# «Three essays on corporate boards R. Øystein Strøm A dissertation submitted to BI Norwegian School of Management for the degree of Dr.Oecon SERIES ...»

Yermack, D. (1996). Higher market valuation of companies with a small board of directors. Journal of Financial Economics 40, 185–212.

The table shows the percentage of ﬁrms having employee directors according to employment categories. N is the number of ﬁrms in the employee directors or the number of employees category. The number of employees category reﬂects the regulations on co-determination (Aarbakke et al., 1999). With more than 200 employees co-determination is compulsory. In the 31 to 200 bracket co-determination is realised if an employee majority demands it, with a larger proportion of representation with a larger workforce. In all categories, including the above 200 employees, ﬁrms in some industries are exempted from the rules.

## CHAPTER 3. CO-DETERMINATION

The table shows the distribution of employee directors across industries. Some or whole parts of the industry may be exempted, for instance the Energy (hydro power and petroleum) sector.

Transport contains the important shipping segment. Media is exempted as well, but in some ﬁrms co-determination comes about through union negotiations.

“Empdir” is short-hand for employee directors.

3.9. TABLES 95

Tobin’s Q is market value divided by book value of assets; Stock return is the raw stock return corrected for dividend and stock split; ROA is accounting proﬁts on book value of assets;

Average wage is the logarithm of total wages divided by the number of employees; Directors’ holdings is the percentage of directors’ ownership; Network is a summary measure of the board’s direct and indirect relations to other ﬁrms through multiple directorships, (see footnote 9);

Size1 is the board size of shareholder elected directors; Gender1 is the fraction of women of the shareholder elected directors; Board index is a summary measure of the above board variables;

Leverage is the book value of debt on book value of equity; Dividend payout rate is dividends on net income; Empdir is the number of employee directors divided by the number of directors;

Empdirfrac is the fraction of employee directors in the total board; Firm size is the natural logarithm of accounting income; Systematic Risk is the company’s exposure to market changes (equity beta); Volatility is the ﬁrm’s total risk measured as its yearly standard deviation.

The “ F sign.” shows the signiﬁcance of the test of the null hypothesis that the two group means are equal, estimated from an analysis of variance (ANOVA). Low values indicate rejection of the null hypothesis. The F value is found by dividing the Between Groups Mean Square by the Error Mean Square (Johnson and Wichern, 1988, p. 235).

## CHAPTER 3. CO-DETERMINATION

The table reports the simultaneous equation estimation of the system of equations in (3.2) with co-determined ﬁrms in the upper part and shareholder determined ﬁrms in the lower part.

The dependent variable is Tobin’s Q, which we measure as the market value of the ﬁrm over its book value. Variables are deﬁned in table 3.3. Each variable is time demeaned in the regressions. For each ﬁrm and each variable, I time demean by subtracting a given year’s observation from the ﬁrm’s overall mean. The table shows the estimates based on the standardized variables, which we construct by deducting each observation from its mean value and dividing by its standard deviation.

Fixed eﬀects estimation in 3SLS framework with standardized variables. All non-ﬁnancial ﬁrms on Oslo Stock Exchange 1989 to 2002.

The Wald test is explained in table 3.4. The test results show that a hypothesis that all coeﬃcients are zero must be rejected in all relations at the 1% level, except one where a 7.7% level is required.

The Chow (Greene, 2003, ch. 7) dummy variable test is an exclusion test for the null hypothesis that variables formed by a co-determination dummy variable interacted with each of the explanatory variables are all zero. Low value indicates hypothesis rejection. The test result shows that the hypothesis that the two sub-samples have equal coeﬃcients must be rejected.

Signiﬁcant results at the 5% (10%) level are marked with ∗∗ (∗ ).

## CHAPTER 3. CO-DETERMINATION

The table reports the simultaneous equation estimation of the system of equations in (3.2) with all ﬁrms larger than 200 employees in the upper part and all co-determined ﬁrms larger than 200 employees in the lower part.

The dependent variable is Tobin’s Q, which we measure as the market value of the ﬁrm over its book value. Variables are deﬁned in table 3.3. Each variable is time demeaned in the regressions. For each ﬁrm and each variable, I time demean by subtracting a given year’s observation from the ﬁrm’s overall mean. The table shows the estimates based on the standardized variables, which we construct by deducting each observation from its mean value and dividing by its standard deviation.

Fixed eﬀects estimation in 3SLS framework with standardized variables. All non-ﬁnancial ﬁrms on Oslo Stock Exchange 1989 to 2002.

The Wald test is explained in table 3.4. The test results show that a hypothesis that all coeﬃcients are zero must be rejected in all relations at the 1% level, except one, where a 4.3% level is required.

Signiﬁcant results at the 5% (10%) level are marked with ∗∗ (∗ ).

3.9. TABLES 99

The table reports the simultaneous equation estimation of the system of equations in (3.2) when the individual variables making out the board index replace the board index. The board index consists of directors’ holdings, network, board size, and gender. The deﬁnition of directors’ holdings is the fraction of ownership for the board as a whole; network is information centrality (Wasserman and Faust, 1994), see footnote 9; the board size is the number of shareholder elected directors; and gender is deﬁned as the number of shareholder elected women over board size.

The dependent variable is Tobin’s Q, which we measure as the market value of the ﬁrm over its book value. Variables are deﬁned in table 3.3. Each variable is time demeaned in the regressions. For each ﬁrm and each variable, I time demean by subtracting a given year’s observation from the ﬁrm’s overall mean. The table shows the estimates based on the standardized variables, which we construct by deducting each observation from its mean value and dividing by its standard deviation.

Fixed eﬀects estimation in 3SLS framework with standardized variables. All non-ﬁnancial ﬁrms on Oslo Stock Exchange 1989 to 2002.

The Wald test is explained in table 3.4. The test results show that a hypothesis that all coeﬃcients are zero must be rejected in all relations at the 1% level.

Signiﬁcant results at the 5% (10%) level are marked with ∗∗ (∗ ).

## CHAPTER 3. CO-DETERMINATION

The table reports the simultaneous equation estimation of the system of equations in (3.2) when all individual variables enter the board index, and not just directors’ holdings, network, board size, and gender. The added variables are Outside owner concentration, Independence, CEO director, Exported and imported directors, and board age dispersion. Outside owner concentration is the sum of squared equity fractions across all the ﬁrm’s outside owners;

Independence is the board tenure of the non-employee directors minus the tenure of the CEO;

CEO director equals 1 if the CEO is a member of his company’s board and zero otherwise;

Exported CEO is the number of outside directorships held by the ﬁrm’s CEO; Imported CEO is the proportion of CEOs from other companies on the board; Board age dispersion is the standard deviation of board age.

The Wald test is explained in table 3.4. The test results show that a hypothesis that all coeﬃcients are zero must be rejected in all relations at the 1% level.

Signiﬁcant results at the 5% (10%) level are marked with ∗∗ (∗ ).

3.9. TABLES 101

The table reports the simultaneous equation estimation of the system of equations in (3.2) when the stock return replaces Tobin’s Q in the upper part and the return on assets replaces Tobin’s Q in the lower part.

The dependent variable is the stock return, deﬁned as the raw stock return adjusted for dividend and stock splits; alternatively, as the return on assets, gauged as the accounting proﬁts on book value of assets. Variables are deﬁned in table 3.3. Each variable is time demeaned in the regressions. For each ﬁrm and each variable, I time demean by subtracting a given year’s observation from the ﬁrm’s overall mean. The table shows the estimates based on the standardized variables, which we construct by deducting each observation from its mean value and dividing by its standard deviation.

The Wald test is explained in table 3.4. The test results show that a hypothesis that all coeﬃcients are zero must be rejected in all relations at the 1% level, except for the average wage, where at least a 1.6% level is needed.

Signiﬁcant results at the 5% (10%) level are marked with ∗∗ (∗ ).

## CHAPTER 3. CO-DETERMINATION

The table reports the simultaneous equation estimation of the system of equations in (3.2) when the dividend payout rate replaces leverage in the upper part and the lagged ﬁrm performance is removed in the lower part.

The dependent variable is Tobin’s Q, which we measure as the market value of the ﬁrm over its book value. Each variable is time demeaned in the regressions. For each ﬁrm and each variable, I time demean by subtracting a given year’s observation from the ﬁrm’s overall mean. The table shows the estimates based on the standardized variables, which we construct by deducting each observation from its mean value and dividing by its standard deviation.

I use ﬁxed eﬀects estimation in 3SLS framework with standardized variables. The sample comprises all non-ﬁnancial ﬁrms on Oslo Stock Exchange 1989 to 2002.

The Wald test is explained in table 3.4. The test results show that a hypothesis that all coeﬃcients are zero must be rejected in all relations at the 1% level, except for the average wage and the dividend payout relations in the upper part, where I cannot reject the hypothesis.

Signiﬁcant results at the 5% (10%) level are marked with ∗∗ (∗ ).

3.9. TABLES 103

The table reports the simultaneous equation estimation of the system of equations in (3.2) when the full sample is sub-divided into informationally intensive industries in the upper part and other industries in the lower. Using the same GICS industry classiﬁcation as in table 3.3, informationally intensive industries are Capital goods, Transport, Consumer articles, Retailing, Food and staples retailing, Health care equipment and supplies, and Telecommunications, while the rest is in other industries.

The dependent variable is Tobin’s Q, which we measure as the market value of the ﬁrm over its book value. Each variable is time demeaned in the regressions. For each ﬁrm and each variable, I time demean by subtracting a given year’s observation from the ﬁrm’s overall mean. The table shows the estimates based on the standardized variables, which we construct by deducting each observation from its mean value and dividing by its standard deviation.

I use ﬁxed eﬀects estimation in 3SLS framework with standardized variables. The sample comprises all non-ﬁnancial ﬁrms on Oslo Stock Exchange 1989 to 2002.

The Wald test is explained in table 3.4. The test results show that a hypothesis that all coeﬃcients are zero must be rejected in all relations at the 1% level, except for a 2.3% level in the average wage relation in the Other industries estimation.

The Chow dummy variable test is explained in table 3.5. The test result indicates that coeﬃcient values are diﬀerent in the two sub-samples.

Signiﬁcant results at the 5% (10%) level are marked with ∗∗ (∗ ).

Board control and departures

**Abstract**

I empirically explore1 CEO turnover and board changes (substitutions and enlargements), their simultaneity, and their relation to outside ownership concentration, ﬁrm performance, board independence and to CEO entrenchment variables on a 14 year-long panel data set of all listed nonnancial Norwegian ﬁrms. The tests may reveal potential agency problems because the sequence of departures of the CEO relative to directors is different when the CEO is in control of the election of directors from the sequence when shareholders are in control. The CEO turnover and board changes provide a natural setting for studying the board endogeneity problem that has not been used earlier. I ﬁnd little evidence of CEO control, since CEO turnover and board changes tend to be simultaneous;

outside ownership concentration inﬂuences the election of directors, but not the CEO; and CEO entrenchment variables are either non-signiﬁcant or have signs pointing away from CEO control. I conclude that joint control of CEO and board is the type of control most typically used in Norwegian ﬁrms.

**Keywords:**