 ** 0.098 **

Female directors*Listed -0.185 * -0.115 -0.119

Inside concentration -0.002 * -0.002 * -0.002 *

Outside concentration -0.076 -0.125 * -0.120 *

Performance -0.002 *** -0.001 ** -0.001 **

Leverage -0.137 ** 0.066 -0.068

Risk -0.001 0.001 0.001

Firm age 0.028 *** 0.000 0.000

Firm size 0.026 ** 0.018 ** 0.017 **

Constant 0.403 *** 0.433 ***

Random firm effects No Yes NoFixed firm effects Yes No NoFixed time effects Yes Yes Nop-value (Wald chi2) (0.000) (0.000)p-value (F) (0.000)R2 0.300 0.220 0.170N 429 429 429

This table shows the base-case estimate from table 8 under technique I. Random-effects ML andOLS with sample averages per firm are used in techniques II and III, respectively. The dependentvariable is board independence, which measures the fraction of outside directors. Female directors isthe proportion of stockholder-elected board members who are women. Listed is a dummy variablewhich is 1 if the firm is public and 0 otherwise. Inside concentration is the ultimate fraction of equityowned by the firm's officers and directors. Outside concentration is the sum of squared ultimateequity fractions in the firm (Herfindahl index). Performance at time t is the average real book returnon assets per year during the last three years. Leverage is total debt divided by total assets. Risk attime t is the standard deviation of performance during the last three years. Firm age is the number ofyears since the firm was founded. Firm size is the log of sales in constant 2009 millions of NOK.Performance, Risk, and Leverage are censored at the 1% and 99% tails. Statistical significance at the1%, 5%, and 10% levels is labeled ***, **, and *, respectively. The sample is all firms exposed tothe Norwegian gender law during the period 2003–2008.

Technique

86

Table 10: Alternative proxies for board independence

Definition of board independence

Determinant Prediction

Female directors (+) 0.576 *** 0.571 *** 0.351 ***

Listed (+) 0.192 *** 0.154 ** 0.125 **

Female directors*Listed (-) -0.294 -0.176 -0.092

Inside concentration (-) -0.001 -0.001 0.000

Outside concentration (+/-) 0.002 0.042 0.043

Performance (-) -0.002 *** -0.004 ** -0.004 **

Leverage (-) -0.106 ** -0.096 ** .0.102 **

Risk (-) -0.097 -0.088 -0.092

Firm age (+) 0.017 ** 0.011 * 0.016 **

Firm size (+) 0.026 ** 0.032 ** 0.027 **

Constant 0.505 ***Random firm effects Yes No NoFixed firm effects No Yes YesFixed time effects Yes Yes Yesp-value (Wald chi2)/(F) (0.000) (0.000) (0.000)R2 0.13 0.28 0.24N 332 429 429

This table shows the estimates using alternative measures of board independence. Model A uses a dummyvariable which equals 0 if the CEO is a board member and 1 otherwise. Model B uses the fraction of outsidedirectors minus the fraction of inside directors, while model C uses the fraction of outside and grey directors.The predicted signs of the coefficients are shown in column 2. Female directors is the proportion of stockholder-elected board members who are women. Listed is a dummy variable which is 1 if the firm is public and 0otherwise. Inside concentration is the ultimate fraction of equity owned by the firm's officers and directors.Outside concentration is the sum of squared ultimate equity fractions in the firm (Herfindahl index). Performanceat time t is the average real book return on assets per year during the last three years. Leverage is total debtdivided by total assets. Risk at time t is the standard deviation of performance during the last three years. Firmage is the number of years since the firm was founded. Firm size is the log of sales in constant 2009 millions ofNOK. Performance, Risk, and Leverage are censored at the 1% and 99% tails. Statistical significance at the1%, 5%, and 10% levels is labeled ***, **, and *, respectively. The sample is all firms exposed to theNorwegian gender law during the period 2003–2008.

A: The CEO is not a director

B: Fraction outside less fraction inside directors

C: Fraction outside and grey directors

87

Table 11: Alternative non-regulatory determinants of board independence

Determinant Prediction (p-value)

Female directors (+) 0.038 ** (0.033)

Listed (+) 0.055 (0.483)

Female directors*Listed (+/-) 0.157 (0.487)

CEO ownership (-) -0.009 *** (0.009)

CEO ownership squared (+) 0.000 ** (0.018)

CEO tenure (-) -0.056 ** (0.011)

CEO tenure squared (+) 2.030 ** (0.042)

Firm age (+) 0.000 (0.999)

Firm age squared (-) 0.000 (0.758)

Board size (+) 0.154 ** (0.016)

Fixed firm effects YesFixed time effects Yesp-value (Wald chi2) (0.000)R2 0.190N 407

This table shows the estimates using alternative proxies for non-regulatorydeterminants of board independence compared to those used for the base case inmodel (1) of the main text. Female directors is the proportion of stockholder-elected board members who are women. Listed is a dummy variable which is 1 ifthe firm is public and 0 otherwise. CEO ownership is the ultimate fraction ofequity held by the firm's CEO. CEO tenure is the number of years since the CEOwas hired. Firm age is the number of years since the firm was founded. Firm sizeis the log of sales in constant 2009 millions of NOK. The predicted signs of thecoefficients are shown in column 2, the coefficient estimates are in column 3, andcolumn 4 reports the corresponding p-values (in parentheses). Statisticalsignificance at the 10%, 5%, and 1% levels is labeled ***, **, and *,respectively. The sample is all firms exposed to the Norwegian gender balancelaw during the period 2003–2008.

Coefficient

88

Appendix 1: Classifying directors as inside, grey or outside: Examples

Employee of Related to a Business Board Full-time Former closely member of relation

Name Gender position employee employee related firms management with the firm Inside Grey Outside

Synnøve Finden ASA, 2007Svein Sundsbø Male Chair xRichard Olav Aa Male Director xMimi K. Berdal Female Director xGro Mykling Female Director xPer Arne Eggen Male EmployeeAnne-Mette Hoel Female Director xGeir Dalsegg Male EmployeeLine Rugsveen Female Employee

Eitzen Chemical ASA, 2006Axel C. Eitzen Male Chair x xJohn G. Bernander Male Director xMai-Lill Ibsen Female Director xJames Stove Lorentzen Male Director xAnnette Malm Justad Female Director x x

Norse Energy Corp. ASA, 2007Petter M. Andresen Male Chair xLise H. Langaard Female Director xJoey S. Horn Female Director xJon-Axel Torgersen Male Director x

Fred. Olsen Production ASA, 2008Per-Oscar Lund Male Chair x xSiv Staubo Female Director xAnette Olsen Female Director x xAngar Gravdal Male Director x x

This table illustrates how we classify a director as inside, grey, or outside. Inside directors are defined as the firm's full-time employees, formeremployees, or employees of closely related firms. Grey directors are related to the firm's management or are likely to have business relationshipswith the firm. Outside directors are neither inside nor grey. The data used to manually classify directors are from Brønnøysundregistrene(www.brrg.no) and Proff (www.proff.no). We obtain supplementary information on director characteristics by manually searching the annualreports. We do not classify employee directors.

89

4.

Determinants of board independence in a free contracting environment

Siv Staubo*

August 28, 2013

Abstract

This paper is the first to explore the demand for monitoring and advice on the board by the owners of firms that are not required by regulation to appoint independent directors. Our focus is on the potential conflict between monitoring and advice and on the idea that the relative value of these two board functions varies across firms. We show that well established, small, and profitable firms with concentrated ownership demand advice from dependent directors more than monitoring of their management from independent directors. Similar results are obtained when we investigate the demand for board independence triggered by the potential conflict between large and small stockholders. Unlike earlier research, we find that female directors are just as likely to be advisors as monitors, particularly in family firms. Our results, which are robust to alternative definitions of board independence and alternative econometric techniques, strongly support the idea that optimal board independence is firm-specific.

Keywords: Corporate governance, Regulation, Board independence, Agency problems,

Ownership concentration, Family firm

JEL classification codes: G30, G38

*Department of Financial Economics, BI Norwegian Business School, N-0442 Oslo, Norway. The email address is siv.staubo@bi.no. I am grateful for valuable comments on earlier drafts from Mike Burkart, Øyvind Bøhren, and seminar participants at BI Norwegian Business School.

90

1. Introduction

The directors’ role as monitors of the CEO is widely described in the research literature

(Shleifer and Vishny 1997, Hermalin and Weisbach 2003, Adams, Hermalin, and Weisbach

2010). In contrast, there is hardly any research on the board’s advisory role. The only

theoretical model that addresses both board roles is Adams and Ferreira (2007), who

recognize the potential conflict between monitoring and advice.28 They show that for CEOs

who dislike monitoring and like advice, the information they provide decreases with the

directors’ monitoring intensity and increases with their advice intensity. Since more

independent directors have stronger incentives to monitor, information production decreases

with increasing board independence. Hence, information provision responds endogenously to

board independence. This relationship means that owners who design board face a tradeoff

between the need for monitoring and the need for advice. Moreover, since the relative value

of monitoring and advice may differ from firm to firm, a board’s optimal independence may

be firm-specific.

This paper explores the determinants of board independence, focusing on the potential

conflict between monitoring and advice. The first premise is the theoretical finding by Adams

and Ferreira (2007) that optimal board independence is firm-specific. The second premise is

the empirical fact that listed firms in most countries cannot freely choose the independence

level of their board because of regulation (ECGI 2012, Sarbane-Oxely 2002). These two

premises suggest that a regulatory floor on board independence is costly for firms as a whole,

and particularly costly for firms when the optimal degree of board independence is

considerably below the mandated one. To ensure that board independence is determined in a

free contracting environment, we analyze non-listed (private) firms. Unlike listed (public)

firms, non-listed firms are not subject to regulation of their board independence. We show

empirically that board independence varies systematically with a series of observable firm

characteristics.

It is a common belief among policy makers and regulators that independent directors reduce

the incidence of corporate scandals. In fact, policy makers and regulators often assume that

good governance more generally requires independent boards (Bhagat and Black 2002, Coles, 28 Other theoretical models also emphasize that directors’ need information in order to be active decision-makers (Raheja 2005, Song and Thakor 2006, Harris and Raviv 2008). However, Adams and Ferreira (2007) provide the first model to address the conflict between information sharing by the CEO and monitoring intensity by the directors.

91

Naveen, and Naveen 2008, Duchin, Matsusaka, and Ozbas 2010). For example, corporate

governance reforms in most countries put particular emphasis on the value of board

independence:

‘The role of independent non-executive directors features prominently in corporate governance codes. The presence of independent representatives on the board, capable of challenging the decisions of the management, is widely considered a means of protecting the interests of stockholders and, where appropriate, other stakeholders.’ European Commission's Recommendation (provisional text), October 6, 2004.

‘The common hallmark of corporate governance reforms proposed by the NYSE,3 NASDAQ4 and AMEX5 is the emphasis placed on a board of directors having the capacity to exercise independent judgment while performing its responsibilities. For example, the NYSE Corporate Accountability and Listing Standards Committee, convened in 2002 to recommend ways to enhance the accountability, integrity and transparency of NYSE-listed companies, stated its belief that having a majority of independent directors would increase the quality of board oversight and lessen the possibility of damaging conflicts of interest.' PriceWaterHouseCoopers: The Sarbanes Oxley Act 2002.

Existing research provides no robust support for these opinions. That is, there is neither

convincing theory nor convincing empirical evidence showing that more board independence

unconditionally improves firm value (Adams, Hermalin, and Weisbach 2010). The reason is

that more board independence brings both benefits and costs, that the costs may dominate the

benefits, and that this relationship between costs and benefits is not constant across firms.

Two recent empirical studies build on the Adams and Ferreira (2007) model, analyzing if

board characteristics like size and independence vary with the firm’s demand for director

expertise on strategic issues (Coles, Naveen, and Naveen 2008, Linck, Netter, and Yang

2008). Both studies find that complex firms have larger boards with more independent

directors, whereas growth firms have smaller boards and a larger demand for advice by

dependent directors. These findings support the idea that board independence is firm-specific.

However, these studies analyze listed firms in the United States, which are required by law to

appoint a majority of independent board members.29 Although the Sarbanes-Oxley Act was

not introduced until 2002, board composition in the United States was influenced by

compliance requirements on board independence introduced many years earlier.30 Similarly,

29 Coles, Daniel, and Naveen (2008) analyze public firms in the United States from 1992 to 2001. Linck, Netter, and Yang (2008) analyze US public firms from 1990 to 2004. 30 The Sarbanes-Oxley Act (SOX) is a federal law that set new or enhanced standards for all boards in public firms. Prior to SOX, board independence was not mandated, but the SEC encouraged firms to increase board

92

European stock exchanges recommend by soft law that listed firms compose boards with at

least 50 percent independent directors. 31 Thus, empirical studies of such regulatory

environments investigate determinants of board independence in firms that are required by

law or recommended by code to appoint a majority of independent directors. That is, the firm

cannot set its optimal board independence level unless this level is at or above the regulatory

minimum, i.e., unless the regulation is not binding relative to the first-best optimum. Since the

firms analyzed in the existing literature are regulated to appoint a majority of independent

directors, this literature may not reveal the relationship between firm characteristics and board

independence that would be observed in a free contracting environment.

Our study avoids this problem. We use a dataset consisting of 16,100 non-listed Norwegian

firms from 2000 to 2011. Board independence is not regulated in these firms in any part of the

sample period. There is a large variation in firm characteristics across the sample. This

heterogeneity produces large cross-sectional variations in the determinants of optimal board

independence. Hence, unlike earlier papers, we explore the determinants of board

independence in firms where stockholders are free to construct the optimal board from a

monitoring and advice perspective. The average fraction of independent directors is close to

30% in the sample period, compared to more than 50% in listed firms, where independence is

regulated.

We make four contributions to the literature. First, unlike existing research, we investigate the

trade-off between monitoring and advice in a setting where board independence is not

regulated. Second, we show that board independence is more important to some firms than to

others. This result suggests that one-size-fits-all regulation is costly.

Our third contribution is to explore the board’s monitoring role not just relative to the CEO,

but also relative to potential conflicts between large and small stockholders. The first of these

two monitoring functions is to reduce the principal-agent problem that arises when managers

exploit their control rights at the stockholders’ expense (Jensen and Meckling 1976). This

situation is called the first agency problem (Agency problem 1) in the literature, and directors

who are independent of management are supposed to be better at reducing this problem

(Villalonga and Amit 2006). The existing literature on board independence focuses independence. For example, Linck, Netter, and Yang (2008) notice that Harold Williams, the SEC chairman from 1971 to 1981, placed significant pressure on NYSE firms to have a majority of outside directors on their boards. 31 A soft law is based on the principle of comply or explain.

93

exclusively on the first agency problem. The board’s second monitoring function is to oversee

the conflict between majority and minority stockholders, which has been called the second

agency problem (Agency problem 2) in the literature (Villalonga and Amit 2006). Directors

who are independent of the main stockholder are supposed to be better at protecting the rights

of minority stockholders.32

We measure board independence relative to the first monitoring function (Agency problem 1)

in terms of the arm’s-length distance between directors and management. We use standard

empirical proxies from existing papers (Carter and Lorsch 2004, Bøhren and Strøm 2010).

Moreover, we establish a new empirical measure related to the monitoring function caused by

the second agency problem. As far as we are aware, we are the first to study the second

monitoring function in a board independence setting.

There is large variation in ownership concentration in our dataset. This fact allows us to

separate Agency problems 1 and 2. Agency problem 1 is expected to be most prominent when

ownership is dispersed, whereas Agency problem 2 is more likely to occur when ownership is

concentrated. Therefore, we divide our data set into two subsamples according to ownership

concentration in order to improve insight into firm-specific demand for board independence.

In the first subsample, the firms have no controlling owner. Hence, the firms in this

subsample are primarily exposed to the first agency problem. The firms in the second

subsample are controlled by one owner. Because these firms are majority owned, there is a

potential conflict between majority and minority stockholders, which creates the second

agency problem. The conflict between owners and management is expected to decrease as the

equity stake of the largest owner increases. That is because the largest owner has strong

power and incentives to control the management. Hence, the first agency problem may not be

as prominent in the second subsample as in the first. We exclude single-owner firms since

these firms have no majority-minority problem and seldom face a separation between

ownership and control.

Our forth contribution is to investigate board independence in family firms. The relationship

between board independence and information provision as addressed by Adams and Ferreira

(2007) needs specific attention in these firms. The owners of family firms are often managers 32 The Norwegian Corporate Governance Codes explicitly address these two dimensions of independence: ”The majority of the stockholder-elected members of the board should be independent of the company’s executive management and material business contracts” and ”At least two of the members of the board elected by stockholders should be independent of the company’s main stockholder(s)” (NUES 2012).

94

and directors, and they are often particularly well informed about the business. The existing

literature suggests that family firms either mitigate or exacerbate agency problems. Anderson

and Reeb (2003b) and Andre and Ben-Amar (2006) argue that families are better monitors of

management than other stockholders because owners and managers are better aligned in

family firms than in non-family firms. However, expropriation of minority stockholders’

wealth by the controlling family is probably the most critical agency cost in family firms

(Claessens et al. 2002, Masulis, Pham, and Zein 2011). Our dataset contains a comprehensive

set of family characteristics, and 82% of the firms in our sample are controlled by a family.33

Our results show that the optimal level of board independence is lower than 50% for more

than half the firms in this study. That is, when we investigate a sample of non-listed firms, we

find that recommending a majority of independent directors to all firms will hurt firms in

which stockholder value advice by the board of directors more than monitoring. In particular,

we show this is the case for well established, small, and profitable firms with concentrated

ownership. Unlike earlier research the relationship between firm age and board independence

is negative and the relationship between female directors and board independent is not

significant in the study. We find similar results when we explore the determinants of board

independence related to the second agency problem. Since we do not know any other studies

that analyses the determinants of board independence related to independence between board

members and the largest owner, we believe that these results are new in literature. Finally, we

find that family firms have less independent boards than non-family firms.

Section 2 in the following reviews the literature and specifies the predictions. The data and

summary statistics are discussed in Section 3, while Section 4 explains the methodology and

presents the results. We estimate the base-case model by alternative econometric techniques

in section 5, where we also use an alternative proxy for board independence as well as

alternative determinants of board independence. Section 6 summarizes and concludes.

2. Theory and predictions

Even though our paper is based on a theoretical model of board composition (Adams and

Ferreira 2007), there is not much theory on this topic compared to the empirics (Hermalin and 33 We define a firm as a family firm if the largest family by ultimate ownership holds more than 50% of the firm’s equity. A family is a group of individuals related by blood or marriage.

95

Weisbach 2003). Hermalin and Weisbach (2003) divide the empirical research on the board of

directors into three groups. The first group studies the impact of board characteristics on

board behavior, such as the relationship between board independence and CEO replacement.

The second group studies whether boards’ actions influence firm performance. Our study

belongs to the third group, which analyzes the determinants of board composition. For

instance, Weisbach (1988), Denis and Sarin (1999) and Hermalin and Weisbach (2003) find

that small, young, and closely held firms tend to have less independent boards. In contrast,

larger, older, and dispersedly owned firms are more likely to have more independent boards.

We want to investigate the impact of these and other determinants on board independence

used in earlier research (Linck, Netter, and Yang 2008, Adams 2009). Unlike earlier studies,

however, our sample firms can choose the level of board independence without concerns for

regulatory restrictions.

Our base-case model is the following:

1 2

3 4 5

6 7 8

+ +

it it it

it it it

it it it

Independence CEO ownership PerformanceLeverage Female directors GrowthInformation costs Firm age Firm size

α β ββ β ββ β β

= + ++ +

+ + + itu (1)

The empirical variables are defined in Table 1.

Table 1

Most of the explanatory determinants used in empirical governance studies are endogenous.

Section 5 addresses endogeneity problems in model (1).

We define measures of board independence in section 2.1. Section 2.1.1 defines the measure

for board independence related to Agency problem 1, whereas section 2.1.2 defines measure

for board independence related to Agency problem 2. We predict how board independence

relates to its determinants in section 2.2.

96

2.1 Measures of board independence

2.1.1 Independence – Agency problem 1 (Independence-A1)

There are several definitions of board independence in the corporate governance literature.

The following definition is from Wikipedia:

An Independent Director (also sometimes known as an outside director or non-executive director) is a director (member) of a board of directors who does not have a material or pecuniary relationship with the company or related persons, except sitting fees. Independent Directors do not own shares in the company.

Earlier research measures board independence in different ways. One of the most common

measures is the fraction of outside directors on the firm’s board (Baysinger and Butler 1985,

Weisbach 1988, MacAvoy and Millstein 1999, Adams and Ferreira 2009b).34 In this context

board independence refers to independence between directors and management. A board is

more independent the higher the fraction of outside directors. An independent board is

assumed to be better at reducing the first agency problem.

We do not have the data to measure the fraction of outside directors. Instead, we follow

earlier research by using a different measure. In the base-case model we use a proxy that

identifies whether or not the CEO is a board member. That is, although the CEO is always

present in the board meetings, this proxy measures whether or not the CEO has voting rights.

A board is expected to be more independent if the CEO has no voting right because he is then

unable to interfere formally with the directors who monitor him (Carter and Lorsch 2004).

We measure board independence by a dummy variable which is 1 if the CEO is not a director

and 0 otherwise.

34 To find the fraction of outside directors, earlier research classifies directors as inside, grey or outside. Inside directors are defined as the firm’s full-time employees, former employees, or employees of closely related firms. Grey (affiliated) directors are related to a member of management, or are likely to have business relationships with the firm, such as investment bankers and lawyers. Outside directors are neither inside nor grey. Moreover, a board is more independent the higher the fraction of outside directors.

(a)1 if the CEO is not a board member

0 if the CEO is a board memberIndependence A1 =

97

A board is more independent if (a) is 1. We use an alternative proxy for Independence - A1

under the robustness checks in Section 5.

2.1.2 Independence – Agency problem 2 (Independence - A2)

The board’s second monitoring function is to protect the minority stockholders. To address

the second agency problem, we are concerned with independence between the board and the

large stockholders. A board is more independent the larger the fraction of directors who are

independent of the largest owner. Such a board is supposedly better at protecting minority

stockholders. We have developed a new measure to capture independence between the

directors and the largest stockholder, using the fraction of directors that do not belong to the

largest family by ultimate ownership.

the number of board members - the number of board members from the largest familyIndependence - A2 = the number of board members

(b)

Independence – A2 measures the fraction of board members not connected to the largest

owner. This measure is justified by the fact that 82% of the firms in our data set have a family

as the largest owner.

The values of (b) will be in the interval [ ]0,1 . A board is assumed to be more independent the

closer (b) is to 1.

2.2 Determinants of board independence

We discuss the determinants of board independence according to the two main roles of the

board, which are monitoring (2.2.1) and advice (2.2.2).

The determinants of board independence described in this section are determinants of board

independence related to the first agency problem. These determinants are used in earlier

research. The existing research on board independence is concerned about independence

between board members and management. When we test board independence relative to the

second agency problem, we use a new empirical measure the captures independence between

board members and the largest owner as the left-hand side variable. However, we use the

same determinants of board independence as we do when we analyze board independence

98

related to Agency problem 1. That is, without support from either theoretical or empirical

research, we test if the same determinants influence board independence between directors

and managers and board independence between directors and the largest owner.

2.2.1 Monitoring

Earlier studies show empirically that board independence declines when the equity holdings

of the CEO increases (Bhagat and Black 1998, Linck, Netter, and Yang 2008). The theoretical

argument is that the need for monitoring is smaller when the CEO is aligned with

stockholders (Morck, Shleifer, and Vishny 1989). We expect CEO ownership to be negatively

related to board independence (H1).

Firms tend to appoint more independent directors after years of poor performance (Hermalin

and Weisbach 1991, Bhagat and Black 2002). We measure performance as the average return

on assets over the last three years, and we expect performance to be negatively correlated to

board independence (H2).

Firms with high debt have less free cash flow. Hence, high debt is a substitute for monitoring

effort by owners and directors (Jensen 1986). We measure debt by leverage, which is total

debt divided by total assets. This argument suggests an inverse relationship between leverage

and board independence. However, using leverage to proxy for firm complexity, Linck, Netter,

and Yang (2008) find that complex firms have more independent boards. Hence, they find

that leverage is positively related to board independence. Due to this contradictory evidence,

the relationship between leverage and board independence is unspecified (H3).

Adams and Ferreira (2009b) argue it is a general belief that diverse boards are more

independent. This view is supported by Beecher-Monas (2007) and Fields and Keys (2003),

who suggest that independence is easier to achieve by increasing ethnic and gender diversity

in the board room. We use the fraction of female directors to proxy for board diversity,

expecting that the fraction of female directors is positively related to board independence

(H4).

2.1.2 Advice

Jensen (1993) argues that it is more costly for large boards to monitor growth firms. Along

the same lines, Linck, Netter, and Yang (2008) find that both board size and board

99

independence decrease with the costs of monitoring. We use sales growth to proxy for growth,

and expect growth to be negatively related to board independence (H5).

Theoretical research by Adams and Ferreira (2007) show that the optimal fraction of

independent directors decreases as the cost of monitoring increases. Such monitoring costs are

particularly high when firms with strong information asymmetry are monitored by

independent directors (Maug 1997). Moreover, empirical research has found that the

information asymmetry is higher the more volatile the firm’s stock returns (Fama and Jensen

1983). Linck, Netter, and Yang (2008) use the standard deviation of stock returns to proxy for

information costs. Similarly, we use the standard deviation of the book return on assets to

proxy for information costs, and expect information costs to be negatively related to board

independence (H6). 35

Earlier research finds that complex firms have more independent boards (Lehn, Patro, and

Zhao 2003, Boone et al. 2007, Linck, Netter, and Yang 2008). Lehn, Patro, and Zhao (2003)

and Boone et al. (2007) use firm age and firm size to proxy for firm complexity. We follow

this earlier work and expect firm age and firm size to be positively correlated with board

independence (H7 & H8).

We extend the base-case model by including family control as a determinant for board

independence. A firm is defined as being a family firm (measured by ‘Family control’) if one

family owns more than 50% of the equity. Family firms differ from other firms in ways that

influence their governance structure (Anderson and Reeb 2003). Family members often hold

the CEO and chair positions. Since the families are in the business for a long time, they are

also often large and committed owners. These characteristics suggest that there are more often

strong ties between management and owners in family firms than in other firms. Therefore,

monitoring of Agency problem 1 is less important. On the other hand, earlier research shows

that the common succession of management positions in family firms may result in

managerial entrenchment, and hence higher agency costs (Gomez-Mejia, Nufiez-Nickel, and

Gutierrez 2001, Anderson and Reeb 2004). We do not pre-specify the relationship between

family control and board independence (H9).

35 We do not have data on stock returns since the firms in our sample are non-listed firms.

100

Summarizing section 2, we predict that the board will be more independent when the firm is

large, old, has many female directors, has low performance, growth, and information costs,

and when CEO ownership is low.

3. Data, sample, and summary statistics

Our sample is non-listed Norwegian firms with limited liability from 2000 to 2011. The data

source is the CCGR database (www.bi.edu/ccgr).36 The data on family relationships is from

the tax authorities. The law mandates a standardized set of full accounting statements certified

by a public auditor regardless of the firm’s listing status, size, and industry. They must also

report the identity of the CEO, directors and owners, as well as these individuals’ equity

holdings in the firm.

There are on average about 200,000 non-listed firms in the population. The vast majority of

the firms are of no interest to this study because they have no activity. We remove these firms

by the filter revenues ≥ 2,000,000 NOK (approximately 260,000 EUR). Next, because listed

firms in Norway are required to have at least three board members, we require a Board size ≥

3. Finally, we exclude firms in which the largest owner holds more than 90% of the equity.37

We are left with 16,100 firms on average each year from 2000 to 2011. Table 2 summarizes

the key properties of the frequency distributions for each of the determinants of board

independence used in this study.

Table 2

According to Table 2, board independence, measured with a dummy variable with a value of

1 if the CEO is not a board member and 0 otherwise is 48%. This is board independence

related to Agency problem 1. One must not be misleading to interpret from this value that

there are 48% independent directors on average in our sample firms. The value of this proxy

for board independence tells that the CEO is not a board member in 48% of the sample firms.

Listed firms are not included in this study. Nonetheless, we compare the values of the

variables in our sample with values of these variables in a sample of listed firms during the

same time period. It is important to notify the differences and similarities between listed and

36 The database includes every limited liability firm registered in Norway from 1994 to present. 37 We use 90% rather than 100% as the upper threshold because 90% is where the corporate law allows the majority to buy out the minority. Furthermore, if the majority owns more than 90%, the minority stockholders may demand the majority to buy them out. See Bøhren and Krosvik (2013) for details.

101

non-listed firm in order to compare our results to existing research and to justify that our

findings are important to all firms, see appendix 1. The CEO has no voting rights in 92% of

listed firms. However, this value is misleading because it exaggerates board independence

because the CEO was not allowed to be a board member in listed firms after 2007. However,

we can avoid this problem by comparing the values for listed and non-listed firms by the

measure we use for board independence for robustness test in Section 5. We find that board

independence is about 25 percentage points higher in listed than in non-listed firms on

average.

Similarly, we find that board independence related to Agency problem 2 is roughly 30

percentage points higher in listed than in non-listed firms.

For a better understanding of the difference in governance problems across firms, the base-

case sample is divided into subsamples. The first subsample is firms with and without a

controlling owner, whereas the second subsample consists of family firms and non-family

firms. The mean values of the variables in the two groups within each subsample are

compared by t-statistics in Table 3, panel A and panel B, respectively.

Table 3

Board independence related to Agency problem 1 is 50% in firms with majority owner and 46%

in firms with no-majority owner. On the other hand, board independence related to Agency

problem 2 is 54% in firms with majority owner and 60% in firms with no-majority owner.

CEO ownership is much greater in majority-held firms than in firms without a majority owner.

The difference in ownership structure in the two subsamples indicates, in accordance with

previous research and our prediction, that boards are more independent in firms with no-

majority owner. In contrast, our predictions are not supported by the findings that firm size

and the fraction of female directors are smaller in firms with no-majority owner. It is

important to point out that these results have to do with the average firm in each subsample,

and that the analysis is univariate. The t-statistics support the notion that board independence

is firm-specific.

The second subsample is family firms and non-family firms. We that 82% of the firms are

family firms. Board independence measured related to Agency problem 1 is 39% in firm with

102

family firms and 88% in non-family firms. The average value of board independence related

to Agency problem 2 is 54% in family firms and 77% in non-family firms. Hence, the

univariate analysis shows that boards in family firms are by far much more independent than

firms in non-family firms. These results can be driven by the large difference in sample size

of family firms and non-family firms and the average values may be strongly influenced by

some firms in the sample of non-family firms. However, family firms perform better, have

higher leverage; are smaller and younger than non-family firms on average.

Table 4 shows bivariate correlation coefficients of the determinants for board independence

included in our base-case model (1). Mulitcollinearity should not be a serious problem,

because the correlation coefficients are far away from the critical limit of 0.8 (Studenmund

2000).

Table 4

4. Research design, methodology, and estimation results

We analyze the estimates of the base-case model according to the two main roles of the board

of directors, which are monitoring and advice. The monitoring role is divided into monitoring

of Agency problem 1 and monitoring of Agency problem 2.

The first estimates of the base-case model (1) are estimates of board independence related to

Agency problem 1. The dependent variable equals 1 if the CEO is not a board member and 0

otherwise. The model is estimated with a logit model and uses GLM. Table 5 shows the

estimates.

Table 5

The estimates for CEO ownership and performance are consistent with our predictions from

H1 and H2. The demand for monitoring carried out by independent directors is not as

prominent when CEO ownership is high because the interests of stockholders and the CEO

are well aligned (H1). When firms perform well, stockholders tend to value advice from the

board of director’s more than monitoring (H2).

103

We did not specify a predicted sign for H3 due to conflicting theoretical arguments for the

impact of leverage on board independence in earlier research. We find that leverage is

negatively related to board independence in the base-case model, indicating that leverage is a

substitute for monitoring rather than a proxy for complex firms. That is, high leverage reduces

the free cash flow and thereby the principal-agent conflict. Hence monitoring by independent

directors is not prominent when leverage is high.

Earlier research suggests that female directors are more likely to be independent directors. A

recent study confirms this presumption, showing a high correlation between board

independence and female directors when boards are subject to gender quotas (Bøhren and

Staubo 2013). The negative and significant coefficient for the fraction of female directors on

board independence in table 5 is new in this literature, and contradicts to H4. One possible

interpretation of this result is that female directors are more often appointed to fill the

advisory role than the monitoring role in this sample of firms where board independence is

not regulated. In addition, unreported statistics show that the distribution of female directors

in our sample is different from the one in listed firms. About half the non-listed firms have no

female directors. The negative relationship between female directors and board independence

is confirmed when the base-case model is estimated in a sample consisting of firms with at

least one female director. Hence, female directors tend to be appointed more often as advisers

rather than monitors.

Growth and information costs are positively related to board independence, but the estimates

are not significant. These results are inconsistent with H5 and H6.

Firm age is negatively related to board independence, indicating that well established firms

demand more advice than monitoring. This is inconsistent with earlier research and our

hypothesis H7. One reason may be that the firms in our sample on average younger than firms

used in earlier research on listed firms (12 years compared to 34 years, 95% of the firms in

our sample are younger than 34 years), see appendix 1. We do not have any theoretical

argument to support this result.

Firm size relates positively to board independence. This result is consistent with H8.

104

Summarizing so far, estimates of the base-case model suggest that board independence related

to Agency problem 1 decreases in CEO ownership, performance, leverage, female directors,

and firm age. Board independence increases in firm size.

Next, we investigate the determinants of board independence relative to the second agency

problem. The dependent variable is a proxy for independence between board members and the

majority owner. This proxy measures the fraction of non-family directors on the firm’s board.

The base-case model is estimated by a random effects panel data regression because the left

hand side variable is a continuous variable. Table 6 shows the estimates.

Table 6

Board independence related to Agency problem 2 decreases in CEO ownership, performance,

female directors, and firm age. Board independence increases in firm size. The results are

similar to the results for the estimates of the determinants of board independence related to

the first agency problem shown in Table 5. The only difference is that leverage is only a

significant determinant for board independence related to the second agency problem at the 10%

level. That is, leverage may not be a substitute for monitoring related to the conflict between

majority and minority stockholders.

Next, we expand the base-case model and include family control as a determinant of board

independence. We define family firms by the variable Family control, which is 1 if the

ultimate family ownership > 50%, and zero otherwise. This criterion classifies 82% of the

sample firms as family firms. We did not pre-specify the relationship between board

independence and family control in the H9.

The estimate of this variable in tables 5 and 6 (rightmost column) shows a negative and

significant impact of family control on board independence. This result suggests that owners

and management are strongly aligned in family firms such that monitoring by independent

directors is less important. Additionally, the family control variable has similar impact on

board independence related to the second agency problem.

Finally, the base-case model is re-estimated in two sets of subsamples. Table 7 shows the

estimates.

Table 7

105

The first set of subsamples is firms with majority owner and firms with no-majority owner.

The results, which are reported in Table 7, Panel A, are consistent with the results from the

full sample. Hence, the same determinants influence board independence related to both

agency problems in the two subsamples.

The second set of subsamples is family firms and non-family firms. The results, which are

reported in Table 7, Panel B, are close to those in the full sample. The only exception is the

fraction of female directors in non-family firms which is positively related to board

independence related to Agency problem 2.

Hence, the inverse relationship between female directors and board independence, which is

new in literature, primarily occurs in family firms. Except for this, the results regarding the

determinants of board independence in subsamples are consistent with those in the full sample.

5. Robustness

We first re-estimate the base-case model in (1) using alternative econometric techniques.

Subsequently, we analyze what happens when we use an alternative proxy of board

independence. Finally, we test the sensitivity to using non-linear empirical proxies in addition

to those used in (1) as determinants of board independence.

5.1 Alternative econometric techniques

First, the base-case model is estimated for board independence related to Agency problem 1.

In addition to the logit regression (the base-case), we apply probit, a standard panel data

method with random effects and with board independence considered as a continuous variable,

a logit panel data method with random effects, and with board independence considered as a

logistically distributed variable. Moreover, we apply pooled OLS with standard errors

adjusted for clustering at the firm level which treats board independence as a continuous

variable. Finally, we estimate model (1) using instrumental variable technique. The results are

reported in Table 8.

Table 8

The results are insensitive to the alternative econometric techniques except for the fraction of

female directors, which is significant only in the logit and probit regressions. Panel data

106

regression and instrumental variable technique control for the potential endogeneity problems

in the base-case model. Endogeneity problems are discussed in more details below. We

believe that estimates of the fraction of female directors in panel data regression and the IV

regression are more reliable than the estimates from the logit and probit models since these

latter models do not control for endogeneity problems. This assumption is confirmed when we

analyze the estimates of the model for board independence related to Agency problem 2. In

addition to standard panel data method with random effects (the base-case), we also apply

standard panel data method with fixed effects, pooled OLS with standard errors adjusted for

clustering at the firm level. 38 Finally, we estimate the model using instrumental variable

technique. The results are reported in Table 9.

Table 9

The results are insensitive to the alternative econometric techniques, except for the fraction of

female directors that is insignificant in the regressions when we apply fixed effects panel data

method and instrumental variable technique. This result suggests that the base-case model is

exposed to endogeneity problems and the fraction of female directors is not related to board

independence. Moreover, females are just as likely to be appointed as advising directors as

monitoring directors.

38 The data set has a panel structure of N x T, where N is the number of firms per year and T is the number of years. There is on average about 16,100 firms per year, which are observed at least once over 11 years. General firm, ownership, ownership, and family characteristics are observed up to 12 times for the same firm. Some firms are observed less than 12 times because they enter the sample after 2000, or left the sample before 2011. Hence, the panel is unbalanced. Compared to studying the cross-section at a given point in time, panel data may improve estimation quality by increasing the number of data points and reducing multicollinearity. Moreover, fixed effects and random effects techniques reduce misspecification problems caused by unobservable determinants of board independence (Hsiao 2003). This property can be seen by separating the error term into one component which is time-invariant and firm-specific, one which is time-variant and firm-independent, and a third which is idiosyncratic and varies within and between firms. In such a context, the model can be specified as:

it it i t itY X uα β δ ψ= + + + + , Yit is the observed board independence measure for firm i at time t, itX is

the vector of observable, time-varying determinants of board independence for i at t, iδ is the unobserved, time-

independent effect of firm i on Yit, ψt is the unobserved, time-varying effect on all firms at time t, and itu is the

idiosyncratic error term. In our case, iδ represents any firm-specific need for monitoring and advice that is not reflected in the firm characteristics we explicitly account for through Xit.

The two alternative ways to account for

iδ are the fixed effect and the random effects approach, respectively. The fixed effects model allows for the unobserved variables to be correlated with the error term, while the random effects model assumes the two are independent and treats the unobservable firm effect as a random variable.

107

As mentioned above, we are aware of the possible endogeneity problems in the regression

models. In fact, we believe that the inconsistent results of the estimates for the fraction of

female directors may be caused by such problems. The problems occur because there may be

causality running from the left hand side variable to some of the explanatory variables. One

way to reduce this problem is to lag the variables we are concerned about. First, consider the

causality between board independence and performance. Firms may tend to appoint more

independent directors after years of poor performance. That is, board independence may

increase if the firm performed bad last year. Therefore, we refer to last year’s performance

because there will always be a time lag between recognition of bad performance and the

opportunity to change the board. Usually, there is one ordinary general meeting a year and a

change in the board has to be accepted by the general meeting. Moreover, to avoid that the

temporary fluctuations in the performance of the firm have any impact on board independence,

we measure performance as average performance over the three last years. By doing so, we

prevent that board independence and performance is contemporaneous.

Second, we are concerned about possible reverse causality between board independence and

CEO ownership. In particular, a firm with a powerful CEO needs advice from the directors

more than monitoring from independent directors. Hence, when CEO’s ownership is large, he

is assumed to be powerful and there is less need for monitoring by the board of directors. To

reduce this causality problem, we study the impact of CEO ownership on board independence

using an instrumental variable (IV). That is, we predict the variation in CEO ownership using

CEO age, which is randomly distributed relative to board independence but not relative to

CEO ownership. The CEO age is limited to an age interval from 17 to 93 years, with an

average close to 47 years and a standard deviation close to 10 years. We define a CEO as

young if his age is lower than the average age of CEOs in our sample, 47 years. A CEO is old

if he is 47 years or older. CEO ownership increases in CEO age, see appendix 2, panel A. We

show that that board independence is consistent for young and old CEOs in appendix 2, panel

B. That is, CEO age is correlated with CEO ownership, but it is uncorrelated with board

independence, which makes CEO age a valid instrument for CEO ownership. Furthermore,

appendix 2, panel B shows that other differences in firm and board characteristics for young

and old CEOs are small compared to the difference in CEO ownership between young and old

CEOs. That is, we find no evidence that firm and board characteristics, other than CEO

ownership, vary as a function of our instrument.

108

5.2 Alternative proxy for board independence

The base-case model is estimated using an alternative proxy for board independence related to

Agency problem 1. This proxy is developed by Bøhren and Strøm (2010) based on the

Hermalin and Weisbach (1998) model where board independence depends on the time of

entry of the directors relative to the entry of the CEO. The theory predicts that a CEO who

runs a firm that performs well is able to recruit dependent directors. Thus, the level of board

independence may depend on whether the directors were appointed before or after the CEO

took office. Board independence in firm i is defined as the difference between the average

tenure of the board’s non-CEO directors and the tenure of the CEO:

1

1Independence - A1 non-CEO director tenure CEO tenuren

n

i ij ij=

≡ −∑ (c)

Non-CEO director tenureij is the number of years since non-CEO director j entered office in

firm i and n is the number of stockholder-elected directors. The board is more independent the

higher the value of (c).

We estimate this model, where the left-hand side variable is continuous, using random effects

regression. Table 10 shows the estimates.

Table 10

The estimates are consistent with the base-case model except for female directors which has

positive but not significant impact on board independence.

5.3 Non-linear determinants of board independence

As mentioned in section 2, earlier research finds that board independence declines in CEO

ownership (Boone et al. 2007, Coles, Naveen, and Naveen 2008, Linck, Netter, and Yang

2008, Shivdasani and Yermack 1999, Weisbach 1988). However, Adams and Ferreira (2009a)

show theoretically that the evidence on an inverse relationship between board independence

and CEO ownership should be reconsidered. The authors also find empirically that board

independence does indeed correlate non-monotonically with CEO ownership since board

independence first decreases and then increases as CEO ownership grows.

109

Futhermore, Boone et al. (2007) argue that complex firms need more independent boards and

they use firm age as a proxy for complexity. But they suggest that firm complexity may not

increase linearly with firm age. Similarly, Linck, Netter, and Yang (2008) suggest that it is

not clear that a firm is more complex once it has matured. They inculde firm age squared as a

determinant of board independence to test if the relationship is nonlinear, and find that the

sign of the linear term changes even though firm age squared is not significant.

To analyse if these findings have any impact on demand for monitoring and advice, we

specify model (2) which includes non-linear determinants.

1 2

3 4 5

6 7 8

+ + costs +

it it it

it it it

it it

Independence CEO ownership CEO ownership squaredPerformance Leverage Female directorsGrowth Information Firm

α β ββ β ββ β β

= + ++

+ +

9 10 it

it it it

ageFirm age squared Firm size uβ β+ + +

(2)

In the first line of the explanatory variables in (2) we follow by Adams and Ferreira (2009a)

by including the squared value of CEO ownership. Lines 2–4 follow the base-case model,

while we also include the quadratic term of firm age as suggested by Linck, Netter, and Yang

(2008).

Model (2) uses the same proxy for the left hand side variable as the base-case model. The

model is estimated as a logit model, using GLM. That is, model (2) estimates determinants for

board independence related to the first agency problem. Table 11 shows the results.

Table 11

Our estimates for CEO ownership and CEO ownership squared are consistent with the

predictions based on earlier research showing that the relationship between CEO ownership

and board independence is not monotonic. That is, board independence first decrease and then

increases in CEO ownership.

The estimates for firm age and firm age squared are inconsistent with our predictions guided

by earlier research on listed firms. The average firm age is lower in our sample of non-listed

firms, (12 years) than in listed firms (34 years). This difference may explain why firm age is

negatively related to board independence, while firm age squared is positively related to board

independence. The estimates of the determinants used in the base-case model, (performance,

110

leverage, female directors, growth, information costs, and firm size) are consistent with the

base-case results.

According to the robustness checks, the base-case results are insensitive to alternative

econometric techniques and to how we define board independence. The only result that needs

to be reconsidered is the estimate for female directors which is negative in the base-case

model. When using other econometric techniques, including instrumental variable technique

(IV), however female directors is not significantly related to board independence. That is, the

gender of the directors has no impact on board independence. This result is new in literature.

The result is confirmed when we test an alternative proxy for board independence related to

the Agency problem 1. We believe that the negative relationship between female directors and

board independence is caused by endogeneity problems in the base-case model. The fixed

effect regression and the IV technique both reduce these problems. Taking into account that

none of the other determinants of board independence change when we use fixed effect

regression and IV technique, we conclude that the fraction of female directors has no impact

on board independence when board independence and gender balance are not regulated.

Finally, the results from the robustness section support earlier research showing that the

relationship between board independence and CEO ownership is non-monotonic, the

relationship is V-shaped.

6. Conclusion

This paper investigates the owners’ demand for monitoring and advice by the board of

directors in a large sample of firms that are not expected by regulation to appoint a minimum

fraction of independent directors. Our motivation is that earlier theoretical research shows that

board independence is firm-specific. That is, the demand for independent directors should

optimally vary from firm to firm.

The existing empirical research on determinants of board independence studies samples of

firms that are required to appoint at least 50% independent directors. Therefore, these studies

may be biased because some firms’ optimal board independence in a free contracting

environment is below the regulatory floor. We avoid this problem, as the firms in our sample

111

are not subject to any laws or corporate governance codes that restrict the board’s

independence.

Our findings show that firms that are well performing, small, well-established, and with high

leverage, are better off if the stockholders appoint directors who focus more on the advisory

role than the monitoring role. These results support the notion that the demand for monitoring

and advice is firm-specific.

Furthermore, we find that the relationship between CEO ownership and board independence

is non-monotonic as shown. In particular, board independence is first declining in CEO

ownership, but when CEO ownership increases the demand for monitoring by the board of

directors increases. Once more, this result supports the idea that the optimal level of board

independence varies across firms and even in a non-linear fashion.

We extend previous research by investigating board independence relative to the second

agency problem that concerns the independence between the directors and the firm’s main

stockholder. The estimates of the determinants for board independence relative to the first and

second agency problem are close to identical. We find that the same firm characteristics seem

to be driving the demand for monitoring the potential agency problem between owners and

managers and the demand for monitoring the potential conflict between majority and minority

owners.

Unlike earlier research, we find that the fraction of female directors is not related to board

independence. This is a surprising result compared to existing research on listed firms that

shows a strong positive correlation between female directors and board independence. We

argue that stockholders are just as likely to appoint female directors to fulfill the advisory role

as the monitoring role when board independence is not regulated and there are no regulations

on gender-quotas.

The results do not support our predictions that the monitoring of potential conflicts between

owners and managers is more prominent when firms have no controlling owner. Similarly, we

do not find evidence that monitoring of the conflict between majority and minority owners is

more prominent when the firm has a controlling owner. These results indicate that concerns

for the potential conflict between owners and managers and majority and minority owners are

just as likely to occur in all firms with multiple owners.

112

Moreover, we find that family firms are less likely to appoint independent directors than non-

family firms. This result suggests that family firms differ from other firms in ways that

influence their governance structure. The evidence is consistent with the notion that there are

strong ties between management and owners in family firms and also between large and small

stockholders. Hence, agency conflicts are not as common in family firms as in other firms.

Overall, our results suggest that board independence is firm specific. When owners face the

trade-off between monitoring and advice in a free contracting environment, they find an

optimal independence level that is lower than the one that is mandated for listed firms. Simply

put, firms value advice more than monitoring by the board of directors.

113

References

Adams, Renée B., and Daniel Ferreira. 2007. "A Theory of Friendly Boards." Journal of Finance no. 62 (1):217–250.

Adams, Renée B., and Daniel Ferreira. 2009a. "Strong Managers, Weak Boards?" CESifo

Economic Studies no. 55 (3–4):482–514. doi: 10.1093/cesifo/ifp023. Adams, Renée B., and Daniel Ferreira. 2009b. "Women in the Boardroom and their Impact

on Governance and Performance." Journal of Financial Economics no. 94 (2):291–309.

Adams, Renée B., Benjamin E. Hermalin, and Michael S. Weisbach. 2010. "The Role of

Boards of Directors in Corporate Governance: A Conceptual Framework and Survey." Journal of Economic Literature no. 48 (1):58–107.

Anderson, Ronald C., and David M. Reeb. 2003. "Founding-Family Ownership and Firm

Performance: Evidence from the S&P 500." Journal of Finance no. 58 (3):1301–1328. Anderson, Ronald C., and David M. Reeb. 2003b. "Founding Family Ownership, Corporate

Diversification and Firm Leverage." Journal of Law and Economics no. 56:653–684. Anderson, Ronald C., and David M. Reeb. 2004. "Board Composition: Balancing Family

Influence in S&P 500 Firms." Administrative Science Quaterly no. 49 (June):209–237. Andre, Paul, and Walid Ben-Amar. 2006. "Separation of Ownership from Control and

Acquiering Firm Performance: The Case of Family Ownership in Canada." Journal of Business Finance and Accounting no. 33 (3–4):517–543.

Baysinger, Barry D., and Henry N. Butler. 1985. "Corporate Governance and the Board of

Directors: Performance Effects of Changes in Board Composition." Journal of Law, Economics, and Organization no. 1 (1):101–124.

Beecher-Monas, Erica. 2007. "Marrying Diversity and Independence in the Boardroom: Just

How Far Have You Come, Baby?" SSRN eLibrary. Bhagat, Sanjai, and Bernard Black. 2002. "The Non-Correlation Between Board Indpendence

and Long-Term Firm Performance." Journal of Corporation Law no. 27 (2):231. Bhagat, Sanjai, and Bernard S. Black. 1998. "Independent Directors." The New Palgrave

Dictionary of Economics and the Law. no. 2:283–287. Boone, Audra L., Laura C. Field, Jonathan M. Karpoff, and Charu G. Raheja. 2007. "The

Determinants of Corporate Board Size and Composition: An Empirical Analysis." Journal of Financial Economics no. 85 (1):66–101.

Brennedsen, Morten, Kasper M Nielsen, Perez-Gonzalez, and Daniel Wolfenzon. 2007.

"Inside the Family Firm: The Role of Families in Succession Decisions and Performance." The Quaterly Journal of Economics no. 122 (2):647–691.

114

Bøhren, Øyvind, and Nils Erik Krosvik. 2013. "The Economics of Minority Freezeouts: Evidence from the Courtroom." International Review of Law and Economics, Forthcoming.

Bøhren, Øyvind, and Siv Staubo. 2013. "Female Directors and Board Independence: Evidence

from Board with Mandatory Gender Balance." Working Paper. Bøhren, Øyvind, and R. Øystein Strøm. 2010. "Governance and Politics: Regulating

Independence and Diversity in the Board Room." Journal of Business Finance & Accounting no. 37 (9–10):1281–1308.

Carter, Colin B., and Jay W. Lorsch. 2004. Back to the Drawing Board: Designing Corporate

Boards for a Complex World, Bosten, MA: Harvard Business School Press. Claessens, Stijn, Simon D. Djankov, Joseph P. H. Fan, and Larry H. P. Lang. 2002.

"Distangling the Incentives and Entrenchment Effects of Large Shareholdings." Journal of Finance no. 57:2741–71.

Coles, Jeffrey L., Daniel D. Naveen, and Lalitha Naveen. 2008. "Boards: Does One Size Fits

All?" Journal of Financial Economics no. 87:329–356. Denis, David J., and Atulya Sarin. 1999. "Ownership and Board Structures in Publicly

Traded Corporations." Journal of Financial Economics no. 52:187–223. Duchin, Ran, John G. Matsusaka, and Oguzhan Ozbas. 2010. "When are Outside Directors

Effective?" Journal of Financial Economics no. 96 (96):215–237. ECGI. 2012. "Index of Codes." European Corporate Governance Institute. Fields, M. Andrew, and Phyllis Y. Keys. 2003. "The Emergence of Corporate Governance

from Wall St. to Main St.: Outside Directors, Board Diversity, Earnings Management, and Managerial Incentives to bear Risk." Financial Review no. 38 (1):1–24.

Gomez-Mejia, Luis R., Manuel Nufiez-Nickel, and Isabel Gutierrez. 2001. "The Role of

Family Ties in Agency Contracts." Academy of Management Journal no. 44:81–96. Harris, Milton, and Artur Raviv. 2008. "A Theory of Board Control and Size." Review of

Financial Studies no. 21 (4):1797–1831. Hermalin, Benjamin E., and Michael S. Weisbach. 1991. "The Effects of Board Composition

and Direct Incentives on Firm Performance." The Journal of the Financial Management Association no. 20 (4):101–112.

Hermalin, Benjamin E., and Michael S. Weisbach. 1998. "Endogenously Chosen Boards of

Directors and their Monitoring of the CEO." American Economic Review no. 88 (1):96–118.

115

Hermalin, Benjamin E., and Michael S. Weisbach. 2003. "Boards of Directors as an Endogenously Determined Institution: A Survey of the Economic Literature." Economic Policy Review no. 9 (1):7.

Hsiao, Cheng. 2003. Analysis of Panel Data, Cambridge University Press. Jensen, Michael C. 1986. "Agency Costs of Free Cash Flow, Corporate Finance, and

Takeovers." The American Economic Review no. 76 (2):323–329. Jensen, Michael C. 1993. "The Modern Industrial Revolution, Exit, and Failure of Internal

Control Systems." Journal of Finance no. 48:831–880. Jensen, Michael C., and William H. Meckling. 1976. "Theory of the Firm: Managerial

Behavior, Agency Costs and Ownership Structure." Journal of Financial Economics no. 3 (4):305–360.

Linck, James S., Jeffry M. Netter, and Tina Yang. 2008. "The Determinants of Board

Structure." Journal of Financial Economics no. 7 (2):308–328. MacAvoy, Paul. W., and Ira. M. Millstein. 1999. "The Active Board of Directors and its

Effect on Performance of the Large Publicly Traded Corporations." Journal of Applied Corporate Finance no. 11 (4):8–20.

Masulis, Ronald W. M., Peter K. Pham, and Jason Zein. 2011. "Family Business groups

around the World: Financing advantages, control motivations, and organizational choises." Review of Financial Studies no. 24:3556–3600.

Morck, Randall, Andrei Shleifer, and Robert W. Vishny. 1989. "Alternative Mechanisms for

Corporate Control." American Economic Review no. 79 (4):842–852. Raheja, Charu G. 2005. "Determinants of Board Size and Composition: A Theory of

Corporate Boards." Journal of Financial and Quantitative Analysis no. 40 (2):283–306.

Sarbane-Oxely. 2002. "Sarbane-Oxely Act; 'Public Company Accounting Reform and

Investor Protection Act' and 'Corporate and Auditing Accountability and Responsibility Act'." United State Federal Law no. 116 (745):107–204.

Shivdasani, Anil, and David Yermack. 1999. "CEO Involvement in the Selection of New

Board Members: An Empirical Analysis." Journal of Finance no. 54 (5):1829–1853. Shleifer, Andrei, and Robert W. Vishny. 1997. "A Survey of Corporate Governance." Journal

of Finance no. 52 (2):737–783. Song, Fenghua, and Anjan V. Thakor. 2006. "Information Control, Career Concerns, and

Corporate Governance." Journal of Finance no. 61 (4):1845–1896. Studenmund, A. H. 2000. Using Econometrics: A Practical Guide. Vol. 4th ed.: Addison

Wesley Longman, Inc.

116

Villalonga, Belen, and Raphael Amit. 2006. "How do Family Ownership, Control and Management Affect Firm Value?." Journal of Financial Economics no. 80:385–417.

Weisbach, Michael S. 1988. "Outside Directors and CEO turnover." Journal of Financial

Economics no. 20:431–460.

117

Variable DefinitionBoard characteristicsBoard independence - A1 0/1 dummy variable that is 1 if the CEO is not a director

and 0 otherwiseBoard independence - A2 The fraction of directors that do not belong to the largest family

by ownershipBoard size The number of board membersFemale directors The proportion of board members who are womenBoD tenure The average number of years since the non-CEO directors were appointed

Ownership characteristicsOutside concentration The sum of squared equity fractions in the firm (Herfinadahl index)Inside ownership Fraction of equity held by the firm's officers and directorsCEO ownership Fraction of equity held by the CEOLargest owner Fraction of equity held by the firm's largest stockholder, counting

each family member as one ownerFamily ownership Fraction of equity held by the family with the largest equity stake

Family characteristicsFamily control 0/1 dummy variable that equals 1 if the largest family owns more

than 50% and 0 otherwiseFamily chair 0/1 dummy variable that equals 1 if the chair belongs to

the largest family by ownership and 0 otherwiseFamily CEO 0/1 dummy variable that equals 1 if the CEO belongs to the

largest family by ownership and 0 otherwiseFamily board The fraction of directors coming from the largest family by ownership

General firm characteristicsPerformance The average real return on assets from t-3 to tLeverage Total debt divided by total assetsGrowth The average percentage increase in real sales from t-3 to tInformation costs The standard deviation of performance from t-3 to tFirm age The number of years since the firm was foundedFirm size Sales in constant 2011 MNOK. Log transformed in regressionsCEO tenure The number of years since the CEO took officeCEO age The number of years since the CEO was born

Table 1: The empirical variables

This table defines the variables used in the empirical analysis. The ownership characteristics are based on ultimate (direct plus indirect) equity holdings.

118

Table 2: Distributional properties of the variables

Variable Mean Std. 0 5 25 50 75 95 100 NBoard characteristicsBoard independence - A1 0.48 0.50 0.00 0.00 0.00 0.00 1.00 1.00 1.00 193,297Board independence - A2 0.57 0.29 0.00 0.00 0.33 0.67 0.75 1.00 1.00 181,651Board size 3.80 1.08 3.00 3.00 3.00 3.00 4.00 6.00 14.00 193,297Female directors 0.15 0.21 0.00 0.00 0.00 0.00 0.33 0.67 1.00 193,297BoD tenure 3.47 2.64 0.00 0.00 1.33 3.00 5.00 8.75 11.00 193,297Ownership characteristicsOutside concentration 0.21 0.10 0.00 0.10 0.16 0.20 0.24 0.37 1.00 193,053Inside ownership 78.30 26.37 0.00 25.00 60.00 93.00 100.00 100.00 100.00 108,240CEO ownership 37.30 17.97 0.00 10.00 25.00 33.33 50.00 70.00 89.90 111,263Largest owner 45.79 17.18 0.00 20.00 33.33 49.00 55.00 78.00 89.99 193,297Family ownership 82.98 25.17 0.00 25.00 70.60 99.95 100.00 100.00 100.00 181,651Family characteristicsFamily control 0.82 0.38 0.00 1.00 1.00 1.00 1.00 1.00 1.00 193,297Family chair 0.47 0.50 0.00 0.00 0.00 0.00 1.00 1.00 1.00 181,651Family CEO 0.53 0.50 0.00 0.00 0.00 1.00 1.00 1.00 1.00 181,651Family board 0.43 0.29 0.00 0.00 0.25 0.33 0.67 1.00 1.00 181,651General firm characteristicsPerformance 8.48 12.62 -31.99 -6.99 1.22 3.42 14.62 35.41 51.99 193,297Leverage 0.74 0.24 0.00 0.32 0.60 0.76 0.89 1.09 2.10 193,297Growth 1.26 1.25 0.41 0.75 0.97 1.03 1.19 2.05 25.78 167,732Information costs 8.58 48.18 0.00 0.08 1.17 5.09 11.00 25.58 14260.00 140,595Firm age 12.28 12.48 0.00 1.00 4.00 9.00 16.00 33.00 162.00 187,487Firm size 28.06 328.98 1.62 2.18 4.04 8.20 19.09 77.61 89907.52 193,297CEO tenure 6.12 4.50 0.00 1.00 2.00 5.00 9.00 15.00 18.00 193,297CEO age 46.64 9.84 17.00 31.00 39.00 46.00 54.00 63.00 93.00 182,274

Percentile, %

This table shows distributional properties of the variables used to measure board, ownership, family, and general firm characteristics. Table 1 defines the variables. The sample is Norwegian non-listed firms, from 2000–2011, where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

119

Table 3: Firm characteristics by ownership

Panel A: Controlling ownerVariable All Majority owner Non-majority owner Difference t-value (p-value)Board independence - A1 0.48 0.50 0.46 0.04 17.58 (0.000)Board independence - A2 0.57 0.54 0.60 -0.06 -44.07 (0.000)CEO ownership 37.30 47.69 28.48 19.21 268.64 (0.000)Performance 8.48 8.42 8.54 -0.12 -2.03 (0.042)Leverage 0.74 0.75 0.73 0.02 17.52 (0.000)Female directors 0.15 0.17 0.14 0.03 30.54 (0.000)Growth 1.26 1.37 1.19 0.18 18.52 (0.000)Information costs 9.29 9.46 9.12 0.34 1.03 (0.299)Firm age 12.28 12.23 12.33 -0.10 -1.77 (0.077)Firm size 28.06 30.78 25.44 5.34 34.51 (0.000)Family control 0.82 0.81 0.83 -0.02 -11.37 (0.000)N 193,297 94,693 98,604

120

Panel B: Controlling family Variable All Family firms Non-family firms Difference t-value (p-value) Board independence - A1 0.48 0.39 0.88 -0.49 -231.83 (0.000) Board independence - A2 0.57 0.54 0.77 -0.23 -136.77 (0.000) CEO ownership 37.30 38.28 19.71 18.57 -93.52 (0.000) Performance 8.48 8.71 7.41 1.30 17.21 (0.000) Leverage 0.74 0.75 0.68 0.07 44.42 (0.000) Female directors 0.15 0.16 0.14 0.02 17.18 (0.000) Growth 1.26 1.17 1.38 -0.21 -11.40 (0.000) Information costs 9.29 9.18 9.74 -0.56 -1.00 (0.032) Firm age 12.28 11.96 13.75 -1.79 -20.82 (0.000) Firm size 28.06 19.15 68.79 -49.64 -110.39 (0.000) N 193,297 158,630 34,667 This table compares the mean values of the variables used in the base case model across subsamples. Panel A compares the mean values of Majority owner and Non-majority owner subsample. Panel B compares the mean values of Family firms and Non-family firms subsample. The differences between mean values, their t-values, and p-values (in parentheses) are reported in the three right-most columns. Table 1 defines the variables. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

121

Table 4: Bivariate correlation coefficients between the determinants of board independence CEO Information Female Firm Firm ownership Performance Growth Leverage costs directors age size Performance 0.016 Growth -0.010 -0.005 Leverage 0.050 -0.111 0.004 Information costs -0.007 0.013 0.001 0.028 Female directors 0.141 -0.018 -0.009 -0.013 -0.007 Firm age -0.004 -0.026 -0.016 -0.200 -0.026 0.070 Firm size -0.126 -0.005 0.006 -0.010 -0.011 -0.011 0.036 Family control 0.402 0.039 -0.013 0.112 -0.030 0.033 -0.055 -0.058 This table shows pairwise Pearson correlation coefficients between the hypothesized determinants of board independence as specified in model (1) of the main text. Table 1 defines the variables. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

122

Table 5: Estimates of the base-case model - Agency problem 1 (owner vs. manager) Base case model Determinant Prediction Base case model with family control CEO ownership (-) -0.072 -0.068 (0.000) (0.000) Performance (-) -0.008 -0.008 (0.000) (0.000) Leverage (+/-) -0.415 -0.380 (0.000) (0.000) Female directors (+) -0.272 -0.239 (0.000) (0.000) Growth (-) 0.000 0.000 (0.881) (0.919) Information costs (-) 0.000 0.000 (0.551) (0.667) Firm age (+) -0.014 -0.014 (0.000) (0.000) Firm size (+) 0.250 0.241 (0.000) (0.000) Family control (+/-) -0.655 (0.000) Constant -3.557 -2.950 (0.000) (0.000) LR chi2 (8) 3,389.66 3,478.44 Prob > chi2 0.000 0.000 R2 0.162 0.167 N 121,403 121,403 This table shows the estimated coefficients of the base-case model and the base-case model with Family control. The relationship is specified in model (1) of the main text. The predicted signs of the coefficients are shown in the second column. Column 3 and column 4 show the estimated coefficients for logit regressions of board, ownership, and general firm on board independence. The p-values of the estimated coefficients are stated in parentheses underneath. The dependent variable is 1 if the CEO is not a board member and 0 otherwise. Table 1 defines the determinants. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

123

Table 6: Estimates of the base-case model - Agency problem 2 (majority vs. minority)

Determinant Prediction Base-case

model Base-case model with Family control

CEO ownership (-) -0.001 -0.001 (0.000) (0.000) Performance (-) -0.002 -0.002 (0.000) (0.000) Leverage (+/-) -0.004 -0.040 (0.070) (0.086) Female directors (+) -0.212 -0.212 (0.000) (0.000) Growth (-) 0.000 0.000 (0.120) (0.120) Information costs (-) 0.000 0.000 (0.192) (0.197) Firm age (+) -0.003 -0.003 (0.000) (0.000) Firm size (+) 0.018 0.180 (0.000) (0.000) Family control (+/-) -0.031 (0.000) Constant 0.371 0.399 (0.000) (0.000) Prob > chi2 0.000 0.000 Random effects Yes Yes R2 0.147 0.153 N 76,856 76,856 This table shows the estimated coefficients of the base-case model and the base-case model with Family control. The relationship is specified in model (1) of the main text. The predicted signs of the coefficients are shown in the second column. Column 3 and column 4 show the estimated coefficients for panel data regressions of board, ownership, and general firm characteristics on board independence. The p-values of the estimated coefficients are stated in parentheses underneath. The dependent variable is the fraction of non-family directors. Table 1 defines the determinants. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

124

Table 7: Estimates of the base case model in subsamples Panel A: Controlling owner Agency problem 1 Agency problem 2 (owner vs. manager) (majority owner vs. minority owner) Determinant Majority owner Non-majority owner Majority owner Non-majority owner CEO ownership -0.050 *** -0.104 *** -0.001 *** -0.003 *** Performance -0.006 *** -0.009 *** -0.001 *** -0.001 *** Leverage -0.129 *** -0.453 *** 0.005 -0.007 ** Female directors -0.621 *** -0.173 ** -0.294 *** -0.138 *** Growth 0.000 0.000 0.000 0.000 * Information costs 0.001 0.000 0.000 0.000 Firm age -0.011 *** -0.017 *** -0.003 *** -0.003 *** Firm size 0.257 *** 0.243 *** 0.016 *** 0.016 *** Constant -4.514 *** -2.676 *** 0.377 *** 0.442 *** LR chi2 (8) 2,614.650 5,546.150 Prob > chi2 0.000 0.000 0.000 0.000 Random effects No No Yes Yes R2 0.157 0.173 0.171 0.124 N 34,769 34,765 34,765 34,765

125

Panel B: Controlling family Agency problem 1 Agency problem 2 (owner vs. manager) (majority vs. minority) Determinant Family firm Non-family firm Family firm Non-family firm CEO ownership -0.067 *** -0.089 *** -0.001 *** -0.004 *** Performance -0.007 *** -0.012 *** -0.001 *** -0.001 *** Leverage -0.297 *** -0.782 *** -0.003 -0.026 ** Female directors -0.290 *** 0.319 -0.217 *** -0.129 *** Growth -0.000 0.000 0.000 0.000 ** Information costs 0.000 -0.001 0.000 0.000 Firm age -0.014 *** -0.016 *** -0.003 *** -0.003 *** Firm size 0.276 *** 0.066 * 0.017 *** 0.014 *** Constant -4.293 *** 0.533 0.373 *** 0.636 *** LR chi2 (8) 6,265.150 710.430 Prob > chi2 0.000 0.000 0.000 0.000 Random effects No No Yes Yes R2 0.146 0.141 0.148 0.128 N 72,863 3,998 72,863 17,040 This table shows the estimated coefficients for regressions of board, ownership, and general firm characteristics on board independence in subsamples. Panel A shows the estimates for Majority owner and Non-majority owner subsamples. Columns 2 and 3 show the estimated coefficients for logit regressions related to Agency problem 1. The dependent variable is 1 if the CEO is not a board member and 0 otherwise. Columns 4 and 5 show the estimated coefficients for panel data regressions related to Agency problem 2. The dependent variable is the fraction of non-family directors. Panel B shows the estimates in Family firm and Non-family firm subsamples. Column 2 and column 3 show the estimated coefficients for logit regression related to Agency problem 1. The dependent variable is 1 if the CEO is not a board member and 0 otherwise. Column 4 and column 5 show the estimated coefficients for panel data regression related to Agency problem 2. The dependent variable is the fraction of non-family directors. The relationship is specified in model (1) of the main text. Statistical significance at 1%, 5%, and 10% levels is labeled ***, **, and *, respectively. Table 1 defines the determinants. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

126

Table 8: Alternative estimation methods - Agency problem 1 (owner vs. manager) Method Determinant Logit Probit Standard panel Logit panel Clustered OLS IV CEO ownership -0.072 -0.033 -0.003 -0.131 -0.004 -0.003 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Performance -0.008 -0.005 -0.001 -0.009 -0.001 -0.001 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage -0.415 -0.206 -0.011 -0.470 -0.037 -0.011 (0.000) (0.000) (0.005) (0.006) (0.006) (0.003) Female directors -0.272 -0.130 -0.004 -0.261 -0.012 -0.004 (0.000) (0.000) (0.563) (0.213) (0.159) (0.493) Growth 0.000 0.000 0.000 0.000 0.000 0.000 (0.881) (0.800) (0.719) (0.960) (0.506) (0.493) Information costs 0.000 0.000 0.000 0.000 0.000 0.000 (0.551) (0.470) (0.625) (0.500) (0.296) (0.611) Firm age -0.014 -0.007 -0.001 -0.031 -0.001 -0.001 (0.000) (0.000) (0.001) (0.000) (0.000) (0.000) Firm size 0.250 0.145 0.025 0.721 0.028 0.025 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Constant -3.557 -2.306 -0.157 -15.021 -0.139 -0.158 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Random effects No No Yes Yes No Yes Prob > chi2 / Prob > F 0.000 0.000 0.000 0.000 0.000 0.000 R2 0.162 0.158 0.090 0.092 0.091 N 121,403 76,861 76,861 76,861 76,861 76,861 This table shows the estimates of the base-case model under six different econometric techniques. The estimated coefficients are shown in columns 2–7, and the corresponding p-values are stated in parentheses underneath. The dependent variable is 1 if the CEO is not a board member and 0 otherwise. Table 1 defines the determinants. The sample is Norwegian non-listed firms, from 2000–2011, where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

127

Table 9: Alternative estimation methods - Agency problem 2 (majority vs. minority) Method Determinant Random effects Fixed effects Clustered OLS IV CEO ownership -0.001 -0.001 -0.003 -0.001 (0.000) (0.000) (0.000) (0.000) Performance -0.002 0.000 -0.001 -0.004 (0.000) (0.292) (0.059) (0.000) Leverage -0.004 0.000 0.005 -0.003 (0.070) (0.039) (0.537) (0.004) Female directors -0.212 -0.154 -0.336 0.002 (0.000) (0.753) (0.000) (0.656) Growth 0.000 0.000 0.000 -0.002 (0.120) (0.059) (0.613) (0.691) Information costs 0.000 0.000 0.001 -0.003 (0.192) (0.000) (0.063) (0.000) Firm age -0.003 -0.002 -0.003 -0.001 (0.000) (0.000) (0.000) (0.001) Firm size 0.018 0.012 0.015 0.013 (0.000) (0.000) (0.000) (0.000) Constant 0.371 0.399 0.511 0.464 (0.000) (0.000) (0.000) (0.000) Random effects/Fixed effects Random Fixed No Random Prob > chi2 / Prob > F 0.000 0.000 0.000 0.000 R2 0.147 0.138 0.154 0.077 N 76,856 72,863 76,856 76,856 This table shows the estimates of the base-case model under different econometric techniques. The estimated coefficients are shown in columns 2–5, where the corresponding p-values are stated in parentheses underneath. The dependent variable is the fraction of non-family directors. Table 1 defines the determinants. The sample is Norwegian non-listed firms, from 2000–2011, where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

128

Table 10: Alternative proxy for board independence - Agency problem 1 (owner vs. Manager)

Definition of board

independence Model A: Model B: Determinant Prediction CEO not director Relative tenure CEO ownership (-) -0.072 -0.037 (0.000) (0.000) Performance (-) -0.008 -0.014 (0.000) (0.000) Leverage (+/-) -0.415 -0.308 (0.000) (0.000) Female directors (+) -0.272 0.032 (0.000) (0.686) Growth (-) 0.000 0.000 (0.881) (0.928) Information costs (-) 0.000 -0.001 (0.551) (0.588) Firm age (+) -0.014 -0.078 (0.000) (0.000) Firm size (+) 0.250 -0.222 (0.000) (0.000) Constant -3.557 3.366 (0.000) (0.000) Random effects No Yes Year fixed effects No Yes Prob > chi2 0.000 0.000 R2 0.162 0.136 N 121,403 76,861 This table shows the estimates of the base-case model (1) using alternative proxies for board independence. Model A is the base-case model, which uses a dummy variable which equals 1 if the CEO is not a board member and 0 otherwise. Model B uses the average tenure of non-CEO directors minus the tenure of the CEO. The relationship is specified in model (1) of the main text. The predicted signs of the coefficients are shown in column 2 and the estimates of model A and model B are reported in columns 3 and 4, respectively, where the corresponding p-values are stated in parentheses underneath. Table 1 defines the determinants. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

129

Table 11: Estimates the base-case model with non-linear determinants - Agency problem 1 (owner vs. manager) Determinant Prediction Estimates CEO ownership (-) -0.144 (0.000) CEO ownership squared (+) 0.001 (0.000) Performance (-) -0.007 (0.000) Leverage (+/-) -0.370 (0.000) Female directors (+) -0.326 (0.000) Growth (-) 0.000 (0.928) Information cost (-) 0.000 (0.602) Firm age (+) -0.013 (0.000) Firm age square (-) -0.000 (0.248) Firm size (+) 0.222 (0.000) Constant -2.302 (0.000) LR chi2 (8) 9055.180 Prob > chi2 0.000 R2 0.183 N 76,861 This table shows the estimates of model (2) of the main text which uses additional determinants of board independence. The dependent variable is 1 if the CEO is not a board member and 0 otherwise. The predicted signs of the coefficients are shown in the second column, and the p-values of the estimated coefficients are reported in the third column are stated in parentheses underneath. Table 1 defines the determinants. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

130

Appendix 1: Characteristics of non-listed firms and listed firms

Variable Mean Median Mean MedianBoard characteristicsBoard independence - A1 0.48 0.00 0.92 1.00Board independence - A2 0.57 0.67 0.93 1.00Board size 3.80 3.00 6.20 6.00Female directors 0.15 0.00 0.21 0.20BoD tenure 3.47 3.00 1.62 1.37Ownership characteristicsOutside concentration 0.21 0.20 0.16 0.10Inside ownership 78.30 93.00 0.12 0.00CEO ownership 37.30 33.33 0.10 8.33Largest owner 45.79 49.00 26.61 22.64Family ownership 82.98 99.95 20.53 14.20Family characteristicsFamily control 0.82 1.00 0.08 0.00Family chair 0.43 0.00 0.12 0.00Family CEO 0.45 0.00 0.08 0.00Family board 0.45 0.00 0.07 0.00General firm characteristicsPerformance 8.48 3.42 5.59 5.02Leverage 0.74 0.76 0.49 0.52Growth 1.26 1.03 1.15 1.08Information costs 8.58 5.09 7.35 4.96Firm age 12.28 9.00 34.61 18.00Firm size 28.06 8.20 885.84 85.58CEO tenure 6.12 5.00 5.09 4.00CEO age 46.64 46.00 47.12 47.00N 178,721 178,721 3,427 3,427

Non-listed firms Listed firms

This table shows the mean and median values of the variables used to measure board, ownership, family, and general firm characteristics. Table 1 defines the variables. The sample is Norwegian non-listed and listed firms, from 2000–2011, where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

131

Appendix 2: Properties of the instrumental variable (IV) Panel A: CEO ownership classified by CEO age Variable All Young CEO Old CEO Difference p-value CEO ownership 37.30 36.16 38.49 -2.33 (0.000) Panel B: Board and firm characteristics classified by CEO age Variable All Young CEO Old CEO Difference p-value Board independence-A1 0.48 0.48 0.48 0.00 (1.000) Board independence-A2 0.57 0.58 0.56 0.02 (0.000) Performance 8.48 8.62 8.27 0.35 (0.000) Leverage 0.74 0.76 0.71 0.05 (0.000) Female directors 0.15 0.15 0.16 -0.01 (0.000) Firm size 16.18 16.13 16.28 -0.15 (0.000) This table shows board and firm characteristics for young and old CEOs. A young CEO is a CEO that is younger than 47 years (the average age of CEOs in our sample), an old CEO is a CEO that is 47 years or older. Panel A shows the ownership of young and old CEOs, the difference is shown in the fifth column, and the p-value is stated in the sixth column. Panel B shows board and firm characteristics for young and old CEOs, the differences are shown in the fifth column and the p-values are stated in parentheses in the sixth column. Table 1 defines the variables. The sample is Norwegian non-listed firms from 2000–2011 where revenue ≥ 2 million NOK, board size ≥ 3, and the largest owner holds < 90% of the equity. Performance is censored at 2% and 98% and leverage is censored at 99%.

132

5. Summary

The findings in this thesis show that regulation, such as the gender balance law and the board independence code, results in unintended effects which may matter for the firm’s behavior. In the first essay we find that stockholders of half the firms that suddenly become exposed to the gender balance law (GBL) choose the only alternative to changing the board, that is to exit into an organizational form where the GBL does not apply.

The second essay finds that the GBL causes a large increase in board independence. Involuntary increase in board independence is a potential problem because there is a trade-off between the board’s monitoring role and advice role. More independent directors are assumed to strengthen the board’s monitoring role, while more dependent directors are better advisors. Finally, we find that recommending a majority of independent directors in every firm by the independence code (IC) may hurt firms that are better off with a lower level of board independence.

This thesis addresses economic consequences of new regulation of board composition. Our results show that profitable, young, and small firms are hurt the most by these regulations. That is, the cost of changing the board, either by increasing the fraction of female directors or by increasing the fraction of independent directors, is particularly costly for such firms.

Recent political signals indicate that the exit option we analyze may soon disappear. In particular, gender balance in corporate boards may be made mandatory not just for ASA firms, but also for some AS firms . If that happens, Norway will not just be special for being the first and only country to mandate a massive, rapid shift in the composition of corporate boards and to punish non-compliers with liquidation. The regulators may also decide to eliminate the option firms currently have to mitigate the costs of regulatory shocks by transforming into organizational forms that are not exposed to the law. Every other country considering gender balance regulation seems to favor the comply-or-explain system or considerably milder sanctions than liquidation. Such regulatory regimes would leave the gender balance choice to the firm’s discretion and hence allow for firm heterogeneity in board design. Our findings suggest that, compared to this more flexible alternative, the mandatory approach, and particularly one without exit options, is a costly way to regulate gender balance of corporate boards.

Although the gender balance law is mandatory, the corporate governance code is not mandatory, but follows the principle of comply-or-explain. Nevertheless, there is a widespread view among policy makers that boards with at least half the directors being independent ensure good governance. Therefore, it is often argued that stockholders should ensure that their firm follows the code. Our evidence suggests that some firms are better off with a lower level of board independence. Hence, one-size-fits-all regulation is costly and should not be complied with by all firms.