This study provides insights in how corporate control is exercised in poorly performing companies of two countries with markedly different equity markets : the U.K. and Belgium.

Since there seems to be little relation between hostile takeovers and poor performance in the U.K., we examine how corporate control is exerted when companies perform poorly. We find an important relation between the composition of corporate boards, share ownership and the exercise of corporate control. This is reflected in a strong relation between board turnover and concentration of share ownership in the sample of poorly performing firms and is particularly pronounced for certain classes of large outside shareholders. However, where there is substantial insider ownership, the incumbent management is more successful in retaining control following poor performance. Corporate control seems to be exerted by coalitions of shareholders. We also report trades in share stakes between shareholders in poorly performing companies. These trades are closely associated with changes in the management of poorly performing companies, which suggests that changes in the pattern of large shareholdings are an important way in which corporate control is exercised in the U.K.

In the second part of the study, we contrast the U.K. findings with the results from the study of corporate control in Belgium. Both the board of directors and large shareholders discipline managerial underperformance, but only when the company reaches critically low profitability. Specific classes of large shareholders, especially those holding majority share stakes or at least stakes of blocking minority size, are involved in disciplining underperforming management. Control is not only exerted by direct shareholders on the first ownership tier, but also by ultimate shareholders.

Furthermore, like in the U.K., poor performance triggers a market in share stakes. Those shareholder classes with superior monitoring abilities increase their shareholdings so that they can substitute errant management. Post-disciplining performance in the form of dividends per share improves over a two year period after management restructuring.

Acknowledgements :

I am most grateful to Julian Franks for his help and stimulating discussions. I am also grateful to Colin Mayer and Michel Habib for comments on earlier drafts of this study. This study was written at the London Business School. Parts of this study have been presented at conferences at the London Business School (July 1994), the London School of Economics (June 1995), the Fondazione ENI Enrico Mattei in Milan (June 1995), the Wissenschaftszentrum Berlin fur Sozialforschung (July 1995), the European Finance Association Meeting in Milan (August 1995), the CEPR European Summer Symposium in Financial Markets in Gerzensee (August 1995) and the ESRC Network for Industrial Economists (November 1995), the CEPR European Summer Symposium in Financial Markets in Gerzensee (August 1996), the European Economic Association annual conference (Istanbul 1996) and the European Finance Association annual conference (Oslo 1996). Useful comments were also provided by participants to the financial seminars at the Universities of Leuven, Tilburg and Antwerp. I am particularly grateful to our discussants Fabrizio Barca, Marco Pagano (Naples), Sigurt Vitols and Jozef Zechner (Vienna). For comments and suggestions I would like to thank Lorenzo Caprio (Milan), Francesca Cornelli (LBS), Marc De Ceuster (Antwerp), Eddy Wymeersch (Ghent), Marc Goergen (Oxford), Jay Dahya (Dundee), Andrea Berardi (LBS), Luis Correia da Silva (Oxford), Ann van Ackere (LBS), Theo Vermaelen (INSEAD), Ian Cooper (LBS) and Kazunori Suzuki (LBS). Financial support was granted by the European Commission.

CHAPTER 1 : Introduction.

PART I : Aspects of corporate control in underperforming companies listed

on the London Stock Exchange.

CHAPTER 2 : The role of large share stakes in poorly performing companies

in the U.K.

2.1 Introduction.

2.2 Hypotheses, Data and Methodology.

2.2.1 Hypotheses.

2.2.2 Data.

2.3 Results.

2.3.1 Board size, composition and turnover.

2.3.2 Concentration of ownership and board turnover.

2.3.3 Sales of share stakes.

2.3.4 Impact of composition of the board, ownership

concentration and the market for share stakes on board turnover.

2.4 Conclusions.

Appendix A

PART II : Corporate governance in Belgian companies listed on the Brussels

Stock Exchange.

CHAPTER 3 : Overview of Corporate Control Issues in Belgium.

CHAPTER 4 : Hypotheses, methodology and data.

4.1 Hypotheses.

4.1.1 Corporate performance and disciplinary corporate

governance actions.

4.1.2 The impact of board composition and structure on the

board's ability to monitor performance.

4.1.3 Ownership concentration, the costs of free riding on

control and superior monitoring abilities.

4.1.4 Dilution of control.

4.1.5 The disciplining role of the market for share stakes.

4.1.6 Post-disciplining corporate performance.

4.2 Performance benchmarks.

4.3 Methodology.

4.4 Data.

4.4.1 Sample description.

4.4.2 Ownership data.

4.4.3 Share price and accounting data.

4.4.4 Data on the board of directors and the management

committee.

CHAPTER 5 : Ownership and control of Belgian listed companies : stylized facts.

5.1 Insider versus outsider ownership and control systems.

5.2 Concentrated ownership in Belgium.

5.2.1 Ownership disclosure legislation.

5.2.2 Voting rights and restrictions, and the rights of the

minority shareholders.

5.2.3 Concentrated direct and ultimate ownership by shareholder class.

5.2.4 Pyramiding and the violation of one share-one vote rule.

5.2.5 Blocking minorities, majorities and supermajorities.

5.2.6 Belgian shareholder classes.

5.2.7 Foreign shareholder classes.

5.2.8 Changes in large shareholdings.

5.3 The board of directors and the management committee.

5.3.1 The board of directors.

5.3.2 The management committee.

5.3.3 Correlations of the turnover of CEO, of executive directors

and of the management committee.

5.3.4 Director interlocks and control.

Appendix B

Appendix C

CHAPTER 6 : Corporate control in Belgium : Empirical results.

6.1 Corporate Performance and the disciplining of management.

6.1.1 Corporate performance and turnover of the board of directors.

6.1.2 Poor company performance and CEO replacement.

6.1.3 Poor performance and management committee turnover.

6.1.4 Poor profitability and top management replacement.

6.2 The impact of the composition of the board and separation of control

on executive turnover.

6.3 Management turnover and ownership concentration.

6.3.1 Disciplining of management, ownership concentration and free

riding on control.

6.3.2 Dilution of control through multiple tier control chains.

6.3.3 Monitoring ability across shareholder classes.

6.3.4 Management turnover and large foreign shareholding.

6.4 The market for share stakes.

6.4.1 Past poor performance and changes in the ownership structure.

6.4.2 Disciplining managerial underperformance and the market for

share stakes.

6.5 An integrated model of managerial disciplining.

6.6 Post-disciplining corporate performance.

6.6.1 Management turnover and subsequent accounting profitability.

6.6.2 Management turnover and subsequent market adjusted share

price returns.

Appendix D

CHAPTER 7 : Conclusions.

Bibliography

Chapter 2

Table 2.1 : Takeover activity and number of bankruptcies in lowest and zero

abnormal return samples for the years 1985-1991.

Table 2.2 : Size distribution of the companies in the lowest and zero abnormal

return samples in 1985.

Table 2.3 : Board turnover and the frequency of chairman and CEO turnover in

the lowest and zero abnormal return samples.

Table 2.4 : The relation between board turnover, performance and the proportion

of non-executives on board.

Table 2.5 : The relation between board turnover and the separation of the role

of CEO and chairman.

Table 2.6 : Aggregate large share stakes of 5% and more held by directors,

industrial companies and institutional investors.

Table 2.7 : Distribution of the size of the share stakes per investor class.

Table 2.8 : The relation between board turnover, performance and large

shareholdings.

Table 2.9 : The relation between board turnover, performance and large

shareholdings of institutional, insider and outsider investors.

Table 2.10 : The relation between performance and changes in large share-

holdings of institutional investors, insider and outsider shareholders.

Table 2.11 : The relation between board turnover, performance and increases

of large shareholdings owned by institutional investors, by outsider

shareholders and by insider shareholders.

Appendix A (to Chapter 2) :

Table A1 : Board turnover in poorly and average performing companies

introduced on the London Stock Exchange respectively before and after 1980.

Table A2 : Aggregate large shareholdings of 5% and more by directors,

industrial companies and institutional investors.

Table A3 : The distribution of the changes in the size of substantial

shareholdings of 5% or more per shareholder class.

Table A4 : Impact of board composition, large shareholdings per investor class and

increases in share stakes on turnover of the board.

Chapter 4 :

Table 4.1 : Sample Description.

Chapter 5 :

Table 5.1 : Number of domestic quoted companies per country and the

market capitalization as a percentage of GDP.

Table 5.2 : Ownership concentration in all Belgian companies listed

on the Brussels Stock Exchange.

Table 5.3 : Largest direct and ultimate levered shareholdings, and

the control leverage factor.

Table 5.4 : Largest direct and ultimate levered shareholdings, and

the control leverage factor by ultimate investor category.

Table 5.5 : Blocking minority, majority and supermajority shareholdings.

Table 5.6 : Size of large shareholdings held by a French ultimate

investor (group).

Table 5.7 : Increases and decreases of large shareholdings over

1989-1992.

Table 5.8 : Size and turnover of the board of directors.

Table 5.9 : Shareholder representation on the board of the Nationale

Portefeuille Maatschappij (NPM).

Table 5.10 : Size and turnover of the management committee.

Table 5.11 : Pearson correlations of the turnover variables.

Table 5.12 : Pearson correlations between director interlocks and

share participations.

Table 5.13 : The relation between director interlocks and share

participations.

Appendix C (to Chapter 5) :

Table C1 : Largest direct and ultimate shareholdings, and the top

level of uninterrupted ownership chains.

Table C2 : Largest direct and ultimate shareholdings, and the top

level of uninterrupted ownership chains.

Table C3 : Changes in large shareholdings.

Table C4 : Size and turnover of the board of directors.

Table C5 : Size and turnover of the management committee. 1

Table C6 : Multiple directorships (positions on the board of directors).

Chapter 6 :

Table 6.1 : Impact of past performance on board turnover in 1989-1992

(pooled data).

Table 6.2 : Impact of past performance on turnover of the CEO in

1989-92 (pooled data).

Table 6.3 : Impact of board structure on executive board turnover

in 1989-92 (pooled data).

Table 6.4 : Impact of board structure on turnover of CEO in 1989-92

(pooled data).

Table 6.5 : Impact of aggregate concentrated ownership on board turnover

in 1989-92 (pooled data).

Table 6.6 : Impact of aggregate large shareholdings on turnover of the

board, the management committee and CEO for industrial and commercial

companies in 198-92 (pooled data).

Table 6.7 : Impact of large shareholders on turnover of the board,

management committee and CEO for 1989-1992 (pooled data).

Table 6.8 : Impact of the ultimate levered ownership on turnover data

for 1989-1992 (pooled data).

Table 6.9 : Impact of large shareholdings per investor class on turnover

of the board and CEO in 1989-1992 (pooled data).

Table 6.10 : Impact of substantial shareholdings held by French investor

groups on board turnover in 1989-92 (pooled data).

Table 6.11 : Impact of performance on increases in substantial shareholdings

in 1989-92 (pooled data).

Table 6.12 : Relation between increases in large shareholdings, performance

and board turnover in 1989-1992 (pooled data).

Table 6.13 : Post-restructuring performance : impact of CEO turnover on

changes in accounting earnings in 1989-92 (pooled data).

Table 6.14 : Post-restructuring performance : impact of CEO turnover on

market adjusted returns in 1989-92 (pooled data).

Appendix D (to Chapter 6) :

Table D1 : Impact of past performance on turnover of the management

committee in 1989-1992 (pooled data).

Table D2 : Impact of substantial shareholdings owned by French investor

groups on turnover of CEO in 1989-92 (pooled data).

Table D3 : Relation between increases in large shareholding, performance

and turnover of board and CEO in 1989-1992 (pooled data).

Table D4 : Post-restructuring performance : impact of board turnover on

market adjusted returns in 1989-92 (pooled data).

Table D5 : Integrative model of managerial disciplining in industrial and

commercial companies.

Figure 1 : Pyramidal shareholding structure of Floridienne.

Figure 2 : Shareholder structure of Cobepa.

CHAPTER 1 : Introduction.

Objective of the study.

Over the last few years, corporate governance debates in, the U.K., France and Belgium have focused on how internal corporate control mechanisms, like the non-executive component of the board of directors, ought to be changed to cope more effectively with corporate underperformance. Legal frameworks of ownership disclosure have recently also been adapted (e.g. in Belgium and Germany). This study aspires to provide insights on how corporate control is exercised in poorly performing companies in two countries with markedly different equity markets : the U.K. and Belgium.

U.K. (and U.S.) research has cast doubt on the role of takeovers on the correction of corporate underperformance. Consequently, we concentrate on alternative mechanisms which correct managerial failure. We find that, in line with the suggestions in e.g. the Cadbury report in the U.K. (Cadbury 1992) and the Viénot report in France (Viénot 1995), the composition of the board of directors and separation of the role of chairman and CEO are positively related to disciplining underperforming managers in the U.K.

Compared to Continental Europe, ownership is more widely held in the U.K., but the aggregate share stake held by shareholders owning stakes of 5 percent or more, still amounts to more than 35 percent. Therefore, we investigated the monitoring role of these large shareholders. We report that the nature of the owner is related to management turnover, when corporate performance is poor, and that coalitions of shareholders actively monitor the company. We also address the question whether poor corporate performance triggers changes in ownership. We find that a market for share stakes results from poor corporate performance and that increases in ownership by specific shareholder classes are related to replacement of management.

We subsequently focus on Belgian corporate control. Belgian equity markets have characteristics typical of other Continental European markets: few companies are quoted, ownership is highly concentrated and pyramiding is used to retain control. Another interesting aspect is the presence of holding companies.

We report whether disciplining of management is preceded by poor corporate performance and which profitability benchmarks trigger corporate control actions. We find that critical performance benchmarks, like negative earnings, dividend cuts and low share price returns are followed by management replacement.

Like in the U.K. study, we address questions on the role of the non-executive board and the impact of separation of the role of the CEO and chairman on monitoring of management. Next, we concentrate on the disciplining of management by large shareholders. We model control in several ways: the importance of direct ownership is contrasted with ultimate shareholder control, and models in which equal weight is given to each of the voting rights are compared to models with threshold shareholdings like blocking minorities, majorities and supermajorities. We find that control is exercised by ultimate shareholders and that large share stakes owned by specific shareholder categories are strongly correlated to management turnover. Poor performance also gives rise to a market of share stakes which is related to corporate control.

Finally, we investigate whether the success of disciplinary actions taken against management by examining the companies' performance after management restructuring. We find that there is a negative relation between CEO turnover and subsequent share price returns and earnings changes. However, CEO turnover precedes dividend increases.

Organization of the study.

Chapter 2 focuses on how corporate control is exerted in U.K. The motivation for the examination of aspects of corporate governance for all Belgian companies listed on the Brussels Stock Exchange starts with Chapter 3. In Chapter 4, the hypotheses are listed and embedded in the relevant literature. In addition, methodology and data sources are discussed. Chapter 5 gives stylized facts regarding ownership concentration in Belgium. The chapter begins with a comparison of ownership structures in different countries and the relation of the shareholder structure to corporate monitoring. A concise overview of recent Belgian ownership disclosure legislation is given, as well as of the laws regarding the protection of minority shareholders. Next, details about ownership concentration, critical shareholdings and control measures are presented. The chapter concludes with an description of size, composition and turnover of the board of directors and of the management committee. Chapter 6 summarizes the empirical tests of the hypotheses advanced in Chapter 4. Finally, Chapter 7 summarizes the findings of the two parts of this study - aspects of corporate control in the U.K. and Belgium - and provides some suggestions for further research.

CHAPTER 2 : The role of large share stakes in poorly performing companies in the U.K.

2.1 Introduction.

In the U.K. and U.S., corporate control is supposed to be closely associated with hostile takeover markets: poorly performing companies are acquired by other firms. Contrary to this conventional view, Franks and Mayer (1995a) in the U.K. and Martin and McConnell (1991) in the U.S. demonstrate that there is little relation between hostile takeovers and poor performance: targets of hostile bids do not appear to be, on average, poorly performing companies. This raises the question of how corporate control is exerted in the U.K. and U.S. in poorly performing companies.

This chapter compares ownership and control in two samples of U.K. quoted firms with markedly different performance; the first sample is drawn from the lowest quintile of performance, the second comes from the middle quintile. We examine and compare how each category of shareholder exerts control over executive management and how both the pattern of ownership and the composition of the board change in the light of poor performance. Shareholders include banks, institutional investors, directors and their families, and industrial and commercial companies.

We find an important relation between the composition of corporate boards, share ownership and the exercise of corporate control. This is reflected in a strong relation between board turnover and concentration of share ownership in the sample of poorly performing firms and is particularly pronounced for certain classes of large outside shareholders. There is more board turnover in poorly performing companies where there is a high proportion of non-executive directors and where there is separation of chairman and chief executive officers. However, where there is substantial insider ownership, the incumbent management is more successful in retaining control following poor performance.

We also report trades in share stakes between shareholders in poorly performing companies. These trades are closely associated with changes in the management of poorly performing companies. Although the chapter finds, like previous studies, no observed relation between performance and takeovers, it finds that changes in the pattern of large shareholdings are an important way in which corporate control is exercised in the U.K.

The results shed light on how control is changed in the U.K. where ownership is less concentrated than in continental Europe. Whereas in Germany, for example, there is frequently a single shareholder with a majority of the voting rights, in the U.K. coalitions of shareholders with stakes greater than 5% own between 35-40% of the equity capital. Findings suggest that substantial changes in these share stakes occur in the absence of tender offers or mergers and without a violation of the U.K. Takeover Code's mandatory offer rule, which requires a full bid to be made to all shareholders. As a result, the market for corporate control may be substantially broader then previously documented.

2.2 Hypotheses, Data and Methodology.

2.2.1 Hypotheses.

Little or no relation has been observed between performance and board turnover consequent on hostile takeovers in the U.K. and U.S. However, Weisbach (1988) finds that top management in U.S. corporations is replaced rapidly after poor share price performance. A similar result is reported by Warner, Watts and Wruck (1988) and Coughlan and Schmidt (1985). This section begins by examining whether there is evidence of disciplining of management of poorly performing companies.

Hypothesis 1 :

There is a higher level of board turnover in poorly performing companies than in average performing companies.

The Cadbury Committee (1992) in the U.K., the Viénot report in France (Jack 1995) and the Bacon study (1993) in the U.S. point to the importance of non-executive directors. As agents of investors, non-executive directors should assist in the monitoring and disciplining of management (Williamson 1983 and 1984). This is consistent with an external labour market for non-executive directors (Fama 1980 and Fama and Jensen 1983). The importance of this market has been emphasized by Kaplan and Reishus (1990) who find that directors of poorly performing companies are likely to lose directorships in their own companies and are unlikely to be offered new directorships in other companies. Similarly, Gilson (1990) finds that directors who left the boards of distressed companies held approximately one-third fewer directorships after their departure.

In the U.S., Weisbach (1988) reports a closer relation of CEO turnover to performance in firms in which non-executive directors dominate the board. Morck, Shleifer and Vishny (1989) argue that internally precipitated executive turnover is more likely to occur in firms that underperform their industry than when the industry as a whole is suffering.

Hypothesis 2 :

There should be a higher level of board turnover in poorly performing companies as the proportion of non-executive directors increases.

A second recommendation made by the Cadbury committee concerns the separation of the role of chief executive officer and chairman.

Hypothesis 3 :

There should be a higher level of board turnover in poorly performing companies which separate the role of chairman and chief executive officer.

Diffuse ownership encourages shirking of monitoring and control responsibilities by owners (Demsetz 1983). Shleifer and Vishny (1986) show that concentrated shareholdings can mitigate free rider problems of corporate control. As a result, we would predict more active corporate governance in the presence of concentrated ownership. Burkart, Gromb and Panunzi (1995) show that optimal ownership structure depends on the performance of a company: when it is performing well a diffuse ownership structure increases managerial discretion and enhances managerial effort, while poor performance necessitates closer monitoring through concentrated ownership.

McConnell and Servaes (1990) examine the relation between ownership concentration and corporate performance. They find that corporate performance, as measured by Tobin's Q, initially rises with low levels of ownership and then falls with high levels of ownership.

Hypothesis 4 :

In a sample of poorly performing companies, there is a greater level of board turnover in the presence of concentrated shareholdings.

Demsetz and Lehn (1985) and Barclay and Holderness (1989, 1991) note that shareholders may attach different values to control derived from concentrated ownership. For example, the exercise of control by corporate investors may be based on superior information or monitoring ability. In contrast, control by insiders, in particular directors and families, may be more heavily influenced by private benefits which are at variance with the interests of outside shareholders.

Hermalin and Weisbach (1991) find that at low levels of ownership, corporate performance increases with managerial ownership up to a level of 1% as managers' and shareholders' interests are more closely aligned. However, it decreases above this level as management is able to insulate itself from disciplinary sanctions.

Hypothesis 5 :

We would expect to observe higher board turnover in poorly performing companies with large shareholdings held by corporate investors and lower board turnover with share stakes held by directors.

In light of the predicted relation between ownership and control, we would expect to observe changing patterns of ownership over time. During periods of good performance, the advantages of managerial discretion may argue for dispersed share ownership as suggested by Burkart, Gromb and Panunzi (1995). However, greater concentrations of share ownership might be expected to emerge during periods of financial difficulty. In particular, we would expect to observe increased concentrations in the hands of those who are best placed to exercise control.

Hypothesis 6 :

In a sample of poorly performing companies, we would expect to observe increasing concentrations of ownership. In particular, we would predict increasing concentrations of ownership in the hands of corporate shareholders and new directors who are best placed to exercise control.

Hypothesis 7 :

The increased concentrations of ownership in the hands of active investors will be reflected in higher board turnover in poorly performing companies.

2.2.2 Data.

Samples of poorly and average performing companies were identified from abnormal share price returns over the period July 1984 to June 1985. Two samples of 80 companies were drawn randomly from the lowest and middle quintile of all industrial and commercial companies listed on the London Stock Exchange in 1985 and ranked according to their abnormal returns. Abnormal returns were calculated from the London Share Price Database (LSPD), which contains data on share prices, returns, market capitalization and risk measures. For the poor performers, abnormal returns, measured using a market model, were -61.35% versus -.7% for the average performers; the raw returns were -39.4% and 19.9% for the two samples respectively. Since complete data for nine companies were unavailable, they were deleted from the samples. As a result, the lowest abnormal return sample consists of 74 companies, of which 19 were taken over during the period 1985-1989 and 3 went into bankruptcy. Table 2.1 shows that, of the 77 firms in the zero return sample, 21 were acquired over the same period and 1 went into receivership. It is striking that the incidence of takeovers in the worst performing sample is similar to the average performing sample.

Table 2.2 shows the size distribution of the 151 companies used in this analysis. The size quartiles are based on 'assets employed', total assets minus short term liabilities, of all non-financial industrial companies listed on the London Stock Exchange in 1985. The average size of companies on the London Stock Exchange was £160.9m. For the 74 companies in the lowest abnormal return sample, the mean of 'total assets employed' is £97.9m, while the average size of the 77 companies in the zero abnormal return sample is £130.2m. The second panel of Table 2.2 shows that the size distribution of companies in the zero abnormal return sample is evenly distributed across the four quartiles. In contrast, 37.8% of the lowest abnormal return sample falls in the lowest size quartile, whereas only 16.3% is in the highest size quartile. Differences in size in part reflect the fact that poor performance shrinks companies' assets.

Of the 74 poorly performing companies, 34 were introduced on the London Stock Exchange over the period 1980-84 - henceforth called recent IPOs, while only 13 companies of the average performing sample were IPOs over the same period. Within the lowest abnormal return sample, half of the recent IPOs (over a period of 1980-84) had a market capitalization lower than the median.

	lowest abnormal return sample		zero abnormal return sample
	M&A	bankruptcy	M&A	bankruptcy
1985	2	0	0	0
1986	7	0	3	0
1987	4	3	7	1
1988	2	0	6	0
1989	4	0	5	0
1990	3	2	1	0
1991	5	2	1	1
total co's in sample	74		77

Source : Risk Measurement Service, London Share Price Database.

	SAMPLES
	Lowest abnormal return sample		Zero abnormal return sample
Size quartiles (1)	NUMBER	%	NUMBER	%
1 Smallest	28	37.8 %	20	25.9 %
2	20	27.0 %	20	25.9 %
3	14	18.9 %	19	24.7 %
4 Largest	12	16.3 %	18	23.5 %
Total	74	100.0 %	77	100.0 %

(1) Total assets employed, total assets minus short term liabilities, is used as size measure.

Source : Datastream.

The lowest abnormal return sample has higher betas and higher variances than the zero abnormal return sample. The Price to Earnings ratio of the poor performers is only 8.1 compared with 16.4 for the average performers. This is consistent with poor performers being higher risk and lower growth stocks than average. Most industry sectors are represented in both samples.

For each of the sample companies, data on the composition of the board of directors and on concentrated ownership were compiled from annual reports, Datastream, the Financial Times and Nexus databases for each year from 1984-91. The proportions of the executive and non-executive board members were collected, including data on the CEO and the Chairman. It was not possible to identify executive and non-executive directors separately in annual reports or other public sources of information for slightly less than half of the poor performers and about one third of the average performers.

We also measured turnover of the board and examined reasons for resignations using the annual reports from 1984 to 1991, press releases on new listings on the Stock Exchange, and the Financial Times and Nexus databases. We were, in particular interested in distinguishing between natural and enforced turnover. A resignation was classified as 'natural' if the director was described as having left the board for reasons of retirement, death or illness. Otherwise the resignation was classified as being enforced. In addition, we recorded whether the managing director and the chairman of the board resigned or whether they relinquished their functions but remained on the board.

We collected ownership data on the size of shareholdings, including stakes of 5% or more of market capitalization. We classified these large shareholdings according to 7 categories: (i) investment trusts, unit trusts and pension funds, (ii) insurance companies and (iii) banks - these three categories are combined into an institutional investor category - (iv) executive and non-executive directors and their families and trusts, (v) venture capital companies, (vi) industrial and commercial companies, and (vii) other major shareholders, mainly individuals. We will refer to directors and their families as 'insiders' and industrial and commercial company and other major shareholders as 'outsiders'.

Some companies had nominee companies as major shareholders. We contacted company secretaries or financial directors of all firms with nominee investors as major shareholders and in almost all cases were able to identify the beneficial shareholder behind the nominees. The beneficiary of the nominee company is usually an institutional investor who amalgamates a number of holdings to reduce administrative costs.

2.3 Results.

2.3.1 Board size, composition and turnover.

Board size and composition.

The size of average and poorly performing companies is similar in 1985, with a median of 7 and a mean of 7.5. It rises slightly to 8 in 1988. Executive directors form a majority of board members of both samples: non-executives on average represent 40% of the board. There is no difference in board composition of the two samples and the average number of executives and non-executive directors remains relatively stable over time at 4.5 and 3.5 respectively, with a median of 4 and 3 over the period 1985-88.

Board turnover.

We calculated the proportional total board turnover and the turnover of executive and non-executive directors for both samples. The turnover data exclude deaths, illness or retirements. Since the focus is on forms of corporate control other than takeovers or bankruptcies, we have excluded turnover of boards in the years of and subsequent to takeovers or receivership.

The first row of table 2.3 records a statistically significantly higher board turnover as a proportion of total board size in poorly performing companies compared with the average. Over a four year period from 1985 to 1988, 46% and 27% respectively of the boards of poorly performing and average companies resign. Rows 3 and 4 show that the difference in turnover is primarily associated with executive rather than non-executive directors. The turnover of the poorly performing sample appears low when compared with the board turnover of more than 80% after takeover reported by Franks and Mayer (1995a) - high board turnover in takeovers may reflect economies of scale in merging two boards of directors. The level of turnover for the average performing sample of 27% over 4 years is equivalent to about 7% per year.

We separated the sample into those firms brought to the market before and after 1980 (recent IPOs) and compared board turnover in poorly and averagely performing companies in the two samples. The results were similar to those reported above: board turnover was higher in poorly than averagely performing companies in both samples. Total board turnover of poorly performing recent IPOs over 1985-88 was, at 50%, statistically different from board turnover of average IPOs at 27% (see table A1). This suggests that high board turnover in poorly performing firms is not merely a feature of recent IPOs.

Executive turnover in poorly performing companies (excluding recent IPOs) over the period 1985-88 amounted to 40% versus non-executive turnover (at 23%). With regard to recent IPOs, however, there was some evidence of a higher proportion of non-executive board turnover suggesting that the executive directors of recent IPOs may be more firmly entrenched than those in companies which have been quoted for longer periods.

Company size is not significantly correlated to board turnover: board turnover in poorly performing companies with a market capitalization below the sample median is similar to that of the larger companies.

For both the lowest and the zero abnormal return samples, the turnover of the board, of the executive directors, of non-executives, of chairmen and of CEOs are shown for the period 1985-88. Sample comparison is performed via t-statistic and the Wilcoxon-Mann-Whitney test. N stands for the number of companies in the sample, STD for the standard deviation and W-M-W p for the p-value of the Wilcoxon-Mann-Whitney test. Calculations are based on data from annual reports.

	LOWEST ABNORMAL RETURN SAMPLE			ZERO ABNORMAL RETURN SAMPLE			STATISTICS
	N	%	STD	N	%	STD	T-stat	W-M-W:p
PANEL A : TURNOVER OF ALL DIRECTORS, EXECUTIVE DIRECTORS AND NON-EXECUTIVE DIRECTORS (proportional to total number of directors, of executives and of non-executive directors respectively).
board turnover in 1985-881	55	45.7	32.9	57	26.6	26.7	3.157	0.002
board turnover in 1985-881 4	23	33.3	29.5	45	25.0	23.8	1.253	0.279
executive turnover in 1985-882	23	35.8	31.1	45	23.1	27.3	1.733	0.077
non-executive turnover in 1985-883	23	27.2	37.8	45	26.9	33.1	0.032	0.974
PANEL B : PERCENTAGE OF COMPANIES IN WHICH AT LEAST ONE, TWO OR THREE CEO(S) LEAVE THE COMPANY.
turnover of one CEO in 1985-88	55	33.3	47.6	57	16.9	37.8	2.057	0.043
turnover of two CEOs in 1985-89	51	9.8	30.0	56	0.0	0.0	2.496	0.017
turnover of three CEOs 1985-89	51	5.9	23.8	56	0.0	0.0	1.856	0.069
PANEL C : PERCENTAGE OF COMPANIES IN WHICH AT LEAST ONE, TWO OR THREE CHAIRMEN LEAVE THE COMPANY.
turnover of one chairman in 1985-88	55	38.6	49.1	57	22.0	41.8	1.958	0.053
turnover of two chairmen in 1985-89	51	8.8	28.5	55	3.4	18.3	1.214	0.228
turnover of three chairmen in 1985-89	51	3.9	19.6	55	1.8	13.4	0.664	0.513

1. turnover is proportional to total number of directors.

2. executive turnover is proportional to total number of executive directors.

3. non-executive turnover is proportional to total number of non-executive directors.

4. sample size was reduced with those companies for which data were not available on which directors had executive or non-executive functions.

CEO and chairman turnover

Table 2.3 shows the turnover of managing director/CEO and chairman for reasons other than retirement, illness and death. It reports that the incidence of resignations of managing directors is significantly higher in the worse performing companies than average performers for the period 1985-88. There are no cases in the averagely performing sample of a CEO being replaced more than once but a significantly positive number of cases in the worst performing sample.

Table 2.3 records a higher turnover of chairmen of poorly performing firms: 39% of chairmen in poorly performing companies leave the board for reasons other than natural retirement over the period 1985 to 1988, while only 22% of chairmen of average performing companies leave. There is therefore prima facie evidence of a disciplining function of both CEOs and chairmen of boards as predicted by hypothesis 1.

The relation between board turnover and performance was investigated further by regressing board turnover in the years 1985 to 1989 on annual abnormal share price returns in 1984-85. For all years apart from 1986, there is a statistically significant negative relation at the 5% level in poorly performing companies. There is a statistically significant relation at the 0.1% level over the whole period 1985-1989 in poorly performing firms but not in average performers. The evidence is consistent with hypothesis 1, that board turnover results from a disciplining process when companies perform poorly.

Non-executives and board turnover

Table 2.4 examines whether there is a relation between board turnover and the proportion of non-executive directors on the board. There is a clear statistically significant relation in the poorly performing sample for three out of the five years from 1985 to 1989. The relation is significant for the whole period at the 5% level. However, there is no statistically significant relation in the average performing sample. Consistent with hypothesis 2, the influence of non-executive directors is therefore particularly pronounced in poorly performing companies. This finding is consistent with the emphasis placed by the Cadbury committee on the presence of non executive directors on the board of companies.

Separation of chairman and chief executive and board turnover

Table 2.5 examines whether the separation of chairman and CEO is an important contributor to corporate governance. There is little relation in the overall sample between separation and board turnover. However, partitioning the samples of poorly performing and average companies reveals a significant relation, independent of company size, over the period 1985 to 1988 in the poorly but not averagely performing companies. The importance of separation of control is limited to poorly performing companies which were not recent IPOs. As described below, this is related to the particular ownership patterns of those companies. Consistent with hypothesis 3, there is a clear relation between separation of functions, disciplining and the performance of firms.

This table shows the results of the relation between the proportion of non-executive directors on board and board turnover. Turnover data are collected from the annual reports and performance data are from the London Share Price Database (LSPD).

1

where i = period of time, TURN_i stands for board turnover and NONEXi represents the percentage of non-executive directors on board over the period 1985-88.

Dependent Variable	Independent Variable
	Intercept	% non-executives on board	Sample Size	R2 adj.	Prob>F
Panel A : Lowest Abnormal Return Sample
1. Board turnover 85-88	0.009 (0.790)	0.175 (0.062)	23	11.1	0.062
2. Board turnover 85-89	-0.009 (0.830)	0.257 (0.017)	20	22.6	0.023
Panel B : Zero Abnormal Return Sample
1. Board turnover 85-88	0.072 (0.003)	-0.022 (0.648)	44	0.0	0.648
2. Board turnover 85-89	0.083 (0.000)	-0.036 (0.444)	43	0.0	0.444

Note : Parameter estimates of the % non-executives in 000, p-values are in parentheses.

This table shows the results of the relation between board turnover and the separation of the functions of CEO and chairman. Turnover data are collected from annual reports, performance data are from the London Share Price Database (LSPD).

2

TURN stands for board turnover. SEPAR_i is a dummy variable which stands for separation of control : 0 means that the functions of chairman and CEO are divided over two directors while 1 represents unitary control over the period 1985-88.

Dependent Variable	Independent Variables		Sample Size	R2 adj.	Prob>F
	Intercept	Separation of control
PANEL A : Lowest Abnormal Return Sample
Board Turnover 1985-88	141.832 (0.000)	-51.422 (0.044)	69	4.5	0.044
PANEL B : Zero Abnormal Return
Board Turnover 1985-88	71.838 (0.000)	23.544 (0.203)	74	0.1	0.203

Notes : All parameter estimates in 000, p-values are in parentheses.

2.3.2 Concentration of ownership and board turnover.

Table 2.6 reports the incidence of concentrated shareholdings of 5% or more. There is no significant difference in overall levels of concentration in the two samples: aggregate ownership stakes of 5% and more amount to between 35 and 42% in each sample (Panel A). Concentrations of shareholdings in Continental European countries are typically much higher. For instance, 84% of Italian companies with more than 1000 employees have a single shareholder owning a majority of the shares (Bianco, Gola and Signorini 1995). In Chapter 4, we show that in 93% of Belgian industrial companies listed on the Brussels Stock Exchange a single shareholder owns a block of at least 25% of voting rights. Franks and Mayer (1995b) report that more than 25% of shares are held by a single shareholder in 85% of the largest German quoted companies.

Subsequent panels record concentrated ownership of the most important shareholding categories. Insiders (managers, directors and their families) own the largest combined ownership stake (panel B): in both samples directors own between 20 and 25% of market capitalization in the period 1985-88. There is no statistical difference in director ownership of the poor and averagely performing sample.

There is, however, a statistically significantly higher level of ownership by industrial companies in poor than in average performers (panel C) over the period 1985-88. Stakes in average companies are between 3 and 4% while they amount to around 10% in poorly performing companies. In contrast, panel D records that institutional ownership is lower in poorly performing companies. Nowhere are holdings by banks found to be significant; on average, banks own less than one percent of reported share stakes. Large stakes held by venture capital companies are also relatively rare: their average reported holdings are less than 0.5%.

This table summarizes the aggregated large shareholdings for the years 1985 and 1988 for both the poorly and averagely performing sample companies. N, MEAN and STD stand for respectively, the number of sample companies, the mean of aggregate concentrated ownership and the standard deviation. W-M-W : p stands for the p-value of the Wilcoxon-Mann-Whitney test. Data are collected from the annual reports.

	LOWEST ABNORMAL RETURN SAMPLE			ZERO ABNORMAL RETURN SAMPLE			STATISTICS
	N	%	STD	N	%	STD	T-STAT	WMW : p
PANEL A : OWNERSHIP OF ALL SHAREHOLDERS WITH A STAKE OF 5 % OR MORE
1985	70	39.021	24.994	75	34.846	25.581	0.994	0.305
19851	56	39.858	25.360	59	35.689	25.689	0.878	0.364
1988	56	42.342	22.905	59	38.361	27.157	0.847	0.358
PANEL B : OWNERSHIP OF DIRECTORS AND THEIR FAMILIES WITH A STAKE OF 5% OR MORE
1985	70	24.211	25.657	75	21.164	23.834	0.740	0.486
19851	56	25.632	26.537	59	20.705	22.733	0.878	0.365
1988	56	20.207	21.632	59	23.167	26.193	-0.659	0.689
PANEL C : OWNERSHIP OF INDUSTRIAL AND COMMERCIAL COMPANIES WITH A STAKE OF 5% OR MORE
1985	70	8.101	16.441	75	3.044	9.260	2.261	0.013
19851	56	8.371	17.093	59	3.869	10.302	1.720	0.085
1988	56	10.608	21.065	59	3.886	90994	2.204	0.050
PANEL D : OWNERSHIP OF INSTITUTIONAL INVESTORS WITH A STAKE OF 5% OR MORE
1985	70	5.042	8.842	75	8.503	13.755	-1.815	0.041
1985	56	4.037	7.486	59	8.864	14.899	-2.166	0.028
1988	59	9.889	12.949	59	9.005	13.004	0.365	0.715

Note : the class of total institutional investors consists of unit trusts, investment trusts, pension funds, insurance companies and banks.

1. The sample in 1985 was reduced to those companies with data available in 1988.

The ownership structures of recent IPOs and of companies brought to the market before 1980 are detailed in table A2 (appendix A). The aggregate ownership of poorly and averagely performing IPOs in 1985 is 46% while on average only about 33% of shares of non-IPOs are held in the form of large stakes (statistical significance within the 1% level). The main reason for ownership discrepancies between IPO's and non-IPO's is the share stakes held by insiders (directors and their families). In 1985, about two thirds of the shares of IPO's are held by directors, whereas insiders only possess 18% of non-IPO shares (1% significance level).

Insiders' ownership is reduced by an average of only 5% between 1985 and 1988 in poorly performing IPO's, but declines sharply from 34% to 22% in averagely performing IPO's. In contrast, the aggregate ownership of directors in averagely performing non-IPO's increases (but not significantly so) from 18% to 23% over a period of 1985-88, whereas the cumulative shareholdings in poorly performing companies declines. The reductions in insider's shareholdings is compensated by increases in ownership by the industrial companies and institutional investors.

In the lowest abnormal return sample, the ownership structure of the smallest companies, defined as firms with a market capitalization below the median of the sample, is different from that of the larger firms. Total aggregate ownership amounts respectively, to 35% and 45%. Whereas average corporate and institutional ownership levels are similar in both subsamples at respectively about 9% and 8%, directors own in 1985 only an average of 17% in the smallest companies versus 31% in the large companies. Differences are reduced over time when insiders' ownership diminishes to 24% (in 1988).

Table 2.7 presents the distribution of ownership stakes per investor class. In about 40% of the poorly performing companies, directors hold share stakes of 25% or more. Most of these stakes are held in recent IPOs. Industrial and commercial companies also hold large stakes while institutional investors (predominantly, investment and insurance companies) typically own shareholdings of less than 10%. Average performers show a similar pattern but with smaller holdings by industrial and commercial companies.

Table 2.8 examines the relation between board turnover and concentrations of shareholdings. It reports the results of regressions of board turnover on performance and concentration. The concentration variable is found to be insignificant in the combined samples. However, there is some evidence of a relation in some years in poorly performing companies and a significant interaction term between concentration and performance. While there is therefore no evidence of a higher concentration of ownership in poorly performing companies, there is evidence of a higher turnover of executives in poorly performing companies in the presence of concentrated shareholdings, a result which is consistent with hypothesis 4. This suggests that disciplining of management is facilitated when free riding on corporate control is limited due to strong ownership concentration. There is evidence in table 2.8 of a negative relation between board turnover and concentration of ownership in average performing companies. Table 2.9 sheds some light on what might be contributing to this. It examines the influence of particular categories of large shareholders on board turnover. Ownership is disaggregated into three classes: institutional investors (banks, insurance companies and pension funds), outsiders (industrial and commercial companies, individual and family investors) and insiders (directors).

This table records the number of companies in which particular investor classes hold shareholdings of a specific size in 1985. The number of IPOs (introduced on the London Stock Exchange over the period 1980-84) with share stakes of a specific size is shown in parentheses.

Total number of companies is 74 for the lowest abnormal return sample (panel A) and 77 in the zero abnormal return sample (panel B). There is a total of 34 IPOs in panel A and of 13 in panel B. Data were collected from the annual reports.

1985	[5%,10%[	[10%,15%[	[15%,25%[	[25%,50%[	[50%,100%]
PANEL A : lowest abnormal return sample
institutional investors	22 (10)	4 (1)	3 (1)	3 (0)	0 (0)
industrial co's	4 (2)	2 (0)	6 (4)	7 (4)	2 (0)
individuals	1 (0)	2 (1)	1 (1)	1 (1)	0 (0)
directors	6 (2)	4 (0)	7 (3)	14 (9)	15 (10)
PANEL B : zero abnormal return sample
institutional investors	29 (3)	8 (3)	7 (1)	3 (0)	2 (0)
industrial co's	3 (0)	1 (1)	0 (0)	6 (0)	0 (0)
individuals	5 (1)	3 (0)	3 (1)	0 (0)	0 (0)
directors	4 (2)	8 (1)	5 (1)	15 (4)	13 (4)

Source : Own calculations based on annual reports.

This table analyzes the relation between board turnover and the presence of a concentrated ownership structure in both the lowest and the zero abnormal return sample.

3

TURN stands for board turnover. PERFORM stands for abnormal returns of 1985. TOTOWN stands for the aggregative large shareholdings (of 5% or more) as a percentage.

Data on turnover and shareholdings are collected from the annual reports. Performance data come from the London Share Price Database (LSPD). parentheses.

Dependent Variable	Independent Variables				Sample Size	R2 adj.	Prob>F
	Intercept	Abnormal Return 1985	Total concentr. ownership over 85-88	(abn. return * tot. concentr. ownership over 85-88
PANEL A : all companies
Board Turnover in 1985-88	86.875 (0.000)	-0.824 (0.000)	-0.334 (0.329)		109	9.6	0.002
Board Turnover in 1985-88	106.254 (0.000)	0.065 (0.896)	-0.855 (0.046)	-0.022 (0.046)	109	12.1	0.001
PANEL B : Lowest Abnormal Return Sample
Board Turnover in 1985-88	97.545 (0.001)		0.640 (0.304)		52	0.1	0.304
PANEL C : Zero Abnormal Return
Board Turnover in 1985-88	107.550 (0.000)		-0.846 (0.029)		56	6.8	0.029

Notes : All parameter estimates in 000, p-values are in parentheses.

Panel B reports that there is evidence of significantly higher turnover of boards of poorly performing companies in the presence of concentrated share stakes held by outsiders. In contrast, directors who have large shareholdings reduce board turnover in both average and poor performing companies. The negative relation between board turnover and concentration in average performing companies would therefore appear to reflect the influence of insiders. This is most pronounced in recent IPOs where, as noted above, holdings by directors are particularly large. This is consistent with directors owning substantial shareholdings impeding board changes to protect the private benefits they derive from control. In the case of non-IPOs, directors appear to be unable to inhibit board turnover in poorly performing companies. However, in recent IPOs, there is a negative relation between board turnover and directors' holdings even in poorly performing companies. There is no significant relation between shareholdings of institutional investors and board turnover in either average or poorly performing companies. We also investigated the interrelation between performance and ownership concentration of the three main shareholder categories by including interaction variables (performance * ownership by each category of investor). Only the interactive term with outside ownership concentration was statistically negatively significant, confirming that board turnover is high in the presence of poor performance and large outside ownership. The interactive term with director ownership was not significant, consistent with the observation that directors impede board turnover in both average and poorly performing IPOs.

In summary, we find clear support for hypothesis 5: board turnover is higher in poorly performing companies in the presence of outsider (non-institutional) shareholders and lower in the presence of insider shareholdings. The influence of insiders is most pronounced in recent IPOs where their holdings are particularly large. Quite strikingly in light of the debate on the role of institutional investors, institutions do not appear to be involved in disciplining poorly performing management.

This table reports whether board turnover is associated to the presence of shareholdings of specific shareholder categories : institutional investors, outsider and insider shareholders.

4

where i stands of the period. PERFORM, TURN, INSTIT, OUT AND INSIDE respectively stand for the performance criterion (abnormal returns over 1985), board turnover, and the concentrated shareholdings of institutional investors, outsider shareholders (industrial and commercial companies and individual large shareholders) and insider shareholders (directors and their families). Concentrated shareholdings of the shareholder classes is averaged over the period 1985-88. IPOs are the companies introduced on the London Stock Exchange in the period 1980-84. Non-IPOs are all sample companies floated before 1980. Turnover and shareholdings are collected from the annual reports. Performance data are from the London Share Price Database (LSPD). All parameter estimates are in 000s, p-values are given between parentheses.

Dependent Variable		Independent Variables					Sample Size	R2 adj.	Prob>F
		Intercept	Abnormal Return 1984-85	Concentr. ownership held by instit. investors	Concentr. ownership held by outsiders	Concentr. ownership held by insiders
Panel A : all companies
Board Turnover in 1985-88	all	82.198 (0.000)	-0.667 (0.003)	-0.398 (0.544)	1.887 (0.010)	-1.599 (0.005)	109	15.9	0.000
Board Turnover in 1985-88	IPOs	117.616 (0.011)	-0.265 (0.575)	-0.718 (0.753)	1.801 (0.181)	-2.535 (0.040)	35	5.5	0.225
Board Turnover in 1985-88	non-IPOs	85.193 (0.000)	-0.738 (0.023)	-0.006 (0.992)	0.604 (0.305)	-1.503 (0.023)	74	13.6	0.006
Panel B : lowest abnormal return
Board Turnover in 1985-88	all	87.436 (0.002)		0.559 (0.691)	1.697 (0.023)	-1.615 (0.022)	52	6.6	0.009
Board Turnover in 1985-88	IPOs	149.967 (0.001)		-2.163 (0.377)	1.977 (0.112)	-2.734 (0.018)	27	11.9	0.112
Board Turnover in 1985-88	non-IPOs	52.668 (0.238)		2.028 (0.346)	2.239 (0.050)	-1.709 (0.182)	25	5.4	0.25
Panel C : zero abnormal return
Board Turnover in 1985-88	all	98.882 (0.000)		-0.059 (0.927)	-0.425 (0.564)	-1.811 (0.025)	57	12.1	0.020
Board Turnover in 1985-88	non-IPOs	104.095 (0.000)		-0.372 (0.607)	0.538 (0.547)	-1.766 (0.075)	48	7.7	0.073

2.3.3 Sales of share stakes.

Although the traditional market for corporate control appears to be unrelated to performance, we find a market in share stakes in poorly performing companies. We divided shareholders into 'old' and 'new' shareholders for each main shareholder category. New shareholders are those who acquire a shareholding of at least 5% in the current year; old shareholders held share stakes of 5% or more in previous year. We investigate three directions of change in ownership patterns: decreases and increases in holdings of 'old' investors and the emergence of 'new' shareholders with reported share stakes.

Over the period 1985-87, old directors decreased their shareholdings in 39 poorly performing companies by more than 5% while they increased their shareholdings in only 6 companies (see table A3 in appendix A). In 16 companies, new directors acquired stakes of more than 5%. The number of companies where outside investors significantly decreased their share stakes was balanced by purchases by industrial and commercial companies. The pattern of share stake sales for the zero abnormal return sample was similar to that of poorly performing companies.

Splitting the sample into recent IPOs and non-IPOs reveals greater sales of director holdings in IPOs. This is particularly pronounced in average performing companies: in 1985 the average size of director holdings was 34% - by 1988 this had fallen to 22%. The share stakes are purchased by industrial companies and institutional investors.

Table 2.10 analyses whether a market for shareholdings is triggered by performance. Panel A records that institutional investors neither reduce nor increase their holdings in poorly performing companies. However, there is some evidence of new large institutional investors emerging in poorly performing companies.

Panel B of table 2.10 reveals that there is no evidence of old outside shareholders increasing their stakes in poorly performing companies. Instead, there is a significant relation between sales by old outside shareholders and purchases by new outside shareholders and poor performance. This suggests that there is a market in share stakes in poorly performing companies with old industrial investors selling out to other companies rather than exercising control themselves.

Panel C of table 2.10 records that insiders sell out of poorly performing companies. It reveals that there is some evidence (at the 10% level) that increased holdings are associated with new directors. This suggests that the sale of share stakes by old directors is part of a change in corporate control in the face of deteriorating corporate performance rather than simply portfolio diversification by directors.

In sum, consistent with hypothesis 6, we observe a market in share stakes in poorly performing companies. Active shareholders - corporate investors and directors - trade shares in poorly performing companies: old corporates and directors sell out to new.

Table 2.11 examines whether increases of ownership stakes have an impact on turnover of board members. We focus on three categories of investors: institutional, outsider and insider investors. The independent variables relate to the combined holdings of old and new shareholdings in the relevant category.

This table reports whether changes in large shareholdings over the period 1985-88 are correlated to abnormal returns in 1984-85.

5

6

7

CHINSTIT, CHOUT AND CHDINSIDE stand respectively for changes in concentrated ownership of institutional investors, of outsider shareholders (industrial and commercial co's and individual and family investors), and of insiders shareholders (directors and their families and trusts). We distinguish among three kinds of changes : 1. decreases in shareholdings by the existing (old) shareholders, 2. increases in shareholdings by the existing (old) shareholders and 3. ownership stakes of the new shareholders (these are the shareholders who had no shares or shareholdings under 5% in previous period). The changes are the average of yearly changes over the period 1985-88. Turnover and shareholdings are collected from the annual reports. Performance are from the London Share Price Database (LSPD). Parameter estimates of abnormal return are in 000s, p-value is given between parentheses.

Dependent Variable	Independent Variable				Sample Size	R2 adj.
	Intercept	p-value of intercept	Abnormal Return 1984-85	p-value of abnormal return
Panel A : Changes in large shareholdings of institutional investors.
decreases	1.652	0.000	7.344	0.296	109	0.0
old increases	0.181	0.003	-0.925	0.466	109	0.0
new increases	1.205	0.000	-16.819	0.017	109	4.1
Panel B : Changes in large shareholdings of industrial and commercial companies, and individual and family investors.
decreases	1.176	0.480	-22.779	0.028	109	3.3
old increases	0.120	0.518	-4.237	0.287	109	0.0
new increases	0.949	0.096	-38.740	0.002	109	7.5
Panel C : Changes in large shareholdings of directors and their families.
decreases	1.381	0.015	-47.194	0.000	109	11.3
old increases	0.599	0.001	2.285	0.560	109	0.0
new increases	1.094	0.013	-14.944	0.100	109	1.3

Table 2.11 records the relation between turnover of the board and increases in concentration of shareholdings in the total sample. The table demonstrates that the relation between board turnover and share ownership changes is particularly pronounced for outsider investors and directors: there is a significantly higher level of board turnover in poorly performing companies where there are increases in large share stakes held by industrial and commercial companies and directors. This suggest that where disciplining of management is required, concentration emerges in the hands of those best placed to exert it. There is little relation between board turnover and increases in institutional investor stakes. Instead, the table reveals that the relation between increased institutional holdings and board turnover is most pronounced in the average performing companies.

Consistent with hypothesis 7, we therefore find evidence of a relation between the exercise of corporate control and increased share ownership by active investors, in particular corporate shareholders and new directors. There may also be a parallel between the high board turnover found by Franks and Mayer (1995a) in hostile takeovers of averagely performing companies and the higher board turnover of averagely performing companies with increased institutional investor stakes. The latter may be motivated by attempts to enhance value in averagely performing companies through changes in corporate control.

2.3.4 Impact of composition of the board, ownership concentration and the market for share stakes on board turnover.

In this section, we have shown that a high number of non-executive directors, high board concentration of specific shareholder classes and increases of share stakes are positively correlated to board turnover in poorly performing firms. In table A4, we investigate, by including those variables into one model, which of these effects prevail. We find that the results of separate analyses remain valid : it seems that when performance is poor, (i) a higher proportion of non-executive directors on the board facilitates the disciplining of underperforming management, (ii) that concentrated ownership held by outside shareholders is positively correlated to board turnover, while directors who own substantial shareholdings impede board turnover, and (iii) that increases in share stakes held by outsiders and new directors coincides with increased turnover.

This table reports the relation between board turnover and increases in the share stakes owned by large shareholders.

8

TURN and PERFORM stand respectively for board turnover (as a percentage of total board size) and the performance criterion (abnormal return 1985). ININSTIT, INOUT and ININSIDE represent increases in concentrated ownership by the existing and new investors of the following shareholder classes: institutional investors, outsider shareholders (industrial and commercial companies and individual holdings), and insider shareholdings (directors and their families and trusts). Data on turnover and shareholdings are collected from the annual reports. Performance data are from the London Share Price Database (LSPD).

Dependent Variable	Independent Variables					Sample Size	R2 adj.	Prob>F
	Intercept	Abnormal Return 1984-85	Concent. ownership of instit. investors	Concent. ownership of outsiders	Concentrated ownership of insiders
PANEL A : all companies
Board Turnover 85-88	55.532 (0.000)	-0.453 (0.045)	3.928 (0.124)	5.356 (0.001)	4.102 (0.036)	109	21.6	0.000
Board Turnover 85-89	51.574 (0.000)	-0.248 (0.351)	6.436 (0.048)	6.498 (0.012)	4.628 (0.079)	70	24.5	0.000
PANEL B : Lowest Abnormal Return Sample
Board Turnover 85-88	39.951 (0.240)	-0.602 (0.228)	-1.058 (0.725)	7.005 (0.000)	9.602 (0.001)	52	33.6	0.000
Board Turnover 85-89	11.395 (0.738)	-0.613 (0.266)	2.725 (0.462)	7.995 (0.008)	17.669 (0.000)	24	64.7	0.000
PANEL C : Zero Abnormal Return
Board Turnover 85-88	56.373 (0.000)	0.746 (0.632)	12.840 (0.007)	1.405 (0.714)	1.031 (0.706)	57	7.2	0.097
Board Turnover 85-89	68.028 (0.000)	0.873 (0.545)	5.875 (0.223)	5.083 (0.208)	-3.059 (0.336)	45	3.2	0.260

Notes : All parameter estimates in 000, p-values are in parentheses.

2.4 Conclusions.

This chapter has reported a strong relation between the exercise of corporate control in the U.K. in the form of board turnover and corporate performance. This result stands in marked contrast to evidence on hostile takeovers in the U.K. and U.S.: the incidence of takeovers in the sample of the worst and averagely performing companies in the U.K. in 1985 is about the same.

If takeovers do not perform a corporate governance function, how is corporate control exercised? The chapter has recorded a number of important influences. Firstly, consistent with recent recommendations about improved corporate governance and the literature on principal-agent relations, the presence of non-executive directors and the separation of the role of chairman and chief executive exert significant influences on corporate governance.

Secondly, consistent with the literature on free rider problems and large share stakes, concentrated ownership is associated with more active corporate governance than dispersed share ownership. However, we also find that the nature of the owner is of critical importance: corporate investors exercise more control than institutional investors and those with private benefits of control, such as directors, may impede the exercise of good governance. Managerial entrenchment is most in evidence in recent IPOs where director shareholdings are particularly high.

But perhaps the most interesting observation relates to the dynamic relation between ownership, control and performance. Where poor performance is observed, sales of share stakes occur between different investors. In particular, there is a market in shares between new and old non-institutional shareholders and directors. These trades in shares are associated with significant changes in boards of poorly performing companies.

There are some interesting parallels between the U.K. and Germany. While levels of concentration of ownership are much greater in Germany than in the U.K., there is more evidence of an influence of concentration of ownership on the disciplining of management in the U.K. than in Germany. This may result from the private benefits of control of large shareholders in Germany impeding the exercise of corporate control in an analogous fashion to the negative influence of directors' holdings on board turnover in the U.K. Takeovers are not the method by which managerial discipline is imposed in either country; instead, partial sales of share stakes are more closely associated with the exercise of corporate governance in both countries.

An important issue concerns the ability of coalition formations in the U.K. to overcome impediments to changes in control from regulatory rules. The Takeover Code, for example, imposes mandatory bid requirements once share stakes of more than 30% have been accumulated. The ability to circumvent such rules through the formation of coalitions may come at the expense of the minority shareholders whom regulatory rules are designed to protect. On the other hand, the ability of large shareholders to exercise control at low cost may be an important contribution to good corporate governance.

	LOWEST ABNORMAL RETURN SAMPLE			ZERO ABNORMAL RETURN SAMPLE			STATISTICS
	N	%	STD	N	%	STD	T-stat	W-M-W: p
PANEL A : TURNOVER OF ALL DIRECTORS, EXECUTIVE DIRECTORS AND NON-EXECUTIVE DIRECTORS : in all companies excluding those IPO's of the period 1980-84.
board turnover in 1985-881	27	41.5	35.2	49	28.1	25.8	1.899	0.061
board turnover in 1985-881 4	14	34.4	33.0	40	25.8	23.7	1.044	0.488
executive turnover in 1985-882	14	40.4	32.6	40	22.7	26.1	1.038	0.043
non-executive turnover in 1985-883	14	22.3	40.4	40	30.2	33.6	-0.721	0.234
PANEL B : TURNOVER OF ALL DIRECTORS, EXECUTIVE DIRECTORS AND NON-EXECUTIVE DIRECTORS : in all companies introduced on the London Stock Exchange in the period 1980-1984.
board turnover in 1985-881	28	49.7	30.7	8	26.2	33.5	1.872	0.073
board turnover in 1985-881 4	9	31.6	24.7	5	18.4	25.9	0.944	0.418
executive turnover in 1985-882	9	28.7	28.8	5	25.3	39.4	0.177	0.728
non-executive turnover in 1985-883	9	34.7	34.2	5	00.0	00.0	2.288	0.032

Source : Own calculations based on annual report.

N : number of companies in the sample.

STD : standard deviation.

W-M-W p : p-value of the Wilcoxon-Mann-Whitney test.

1. turnover is proportional to total number of directors.

2. executive turnover is proportional to total number of executive directors.

3. non-executive turnover is proportional to total number of non-executive directors.

1. sample size was reduced with those companies for which data were not available on which directors had executive or non-executive functions.

This table shows the average shareholdings for both recent IPOs and non-IPOs. The mean stands for the average percentage of concentrated ownership (> 5%) of the sample companies per investor class. IPOs are for those companies that were introduced on the London Stock Exchange during 1980-1984. Non-IPOs were floated before 1980.

	LOWEST ABNORMAL RETURN SAMPLE			ZERO ABNORMAL RETURN SAMPLE			STATISTICS
	N	%	STD	N	%	STD	T-STAT	WMW:p
PANEL A : OWNERSHIP OF ALL SHAREHOLDERS WITH A STAKE OF 5 % OR MORE
1985 non-IPOs	38	33.310	26.645	62	33.200	25.475	0.020	0.983
1988 non-IPOs	27	37.555	24.482	51	38.647	28.375	-0.169	0.981
1985 IPOs	32	45.796	21.345	13	42.661	25.616	0.421	0.782
1988 IPOs	29	46.800	20.767	8	36.537	18.847	1.259	0.260
PANEL B : OWNERSHIP OF DIRECTORS AND THEIR FAMILIES WITH A STAKE OF 5% OR MORE
1985 non-IPOs	38	17.463	23.554	62	18.422	22.506	-0.203	0.999
1988 non-IPOs	27	12.900	17.544	51	23.394	27.106	-1.817	0.133
1985 IPO's	32	32.225	26.095	13	34.238	26.531	-0.233	0.706
1988 IPOs	29	27.010	23.100	8	21.725	20.828	0.583	0.838
PANEL C : OWNERSHIP OF INDUSTRIAL AND COMMERCIAL COMPANIES WITH A STAKE OF 5% OR MORE
1985 non-IPOs	38	8.623	18.913	62	3.493	10.037	1.770	0.076
1988 non-IPOs	27	10.496	22.755	51	3.639	9.928	1.852	0.089
1985 IPOs	32	7.481	13.189	13	0.900	3.244	1.766	0.114
1988 IPOs	29	10.713	19.769	8	5.462	10.970	0.716	0.481
PANEL D : OWNERSHIP OF INSTITUTIONAL INVESTORS (1) WITH A STAKE OF 5% OR MORE
1985 non-IPOs	38	6.613	10.343	62	9.324	14.706	-0.994	0.234
1988 non-IPOs	51	12.485	15.242	51	9.096	13.044	1.029	0.354
1985 IPOs	32	3.171	6.299	13	4.584	6.867	-0.664	0.473
1988 IPOs	29	7.472	10.054	8	8.425	13.612	-0.219	0.968

Source : Own calculations based on annual reports.

Note : the class of total institutional investors consists of unit trusts, investment trusts, pension funds, insurance companies and banks.

N : number of companies in the sample.

STD : standard deviation.

W-M-W p : p-value of the Wilcoxon-Mann-Whitney test.

This table shows the number of increasing or decreasing share stakes over the period 1985-87. Old investors are defined as investors who owned a substantial share stake of at least 5% in the previous year. New shareholders did not hold a share stake of 5% or more in previous year but acquire shareholding so that their shareholding reaches the 5% ownership notification threshold. The class of total institutional investors consists of unit trusts, investment trusts, pension funds, insurance companies and banks.

1985-87		[0.2%,5%[	[5%,10%[	[10%,25%[	[25%,50%[	[50%,100%]
PANEL A : lowest abnormal return sample
increase old investors	instit. investors	15	2	1	0	0
	industrial co's	2	3	0	0	1
	individuals	2	0	0	0	0
	directors	8	4	1	1	0
increase new investors	instit. investors	0	7	9	2	0
	industrial co's	0	11	6	6	4
	individuals	0	4	3	1	0
	directors	0	2	8	5	1
decreases old investors	instit. investors	14	19	8	1	1
	industrial co's	10	12	11	5	1
	individuals	1	2	4	0	0
	directors	35	12	20	4	3
PANEL B : zero abnormal return sample
increase old investors	instit. investors	20	1	0	0	0
	industrial co's	6	2	0	0	0
	individuals	6	0	0	0	0
	directors	29	4	1	0	0
increases new investors	instit. investors	0	2	9	1	0
	industrial co's	0	3	3	2	0
	individuals	0	5	2	0	0
	directors	0	3	4	1	1
decreases old investors	instit. investors	18	22	11	2	0
	industrial co's	2	3	1	3	0
	individuals	5	6	3	0	0
	directors	46	13	5	2	1

Source : Own calculations based on annual reports.

This table shows, for both averagely and poorly performing sample companies, the regression of board turnover on the following independent variables : 1. a company size variable (log of total assets), 2. the proportion of non-executive directors on board, 3. separation of control (dummy equals 1 if the functions of CEO and chairman are combined), 4. the aggregate share stakes for the investor classes of institutional investors (banks, insurance companies and investment companies), outsiders (industrial and commercial companies and individual investors) and insiders (directors and their families) and 5. increases in share stakes for the same investor classes.

Between brackets, under the parameter estimates of the regression, the standard error and the corresponding p-value are given.

INDEPENDENT VARIABLES (averages 1985-88)	DEPENDENT VARIABLE : Board Turnover over 1985-88.
	Poor performers	Good performers	Poor performers	Good performers
Sample Size	23	44	52	56
Intercept	0.058 (0.074,0.45)	0.126 (0.047,0.01)	0.101 (0.041,0.02)	0.129 (0.041,0.00)
Company size	-0.008 (0.008,0.30)	-0.006 (0.007,0.37)	-0.011 (0.005,0.06)	-0.008 (0.006,0.19)
Non-executives (proportion of total board)	-0.015 (0.084,0.85)	-0.029 (0.050,0.57)
Separation of control (yes=1, no=0)	-0.043 (0.022,0.07)	-0.005 (0.020,0.78)	-0.058 (0.017,0.00)	0.008 (0.017,0.65)
Institutional investors' stakes (banks, insurance, investm. co's)	-0.001 (0.001,0.42)	-0.0003 (-0.390,0.69)	0.0001 (0.001,0.96)	-0.0006 (0.0007,0.38)
Outsiders' stakes (ind. co's and individuals)	0.002 (0.0005,0.03)	-0.0014 (0.0009,0.12)	0.001 (0.006,0.10)	-0.001 (0.0009,0.14)
Insiders' stakes (directors)	-0.002 (-2.782,0.01)	0.0002 (0.0008,0.81)	-0.001 (0.006,0.02)	-0.0002 (-0.299,0.76)
Increases in Institutional investors' stakes (banks, insurance, investm. co's)	0.006 (0.004,0.13)	0.004 (0.004,0.40)	0.00003 (0.003,0.99)	0.005 (0.004,0.22)
Increases in Outsiders' stakes (ind. co's and individuals)	0.004 (0.002,0.10)	0.006 (0.004,0.14)	0.005 (0.002,0.01)	0.001 (0.375,0.70)
Increases in Insiders' stakes (directors)	0.008 (2.501,0.02)	0.0005 (0.003,0.85)	0.011 (0.002,0.00)	0.002 (0.002,0.33)
F-test	0.00	0.26	0.00	0.06
adjusted R squared	0.68	0.06	0.50	0.14

Source : Own calculations based on annual reports.

CHAPTER 3 : Overview of Corporate Control Issues in Belgium.

While in the U.S. and the U.K., managerial performance is maintained by the complementary intervention of both internal and external control mechanisms, the impact of the takeover market as a corporate governance device in Belgium, and most other Continental European countries, is limited. Recent Belgian legislative changes with regard to ownership disclosure laws and anti-takeover procedures have further reduced the likelihood of takeovers as a corporate control mechanism. Consequently, as in the French Viénot report (Jack 1995, Viénot 1995) and Cadbury report in the U.K. (Cadbury 1992), the Belgian policy debate on corporate governance currently focuses on the effectiveness of internal corporate control mechanisms and the role of large shareholders in the corporate governance monitoring process.

The main objective of this part of the study is to investigate whether poor corporate performance triggers executive board turnover and whether disciplining actions are initiated by the non-executive directors, usually appointed by and representing the large shareholders. This paper also investigates whether the accumulation of shares into large blocks of shares mitigates problems of free riding in corporate control, permitting control to be exerted more effectively. In addition, we examine whether the presence of particular types of major shareholders is associated with increased incidence of disciplinary turnover when corporate performance is poor. If this is the case, we can assume that such large shareholders are more effective monitors.

We also analyze whether, when company performance is poor, a market for share stakes arises. In Continental Europe, such a market might have an equivalent role of the external market for corporate control in the U.K. and the U.S. We hypothesize that, if a company underperforms and if there is no absolute majority control, able monitors will increase their voting rights in order to reach a control level allowing them to nominate a new management team. Finally, for corporate governance mechanisms to be effective, there should not only be a greater incidence of top management changes in poorly performing firms, but also improvements in firm performance following management restructuring.

This part of the study also describes ownership patterns of all Belgian companies listed on the Brussels Stock Exchange over the period 1989-1992, and details the ownership structure of subsamples consisting of all the Belgian holdings companies, the industrial and commercial companies, and financial firms. We also report the relative importance of different investor categories, the occurrence of blocking minorities and majorities, and control leverage via ownership pyramids.

The paper has 6 major findings. First, we document that poor company performance precedes increased turnover of the executive directors, of the management committee and of the CEO and executive chairman. These findings are consistent with the board of directors and major shareholders serving an important role in monitoring and disciplining poorly performing companies. Similar relations were reported by Denis and Denis (1994) and Warner, Watts and Wruck (1988) for the U.S., by Franks and Mayer (1995b) for Germany and in Chapter 2 for the U.K. We use a number of different criteria to measure performance : operating income, earnings after tax, share price returns, dividend changes and negative earnings. The levels of and changes in the performance criteria are standardized by total assets or total equity and corrected for industry effects. Market adjusted returns over both short (one to three years) and long term periods (up to ten years) are employed. Statistically significant corporate control relations are only found for the inability to generate positive earnings, for substantial decreases in dividends and for market adjusted returns, but the other performance benchmarks are still, as expected, negatively correlated with board turnover.

Second, the structure of the Belgian board of directors has an important impact on the functioning of the internal corporate control system, since a high number of non-executive directors and separation of the functions of CEO and chairman of the board increases turnover of executive directors of underperforming companies. Weisbach (1988) reports similar results for the U.S. : outside directors play a larger role in monitoring management than inside directors. For Japan, Kaplan and Minton (1994) show that board appointments of directors representing banks and corporations are followed by increases in top management turnover. In Chapter 2 of this study, we found that a high proportion of non-executive directors serving on the boards of poorly performing U.K. companies is positively correlated with board turnover. Franks and Mayer (1995b) show that, in German companies with concentrated ownership, supervisory board representation goes hand in hand with ownership or large shareholdings.

Third, consistent with Shleifer and Vishny (1986) and Grossman and Hart (1980), we find higher board turnover related to increasing concentration in ownership since the costs of free riding in control are reduced. We also find that disciplining of underperforming management depends on the presence of specific categories of large shareholders. For instance, while industrial and commercial companies, family shareholders and holding companies, replace directors and management when the company's profitability is low, large institutional investors do not seem to be involved in monitoring the corporation.

Fourth, the ownership structure of Belgian companies is complex with multiple owners and stakes held through multiple tiers of ownership. The decision to substitute top management of poorly performing companies is taken by ultimate shareholders who control either directly or indirectly, via affiliated companies, a large percentage of the voting rights. The presence of direct large ownership stakes is only weakly correlated to board turnover, whereas regressions of the aggregate of the direct stakes controlled by the same ultimate shareholder on board turnover shows a significant positive relation. However, when an ultimate shareholder has invested in a company via a pyramid of intermediate companies (through multiple ownership tiers) which he controls with less than 100% of the shares, the association between board turnover and large shareholdings is weaker. This yields some evidence of control dilution throughout the ownership tiers of a control chain.

Fifth, we find an active market in share stakes following poor performance. Specific shareholder classes with superior monitoring abilities or with private benefits of control, increase their percentage of voting rights in order to be better positioned to replace management. Such a market for blocks of control also exists in the U.K. and in Germany, as respectively detailed in Chapter 2 and Franks and Mayer (1995b). Shareholders who increase their holdings do so with a clear intention to assume an active monitoring role since management turnover significantly increases in subsequent periods.

Sixth, corporate control actions leading to the replacement of management are followed by an improvement of corporate performance in the form of increases in dividends per share over a period of two years after turnover. However, replacement of CEO and executive directors is followed by decreases in earnings. This may be the result of new management's decision to expense large amounts of costs so that the reduction in earnings can still be attributed to their predecessors and the lower results allows for substantial improvements in subsequent years. This result is analogous to the findings of, among others, Murphy and Zimmerman (1993).

The remainder of this part of the study is organized as follows. In Chapter 4, the hypotheses and methodology are discussed and the sample data are described. Chapter 5 details the ownership structure and the importance of investor groups, foreign investment and the size and composition of the board of directors and the management committee of all Belgian listed companies. Chapter 6 exhibits the main results of the hypotheses. Chapter 7 concludes.

CHAPTER 4 : Hypotheses, methodology and data.

4.1 Hypotheses.

4.1.1 Corporate performance and disciplinary corporate governance actions.

Since one of the principal responsibilities of the board of directors is monitoring the company's performance, the most striking, observable actions taken by directors or major shareholders are the replacement of the CEO, of members of the management committee and of executive directors when companies underperform.

Hypothesis 1 :

Disciplining of top management is triggered by poor company performance : directors, CEOs, top managers and executive chairmen are replaced following poor share price performance and/or low operating income and net earnings.

Boards must assume the task of extracting information about true managerial performance from noisy financial performance realizations. Both accounting and market returns (see section 4.2) are determined in part by factors beyond the control or influence of the firm's managers. However, to the extent that these returns are also influenced by the quality of managerial inputs and actions, they may provide useful information on managerial performance (Joskow and Rose 1994).

Testing hypothesis 1 also yields an answer to the question about who is held accountable for the poor performance : CEO, executive directors, or those members of the management committee who do not serve on the board. Unlike in Germany, a Belgian management committee is not a collegial council; it consists of the top management of the company and is chaired by the CEO (delegated director). The most senior members of the management committee, including the CEO, usually serve on the board of directors.

Furthermore, we investigate whether the non-executive directors and the non-executive chairmen are replaced following poor performance, which would either be an indication that poor monitors are substituted or that changes in the shareholder structure have occurred.

4.1.2 The impact of board composition and structure on the board's ability to monitor performance.

The board of directors which has the power to hire, fire and compensate senior management, serves to resolve conflicts of interest among decision makers and residual risk bearers, the shareholders (Williamson 1983 and 1984). The existence of a balanced board including both executives and non-executives, reduces the transaction or agency costs associated with the separation of ownership and control. Executive directors and members of the management committee, who bear responsibility for their company's results are not likely to discipline themselves. Moreover, their careers are tied to the CEO's, which discourages them to remove incumbent CEOs (Vance 1987, Mace 1986). Consequently, the task of evaluating senior management is likely to fall mainly on the non-executive directors who have several incentives to remove underperforming CEOs and top management :

Firstly, some of the non-executive directors represent the (large) shareholders who have delegated their monitoring task and who might replace directors who do not assume their monitoring tasks. In other words, the monitors on the board are in turn monitored by large shareholders. The ownership structure in Belgium is highly concentrated : in more than 85% of all quoted companies a large shareholder owns a share stake of at least 25%. A voting rights majority is held by a shareholder in more than half of the companies (see Chapter 5). Therefore, in many cases, even 'independent experts' who serve on the board of directors are appointed with the consent of major shareholders. Consequently, the concentration of ownership ensures that the relation between major shareholders and the board of directors in Belgium is much stronger than that in an Anglo-American corporate governance system. Hirshleifer and Thakor (1992) model a multi-layered principal-agent relationship in which shareholders delegate the task of monitoring managerial quality to the board of directors and rely on the external take-over market to provide additional disciplining of poorly performing managers. However, their U.S. model is not an appropriate description of the corporate governance mechanism in Continental Europe, where ownership structure and legislation limits the (potentially) disciplining role of such a (hostile) takeover market much more than in Anglo-American capital markets.

Secondly, non-executive directors are usually respected business leaders whose reputations suffer when they are directors of faltering companies. Non-executives have incentives to develop reputations as experts in decision control whose human capital depends on their performance as internal decision managers and monitors in other organizations (Fama and Jensen 1983, Fama 1980). Consequently, directors face an external labour market which provides some form of disciplining of passive leadership. The importance of this market and its reputation signal has been emphasized by Mace (1986) : managers accept outside directorships to signal that they have been accepted by their peers. In the case of the U.S., Kaplan and Reishus (1990) report that managers of poorer performers are likely to lose directorships in their own company, but they do not seem to lose directorships in other companies. However, failing managers will rarely be offered new directorships. Directors who left the boards of distressed U.S. companies, of firms that filed for bankruptcy or restructured their debt, held approximately one-third fewer directorships three years after their departure (Gilson 1990).

A third reason for the non-executive board to monitor the company's performance actively, is that directors have legal obligations to the shareholders and they can be held liable for damages. In Belgium, minority shareholders who own, individually or collectively, a minimum of 1 percent of the voting rights can sue the board if the board has violated the rights of the minority shareholders.

Hypothesis 2 :

The composition of the board of directors determines the board's monitoring capabilities. The greater the proportion of non-executive directors, the lower potential board domination by management and the higher the monitoring ability of the non-executive directors as observed in turnover of executive directors, of the CEO and of the management committee.

One of the recommendations of the U.S. Bacon study (1993), of the U.K. Cadbury Committee (1992) and of the French Viénot report (Viénot 1995, Jack 1995) focused on the importance of separation of the role of CEO and of non-executive chairman. As a result of this direction, conflicts of interest for the CEO who would be both chairman of the management committee and of the board of directors are avoided and the possibility that a strong CEO dominates the board is reduced. The recommendation is based on the idea that a non-executive chairman could set the agenda of board meetings more independently of management and that this would strengthen the monitoring ability of the non-executive directors.

Hypothesis 3 :

The separation of the functions of CEO and of chairman of the board, facilitates disciplining of underperforming management. Therefore, with dual control, we would expect to see higher turnover.

4.1.3 Ownership concentration, the costs of free riding on control and superior monitoring abilities.

Grossman and Hart (1980 and 1988) persuasively argue that outsiders without a share in a diffusely held corporation would never take over that company in order to improve its performance. If such a shareholder can gain only on the shares he already owns and has to pay all the monitoring and takeover costs, the deal may not be worthwhile. The atomistic incumbent shareholders will hold out unless they are offered a price which equals their estimate of the post-restructuring value of the company. For the same reason, monitoring management and disciplining in the case of company underperformance, may be prohibitively expensive for small shareholders. Monitoring shareholders pay the costs related to their corporate control efforts but they only benefit in proportion to their shareholding (Demsetz and Lehn 1985, Demsetz 1983, Easterbrook and Fischel 1983). Therefore, monitoring of management will only be effective if a single party becomes large enough to internalize the costs of corporate control. The disadvantage of greater diffuseness in ownership structure is shirking of control by the owners. A shareholder's benefit of free riding on control is the ability to use his time on other tasks while the costs of shirking are shared by all the shareholders. The inefficiencies implied by these externalities highlight the advantage of concentrated ownership with regard to corporate governance.

Shleifer and Vishny (1986) focus on the ways in which large shareholders bring about value-increasing changes in corporate policy. They model that in the process of tender offers, proxy fights and internal management shake-ups, the presence of a large minority shareholder provides a partial solution to the free-rider problem. Burkart, Gromb and Panunzi (1995) argue that the ownership structure of a firm acts as a commitment device to delegate a certain degree of authority from the shareholders to the management. They show that when long-term profits are important, a large shareholder may be desirable. Their model also shows that, on the one hand, monitoring ensures that managers' and shareholders' interests are aligned and reduces the risk that a bad manager continues to provide low returns. On the other hand, close monitoring reduces managerial discretion and hence management's current effort. In fact, the authors suggest that, depending on the performance of the company, an optimal size for a large share stake exists so that conflicting effects of monitoring are balanced.

Hypothesis 4 :

The presence of large shareholdings in the ownership structure is positively correlated with higher board turnover when performance is poor.

The incentives to monitor and correct managerial failure depend not only on the concentration of ownership, but also on the monitoring ability of major shareholders. Since different classes of shareholders might have different information, monitoring competencies and incentives, we categorized all shareholders with a stake of 5 percent or more into 8 classes : (i) holding companies, (ii) banks, (iii) investment companies (pension funds, investment funds), (iv) insurance companies, (v) industrial and commercial companies, (vi) families and individual investors, (vii) federal or regional authorities, (viii) realty investment companies. Each of these shareholder classes is subdivided into Belgian and foreign investors.

Like Shleifer and Vishny (1986), we do not expect monitoring actions by investment companies, banks, insurance companies and realty investors. In Belgium, these investors have so far taken a passive stance with regard to monitoring; they are often affiliated to holding companies or financial institutions and want to avoid conflicts of interest. No such impediments hinder monitoring by holding companies, industrial and commercial companies, individual investors and the government whose major shareholdings are, consequently, expected to be positively correlated with turnover.

Burkart, Gromb and Panunzi (1995) argue : "In absence of private gains, blocks of shares ought to be sold at a discount due to the greater risk exposure and due to the monitoring costs. However, blocks are usually sold at a premium which suggests the presence of private gains. Private gains may be pecuniary or non-pecuniary, and may stem from taking decisions which actually reduce the security benefits, e.g. use the firm's structures for personal purposes, or engage into sweetheart deals. (p.26)" Demsetz and Lehn (1985) and Barclay and Holderness (1989, 1991) also note that concentrated ownership and control is valued differently by diverse classes of U.S. shareholders who have different attitudes towards monitoring. The holding companies' incentives to monitor the company or to acquire a major shareholding are manifold and include e.g. capturing tax reductions by facilitating intercompany transfers, reducing transaction costs by offering economies of scale or an internalized form of capital market (Leleux, Vermaelen and Banerjee 1995). Corporate shareholders with customer or supplier relations might hold substantial share stakes in order to have a board representative who could try to influence management's strategic decisions favourably for the investor. Individuals or family shareholders whose stakes give them the right to an executive or non-executive board seat are likely to value opportunities to consume perquisites more highly than will corporate blockholders.

Thus, the incentives to monitor and discipline underperforming management are not uniquely based on an improvement of the financial benefits related to the shares (dividend income and capital gains), but also on private benefits of control accruing to a major shareholder. Such benefits of control are unique to a shareholder and not tradeable. Manne (1964) and Jensen and Meckling (1976) report that control is valuable and the source of value is the additional compensation and perquisites that the controlling security holders can accord themselves.

Hypothesis 5 :

Disciplining of underperforming management is accomplished by large shareholders with superior monitoring abilities.

4.1.4 Dilution of control.

As ownership structures are frequently complex and/or pyramidal (see Chapter 5), the question arises as to whether decisions about disciplining management of the sample company are taken by direct investors (at ownership tier 1) or by 'ultimate shareholders' who control these direct shareholders directly or via intermediate companies through multiple tiers of ownership. An investor is considered to be the 'ultimate shareholder' in an ownership-control chain if control is maintained through multiple tiers of ownership. Interlocking ownership via a holding company or through a more elaborate stock pyramid enables a given investor to own different quantities of voting and cash flow rights. For instance, 50.1 percent of ownership (and voting rights) held by the ultimate shareholder in an intermediary holding company which, in turn, owns 50.1 percent of an operating subsidiary could guarantee majority control on the subsidiary's board with only a 25.1 percent interest in its common stock cash flows.

Sequences of majority control in the form of e.g. stakes of 50.1% throughout the pyramid might not guarantee the same control rights as a first tier majority holding would give, unless there is board representation on each level of the ownership structure. As the number of ownership tiers increases, control dilution might occur because of agency costs. In addition, the more the share stakes in intermediate companies of an ownership pyramid deviate from full control (100%), the higher the potential dilution of control. A shareholding of 50% owned by an ultimate investor at the second tier in an intermediate company that holds in turn 50% of the voting rights of a sample company might not give the same control as a direct stake of similar size owned by a company with a dispersed ownership structure. A sequence of vetos does not ensure that the ultimate shareholder can veto decisions at the board level of the target.

Hypothesis 6 :

When a sample company's performance is poor, the influence of an ultimate shareholder on managerial disciplining is reduced when controlling stakes are held through multiple tiers of ownership.

4.1.5 The disciplining role of the market for share stakes.

When performance is poor, a market for share stakes may result. Decisions to build up a substantial shareholding, to increase a shareholding to a critical ownership threshold (e.g. 25% or 50%) or to expand a toehold share stake are motivated by future performance improvements after the failing management team and/or the board is restructured. Moreover, stakeholders can obtain or safeguard private benefits of corporate control at relatively low cost when performance is poor. Still, Barclay and Holderness (1989) point out that large blocks of shares are typically priced at substantial premiums (of about 20 percent) over the stock exchange price in the U.S.

Poor performance may reflect not simply poor management but may also be the result of ineffective monitoring and control. If this is the case, we may also expect that poor performance is accompanied by sales of stakes and changes of concentration of ownership. Low quality monitors sell out to shareholders with a managerial alternative. Large share stakes usually change hands through negotiated deals ex exchange. Shleifer and Vishny (1986) state that once a block of shares is assembled, the position is unlikely to be dissipated. It is in the large shareholder's interest to wait until someone who can monitor effectively expresses interest in his shares. For if he sells his shares on the open market, he loses that part of the firm's value that comes from the possibility of a value-increasing monitoring. This suggests that large blocks of shares will tend to be passed on rather than dissipated.

We analyze the changes in ownership by distinguishing among increases in ownership of the old (existing) shareholders and increases in ownership of 'new' investors. These 'new' investors are investors who build up a shareholding of 5 percent or more and did not hold any shares in the previous year or had a toe-hold stake under 5 percent.

Hypothesis 7 :

In companies without sufficiently large shareholders or with shareholders who take a passive stance with regard to monitoring, poor performance gives rise to changes in the ownership pattern.

When a market of share stakes originates from poor performance and for control purposes, we might expect disciplining of management. Larger ownership stakes held by high quality monitors yield a more powerful control position. Therefore, increasing shareholdings owned by 'old' shareholders and share stakes acquired by new shareholders coincide with or are followed by changes in management.

Hypothesis 8 :

Increases in shareholdings are associated with higher managerial and board turnover in the same year or the year following the monitors' disciplinary actions.

4.1.6 Post-disciplining corporate performance.

For internal and external control mechanisms to be effective, the greater incidence of replacement of top management and directors should be followed by improvements in firm performance.

Hypothesis 9 :

Management and board restructuring triggered by poor performance results in improvements of company performance.

Both share price returns and accounting measures are used as performance measures. Changes in dividends per share are explained by the permanent component in earnings levels and the company's target dividend level (Swanson and Alltizer 1995, Fama and Babiak 1968, Lintner 1956) and dividend reductions tend to have a permanent character (see section 4.2). Therefore, increases in dividends per share are good indicators of performance improvements when dividends were reduced in the past.

However, it is not certain when improvement of accounting measures in the year of the turnover and the subsequent year is to be expected. Following management changes, asset write-offs (Strong and Meyer 1987), changes to income reducing accounting methods changes (Moore 1973) or income reducing accounting accruals (Pourciau 1993) frequently occur. Murphy and Zimmerman (1993) conclude that a 'big bath' is more likely to occur if the outgoing CEO is terminated following poor performance since in these situations it is more credible for the new CEO to blame the previous CEO for past mistakes. Moreover, by constantly overstating losses attributable to predecessors, management improves accounting expectations about the future and lowers the benchmark against which its own accounting performance will be measured (Elliott and Shaw 1988). Consequently, we only expect increases in earnings as of the second or third year after the replacement of management.

Expectations about future performance of a new management team, CEO or new directors will be reflected in the share price return at the latest at the announcement of the replacement. Previous studies examining the wealth effects of changes in the top management team have produced mixed effects. Bonnier and Bruner (1989), Furtado and Rozeff (1987) and Weisbach (1988) detect significant positive price effects while insignificant price reactions are found by Reinganum (1985), Warner, Watts and Wruck (1988) and Dennis and Dennis (1994).

4.2 Performance benchmarks.

To evaluate whether poor performance triggers corporate control actions (hypothesis 1), we focus on several different performance criteria. We analyze whether monitors react quickly when performance declines or whether disciplinary actions are only taken against management when performance reaches critically low levels.

Share price returns

The first performance measure used is the return on the company's stock, adjusted for the return on the value-weighted market index of the Brussels Stock Exchange. To explore how quickly boards or major shareholders react to poor share price performance, market adjusted returns over short pre-turnover periods (one or two years) are examined. Warner, Watts and Wruck (1988) and Coughlan and Schmidt (1985) report that U.S. boards react quickly to poor performance in their decision to replace the CEO : share performance lagged up to two calendar years helps predict current-calendar-year management changes. However, if the CEO and/or top management dominates the board of directors or if top management has built up an excellent managerial record, underperformance over a short period of time might not trigger any disciplinary action (Yungsan 1993). Therefore, returns over longer periods (three to ten years) are also used to verify the presence of managerial entrenchment.

Accounting earnings

A second benchmark of corporate performance is accounting earnings. These earnings data have an advantage over share price data for the purposes of measuring the performance of top management. The share price reflects the present discounted value of the expected future cash flows of the company and, therefore, incorporates the market's estimate of the probability that bad management will be replaced. Consequently, the share price of poorly performing companies is higher than it would be if the management could not be fired. Using share price returns as a performance measure might lead to an underestimation of the turnover-performance relation (Weisbach 1988). Accounting earnings, on the other hand, also have limitations since it is a variable that management can manipulate to some extent. Large firms are more easily able to smooth earnings (Schipper 1989, Moses 1987). Several earnings definitions are used : operating income (earnings before financial and extraordinary results and before taxes, EBIT), earnings after financial results but before extraordinary results and taxes (EBT), and earnings after extraordinary results and taxes (EAT), all normalized by total assets or by equity. Earnings before interest and taxes are not sensitive to changes in capital structure of special tax treatments. Correcting earnings for depreciation and non-cash items yields a measure of cash flow.

As it is a priori not certain whether the monitors focus on absolute or on relative levels of performance, both earnings levels, standardized by total assets or total equity, and changes in earnings over the period of one to two years before management replacement, are taken into consideration. If the decision to replace top management is related to unanticipated changes in performance, changes in accounting earnings is the appropriate benchmark (Weisbach 1988). Earnings (but not changes in earnings) are reasonably approximated by a random walk which implies that changes in earnings are an unbiased estimate of unexpected earnings (Dechow 1992, Ball and Watts 1972, Foster 1978).

Interviewed Belgian executive and non-executive directors pointed out that absolute performance benchmarks are more likely to trigger management and director turnover.

Industry effects

To control for industry effects, the average standardized absolute level of and the change in operating income of a specific industry is subtracted from, respectively, the standardized operating income levels and the changes for each firm within that industry. A similar industry corrected variable was constructed for the levels of earnings after tax. Morck, Shleifer and Vishny (1989) find that when a firm significantly underperforms its industry, the probability of complete turnover of the top management team rises. They report that boards are more successful in addressing firm-specific difficulties while industry-wide problems are more likely to trigger hostile takeovers in the U.S.

Critical earnings levels

The only monitoring acts we can observe are the rather drastic actions of replacing underperforming management. It is possible that when performance is declining, the monitors try to assist management in their attempts to improve the company's declining profitability. Only when the non-executive directors or large shareholders doubt current management's competence to make the company more profitable, the monitors will replace them. Therefore, management's inability to generate profits might be the benchmark which triggers a disciplining action by the monitors. Dummy variables referring to negative operating income, negative earnings before tax and negative earnings after tax are used as proxies for poor performance, as does Kaplan (1994a, 1994b).

Dividend changes

Reductions in dividends per share are also used as a measure for poor performance since they are widely followed and reported in the financial press and are not likely to be reversed (Marsh and Merton 1987). Management is generally reluctant to reduce dividends unless a reduction is unavoidable (Baker, Farrelly and Edleman 1985). Bonnier and Bruner (1989) show that a dividend cut tells the market that the problems responsible for the share price drop before the dividend cut are not temporary. Furthermore, dividend cuts are associated with unusually poor stock-price and earnings performance (Healey and Palepu 1988, Ofer and Siegel 1987).

4.3 Methodology.

The control of companies is examined via OLS regressions with the dependent variable being the turnover of the board proportional to board size or the turnover of members of the management committee proportional to committee size. For CEO and chairman turnover, logistic models are used to predict the probability of turnover when performance is poor. The firm-years are pooled with each firm-year over the four year period (1989-1992) representing a separate observation. Using panel data data on ownership, we are able to control for possible simultaneity between performance and turnover data.

4.4 Data.

4.4.1 Sample description.

The sample consists of all Belgian companies listed on the Brussels Stock Exchange during July 1989 and August 1994. In total, 192 firms are included in the sample; some of these went bankrupt in the period under consideration, while others were introduced after 1989. In 1989 and 1994, respectively, 186 and 165 companies were listed. Sector codes, dates of introduction and of delisting are provided by the Documentation and Statistics Department of the Brussels Stock Exchange. In the analysis, the sample size was reduced by 9 companies in 1989 and by 10 in 1994 as these listed firms, all in coal mining and steel production, were involved in a long liquidation process but were still listed.

Table 4.1 : Sample Description

1989 1994

All listed sample companies¹ 177 155

Holdings² 71 64

Financial sector 23 19

Industrial and Service companies 83 72

Financial Sector

Banks 8 7

Insurance 7 5

Real estate 8 7

Industry

energy³ 6 5

materials⁴ 4 26

capital equipment⁵ 13 12

consumer goods⁶ 19 16

Services 11 13

¹ For 1989 and 1994, respectively, nine and ten listed companies that have been in liquidation for years, were not included in the sample. These companies are all in coal mining and steel production. The number of delistings in the period 1989-1994 surpasses the number of new introductions due to mergers, industry restructurings (e.g. in the energy sector) and the policy of the stock exchange to delist infrequently traded companies with tiny market capitalizations.

² The holding companies have multi-industry investments. The categorization is based on the NACE classification of the National Bank and the classification of the Bank Brussel Lambert.

³ mainly petrochemical and electricity production.

⁴ ferro, non-ferro, chemicals, building, paper, glass.

⁵ electricals, electronics, construction, machine building.

⁶ mainly food, pharmaceuticals and retail.

Table 4.1 shows that 40 percent of the Belgian listed companies are holding companies with multi-industry investments, 13 percent are in the financial sector (banking, insurance and real estate) and 47 percent are industrial and commercial companies.

4.4.2 Ownership data.

Data on the ownership structure over the period 1989-1994 were collected from the Documentation and Statistics Department of the Brussels Stock Exchange. Ownership data are only available since 1989, following the introduction of the Ownership Disclosure Legislation (see Chapter 5). The Documentation Department maintains a daily updated database BDPart (Bourse Data Participations) of the shareholding structure of Belgian listed companies. BDPart provides data on the first level of shareholding (direct ownership) in all Belgian listed companies, such as the names of the investors, the number of shares declared, number of shares issued and the percentage of ownership. Apart from voting rights linked to the shareholdings, BDPart also displays potential voting rights linked to securities that will represent voting rights when converted or exercised (e.g. convertible bonds, warrants). Previous ownership positions in the BDPart database are overwritten once new ownership information becomes available. To capture a company's ownership position at the end of its fiscal year since 1989 and changes in shareholdings during each year, about 5000 hardcopy Notifications of Ownership Change from 1989 till 1994 were consulted. These Notifications were sent by the target to the Brussels Stock Exchange which published this information in the official Stock Exchange newspaper Cote de la Bourse. Apart from details on voting rights, the investors' status (independent, affiliated or acting in concert with other investors) was compiled from the Notifications. With this information about major direct shareholdings and indirect control, the multi-layered ownership structure was reconstructed for each company over the period 1989-1994. The shareholding data from BDPart and the Notifications of Ownership Change were verified with ownership data of the database of the National Bank which is based on annual reports.

The 1988-1994 yearbooks of Trends 20,000, which comprise industry sector classification and financial data for most listed and non-listed Belgian companies, were used to classify all Belgian investors into the following categories : (i) holding companies, (ii) banks, (iii) institutional investors, (iv) insurance companies, (v) industrial companies, (vi) families and individual investors, (vii) federal or regional governments and (viii) real estate investors. Foreign companies owning a large share stake in Belgian companies were classified with information from Kompass.

4.4.3 Share price and accounting data.

Monthly (from 1980) and weekly (from 1986) share price returns, corrected for stock splits and dividend pay-outs, and a value-weighted index of all companies listed on the Brussels Stock Exchange were provided by the Generale Bank. The number of outstanding shares were collected from the yearbooks Memento der Effecten for the years 1988 till 1994.

Accounting data (total assets, equity, operating income, earnings after tax, dividends per share) were collected from annual reports and from the database of the Balans Centrale (Central Depository of Balance Sheets) of the National Bank of Belgium.

4.4.4 Data on the board of directors and the management committee.

The database of the National Bank of Belgium also contains data on the boards of directors: the names of directors, of the chairman and of the 'delegated director' (equivalent to the CEO) and the fiscal years during which these directors served on the board. The reasons for a director, chairman, CEO or manager to leave the company were collected from the notes in the annual reports. Natural turnover due to retirement, death or illness is often reported. However, since the usual retirement age is 65, early retirement is only accepted as natural turnover if the director or manager was 63 years of age or older. This way, we eliminate most of the non-linearity in the turnover-age relationship. Other reasons for turnover are rarely mentioned in either the annual reports or the financial press. Resignations related to a merger process were also eliminated. When no grounds or non-informative reasons were given for turnover, forced turnover due to disciplining actions or due to company policy disputes was assumed. This turnover uncertainty inserts an unavoidable bias in the data on forced turnover.

Data on size and turnover of the management (or direction) committee were gathered from the annual reports. The number of directors who were (are) also a member of the management committee was also compiled. If the annual report did not mention explicitly the existence of a management committee, the yearbooks Memento der Effecten and the Jaarboek der Bestuurders (Yearbook of Directors) for the years 1988 till 1993 were consulted to determine whether or not directors had executive functions.

CHAPTER 5 : Ownership and control of Belgian listed companies : stylized facts.

This chapter provides an overview of the main characteristics of the ownership structure of the Belgian companies quoted on the Brussels Stock Exchange. Prior to the changes in corporate law regarding ownership disclosure in 1989, as described in section 5.2.1, little was known about ownership and control and, so far, no comprehensive description has been composed. We detail ownership concentration, the importance of different shareholder classes, the violation of the one share-one vote rule via pyramidal ownership structures, and the corporate control market for share stakes. We also describe management representation on the board of directors, the organization of executives in management committees, and turnover of board and committee. We begin this chapter with a summary of the main aspects of the Belgian capital market which are compared with Anglo-American and other Continental European markets.

5.1 Insider versus outsider ownership and control systems.

According to Berle and Means (1932), dispersed ownership has given rise to separation of ownership and control. Demsetz and Lehn (1985) argue that ownership patterns reflect a trade off of the risk to investors of concentrated investments in large firms and the control potential of the firm. Diversified shareholdings are useful from the point of view of risk reduction but discourage active participation of investors. As Franks and Mayer (1995c) point out, it is puzzling that the resolution of this trade off has taken such different forms in different countries. German and French equity markets can be characterized by few listed companies, an illiquid capital market where ownership and control is infrequently traded and complex systems of intercorporate holdings (Mayer 1993, Franks and Mayer 1992). Consequently, these structures are appropriately described as insider systems in which the corporate sector has controlling interests in itself; outsider investors, while able to participate in equity returns through the stock market, are not able to exert much control. In contrast, the Anglo-American system is a market oriented or outsider system and is characterised by a large number of listed companies, a liquid capital market where ownership and control rights are frequently traded and few intercorporate holdings. There are few large, controlling shareholdings and these are rarely associated with the corporate sector itself.

The main characteristics of the Belgian corporate ownership and equity market can be summarized as follows : (i) few Belgian companies are listed, (ii) there is a high degree of ownership concentration, (iii) holding companies and families, and to a lesser extent industrial companies, are the main investor categories, (iv) control is levered by pyramidal and complex ownership structures and (v) there is a market for share stakes. Properties (i) to (iv) imply that Belgium can be portrayed as a German-French 'insider system' rather than an Anglo-American system. However, typical for Belgium is the importance of holding companies which are often part of pyramidal ownership chains and are used to lever control (see section 5.2.3).

Table 5.1 shows the number of quoted companies per country and the total market capitalization as a percentage of GDP. The U.K., U.S. and Japan are characterised by a large number of quoted companies; respectively 1878, 6342 and 1627 in 1992. The market capitalization of companies quoted on the London Stock Exchange is around 81 percent of the U.K. GDP. Companies quoted on the Tokyo Stock Exchange have a value of 89 percent of the Japanese GDP while the value of corporations listed on the New York Stock Exchange and NASDAQ amounts to 56 percent of U.S. GDP. The capital markets of France, Germany, Belgium and Spain and of most of the remainder of continental Europe, present a different situation: they have many less quoted companies with a market capitalization as a percentage of GDP which is lower than 32 percent.

Compared to the shareholding structure of Continental European corporations, ownership in the U.S. and the U.K. is much less concentrated. For the U.S., the average shareholding of the five largest shareholders in a sample of Fortune 500 companies is 15.4 percent and 23 percent of these companies do not have a shareholder with a share stake over more than 5% (Shleifer and Vishny 1986, Demsetz and Lehn 1985). These two percentages compare to respectively 60 percent and to 1 percent for Belgium. The large shareholders with a stake of at least 5% in the U.S. are mostly families, pension an profit-sharing plans as well as banks, insurance companies and investment funds. About two-thirds of the market capitalization are held by individual investors and institutional investors on behalf of individuals in U.S. and U.K. quoted companies, but the U.S. has a far higher proportion of equity owned directly by individuals. However, Davies and Stapledon (1994) report the enormous growth in the percentage (by value) of equity held by institutional investors in the U.K. and a decline in the percentage held by individuals.

The numbers of quoted companies refer to 1992, but to 1991 for the U.S. and Japan. For each country, only domestic companies listed on the main stock exchanges have been considered : New York and NASDAQ combined, London, Tokyo, Paris, Frankfurt, Madrid, Amsterdam, all Swiss exchanges, and Brussels.

Country	number of domestic quoted companies	equity of quoted co's as % of GDP
U.S.	6,342	56 %
U.K.	1,878	81 %
Japan	1,627	89 %
France	786	26 %
Germany	665	18 %
Spain	433	20 %
Netherlands	314	44 %
Switzerland	180	78 %
Belgium	171	31 %

Source : Own calculations for Belgium and the U.K. are based on data from the Brussels Stock Exchange and the Department of Trade and Industry in London, Wymeersch (1994b) for the Netherlands, Germany, France and Switzerland, Goergen (1993) for Spain, Franks and Mayer (1992) for the U.S. and Japan.

Germany, like Belgium, has few widely held listed companies : only 15 percent of a German sample of the 171 largest companies do not have any shareholder with an equity stake of 25 percent or more (Franks and Mayer 1995b and 1995c). Other German companies and families own the largest share stakes. Trusts and institutional investors are sometimes large shareholders but their stakes are rarely majority holdings. The same holds for banks. However, the significance of banks is greater than their direct equity holdings would suggest : as holders of bearer shares they are able to exercise proxy votes on behalf of dispersed shareholders. Control is maintained at low cost via complex and pyramidal structures : the average tier of company holdings is 2.2 compared with 3.1 for families and 4.2 for banks.

In a French sample of the largest 155 quoted companies, almost 89 percent have a shareholder with an equity stake of 25 percent or more. The major shareholders in the French sample are predominantly other industrial companies (Goergen 1993). So, in France, like in Germany, the corporate sector is by far the single largest group of shareholders. Foreign companies, families and banks are the other large shareholders. Corporations who hold equity stakes in each other are often in related industries or in the same industry (Franks and Mayer 1995c). Furthermore, in most cases, these companies are not trading partners.

The Italian shareholding structure is characterized by high concentration of ownership, the presence of family owners and the pervasive role of the state (see Bianco, Gola and Signorini 1995). About 95 percent of the largest 500 non-financial companies are controlled with absolute majority (Bianchi and Casavola 1995). Contrary to what one would expect, the concentration of direct ownership is greater in larger firms. Controlling shareholders hold via pyramids and coalitions, 88 percent of the largest companies.

Japanese ownership is, similar to Continental Europe, highly concentrated. Financial and industrial groups (keiretsu), represent about 61 percent of the market capitalization of the Tokyo Stock Exchange (Lichtenberg and Pushner 1992). Average ownership in quoted companies held by financial groups has risen to 30 percent in 1989, while average corporate ownership remained stable over the period 1975-1989 at 43 percent.

Franks and Mayer (1995c) argue that the theories of ownership and corporate control do not provide adequate explanations for the organization and operation of Anglo-American, Japanese and Continental European capital markets. They advance the hypothesis that the patterns of ownership are associated with different forms of corporate control that allow for different types of correction. Concentrated ownership allows relations involving commitment on the part of investors to be sustained. Large shareholders who face limited free riding costs of control, can give a long-term commitment to the firm, while allowing a large number of small shareholders to trade in investment opportunities without having any effect on control. Dispersed ownership gives management more discretionary power but permits restructuring of management (e.g. by takeovers or by a market for share stakes as shown in chapter 2) even in the absence of past failure, largely because owners are unable to commit. Consequently, it could be expected that different forms of ownership would be suited to promoting different types of activity. Concentrated ownership is needed where investment by other stakeholders is important and cannot be promoted contractually. When little investment is required by other parties or adequate contracts can be written, dispersed ownership will be advantageous.

5.2 Concentrated ownership in Belgium.

5.2.1 Ownership disclosure legislation.

Up to 1989, little was known about the ownership structure of companies listed on the Belgian stock exchanges, given the general use of bearer shares and the lack of ownership disclosure obligation. Following the takeover battle in 1988 between the French Compagnie Financière de Suez and the de Benedetti group for the largest Belgian holding company, Generale Maatschappij van België (Société Générale de Belgique), new legislation concerning corporate control and ownership was initiated. An Ownership Disclosure Law was introduced in 1989 and amendments to the company law with regard to takeovers were made in 1991.

The Ownership Disclosure Law requires all investors, both individuals and companies, to reveal their share stakes in those companies governed by Belgian law, all or part of whose securities conferring voting rights are officially listed on a stock exchange located in a Member State of the European Union. Notification is obligatory if a shareholding equals or exceeds 5 percent. Furthermore, shareholders have to declare any increases and decreases in ownership and their new ownership position if their stake exceeds a multiple of 5 percent of the voting rights or falls below such a threshold. For instance, a company that has revealed that it owns a stake of 11 percent will have to notify the Banking Commission again once this ownership stake reaches 15 percent or more, or decreases below the 10%-threshold.

The notification percentages refer to real and potential voting rights. As a result, ownership of securities convertible into shares (convertible bonds, warrants, etc) is treated in the same way as shares in the company. So, when investors make voting rights declarations, they include : (i) the percentage of the actual total voting rights they own proportional to all the actual voting rights outstanding, (ii) the potential voting rights, as a percentage to the aggregate of all potential voting rights and (iii) the percentage of cumulative actual and potential voting rights in the company based on the aggregate number of the voting rights associated with all outstanding shares and convertible instruments.

Furthermore, the law applies not only to the direct owners of the voting rights, but also to those investors who control voting rights indirectly via a pyramid structure of intermediate companies. Investors are obliged to reveal whether they are affiliated to a group of companies or whether they act in concert with other investors. If the real or potential voting rights of the individual investor or of the investor group exceed or fall below the notification thresholds, they have to reveal their cumulative and individual direct and indirect ownership positions and changes in shareholdings. The Banking Commission suggests that the ultimate shareholder of an investor group assume notification responsibility for voting rights of its own direct and indirect holdings and for those share stakes held by investors this 'reference shareholder' is affiliated to or acts in concert with. In addition, once the stake of an investor (or of the investors belonging to the same investor group) reaches 20 percent of the voting rights of the company, the strategic policy with regard to the target has to be declared to the Banking Commission and the target.

With regard to timing of notification, the investor who purchases or sells shares (voting rights) has to disclose his shareholding and the changes in his position to the target and to the Banking Commission in Brussels at the latest on the second working day after the transaction, if a notification threshold has been passed. The target who has been notified about changes in ownership by substantial investors, has a maximum of one working day after disclosure to pass on this information to the Documentation and Statistics Department of the Brussels Stock Exchange (Maertens 1994). This department updates its on-line ownership database BDPart and makes this information available ad valvas on the trading floor (parquet). The following day, the Documentation department publishes the information in the Cote de la Bourse, a Stock Exchange publication that is inserted in the two Belgian financial newspapers, De Financieel Economische Tijd and L'Echo de la Bourse. The same notification timing applies to disclosure of investors' policies (20 percent ownership rule).

An investor's failure to disclose a substantial shareholding may lead to an interdiction for the investor in question to participate to the annual meeting, to a cancellation of the annual meeting which has been called for, to a suspension of the exercise of all or part of the rights pertaining to the securities for a certain period and to liability to penalties. The voting rights of recently acquired major shareholdings (5 percent and more) can only be exercised 45 days after notification.

5.2.2 Voting rights and restrictions, and the rights of the minority shareholders.

In principle, the general assembly takes decisions based on a simple majority of the voting rights. Since 1991, the balance of corporate power has shifted to the controlling shareholders who have been given legal instruments to entrench their position in the company and to protect themselves against undesired takeovers. Anti-takeover instruments, like share repurchase schemes or issuance of warrants, are valid for a maximum of 5 years but can be reinstated for a similar period by the general assembly (Wymeersch 1994a). Such measures have further reduced the likelihood of hostile takeovers in Belgium.

However, to provide more protection to small shareholders a supermajority of 75 percent of the voting rights voted at the general assembly, is needed with regard to decisions about changes in the acts of incorporation, increases of the equity capital, limitations or changes in the preferential rights of existing shareholders to purchase shares in new equity issues, changes in the rights of different classes of shareholders, repurchases of shares and changes in the legal form of the corporation (Lievens 1994).

Since 1991, minority shareholders or a group of minority shareholders owning at least 1 percent of the equity capital or shares with a value of not less than BEF 50 million, can appoint one or more experts who can scrutinize the company's accounting and its internal operations. The appointment of experts is conditional on indications that the interests of the company are threatened. Shareholders owning at least 1 percent of the votes can initiate a minority claim against the directors for the benefit of the company, if it can be proven that the directors have managed or supervised the company poorly and if the minority shareholders have voted against the directors' discharge at the annual meeting. For instance, a minority claim would be justified when directors ensured that the company paid out benefits to large shareholders they represent at the detriment of the company.

Another important change, since the law of 1991, is the abolition of automatic voting rights restrictions. This abolition was motivated by the fact that the restrictions could be easily evaded by redistributing the shares to family members, friends and subsidiaries (Van Nuffel 1994). Still, as in Germany, individual companies can still apply voting right restrictions by including such clauses in the acts of incorporation. While automatic voting restrictions are abolished, voting agreements among shareholders for (renewable) periods of 5 years are allowed since 1991 if these agreements do not limit the responsibilities of the directors or are used to create different classes of voting rights.

5.2.3 Concentrated direct and ultimate ownership by shareholder class.

The structure of substantial shareholdings in all Belgian companies listed on the Brussels Stock Exchange in August 1994 is presented in table 5.2. On average, the sum of the direct share stakes held by large shareholders (who own at least 5 percent of the outstanding shares) amounts to more than 65 percent (panel A). Cumulative direct ownership is higher, almost 70 percent in the financial sector (panel C), and around 65 percent for both holding companies (column 1 of panel B) and industrial and commercial companies (panel D). It is clear that the concentrated ownership structure does not facilitate hostile takeovers if the acquirer does not initially have a large toehold. In their analysis of the Belgian market for corporate control over the period 1970-1985, Van Hulle, Vermaelen and de Wouters (1991) confirm that tender offers made directly to the public were characterised by substantial initial toehold interests.

Table 5.2 also reports the cumulative ownership of the three most important investor classes: holding companies, families and individual investors, and industrial and commercial companies. From panel A can be concluded that holding companies are the largest direct investors; they hold on average 33 percent of the shares and account for half of the substantial ownership stakes in Belgian companies. Domestic and foreign holding companies have invested more in the Belgian holding companies than in the industrial and in the financial sector. Direct investment of industrial and services companies (panel A) totals almost 15 percent and is focused on other industrial and commercial companies (panel D). Families' direct investment is of less importance with an average stake of about 4 percent.

A substantial number of share stakes are held by other companies which in turn are held by other shareholders. Therefore, if we want to answer the question who actually owns and controls a sample company, pyramidal and complex ownership structures should be taken into account. Examples of pyramidal and complex ownership structures are illustrated in figures 1 and 2. Figure 1 shows part of the ownership structure of Floridienne, a company in the chemical and food industry, at the end of 1994. On the direct investment level, Mosane and its fully owned subsidiary Cippar hold 25 percent of Floridienne's voting rights. Ultimate minority control lies with the Paribas group which controls its Belgian subsidiary Copeba. Ultimate minority control exists when there is a continuous chain of at least 25 percent if there are no other shareholders with large stakes available at any ownership tier. A continuous chain of holdings of at least 50 percent is called ultimate majority control while supermajority control arises when an uninterrupted chain of 75 percent is in place. The most important reason for the use of pyramids in Belgium is leverage (Wymeersch 1994a) : external equity can be raised while retaining control. The Paribas group controls the blocking minority in Floridienne with an interest in cash flow rights of merely 11 percent (60% x 74% x 25%). In fact, Paribas exercises pyramidal or levered control over Mosane. It is clear that, although the one share-one vote rule applies to each individual ownership tier, pyramidal or levered control constitutes a violation of the one share-one vote rule if control extends throughout multiple ownership tiers.

This table reports the aggregate of individual shareholdings of 5% and more¹ for the main ownership categories. The shareholder classes (holding companies, industrial and commercial companies, and families) consist of both Belgian and foreign investors. Direct stands for the direct shareholdings. Ultimate refers to the fact that the direct shareholdings were classified according to the shareholder class of the ultimate investor and these direct shareholdings belonging to the same ultimate investor group were subsequently summed. Ultimate control is control based on (i) a majority control (minimal 50% of the voting rights) on every ownership tier of the ownership pyramid or (ii) shareholdings of at least 25% on every tier in the absence of other shareholders holding stakes of 25% or more. A chain of fully owned subsidiaries are considered as one single shareholder.

Aug. 1994	all investors	holding co's	families	industr. co's	Belgian investors	foreign investors
PANEL A : ALL SAMPLE COMPANIES (N=155)
Direct	65.38	32.71	3.90	14.60	49.38	16.00
Ultimate	65.38	26.68	15.59	10.84	39.60	24.35
PANEL B: ALL HOLDING COMPANIES (N=64)
Direct	63.92	36.73	5.15	13.11	46.85	17.07
Ultimate	63.92	34.43	14.12	8.33	36.08	27.97
PANEL C : FINANCIAL SECTOR (BANKS, INSURANCE, REAL ESTATE) (N=19)
Direct	69.96	26.45	1.18	5.45	55.00	14.96
Ultimate	69.96	26.22	5.31	5.41	38.40	23.63
PANEL D : INDUSTRIAL AND COMMERCIAL COMPANIES (N=72)
Direct	65.48	30.80	3.50	18.34	50.16	15.32
Ultimate	65.48	20.02	19.70	14.52	43.01	21.36

Source : Own calculations based on information from the BDPart database of the Brussels Stock Exchange and Ownership Notifications of the Documentation Centre of the Brussels Stock Exchange.

¹ In line with the Ownership Disclosure Legislation, substantial shareholdings are defined as share stakes that equal or exceed 5% (of the voting rights), unless investors with smaller shareholdings are affiliated to or act in concert with major shareholders, in which case small stakes ought to be revealed as well. The 5% threshold can be reduced to 3% if the company states this in its acts of constitution.

Cobepa, a Belgian holding company, is also listed on the Brussels Stock Exchange and its organization chart is exhibited in figure 2. Within the ownership chain, Swiss, French and Dutch companies and banks belonging to the Paribas group control the underlying levels with almost 100 percent of the voting rights. This complex ownership structure, however, is not an example of an ownership pyramid, but is a case of majority control where there is hardly any control leverage. Basically, 60 percent of Cobepa's voting rights are held by one major shareholder, the Compagnie Financière Paribas.

Previous examples clarified that the true owners of the Belgian sample companies are mostly not the direct shareholders (at ownership level 1), but that control is exercised by an ultimate shareholder on a higher ownership tier in the pyramid. It is important to identify these ultimate shareholders so that the percentages of voting rights held by direct or first-level shareholders controlled by the same ultimate investor can be aggregated into investor groups. Such investor group is named after and classified according to the identity and shareholder class of the ultimate shareholder. Control exerted by an ultimate shareholder on a sequence of intermediate companies and, ultimately, on the sample company exists if (i) there is a series of uninterrupted majority shareholdings on every ownership tier throughout the pyramid or (ii) if there is a large shareholding of at least 25 percent on every ownership level in the absence of other shareholders with stakes of blocking minority size or larger. Applying this criterion, henceforth called the ultimate shareholder criterion, to the example (figures 1 and 2), the direct shareholdings of Mosane (18.9%) and Cippar (6.1%) are summed to 25% and classified according to the shareholder category of the ultimate shareholder (Paribas), namely, a holding company.

Table 5.2 also details the aggregate large share stakes of the main investor classes after applying the ultimate shareholder criterion. Although holding companies remain the most important shareholder class in Belgian listed companies, their average cumulative shareholding on an ultimate control basis decreases to 26.7 percent from an average direct shareholding of 32.7 (panel A, table 3.2). The differences are explained by the fact that family controlled holding companies are now classified according to the identity of the ultimate investors, namely, families and individuals. The average shareholding held by industrial and commercial companies decreases to 11 percent for similar reasons. Industrial and commercial companies seem more inclined to hold substantial stakes in other industrial firms (panel D). Individual and family investors frequently do not hold shares directly in Belgian companies, but use intermediate companies as their average concentrated ownership amounts to almost 16 percent, while direct stakes held by individual and family investors average only 4 percent (panel A). Family shareholdings are most distinctly present in the ownership structure of industrial and commercial companies (panel D) with an average substantial shareholding of nearly 20 percent.

The relative importance of domestic and foreign investors is examined in the last two columns of table 5.2. More than 75 percent of the direct large shareholdings (or an average of 49.4 percent of the voting rights) are held by Belgian investors, while foreign investors' direct investments account for an average of 16 percent. This proportion is similar for holding companies (panel B) and the industrial firms (panel D), but for the financial sector, domestic investments are higher with an average of 55 percent (panel C). When applying the ultimate shareholder criterion and taking account of the nationality of the ultimate shareholders, columns 5 and 6 show that foreign investors often use Belgian intermediary companies to control Belgian listed companies. Domestic ownership in a Belgian company amounts to nearly 40 percent; slightly lower (36%) in holding companies, and somewhat higher (43%) in industrial and service companies. Foreign investors hold about 38 percent of the substantial shareholdings (or an average of 24.3 percent of the total number of shares) in Belgian listed companies.

A comparison of the size of means and medians of concentrated cumulative ownership in 1994 and 1989 via parametric and non-parametric tests reveals that neither the total ownership concentration nor the average shareholding by shareholder class has changed significantly over time. This suggests that stakes are mostly sold to investors of the same shareholder class with whom the seller has a priority purchase agreement or to investors who belong to the same investor group.

5.2.4 Pyramiding and the violation of one share-one vote rule.

The ultimate shareholder criterion served to determine control relations through the pyramidal ownership structures. In previous section, we aggregated direct shareholdings which belonged to the same investor group and reclassified the aggregate share stake according to the investor class of the ultimate shareholder. In the example of figure 1, we found that the Paribas controlled 25 percent of the shares of Floridienne. In this section, we examine pyramiding by estimating deviations from the one share-one vote rule. These deviations have potentially important implications with regard to dilution of control. For instance, it is not certain whether a sequence majority control with e.g. 50% at every ownership tier, yields a determining voice in board decisions of the target sample company (level 0).

Table 5.3 shows the average ultimate ownership level (ultimate shareholder criterion). Direct share stakes are defined as level 1-shareholdings. The level from which ultimate control is exercised is, on average, 2.2 and only slightly decreases to 2.1 over the four year period.

This table presents the ultimate ownership level, defined as the highest level of ownership in an uninterrupted control chain (direct shareholdings are level 1). Ultimate control is control based on (i) a majority control (minimal 50% of the voting rights) on every ownership tier of the ownership pyramid or (ii) shareholdings of at least 25% on every tier in the absence of other shareholders holding stakes of 25% or more. A chain of fully owned subsidiaries are considered as one single shareholder.

The direct largest shareholding is the average direct largest share stake of at least 25%. The ultimate levered shareholding is calculated by multiplying the share stakes of subsequent ownership tiers. The control leverage factor is the ratio of the direct shareholding divided by the ultimate levered shareholding. For instance, company A, whose shares are widely held, owns 40% of company B which, in turn, owns 40% of company C. The ultimate shareholder level is 2, the direct largest shareholding (of B in C) is 40%, the ultimate shareholding is 16% (40% x 40%), and the leverage factor is 2.5 (40/16).

There was no direct shareholding of at least 25% in 17 sample companies, which were not included in this table. Standard deviation in parentheses.

	1989	1990	1991	1992
sample size	160	156	156	156
ultimate ownership level	2.2 (1.364)	2.2 (1.290)	2.1 (1.188)	2.1 (1.159)
direct largest shareholding	55.1 (19.737)	56.4 (19.509)	57.2 (19.923)	57.8 (20.632)
ultimate levered shareholding	38.0 (22.524)	38.5 (22.906)	40.3 (23.988)	41.7 (24.600)
control leverage factor (direct/ultimate shareholding)	3.6 (8.391)	3.6 (8.650)	3.0 (6.756)	2.9 (6.710)

Source : Own calculations based on data from the BDPart database and the Notifications of Ownership.

As a proxy for the control leverage effect of the pyramid structures, we define the control leverage factor as the ratio of the direct largest shareholding and its ultimate levered shareholding. The average of the largest direct stake per investor group amounts to about 58% in 1992. The ultimate levered shareholding is calculated by multiplying the consecutive controlling shareholdings. For example, the ultimate levered shareholding of Paribas in Floridienne (see figures 1 and 2) amounts to 11 percent (60%*74%*25%) while the largest direct shareholding of the Paribas group is 25 percent. Consequently, the control leverage factor is 2.27 (25%/11%). The smaller the shareholdings with which control is maintained throughout intermediate levels and the more intermediate ownership tiers, the higher the control leverage factor or the more considerable the violation of one share-one vote. Table 5.3 discloses that the control leverage factor in 1989 was 3.6 and decreases to 2.9 in 1992. Since the average ultimate ownership level and the ultimate levered shareholding do not change significantly over time, the decline of the control leverage factor indicates that control on intermediate levels becomes more concentrated. The average direct largest shareholding for companies with a direct share stake of at least 25 % amounts to 57 percent while the ultimate levered shareholding is 41 percent.

There are substantial differences in pyramiding among the subsamples of the listed Belgian holding companies, financial firms and industrial and commercial companies. In 1992, the ultimate ownership level for financial firms amounted to 2.6 versus 1.9 for industrial companies. Moreover, the control leverage factor for financial firms was 7.1, 3.0 for holding companies and only 1.9 for industrial companies. This reveals that control of holding companies and financial firms is more levered than that of industrial firms.

We also investigate the control leverage established by the different classes of ultimate investors (table 5.4). Of the 156 sample companies in 1992, 64 ultimate investors were holding companies, 49 were families and 27 were industrial companies. Both the ultimate ownership level and the control leverage factor point out that holding companies, insurance companies and families use more intermediate companies and smaller intermediate share stakes to ascertain control than industrial companies. Hence, the deviation of the concept of one share-one vote is considerable for investing holding companies and, consequently, the potential for dilution of control increases (see chapter 6).

This table presents the ultimate ownership level, defined as the highest level of ownership in an uninterrupted control chain (direct shareholdings are level 1). Ultimate control is control based on (i) a majority control (minimal 50% of the voting rights) on every ownership tier of the ownership pyramid or (ii) shareholdings of at least 25% on every tier in the absence of other shareholders holding stakes of 25% or more. A chain of fully owned subsidiaries are considered as one single shareholder.

The direct largest shareholding is the average direct largest share stake of at least 25%. The ultimate levered shareholding is calculated by multiplying the share stakes of subsequent ownership tiers. The control leverage factor is the ratio of the direct shareholding divided by the ultimate levered shareholding. For instance, company A, whose shares are widely held, owns 40% of company B which, in turn, owns 40% of company C. The ultimate shareholder level is 2, the direct largest shareholding (of B in C) is 40%, the ultimate shareholding is 16% (40% x 40%), and the leverage factor is 2.5 (40/16).

There was no direct shareholding of at least 25% in 17 sample companies, which were not included in this table.

Standard deviation in parentheses.

1992	ULTIMATE SHAREHOLDERS
	holding co's	investment co's	insurance co's	industrial co's	families	government
sample size	64	5	5	27	49	6
ultimate ownership level	2.3 (1.270)	1.4 (0.489)	2.4 (1.496)	1.7 (1.116)	2.0 (0.868)	1.7 (1.105)
direct largest shareholding	57.0 (17.906)	44.6 (12.116)	75.2 (23.961)	60.4 (23.584)	54.6 (20.649)	63.3 (18.607)
ultimate levered shareholding	35.1 (21.741)	31.2 (12.023)	43.6 (27.659)	50.8 (25.277)	41.5 (23.997)	63.3 (21.116)
control leverage factor (direct/ultimate shareholding)	4.3 (9.959)	1.7 (0.877)	3.0 (3.121)	1.5 (1.387)	2.9 (1.387)	1.1 (1.185)

Source : Own calculations based on data from the BDPart database and the notification of ownership disclosure.

5.2.5 Blocking minorities, majorities and supermajorities.

Table 5.5 examines control patterns and gives the percentage of Belgian companies with an ownership structure characterized by the presence of blocking minorities, majorities and supermajorities. When a shareholder possesses more than 50 percent of the voting rights, he can dominate the agenda at the annual meeting and control the selection and hiring process of the board members and the delegated director (CEO). In practice, less than 50 percent of the voting rights will be needed to have a majority on the annual meeting because some - predominantly the small - investors usually choose not to be involved in active monitoring and will only use their voting rights under special circumstances e.g. in the case of a potential acquisition. The importance of a blocking minority of at least 25 percent the voting rights was clarified in section 5.2.2. Therefore, table 5.5 shows the percentage of sample companies with the critical threshold stakes of 25%, 50% and 75%. Both the direct threshold shareholdings are presented and the threshold shareholdings per investor group. Panel A reveals that a voting rights majority exists in more than half (56%) of the Belgian listed companies based on the ultimate shareholder criterion. In 18 percent of the Belgian companies, a supermajority gives absolute control to one shareholder or a group of shareholders as blocking minorities cannot be formed. Shareholdings of 25 percent or more are present in 85 percent of all companies. The concentrated ownership pattern is similar in all subsamples. Share stakes of more than 25 percent exist in more than 80 percent of the holding companies (panel B) and the financial firms (panel C) and even in 93 percent of the industrial and commercial companies (panel D). We find that ownership concentration in very strong in most companies within each subsample. Consequently, as, to large extent, takeovers have to be ruled out as a corporate control mechanism, large shareholders bear responsibility for monitoring management's performance.

Holding companies, both Belgian and foreign, are the main ultimate investors since they dominate with voting rights majorities 26 percent of the Belgian firms (panel A). Holding companies invest mainly in other Belgian and foreign holding and companies (see panels B and D). Family and individual investment (panel A) is high (on ultimate control basis) since they hold stakes of at least 25 percent in almost one fourth of all Belgian listed companies and majorities in 14 percent. This shareholder class owns large stakes (of over 25%) in 29 percent of the industrial and commercial sector (panel D) and has absolute control in 18 percent. The industrial shareholders predominantly hold share stakes of minimum blocking minority size in other industrial companies (panel D).

Total Belgian and foreign ownership concentration based upon direct shareholdings gives a different picture when ultimate control is considered. The proportion of about 75%-25% of the sample companies with direct share stakes of at least blocking minority size held by respectively Belgian and foreign shareholders, changes to a 60%-40% ratio on an ultimate shareholder basis. This fact reconfirms that foreign investors predominantly control stakes in Belgian companies via Belgian intermediaries.

With regard to absolute control in the form of supermajorities, foreign investors control 10 percent of the companies while Belgian investors only control 9 percent (panel A). Table 5.5 also reveals that Belgian and foreign investors each hold majority stakes in 30 percent of the Belgian listed holding companies. Consequently, the proportion domestic versus foreign ultimate investors has changed to a 50%-50% proportion. The majority of Belgian industrial and services companies (panel D) is still dominated by Belgian investors.

This section has disclosed that over the period 1989 till 1994, Belgian ownership was highly concentrated with more than half of the listed companies controlled with majority stakes. The average substantial stakes held by the different ownership classes has remained relatively stable.

Percentage of the sample companies with a minority, majority or supermajority shareholdings held by the main shareholder categories.

MIN = % of companies with a stake of 25% or larger,

MAJ = % of companies with a stake of 50% or larger,

SUP = % of companies with a stake of 75% or larger.

Direct stands for the direct shareholdings. Direct stands for the direct shareholdings. Ultimate refers to the fact that the direct shareholdings were classified according to the shareholder class of the ultimate investor and these direct shareholdings belonging to the same ultimate investor group were subsequently summed. Ultimate control is control based on (i) a majority control (minimal 50% of the voting rights) on every ownership tier of the ownership pyramid or (ii) shareholdings of at least 25% on every tier in the absence of other shareholders holding stakes of 25% or more. A chain of fully owned subsidiaries are considered as one single shareholder.

AUG. 1994	all investors			holding co's			families			indus. co's			Belgian investors			foreign investors
	MIN	MAJ	SUP	MIN	MAJ	SUP	MIN	MAJ	SUP	MIN	MAJ	SUP	MIN	MAJ	SUP	MIN	MAJ	SUP
PANEL A : ALL SAMPLE COMPANIES (N=157)
Direct	82	45	14	48	23	5	2	1	1	21	12	5	63	36	9	19	9	5
Ultimate	85	56	18	41	26	6	23	14	3	15	8	5	51	33	9	34	23	10
PANEL B : HOLDING COMPANIES (N=64)
Direct	79	39	14	50	23	8	5	2	2	17	9	2	59	31	11	20	8	3
Ultimate	83	59	20	50	36	13	22	13	2	9	6	3	45	30	11	38	30	13
PANEL C : FINANCIAL SECTOR (BANKING, INSURANCE, REAL ESTATE) (N=20)
Direct	75	50	10	35	15	0	0	0	0	5	5	5	62	40	10	13	10	0
Ultimate	80	55	15	40	15	0	5	5	0	5	5	5	48	33	10	32	22	5
PANEL D : INDUSTRIAL AND COMMERCIAL COMPANIES (N=73)
Direct	86	47	15	48	25	4	0	0	0	28	15	8	66	37	7	20	10	8
Ultimate	93	55	16	34	19	3	29	18	4	24	11	7	61	37	8	32	18	8

Source : Own calculations based on BDPart and Ownership Notifications.

5.2.6 Belgian shareholder classes

Of the Belgian shareholder classes, the dominant stake holders are families and holding companies. These two shareholder groups hold most of the controlling stakes (in respectively 12% and 11% of all the sample companies) and each shareholder class holds share stakes of more than 25 percent in about 20 percent of the sample companies.

Family shareholders.

Belgian families own a voting rights majority in 15 percent of the industrial and commercial companies and hold 26 percent of the shareholdings of at least 25%. Families also often use the holding companies as investment vehicles to control indirectly a variety of listed and non-listed companies in different industries.

Holding companies.

Belgian holding companies are substantial investors in all sectors : in other Belgian holding companies, in the financial sector and in industrial and commercial companies. The importance of the Belgian holding companies and the lack of large share stakes held by banks should be understood in its historic framework : banking and investment business had to be separated by law in 1934. This resulted in the creation of large financial holding companies which became the major shareholders in the financial institutions and diversified their investments over a wide gamut of industrial and commercial sectors. As clarified in figure 1, pyramidal ownership structures allowed holding companies to exercise levered control with relatively small share stakes.

Financial Institutions.

As of 1934, 'credit institutions' were prohibited from taking share participations in industrial companies. Only since the 1993 Credit Institutions Act which implemented the Second Banking Directive of the European Union, are credit institutions (banks, savings banks and other financial institutions) entitled to hold shares in industrial corporations and holding companies. Currently, credit institutions are allowed to hold up to 10 percent of their equity in Belgian shares. There is no limitation with regard to the percentage of the outstanding shares of an individual company a credit institution is allowed to own.

In practice, banks still do not invest much in shares of non-financial companies to avoid conflicts of interest :

- According to Belgian law, banks are held liable towards creditors of bankrupt companies, if the banks granted credit to these companies at times when a reasonably prudent banker should not have granted nor maintained the credit. A substantial shareholding in a financially distressed company by a bank might influence that bank's decision with regard to ceasing additional credit.

- Since most banks are controlled by a holding company which might be a substantial shareholder in a company, it is doubtful whether banks would be able to make independent decisions with regard to a shareholding in that company or the loans granted to a company (Verwilst 1992).

- Most investment and pension funds are managed by a bank that ensures the distribution of the investment fund's certificates (shares). Legally, investment and pension funds' management should use the voting rights associated with the shares of a company they have invested in, independent of the managing bank.

The Government .

In principle, the federal state does invest in listed Belgian companies. But it owns 50 percent of the shares of the National Bank, of which the shares are listed in the Brussels Stock Exchange, and 50 percent of the 'public credit institutions'. The role of the public credit institutions has been broadened to that of a bank and they are being privatised. The 'public investment companies', owned by the regional governments hold blocks in shares of a few listed companies. Those investments were made either to save ailing companies or to provide risky companies with growth capital so as to stimulate and support entrepreneurial and industrial expansion activities. In general, in contrast to France, federal and regional governments have not considered their shareholdings in companies as a long term financial investment. Only in two percent of the listed companies, the state still holds a share stake via the regional investment companies.

Employee shareholdership.

Since 1991, mechanisms of beneficial acquisition of shares by employees have been introduced. In general, employee ownership in most companies remains low. For instance, employees of Petrofina own 5.4 percent of the shares; in de Bank Brussels Lambert, employees hold 7%; in Creyf's Interim 0.9%; in Desimpel Kortemark 0.5%; in Royale Belge, 0.69% (Wymeersch 1994a).

Institutional investors.

Belgian institutional investors (insurance companies, pension funds, credit institutions, investment funds and investment companies) usually hold small share stakes (of under 5 percent), but own in aggregate about 22 percent of the shares in Belgian listed companies. For instance, the average shareholding of all Bevek/Sicav-investment funds in the 60 most traded Belgian companies, amounted to 1.5 percent in 1994 and the average shareholding of pension funds measures about 4 percent (B.B.L. 1994). Insurance companies are legally allowed to invest up to 25 percent of their reserves in shares listed on the Belgian stock exchanges, but owned only about 12 percent of the Belgian shares over the period 1986-1991. Most institutional investors reinforce the present majority's power by systematically voting in favour of management or, more commonly, by not taking part in the general assembly.

5.2.7 Foreign shareholder classes

Of the foreign investors, it is primarily the holding companies that hold large share stakes and control with a majority stake in 15 percent of all the Belgian listed companies. Foreign holding companies invest predominantly in Belgian holding companies, one fourth of which they control with a majority of the voting rights. This way foreign holding companies also indirectly invest in unlisted Belgian companies with shares held in the investment portfolios of Belgian holding companies. Foreign industrial companies prefer Belgian industrial companies as long term investments, while foreign banks and insurance companies are substantial shareholders in the Belgian financial and insurance sector. Foreign institutional investors do not rely heavily on the Belgian stock market.

Although shareholders from a wide variety of countries are present in the ownership structure of Belgian listed companies, the main investors are from the neighbouring European countries. Dutch investors own an average direct share stake of 3.8 percent and invest predominantly in Belgian industrial and commercial companies. German direct average ownership is low. German industrial companies mainly invested in the concrete industry via e.g. Heidelberger Zement. Investors from Luxembourg own, on average, directly 4.1 percent of Belgian companies, and have invested mainly in industrial and commercial companies. But, companies from Luxembourg are almost never the ultimate investor and are used as intermediary investment vehicles by e.g. French companies. U.K. and North American shareholders hold large stakes in only 3 companies. Only one large shareholding of a Belgian listed company is Japanese: Ashaki acquired a majority stake in the glass manufacturer Glaverbel. The average French direct average shareholding is higher and close to 4.3 percent.The single most important foreign ultimate investors are French; their accumulated substantial shareholdings amount on average to almost 13 percent (table 3.4). They invest mainly in the Belgian holding companies of which they own an average stake of 19 percent and in the financial sector in which they hold an average of 14 percent of the voting rights. Via controlling participations in Belgian large holding companies, French investors control a substantial part - estimated at 30% (Wymeersch 1994a) - of all the listed and unlisted industrial companies in Belgium. Columns 2 to 5 of table 5.6 reveal that it is the French holding companies, rather than French family investors or industrial companies that have acquired substantial stake of the Belgian listed companies. French insurance companies own significant shareholdings in the Belgian banks and insurance companies.

Table 5.6 : Size of large shareholdings held by a French ultimate investor (group).

This table reports the aggregate substantial shareholdings¹ owned by the main French investor groups.

Ultimate refers to the fact that the direct shareholdings were classified according to the shareholder class of the ultimate investor and these direct shareholdings belonging to the same ultimate investor group were subsequently summed. Ultimate control is control based on (i) a majority control (minimal 50% of the voting rights) on every ownership tier of the ownership pyramid or (ii) shareholdings of at least 25% on every tier in the absence of other shareholders holding stakes of 25% or more. A chain of fully owned subsidiaries are considered as one single shareholder.

AUG.1994	SHAREHOLDINGS OWNED BY ULTIMATE FRENCH INVESTORS					SHAREHOLDINGS EXCLUDING SUEZ AND PARIBAS
	all investors	holding co's	insurance co's	indus. co's	families	all investors	holding co's
PANEL A : ALL SAMPLE COMPANIES (N=157)
MEAN	12.89	9.37	1.05	1.41	0.45	6.32	2.80
STD	25.17	22.27	8.53	8.91	5.67	19.39	13.91
t-stat3	­1.7754	­1.7404	­0.453	0.125	­0.600	­0.670	­0.513
PANEL B : ALL HOLDING COMPANIES (N=64)
MEAN	18.82	15.28	0.16	2.28	1.11	9.21	5.67
STD	31.09	29.11	1.25	12.40	8.88	24.30	20.07
t-stat3	0.040	­0.015	0.120	0.472	­0.064	0.050	­0.025
PANEL C : FINANCIAL SECTOR (BANKS, INSURANCE, REAL ESTATE) (N=20)
MEAN	13.96	5.72	7.76	0.00	0.00	11.61	3.37
STD	25.82	15.04	23.19	0.00	0.00	26.01	13.98
t-stat3	­1.253	­0.933	­0.408	0.000	­1.000	­0.729	­0.080
PANEL D : INDUSTRIAL AND COMMERCIAL COMPANIES (N=73)
MEAN	7.39	5.19	0.00	1.04	0.00	2.33	0.13
STD	17.00	14.87	0.00	6.00	0.00	9.38	0.84
t-stat3	­2.2745	­2.4845	­0.998	­0.384	0.000	­0.783	­1.511

Source : Own calculations based on BDPart and Ownership Notifications.
¹ In line with the Ownership Disclosure Legislation, substantial shareholdings are defined as share stakes that equal or exceed 5% (of the voting rights), unless investors with smaller shareholdings are affiliated to or act in concert with major shareholders, in which case small stakes ought to be revealed as well. The 5% threshold can be reduced to 3% if the company states this in its acts of constitution.

² The direct shareholdings are accumulated if they are directly owned or (indirectly) controlled by a French ultimate investor (group)

³ The t-stat. tests the difference between the ownership means in 1994 and 1989. Non-parametric tests give similar results.

⁴ Statistical significance at 10%.

⁵ Statistical significance at 5%.

The French Suez group controls the Generale Maatschappij van België (Société Générale de Belgique) and the Paribas group dominates the Belgian Cobepa holding. To investigate the prominence of these two large French holding companies, the average substantial shareholdings held by French investors excluding the Suez and the Paribas group are presented in columns 6 and 7 of table 5.6. A comparison of the aggregate concentrated French ownership including and excluding Suez and Paribas reveals that these holding companies account for more than half of the substantial French investments in Belgian listed companies (holding and industrial companies). The average large share stake held by the French holding companies falls from 9.4% to 2.8% after exclusion of the Suez and Paribas holding companies (columns 2 and 7). The 9.4% average shareholding is equivalent to majority control in 10 companies and the 2.8% represents control in 2 companies. Apart from controlling stakes, Suez and Paribas are present with minority stakes in 45 listed companies. Panel D (column 7) shows that the French holding companies other than Suez and Paribas, control virtually no voting rights directly in the Belgian industry.

The French average shareholding slightly decreases from 1989 to 1994 mainly due to a reduction of ownership by the French holding companies. An important reason is the restructuring of the Generale Maatschappij van België (Société Générale de Belgique) after the takeover by Suez. Since then, the Generale focuses on eight core strategic sectors and has reduced its shareholdings in others.

5.2.8 Changes in large shareholdings.

We have shown that the aggregated large shareholdings per shareholder category remained stable over time. As selling activity of stakes within shareholder categories is not reflected in the aggregate ownership data, table 5.7 examines these changes in large shareholdings. Over the period 1989-1992, there were 238 shareholding increases of more than 1 percent, while 247 stakes were sold. Of these changes in ownership, there were 120 increases of a magnitude between 5% and 24.9%, versus 110 decreases of similar size. In 16 cases, majority shareholdings were acquired and 28 blocks of blocking minority size were purchased. Thirty-three blocking minorities were sold, in addition to 28 majority stakes. It should be noted that the changes are corrected for shareholding restructuring within investor groups. For example, a redistribution of share stakes in a sample company held by two companies which are controlled by the same ultimate investor, has a limited impact on control and is consequently not included in the changes of large shareholdings.

These observations suggest that this market for share stakes is not insignificant : in one fourth of the sample companies, share stake changes of 5 percent or more occur in the period 1989-1992. The relevance of this market as a an external corporate control mechanism will be investigated in the following chapter.

Table 5.7 discloses that the holding companies are the main sellers and purchasers of share stakes. Institutional investors, mainly banks and insurance companies, acquire 38 shareholdings of more than 5 percent and sell 30 stakes of similar sizes. Families sell 15 stakes of blocking minority size and more, while 8 such stakes are bought by this shareholder category. Most of the exchanges of blocks of shares are negotiated deals and take place ex exchange.

Table 5.7 : Increases and decreases of large shareholdings over 1989-1992.

This table gives the size distribution of increases and decreases of large shareholdings over the period 1989-1992. Increases and decreases

were calculated by comparing the share stakes of a shareholder category of a fiscal year to the shareholdings of previous year.

1989-1992	Number of increases and decreases stakes
	[1%-5%[	[5%-10%[	[10%-25%[	[25%-50%[	[50%-100%[	Total
PANEL A : INCREASES FOR ALL SAMPLE COMPANIES (number of observations : 693)
Increases : all shareholders	74	72	48	28	16	238
Increases : holding companies	34	35	17	16	2	104
Increases : institutional investors	24	17	12	4	5	62
Increases : industr. & commerc. co's	5	9	8	4	5	31
Increases : families	11	11	11	4	4	41
PANEL B : DECREASES FOR ALL SAMPLE COMPANIES (number of observations : 693)
Decreases : all shareholders	76	51	59	33	28	247
Decreases : holding companies	26	31	34	12	18	121
Decreases : institutional investors	31	8	11	9	2	61
Decreases : industr. & commerc. co's	3	2	6	1	4	16
Decreases : families	16	10	8	11	4	49

Source : Own calculations based on BDPart and Ownership Notifications.

A more detailed overview of the selling and purchasing is presented in table C3 (appendix C). For the subsamples of the holding companies, financial firms and industrial and commercial companies, holding companies remain the most active sellers and buyers. Not many large stakes of 25 percent or more in holding companies and financial firms are sold, but 31 such stakes change hands in industrial and commercial firms.

If this market for share stakes proves to be a relevant corporate control mechanism (see chapter 6), the findings of table 5.7, suggest that the holding companies are fulfilling a monitoring role, but monitoring is limited to the industrial and commercial sector.

5.3 The board of directors and the management committee.

5.3.1 The board of directors.

Role and size of the board of directors.

The power of the board is 'residual', which means that the board holds the authority that has not been explicitly conferred by law to the annual general assembly. The articles of association, however, can redefine the respective responsibilities of the board and of the general assembly. Usually, the general assembly at the annual meeting is responsible for, apart from the election of directors and auditors, the approval of accounts and the changes of the articles of association. Consequently, the board possesses wide legal powers and has substantial freedom in decision making. Still, the board acts under supervision of the general assembly: board members cannot be appointed permanently but can be dismissed at will by the general assembly, without notice or indemnity. However, case law concerning conflicts between board and annual meeting or concerning conflicts interests of directors on the board is rare. Sensitive issues, like e.g. conflicts or discussions about directors' renumeration, never reach the outside world due to the lack of an investigative press and a tradition of secrecy (Wymeersch 1994a).

The general assembly at the annual meeting has the authority to appoint the board of directors. As a result, the composition of the board to a certain extent reflects balance of power of the shareholders : the major shareholders are well represented whereas small shareholders do not usually have board representatives. Unlike the German two-tier board system, a system of co-determination was not created in Belgium.

The board of directors consists of almost ten members over the period 1989-92 (table 5.8) with a median of nine. Whereas the board of holding companies and industrial and commercial corporations comprises about 9 directors, the boards of banks and of utilities have much larger boards with an average of about 19 directors. Average board size is relatively small for insurance companies with 5 directors and for the services industry (retail, leasing, etc) which has a mean and median of 7 directors. The t-test on the means and non-parametric tests (not shown) indicate that the average board size has not changed substantially over the period 1989-1992.

This table presents the size of the board of directors as well as average yearly turnover of total board, of executive and of non-executive directors, and the percentage of sample companies of which the chairman or of CEO is replaced.

In parentheses : number of sample companies and the standard deviation.

ALL SAMPLE COMPANIES	1989	1990	1991	1992	average 1989-1992
Size of the board	10.034 (175,6.495)	10.005 (175,6.529)	9.872 (173,6.066)	9.738 (168,6.185)	9.914 (691,6.311)
Board turnover1	0.110 (175,0.157)	0.103 (175,0.162)	0.106 (173,0.153)	0.097 (168,0.154)	0.104 (691,0.156)
Executive director turnover2	0.312 (173,0.611)	0.403 (173,1.046)	0.252 (171,0.602)	0.277 (165,0.637)	0.312 (682,0.750)
Non-executive director turnover3	0.081 (173,0.144)	0.066 (173,0.126)	0.074 (171,0.145)	0.058 (165,0.129)	0.070 (682,0.136)
Chairman turnover45	0.127 (173,0.334)	0.093 (172,0.291)	0.064 (171,0.246)	0.083 (167,0.277)	0.092 (682,0.289)
CEO turnover5	0.224 (174,0.418)	0.217 (175,0.413)	0.121 (173,0.327)	0.154 (168,0.362)	0.180 (690,0.384)

Sources : Own calculations based on annual reports, the database of the National Bank of Belgium.

All turnover data are corrected for 'natural turnover', turnover resulting from retirement, illness or death.

1. Proportional to board size.

2. Proportional to total number of executives on the board.

3. Proportional to total number of non-executives on the board.

4. Non-executive chairman turnover only.

5. Percentage of sample companies with chairman or CEO turnover.

Shareholder representation on the board of directors.

In Germany, the Vorstand (management board) consists of 7 senior managers, while representatives of the major shareholders (banks, families and affiliated companies) and of unions and employees have seats on the supervisory board (Aufsichtsrat) (Kaplan 1994b). The German two-tier system integrates otherwise unrepresented or underrepresented interests into the governance structure. For instance, employee representatives constitute half the supervisory board on which executives are prohibited to reside. The true power of the supervisory board stems from the system of cross-holdings and from the banks' right to use the voting rights of the shares deposited with them. French, Spanish and Italian company ownership structures are characterised by controlling interest of a single family or individual, a core group of shareholders or the state.

In Japan, the board of directors is responsible for managing the corporation, but few are outside directors (Gerlach 1993). Most of the directors are executive managers who are long-term employees. Among the directors, the president of a Japanese company is the most important; he generally fulfils the role of CEO. In addition to the president, two or three directors are given the legal right to represent the company (Kaplan & Minton 1994, Kaplan 1994a, Aoki 1990). The outside directors, who form a small minority on the relative large Japanese boards, hold no representative rights but generally represent either the main bank of the company or an affiliated company of the keiretsu and collect information for them.

Anglo-American companies rely on a strong monitoring role for non-executive directors and on the appointment of a greater number of independent non-executives to boards (Oxford Analytica 1993). Moreover, audit committees consisting of a majority of or completely of outside directors, are present in a vast majority of quoted companies in the U.S., Canada and the U.K. Other committee structures like nomination and remuneration committees are increasingly established to reduce the CEO's potential board domination and to facilitate challenging his authority.

The need for independent directors has been emphasized in France, the U.K. and North America, but has not been incorporated into Belgian corporate governance legislation. Some Belgian companies, however, nominate foreign business associates, former politicians or leading personalities to the board (Wymeersch 1994a). Bankers are not present on the board due to banking restrictions and potential conflicts of interest. Institutional investors who usually only hold small share stakes are seldom represented on the board.

Data on board representation in Belgium are not publicly available. As an example of board representation, we focus on the financial holding company, Nationale Portefeuille Maatschappij (Compagnie Nationale à Portefeuille). During the fiscal year 1994, the board of directors of NPM consisted of 15 members of whom only 2 had executive tasks (see table 5.9). The sole large shareholder, the holding company Frère-Bourgeois Group, owned 50.3 percent of the shares and had 2 board representatives. Via intermediate holding companies NPM controls the Group Brussels Lambert (GBL). Some of the companies controlled by GBL, like Royale Belge and Electrabel, who own a minor share stake in NPM have a representative on the NPM board. The insurance company Royal Belge owns 4.5 percent and appoints 1 director. Electrabel, an energy holding company, has 1 board representative. Mosane, a financial holding company and a subsidiary of Cobepa (Paribas), holds a stake of 2.8 percent (over the last 4 years reduced from 10 percent) has 1 board representative. Two board members represent the minority shareholder Paribas. An institutional investor who held 2 percent of the voting rights nominated 1 director. Of the four directors who were considered to be independent, one was a former chairman of a large Belgian bank and another a former politician. One director was appointed by the large shareholders to represented the small (atomistic) shareholders and oversee whether their rights were respected. This table shows that, despite the fact that large shareholders appoint the directors, their direct representatives do not form a board majority. However, it is still uncertain to what extent the direct representatives of the majority shareholder dominate board and to what

Table 5.9 : Shareholder representation on the board of the Nationale Portefeuille Maatschappij (NPM).

1994	Size of shareholding in NPM	Number of board representatives
Large shareholders
Group Frère-Bourgeois	47.8% (and 2.5% indirect control)	2
Minority shareholders
Mosane	2.8	1
UAP Paris and Elf Acquitaine (via Figexsa, Safrep)	12.2 (sum)	2
Institutional investor	2	1
Minority shareholders, which NPM controls indirectly (via GBL group)
Royal Belge	4.5	1
Electrabel	0	1
Independent and Executive directors and minor shareholders' representatives
Minor shareholders (atomistic)	28.8	1 (appointed by large shareholders)
Independent directors	---	4 (1 politician, 1 former bank chairman, two 'experts')
Executive directors	---	2
Total		15

Source : annual report of Nationale Portefeuille Maatschappij and information provided by NMP's CEO.

degree minority shareholder representatives and independent experts are involved in actual decision making and monitoring of management.

Turnover of directors, CEO and chairman.

Table 5.8 shows that total yearly board turnover amounts to 10 percent and is relatively stable over time. On average, 75 percent of the directors do not exercise executive functions in the company. Industrial and commercial companies comprise a higher proportion of executive directors, namely about 37 percent. Average executive turnover proportional to the total number of executives on the board, is high and amounts to an average of 31 percent over the period 1989-1992 and increased even to 40 percent in 1990. Executive turnover in industrial and commercial companies amounts of 28 percent, but is higher in holding companies (31%) and in the financial sector (42%). Non-executive turnover is much lower and is on average 7 percent (see table 5.8). In holding companies and financial firms, average non-executive turnover amounts to 5 percent and 7 percent respectively and is 8 percent in industrial and commercial companies.

Seven percent of the sample companies lose their non-executive chairman. Over a four year period in about 36 percent of the companies the non-executive chairman leaves or is replaced. Dual control, separation of the functions of CEO and chairman, is established in 55 percent of the sample companies. CEO turnover is high: in almost 70 percent of the companies the CEO leaves within the period 1989-1992, equivalent to an 18 percent yearly turnover. For a sample of 46 U.S. companies that had negative earnings during at least one year followed by three years of positive earnings, John, Lang and Netter (1992) report a CEO turnover rate of about 95 percent over a four year period : in half of their sample companies the CEO leaves in the year of negative earnings and the average turnover rate amounts to 15 percent over the three following years. Coughlan and Schmidt (1985) detect an annual turnover rate of 13 percent, Weisbach (1988) attains an 8 percent resignation rate and Klein and Rosenfeld (1988) report more than 9 percent. Gilson (1990) finds a forced CEO turnover rate of 83 percent in companies in financial distress. Martin and McConnell (1991) show that the turnover of the top manager increases from approximately 10 percent in each of the five years preceding a takeover to 41.9 percent in the year after a takeover. The Belgian turnover data are stable over time. Less directors are joining the company following turnover, which explains slightly smaller board size over the period 1989-1992.

5.3.2 The management committee.

Although the creation of a management or direction committee is not a legal requirement, 60 percent of the listed companies explicitly mention the existence of such a committee in the annual reports. More then two thirds of the industrial and commercial companies organize top management into a formal management committee. The committee consists of the CEO, executive directors and the top senior managers. The chairman of the management committee usually is the delegated director (CEO), who serves on the board. The annual general meeting sometimes appoints several delegated directors : the average number of delegated directors is 1.2 (with a median of 1). Banks, for instance, appoint on average of 2.7 delegated directors. A management committee has on average 3.6 members (see table 5.10). The number of executives on the management committee of holding companies totals 2.8 whereas that of industrial and commercial companies amounts to 4.

The turnover of the management committee, excluding CEO turnover, amounts to 13 percent and is low compared to a yearly average of 31% for the executive directors . After excluding turnover of the executive directors, the turnover of top managers who serve on the management committee but not on the board, is even more reduced. This is consistent with the fact that it is the most important members of the management committee who usually get a seat on the board of directors. Only these executives, who bear most responsibility for the company's results and strategy, will face disciplinary actions when the company underperforms or will resign as a result of policy conflicts.

Table 5.10 also shows that an average of about 2.5 executive directors (or about 25 percent of the board) are a member of the management committee. In holding companies and firms in the financial sector, a similar average and proportion of the board acts as executive director, while in the industrial and commercial companies more than 3 board members (or about 38 percent) have managerial functions.

This table presents the size of the management committee, the number of directors serving on the committee and yearly turnover data.

In parentheses : number of sample companies and the standard deviation.

ALL SAMPLE COMPANIES	1989	1990	1991	1992	average 1989-1992
Size of management committee	3.560 (175,3.358)	3.451 (175,3.145)	3.537 (173,3.146)	3.607 (168,0.362)	3.538 (691,3.208)
Management committee turnover1	0.144 (172,0.327)	0.186 (172,0.593)	0.086 (170,0.233)	0.110 (165,0.291)	0.132 (679,0.389)
Number of directors member of mgt committee2	2.508 (175,2.167)	2.440 (175,2.094)	2.433 (173,1.989)	2.428 (168,1.950)	2.452 (691,2.049)

Sources : Own calculations, based on annual reports, the database of the National Bank of Belgium, Memento der Effecten.

Turnover data are corrected for 'natural turnover', turnover resulting from retirement, illness or death.

1. Turnover data exclude turnover of CEO.

2. The number of managers who serve on the management committee and also hold an executive directorship.

5.3.3 Correlations of the turnover of CEO, of executive directors and of the management committee.

Replacement of an underperforming management team consisting of the CEO, executive directors and members of the management committee may result from disciplinary actions taken by actively monitoring non-executive directors or large shareholders. An investigation of the correlations between these turnover variables, our proxies for disciplining, gives some preliminary insights in process of corporate control. High correlations would indicate that when the CEO is replaced, executive directors and the members of the management team are disciplined as well. Table 5.11 reports that the correlation between CEO and executive turnover (excluding CEO turnover) is not high and amounts to 0.125. The coefficient would equal 1 if CEO departure coincided with a complete turnover of the executive directors. The low correlation indicates that when the CEO departs, not many other executive directors leave the company. The Pearson correlation between turnover of the management committee (excluding CEO) and the executive directors (excluding CEO) is expected to be high since the variables partially overlap; the executive directors are members of the management committee. However, the correlation is at 0.301 rather low. This suggests that when executive directors are replaced, this is hardly the case for those members of the management committee who do not hold a directorship. The departure of the CEO coincides more frequently with the replacement of members of the management committee (excluding CEO): the Pearson coefficient equals 0.423. These findings give some preliminary evidence that it is either the CEO with some members of this management committee or the executive directors who are replaced following disciplinary actions or policy conflicts.

The correlation between the turnover of the executive directors and the non-executive turnover is also exhibited in table 5.11. If the hypotheses of previous chapter are supported - the non-executive directors and large shareholders discipline underperforming management - we would expect a low correlation between forced executive and non-executive turnover. The correlation coefficient is only 0.143. For the positive sign two reasons can be given within this corporate control framework : (i) in poorly performing companies, the large shareholder replaces his representative non-executive director who is not fulfilling his monitoring tasks adequately or (ii) large shareholders with low monitoring abilities sell their stake and lose board representation. The relations of poor performance, management turnover, and non-executive and large shareholder disciplining are examined in detail in Chapter 6.

The table presents the Pearson correlation coefficients of the turnover of the board (proportional to board size), the non-executive turnover (exclusive of chairman turnover and proportional to total number of non-executives on board), the executive turnover (exclusive of CEO turnover and proportional to the executives on board), the management committee turnover (proportional to management committee) and the turnover of non-executive chairman turnover and CEO turnover.

Between brackets : p-values corresponding to the null hypothesis that correlation coefficient equals 0.

Number of observations : 702, Period : 1989-1992

TURNOVER OF :	Board of directors	Executive directors1	Non-executive directors2	Management committee3	Chairman4	CEO
Board of directors	1.000
Executive directors1	0.602 (0.00)	1.000
Non-executive directors2	0.693 (0.00)	0.143 (0.00)	1.000
Management committee3	0.255 (0.00)	0.301 (0.00)	0.152 (0.00)	1.000
Chairman4	0.286 (0.00)	0.184 (0.00)	0.133 (0.00)	0.252 (0.00)	1.000
CEO	0.290 (0.00)	0.125 (0.00)	0.205 (0.00)	0.423 (0.00)	0.192 (0.00)	1.000

Source : Own calculations based on a database from the Balance Depository of the National Bank and on annual accounts.

1. Exclusive of CEO turnover and proportional to total number executive directors.

2. Exclusive of chairman turnover and proportional to total number non-executive directors.

3. Exclusive of CEO turnover and proportional to total management committee.

4. Turnover of non-executive chairmen only (corrected for non-separation of control between CEO and chairman).

5.3.4 Director interlocks and control.

Control is usually linked to voting rights since a majority shareholding gives an investor the right to nominate and appoint the directors. However, a company can also have some influence on another company's board via directors interlocks. As mentioned before, data on the representation of shareholders by specific directors, are generally not available. However, the relation between a substantial shareholding in a listed Belgian company held by another listed Belgian holding company, financial firm or industrial and commercial company, on the one hand, and the number of directors these listed corporate investors and targets have in common, on the other hand, can be examined.

The correlation between director interlocks in Belgian companies and ultimate shareholdings for the years 1989 and 1992 is about 0.35 (table 5.12). A correlation coefficient of 1 would indicate that e.g. companies which own 50 percent and 20 percent appoint respectively half and one fifth of their board of directors to the target's board of directors. In practice, as reported in section 5.3.1, a majority shareholder with stake of 50 percent will usually not appoint more than half of the directors because, even if the minority shareholders demand board representation, the direct representatives of large shareholders will have a more powerful voice on board meetings. Within this context, the correlation coefficient of 0.354 is high. Table 5.13 confirms the strong relation between director interlocks and share stakes. We report that a listed Belgian company that owns a share stake of 30 percent (an on ultimate control basis), generally delegates one of the directors of its own board to the board of the target company. Analysis with direct share participations gives similar results.

In total, there were 1582 directorships (positions on the boards of directors) in the 177 sample companies in 1989 and 1587 in 168 companies in 1992. The average number of directorships in listed Belgian companies per director amounts to 1.43. In both years, 78 percent of the directors occupied only one directorship in a listed Belgian company, representing 54 percent of the number of positions. Ten per cent of the directors hold three directorships or more or 28 percent of the total number of directorships in both 1989 and 1992.

The director interlocks are the number of directors two listed companies have in common. The share participations are the share stakes owned by listed Belgian companies in other Belgian listed companies on a 'levered' ultimate shareholder basis. For instance, company A, whose shares are widely held, owns 40% of company B which, in turn, owns 40% of company C. The direct largest shareholding (of B in C) is 40%, the 'levered' ultimate shareholding is 16% (40% x 40%).

The causality with regard to nominations of directors to the board of another company is based on shareholdings : if company A and B have 2 directors in common and company A owns 20% of B's shares while B has no shares in A, company A is assumed to appoint the 2 common directors on B's board of directors.

In parentheses : the p-value expressing the probability that the correlation coefficient differs from zero.

Number of observations : 33,442 (for a 194 by 193¹ matrix of 194 listed companies).

Correlation Matrix	director interlocks 1989	director interlocks 1992	share participation 1989	share participation 1992
director interlocks 1989	1.000
director interlocks 1992	0.662 (0.00)	1.000
share participation 1989	0.354 (0.00)	0.327 (0.00)	1.000
share participation 1992	0.337 (0.00)	0.354 (0.00)	0.783 (0.00)	1.000

Source : Own calculations based on data from annuals reports and Notifications of Ownership.

¹ Diagonal of companies' matrix was deleted.

The share participations (SHARES) are the levered ultimate shareholdings owned by listed Belgian companies in other Belgian listed companies. For instance, company A, whose shares are widely held, owns 40% of company B which, in turn, owns 40% of company C. The direct largest shareholding (of B in C) is 40%, the 'levered' ultimate shareholding is 16% (40% x 40%). The director interlocks (DIRLOCK) are the number of directors two listed companies have in common. The causality with regard to nominations of directors to the board of another company is based on shareholdings : if company A and B have 2 directors in common and company A owns 20% of B's shares while B has not share stake in A, company A is assumed to appoint the 2 common directors on B's board of directors.

9

Number of observations : 33,442 (for a 194 by 193¹ matrix of 194 listed companies).

In parentheses : p-values.

Dep. variable	sample size	intercept	shareholdings 1989	shareholdings 1992	p-value of F-test	R-sq. adj.
directors interlocks 1989	37,441	0.022 (0.00)	0.035 (0.00)		0.00	0.13
directors interlocks 1992	37,441	0.021 (0.00)		0.036 (0.00)	0.00	0.13

Source : Own calculations based on data from annuals reports and Notifications of Ownership Disclosure.

¹ Diagonal of companies' matrix was deleted.

To describe the ownership relation between companies, we can distinguish among affiliation, association and participation. Two companies are affiliated if one owns at least 50% in the other company (the subsidiary). When two companies are associated, one of these companies holds a stake of more than 25% in the other company. Note that 25% is the blocking minority threshold. If a company X owns a stake of less than 25% in company Y, there is a 'participating relationship' between them.

- A, C and D are affiliated. The control percentage of A in D is 60%, while its percentage of interest on a levered basis amounts of 42% (70% * 60%).

- A and E are associated. The control percentage of A in E is 25%, whereas its percentage of interest on a levered basis is only 11% (70% * 60% * 25%).

- A has a participation in B ; percentage of control and interest is 10%.

Cross shareholdings

It is possible that there is a reciprocal shareholdership between two companies. For instance, company P (parent) owns 75% of the shares of company S (subsidiary) while company S owns 5% of company P.

--- 75% ->

P S

<-- 5% ---

To calculate the percentage of interest of P in S, let us assume that a=75% and b=5%. P's shareholders own (1-b) of the share capital of P the remaining b% is held by S. The direct interest of P's shareholders in S is [(1-b) * a]. Indirectly - this is via the shareholdings of S in P, they possess : (1-b) * a * b * a.

If this circular reasoning is repeated several times, the total interest of P in S can be expressed as follows :

direct holding of : (1-b) * a

plus an indirect holding of :

(1-b) * a + (1-b) * a²b + ... + (1-b) * aⁿ⁺¹bⁿ

This sum is a geometrical progression :

(1-b) * a * (1 + ab + a²b² + ... +aⁿbⁿ)

= (1-b) * a * (1 - aⁿbⁿ)/(1 - ab)

And since aⁿbⁿ converges to one for a large n, we can write P's interest in S as :

(1-b) * a / (1 - ab)

Applying this result to our example, we conclude that P's ownership in S amounts to 74,03%.

Via a similar reasoning, we find that the percentage of interest of S in P can be formalized (Uytterschaut 1989) :

(1 - a)/(1 - ab)

Applied to our example, we find that S owns 5,97% of the share capital of P.

Since the shareholdings of the 'subsidiary' are limited to 10% of the share capital of the parent company, the difference between the percentage of interest of the 'parent' company in the subsidiary with and without considering the cross shareholding of the subsidiary will not be substantial.

Table C1 : Largest direct and ultimate shareholdings, and the top level of uninterrupted ownership chains.

This table presents ultimate control, defined as control which is uninterrupted throughout the pyramid if there is a majority shareholding or if there is a large shareholder with at least 25% of the voting rights in the absence of other shareholders with stakes of 25% and more.

The ultimate ownership level defined as the highest level of ownership in an uninterrupted control chain, whereby direct shareholdings are at level 1.The direct largest shareholding is the average direct largest share stake. The ultimate levered shareholding is calculated by multiplying subsequent share stakes.

The control leverage factor is the ratio of the direct shareholding divided by the ultimate levered shareholding. For instance, company A, whose shares are widely held, owns 40% of company B which, in turn, owns 40% of company C. The ultimate shareholder level is 2, the direct largest shareholding (of B in C) is 40%, the ultimate shareholding is 16% (40% x 40%), and the leverage factor is 2.5 (40/16).

A chain of fully owned subsidiaries are considered as one single shareholder.

There was no direct shareholding of at least 25% in 17 sample companies, for which the ownership structure of the largest holding was taken into account.

Standard deviations in parentheses.

	1989	1990	1991	1992
sample size	177	173	173	170
ultimate ownership level	2.3 (1.471)	2.1 (1.330)	2.1 (1.312)	2.1 (1.300)
direct largest shareholding	50.4 (22.898)	51.5 (22.943)	52.6 (23.073)	53.6 (23.453)
ultimate levered shareholding	34.8 (22.131)	35.3 (24.544)	37.1 (24.544)	38.6 (25.222)
control leverage factor (direct/ultimate shareholding)	3.5 (7.956)	3.4 (8.917)	3.0 (6.535)	2.9 (6.555)

Source : Own calculations based on data from the BDPart database and the Notifications of Ownership.

This table presents ultimate control, defined as control which is uninterrupted throughout the pyramid if there is a majority shareholding or if there is a large shareholder with at least 25% of the voting rights in the absence of other shareholders with stakes of 25% and more.

The ultimate ownership level defined as the highest level of ownership in an uninterrupted control chain, whereby direct shareholdings are at level 1.The direct largest shareholding is the average direct largest share stake of at least 25%. The ultimate shareholding is calculated by multiplying subsequent share stakes. The control leverage factor is the ratio of the direct shareholding divided by the ultimate levered shareholding. For instance, company A, whose shares are widely held, owns 40% of company B which, in turn, owns 40% of company C. The ultimate shareholder level is 2, the direct largest shareholding (of B in C) is 40%, the ultimate shareholding is 16% (40% x 40%), and the leverage factor is 2.5 (40/16). A chain of fully owned subsidiaries are considered as one single shareholder.

There was no direct shareholding of at least 25% in 17 sample companies, which were not included in this table.

Standard deviations in parentheses.

	1989	1990	1991	1992
PANEL A : HOLDING COMPANIES (sample size = 60)
ultimate ownership level	2.2 (1.313)	2.2 (1.330)	2.1 (1.202)	2.0 (1.197)
direct largest shareholding	51.8 (16.125)	51.7 (16.491)	53.3 (16.569)	55.3 (18.722)
ultimate levered shareholding	37.213 (20.903)	37.4 (21.604)	38.4 (21.457)	40.7 (23.053)
control leverage factor (direct/ultimate shareholding)	3.7 (9.253)	3.8 (9.498)	3.0 (7.107)	3.0 (7.150)
PANEL B : FINANCIAL SECTOR (sample size = 20 in 1989 and 17 in other years)
ultimate ownership level	2.9 (2.021)	2.6 (1.606)	2.6 (1.603)	2.6 (1.610)
direct largest shareholding	55.7 (19.606)	57.8 (19.746)	61.6 (20.322)	61.5 (20.654)
ultimate levered shareholding	29.8 (23.313)	32.4 (22.654)	33.8 (27.109)	34.8 (28.220)
leverage factor (direct/ultimate shareholding)	7.5 (13.597)	6.4 (13.535)	6.9 (13.535)	7.1 (13.841)
PANEL C : INDUSTRIAL AND COMMERCIAL COMPANIES (sample size = 78 in 1989 and 76 in other years)
ultimate ownership level	2.2 (1.117)	2.1 (1.152)	2.0 (1.018)	1.9 (0.958)
direct largest shareholding	57.3 (21.845)	58.9 (21.113)	59.272 (21.656)	59.012 (21.826)
ultimate levered shareholding	38.8 (23.126)	40.779 (23.614)	43.2 (24.657)	43.9 (24.634)
leverage factor (direct/ultimate shareholding)	2.7 (4.847)	2.8 (5.908)	2.1 (2.337)	1.9 (1.642)

Source : Own calculations based on data from the BDPart database and the Notifications of Ownership.

This table presents the size distribution of increases and decreases of large shareholdings over the period 1989-1992. Increases and decreases were calculated by comparing the share stakes of a shareholder category of a fiscal year to the shareholdings of previous year. The changes in shareholdings per size class over the period 1989-92 are summed.

1989-1992	Number of increases and decreases stakes
	[1%-5%[	[5%-10%[	[10%-25%[	[25%-50%[	[50%-100%[	Total
PANEL A : CHANGES FOR THE HOLDING COMPANIES (number of observations : 273)
Decreases : all shareholders	28	35	27	6	1	97
Decreases : holding companies	13	18	14	3	0	48
Decreases : institutional investors	7	6	4	0	0	17
Decreases : industr. & commerc. co's	2	3	2	2	0	9
Decreases : families	6	8	7	1	1	23
Decreases : all shareholders	34	25	29	12	6	106
Decreases : holding companies	9	18	23	4	3	57
Decreases : institutional investors	14	2	4	7	2	29
Decreases : industr. & commerc. co's	1	1	0	0	0	2
Decreases : families	10	4	2	1	1	18
PANEL B : CHANGES FOR THE FINANCIAL SECTOR (number of observations : 91)
Increases : all shareholders	21	13	2	2	4	42
Increases : holding companies	9	4	0	1	1	15
Increases : institutional investors	10	8	2	1	2	23
Increases : industr. & commerc. co's	2	1	0	0	1	4
Increases : families	0	0	0	0	0	0
Decreases : all shareholders	13	6	9	5	7	40
Decreases : holding companies	6	2	2	0	5	15
Decreases : institutional investors	7	3	6	0	0	16
Decreases : industr. & commerc. co's	0	0	0	0	2	2
Decreases : families	0	1	1	5	0	7
PANEL C : CHANGES FOR THE INDUSTRIAL AND COMMERCIAL COMPANIES (number of observations : 329)
Increases : all shareholders	25	24	19	21	11	100
Increases : holding companies	12	13	3	12	1	41
Increases : institutional investors	7	3	6	3	3	22
Increases : industr. & commerc. co's	1	5	6	3	4	19
Increases : families	5	3	4	3	3	18
Decreases : all shareholders	29	20	21	16	15	101
Decreases : holding companies	11	11	9	8	10	49
Decreases : institutional investors	10	3	1	2	0	16
Decreases : industr. & commerc. co's	2	1	6	1	2	12
Decreases : families	6	5	5	5	3	24

Source : Own calculations based on BDPart and Ownership Notifications.

In parentheses : number of sample companies and standard deviation.

	1989	1990	1991	1992	average 1989-1992
Panel A : HOLDING COMPANIES
Size of the board	9.426 (68,5.953)	9.550 (69,5.942)	9.275 (69,5.606)	9.132 (68,5.719)	9.346 (274,5.777)
Board turnover1	0.092 (68,0.158)	0.095 (69,0.141)	0.075 (69,0.111)	0.092 (68,0.162)	0.089 (274,0.143)
Executive director turnover2	0.352 (67,0.712)	0.384 (68,1.050)	0.203 (68,0.524)	0.338 ((66,0.773)	0.312 (269,0.788)
Non-executive director turnover3	0.064 (67,0.111)	0.056 (68,0.120)	0.056 (68,0.113)	0.039 (66,0.090)	0.054 (269,0.110)
Chairman turnover4	0.089 (67,0.287)	0.058 (68,0.237)	0.028 (69,0.168)	0.029 (68,0.170)	0.051 (272,0.221)
CEO turnover	0.176 (68,0.384)	0.217 (69,0.415)	0.043 (69,0.205)	0.147 (68,0.356)	0.145 (274,0.353)
Panel B : FINANCIAL SECTOR
Size of the board	13.047 (21,7.946)	13.454 (22,8.545)	13.000 (20,8.194)	13.500 (18,8.590)	13.246 (81,8.163)
Board turnover1	0.134 (21,0.103)	0.191 (22,0.251)	0.115 (20,0.111)	0.109 (18,0.096)	0.139 (81,0.159)
Executive director turnover2	0.356 (20,0.447)	0.669 (21,0.887)	0.222 (19,0.372)	0.422 (17,0.579)	0.423 (77,0.661)
Non-executive director turnover3	0.072 (20,0.085)	0.101 (21,0.113)	0.063 (19,0.094)	0.050 (17,0.072)	0.072 (77,0.093)
Chairman turnover4	0.200 (20,0.410)	0.095 (21,0.300)	0.000 (19,0.000)	0.235 (17,0.439	0.129 (77,0.338)
CEO turnover	0.380 (21,0.497)	0.409 (22,0.503)	0.100 (20,0.307)	0.222 (18,0.427)	0.283 (81,0.453)
Panel C : INDUSTRIAL AND COMMERCIAL SECTORS
Size of the board	9.779 (86,6.401)	9.476 (84,6.194)	9.619 (84,5.692)	9.414 (82,5.724)	9.574 (336,5.989)
Board turnover1	0.118 (86,0.168)	0.087 (84,0.145)	0.129 (84,0.185)	0.098 (82,0.160)	0.108 (336,0.165)
Executive director turnover2	0.277 (86,0.552)	0.351 (84,1.060)	0.297 (84,0.697)	0.196 (82,0.503)	0.281 (336,0.737)
Non-executive director turnover3	0.096 (86,0.173)	0.065 (84,0.133)	0.090 (84,0.174)	0.075 (82,0.160)	0.082 (336,0.160)
Chairman turnover4	0.141 (85,0.350)	0.120 (84,0.327)	0.108 (83,0.312)	0.097 (82,0.298)	0.117 (333,0.322)
CEO turnover	0.223 (85,0.419)	0.166 (84,0.374)	0.190 (84,0.395)	0.146 (82,0.355)	0.182 (335,0.386)

Sources : Own calculations based on annual reports, the database of the National Bank of Belgium.

All turnover data are corrected for 'natural turnover', turnover resulting from retirement, illness or death.

1. Proportional to board size.

2. Proportional to total number of executives on the board.

3. Proportional to total number of non-executives on the board.

4. Non-executive chairman turnover only.

In parentheses : number of sample companies and standard deviation.

	1989	1990	1991	1992	average 1989-1992
PANEL A : HOLDING COMPANIES
Size of management committee	2.691 (68,2.234)	2.797 (69,2.392)	2.855 (69,2.396)	2.823 (68,2.374)	2.791 (274,2.338)
Management committee turnover1	0.117 (66,0.288)	0.143 (67,0.312)	0.058 (67,0.261)	0.100 (66,0.271)	0.105 (266,0.284)
Number of directors member of mgt committee	2.191 (68,1.870)	2.333 (69,2.048)	2.376 (69,2.015)	2.279 (68,1.819)	2.295 (274,1.932)
PANEL B : FINANCIAL SECTOR
Size of management committee	3.952 (21,3.368)	3.590 (22,3.142)	3.650 (20,3.183)	4.222 (18,3.797)	3.839 (81,3.310)
Management committee turnover1	0.270 (21,0.382)	0.354 (22,0.413)	0.075 (20,0.115)	0.128 (18,0.319)	0.213 (81,0.344)
Number of directors member of mgt committee	3.190 (21,2.522)	2.909 (22,2.136)	2.900 (20,2.125)	3.222 (18,2.414)	3.049 (81,2.263)
PANEL C : INDUSTRIAL AND COMMERCIAL SECTOR
Size of management committee	4.151 (86,3.936)	3.952 (84,3.596)	4.071 (84,3.579)	4.121 (82,3.546)	4.074 (336,3.654)
Management committee turnover1	0.134 (85,0.339)	0.177 (83,0.777)	0.110 (83,0.229)	0.113 (81,0.303)	0.134 (332,0.463)
Number of directors member of mgt committee	2.593 (86,2.271)	2.404 (84,2.129)	2.369 (84,1.943)	2.378 (82,1.928)	2.437 (336,2.068)

Sources : Own calculations, based on annual reports, the database of the National Bank of Belgium, Memento der Effecten.

Turnover data are corrected for 'natural turnover', turnover resulting from retirement, illness or death.

1. Turnover data are proportional to total committee size and exclude CEO turnover.

These data are for all sample companies (or 98% of the Brussels Stock Exchange). Number of companies in 1989 is 177, and 168 in 1992. The companies in liquidation were deleted.

number of directorships held by a director	1989		1992
	number of directors	number of positions	number of directors	number of positions
1	869	869	864	864
2	136	272	137	274
3	51	153	62	186
4	17	68	18	72
5	11	55	11	55
6	8	48	4	24
7	6	42	4	28
8	3	24	4	32
9	1	9	2	18
10	3	30	1	10
11	0	0	0	0
12	1	12	2	24
total number of positions	1106	1582	1109	1587
average number of directorships per director		1.4		1.4

Source : Own calculations based on information from a database of the National Bank.

CHAPTER 6 : Corporate control in Belgium : Empirical results.

In this chapter, we test the corporate control hypotheses advanced in Chapter 4. We will discuss whether disciplinary actions against management are taken when corporate performance is poor. The impact of the composition of the board of directors and of the presence of specific shareholder classes on turnover of management is analyzed. We also investigate whether a market for share stakes results from poor performance and whether this market is related to the exercise of corporate control. Finally, we examine the performance after the management restructuring.

6.1 Corporate performance and the disciplining of management.

Hypothesis 1 states that poor share price performance and low accounting earnings trigger disciplinary actions against (i) the executive directors, (ii) the CEO and (iii) the management committee. The three following sections will focus on one of these turnover variables.

6.1.1 Corporate performance and turnover of the board of directors.

Share price performance.

Share prices reflect the current profitability of the firm and expected future opportunities including the expected managerial performance and the consequences of a potential substitution of top management in case of underperformance. Table 6.1 presents the relation between board turnover proportional to board size and corrected for natural turnover, and share price returns. In the regressions, executed on pooled data over the period 1989-1992, we correct for size, proxied by the logarithms of total assets or of the market capitalization. A dummy variable referring to each year of the period 1989-1992 is also included.

The annual market adjusted share price data are averaged for periods of one to ten years preceding the year of board turnover. As such we investigate not only short but also long term share price underperformance because, in some companies delays in replacing an underperforming management team might occur if executives can entrench themselves by dominating the board (Molz 1988, Mizruchi 1983). Furthermore, excellent past performance records might insulate executive directors with long tenure against the threat of disciplining following short term poor performance. Consequently, managerial disciplining might be slow and will only take place after prolonged poor profitability. Another explanation is that the whole managerial record of executives with long tenure is evaluated so that short term low profitability does not immediately affect their position.

Panel A of table 6.1 shows, that both long term underperformance (market adjusted return over a period of 5 or 10 years before turnover) and short term underperformance (one year before turnover) are negatively correlated to board turnover (at 5% statistical significance). Board turnover includes both turnover of executive directors, who assume direct responsibility for the firm's profitability, and of the non-executives, who represent the large shareholders and are accountable to them for their monitoring accomplishments. An investigation of the relation between executive director turnover and past stock performance yields a strong negative relation, while the replacement of non-executive directors is not preceded by low share price returns. This confirms hypothesis 1 : the lower short and long-term returns, the higher the subsequent executive board turnover.

An analysis of the subsamples reveals that, in industrial and commercial companies, poor stock price performance for two to five years leads to a high proportion of the directors leaving the company (panel C). Turnover in financial firms is preceded by long term poor share performance (over a 10 year period) only. In holding companies, the replacement of directors does not seem sensitive to share price returns.

Accounting earnings.

In table 6.1, we also analyze the importance of several accounting-based profitability measures. Accounting measures can portray a biased picture of the company's profitability since they can be temporarily manipulated by management. Outgoing managers may have an incentive to increase reported earnings in order to safeguard their position. Similarly, incoming managers may be tempted to reduce reported earnings immediately following taking office so that predecessors might be blamed for poor performance. Moreover, a reduction in reported earnings creates more potential for impressive performance improvements for which incoming management can take credit. Therefore, we do not consider the accounting earnings in the year of management changes, but over periods of one and two years preceding turnover .

We use three profitability benchmarks. The first measure is operating income which is defined as earnings before financial and extraordinary results and taxes (EBIT/TA), as used by Weisbach (1988). The advantage of this measure is that it reflects a clearer picture of the true profitability as it are not sensitive to financing policy, tax regime, windfall profits or extraordinary losses. The use of operating income rather than net earnings after tax reduces the impact of the described 'earnings management' (Dennis and Dennis 1994). The second and third accounting yardsticks are respectively earnings after financial but before extraordinary results and taxes (EBT), and earnings after extraordinary results and taxes (EAT). All accounting measures are standardized by total assets (TA) or equity. In our regression and logistic models, both absolute income levels and changes over the current year and over one and two years preceding the year of management turnover are utilized.

We find that the parameter estimates of the accounting profitability measures over the current year and the changes in profitability over periods of one and two years preceding the year of turnover are all negative (not shown). This implies that the lower the operating income or the lower EBT/TA or EAT/TA, the higher the board turnover. However, these relations are not statistically significant within the 5% level for all sample companies and for the subsamples of industrial companies and of financial firms. In holding companies, the replacement of directors is sensitive to a relative earnings benchmark: we find statistically significant negative parameter estimates for changes in earnings before and after tax (EBIT/TA and EAT/TA) over a one or two year period before the year of management replacement.

We conclude that we only find weak evidence that the levels of and changes in accounting earnings precede the replacement of (executive) directors. Consequently, hypothesis 1 is only weakly supported.

10

TURN stands for the turnover of all members of the board proportional to board size. SIZE represents the logarithm of the total assets. PERFORM stands for performance variables : share price, accounting returns and changes in dividends per share. OLS regressions are estimated. The market adjusted returns are calculated over a periods of 1, 2, 3, 5 and 10 years before the year of turnover. T stand for the year of turnover in the period 1989-1992. (T-1) and (T-2) represent respectively 1 and 2 years before the year of turnover. EBIT/TA : earnings before financial and extraordinary results and taxes / total assets, EBT/TA : earnings before extraordinary results and taxes / total assets, EAT/TA : earnings after taxes / total assets. EBIT/TA (T-1,T), EBT/TA (T-1,T), EAT/TA (T-1,T), EBIT/TA (T-2,T-1), EBT/TA (T-2,T-1) and EAT/TA (T-2,T-1) are dummy variables indicating whether the respective earnings were negative (dummy equals 1) in at least one of the years of period (T-1,T) or (T-2,T-1). DVD/SH (T-1,T) is a dummy variable indicating whether there was a reduction in dividends per share of 25% or more over the period (T-1,T) or whether dividends remained at zero over this period (dummy equals 1).

	MARKET ADJUSTED RETURNS					OPERATING INCOME		EARN. BEFORE TAX		EARN. AFTER TAX		DIV/SH
	1 year period	2 year period	3 year period	5 year period	10 year period	EBIT/TA (T-1,T)	EBIT/TA (T-2,T-1)	EBT/TA (T-1,T)	EBT/TA (T-2,T-1)	EAT/TA (T-1,T)	EAT/TA (T-2,T-1)	DIV/SH (T-1,T)
PANEL A : ALL SAMPLE CO'S
sample size	622	612	598	563	533	594	580	606	592	653	642	567
betacoeff.	-0.020	-0.006	-0.005	-0.005	-0.001	0.019	0.017	0.064	0.060	0.057	0.057	0.043
t-stat.	-2.095	-1.213	-1.258	-1.890	-1.948	1.444	1.284	4.269	3.624	3.938	3.586	3.475
p-value	0.03	0.22	0.20	0.05	0.05	0.14	0.19	0.00	0.00	0.00	0.00	0.00
R-sq. adj.	0.05	0.01	0.01	0.01	0.01	0.01	0.01	0.03	0.02	0.02	0.02	0.02
PANEL B : ALL HOLDING COMPANIES
sample size	259	255	252	249	240	228	220	240	232	250	244	223
betacoeff.	-0.014	-0.001	-0.001	-0.001	-0.001	0.001	-0.021	0.072	0.064	0.068	0.060	0.052
t-stat.	-1.249	-0.142	-0.161	-0.390	-0.838	0.056	-0.912	3.553	2.735	3.326	2.597	0.789
p-value	0.21	0.88	0.87	0.69	0.40	0.95	0.36	0.00	0.01	0.00	0.01	0.01
R-sq. adj.	0.01	0.00	0.00	0.00	0.00	0.00	0.00	0.05	0.03	0.04	0.03	0.03
PANEL C : INDUSTRIAL AND COMMERCIAL COMPANIES
sample size	283	278	272	251	233	330	325	330	325	330	325	277
betacoeff.	-0.031	-0.021	-0.018	-0.010	-0.001	0.045	0.052	0.067	0.066	0.059	0.066	0.053
t-stat.	-1.512	-1.706	-1.779	-1.886	-1.265	2.465	2.779	3.271	2.948	2.868	2.948	2.936
p-value	0.13	0.08	0.07	0.06	0.20	0.01	0.01	0.00	0.00	0.00	0.00	0.00
R-sq. adj.	0.01	0.02	0.02	0.03	0.01	0.02	0.02	0.03	0.03	0.02	0.02	0.03

Source : Own calculations based on data of the National Bank and annual reports.

Industry corrections.

The relation between board turnover and companies' performance relative to the performance of their industry was also examined. The average EBIT/TA, EBT/TA and EAT/TA was calculated for each industrial sector and subtracted from the sample companies' respective profitability measures. Board turnover increases when earnings levels, corrected for industry and size, are low and when adjusted changes in earnings levels are negative. However, the relation is not statistically significant.

Critical performance measures.

As the replacement of management is a rather radical governance action, it is expected to arise only after considerable corporate underperformance. Warner, Watts and Wruck (1988), for instance, confirm that unless performance is extremely good or bad, their logit models for U.S. CEO turnover have no predictive value. Therefore, in our regression and logistic models, a dummy variable was included which equals 1 when operating income is negative in at least one of the years in a two year period before replacement of executive directors. Similar dummy variables are constructed for negative EBT/TA and EAT/TA.

Table 6.1 exhibits the parameter estimates and significance levels of the negative earnings dummy variables. Since this variable equals 1 when earnings are negative, a positive sign is expected for hypothesis 1 to be supported. Panel A discloses that negative operating income is not followed by increased board turnover. However, negative earnings before tax or after tax (EBT and EAT) is strongly correlated to high subsequent board turnover (0.1 % significance level).

Consequently, we conclude that hypothesis 1 is strongly supported by our evidence: the inability to generate earnings is followed by executive board turnover. An analysis on the subsamples shows that this relation is corroborated for the holding companies (panel B) and the industrial and commercial companies (panel C), but not for the financial sector.

Dividend changes.

Another critical performance measure is substantial changes in dividends. Management will usually postpone a reduction in dividends per share as long as possible because such a reduction emits a strong signal of poor performance. We define a considerable reduction of dividends per share as a decrease of more than 25 percent. Table 6.1 shows that a substantial dividend cut in the same year as the director replacement is positively correlated by high director turnover (panel A) at a significance level of 1%. This relation can also be observed for holding companies (panel B) and industrial and commercial companies (panel C).

The parameter estimate of size variable (not shown) is in most regressions of this section significant at the 5% level and has a negative sign implying that there is more resistance to turnover in larger companies than in smaller ones.

We conclude that hypothesis 1 is corroborated when critical accounting measures are reached. When management faces negative earnings or has to reduce dividends substantially, there is a high probability that it will be replaced. In comparison, German and U.K. board turnover is also sensitive to poor performance measured by the incidence of earnings losses and abnormal returns (see Franks and Mayer (1995b) and Kaplan (1994a) for Germany and Chapter 2 for the U.K.).

6.1.2 Poor company performance and CEO replacement.

While previous section concentrated on executive director turnover, table 6.2 examines the relation between the replacement of the CEO and corporate performance. Identical profitability yardsticks are employed. The results of the logit models indicate that poor share price performance over both a short and over longer period (2 to 10 years) is associated with a high probability of CEO turnover (panel A). However, this relation is only valid for industrial and commercial companies (panel C). It seems that the position of the delegated director of a holding company or a financial firm is relatively insensitive to stock price performance.

When the levels of earnings and cash flows are low, or earnings, cash flows and dividends decrease, CEO and executive chairman substitution increases, but the relation is not statistically significant. Similar results are found for industry corrected performance variables.

From table 6.2 can also be deduced that CEO (and executive chairman) turnover is significantly related to critically low profitability measures, namely to negative earnings before tax (EBT) and negative earnings after tax (EAT). Panel A shows that the absolute benchmark of negative earnings (over a period of two years before CEO turnover) causes CEOs to be disciplined. Analysis of the subsamples of holding companies shows that the result is valid for both the subsamples of holding companies (panel B) and industrial firms (panel C). Substantial decreases in dividends in the year before the replacement of the CEO provide support for hypothesis 1 as well, but only for industrial and commercial companies (panel C).

In each of the logistic regressions a size factor in the form of the logarithm of the book value of total assets or the logarithm of market capitalization was included. This statistically significant size variable indicated that, in contrast to the size factor in the board turnover regressions, that the probability of CEO turnover following poor performance increases for larger companies. Reasons might be that larger companies have a larger internal managerial labour market which might create greater internal pressure and instigate political struggles for the top positions as Harrison, Torres and Kukalis (1988) suggest. Another reason is that large firms can attract more outside job offers because of their high visibility and CEO turnover might be less disruptive in large organizations because they are more formalized and decentralized (Puffer and Weintrop, 1991).

We conclude that hypothesis 1 is strongly supported : the CEO of an industrial company is disciplined when share price performance is low, when earnings are negative and when dividends are cut considerably. The probability that delegated directors of holding companies depart is significantly correlated to negative earnings.

CEO turnover has been researched extensively in the U.S. Most of those results are in line with the findings of this paper. Jensen and Murphy (1990) measure the strength of the performance-turnover relation and also use net-of-market share price returns over a period of two years. They find that CEO turnover increases when performance is poor. Weisbach (1988) includes in his logit models the current and one year lagged stock returns and accounting earnings and also finds that the lower the company's profitability, the higher CEO turnover. Relative measures of performance are utilized by Gibbons and Murphy (1990). They include market average stock returns and/or industry average stock returns together with the firm's own stock return. Since the industry return has a significant and positive parameter estimate, they conclude that companies take other firms' performance into account in evaluating their CEOs. Harrison, Torres and Kukalis (1988) employ return on assets, return on equity, profit margin and dividend yield of the current year as measure of performance. Only return on assets has a significant effect on CEO turnover. The sum of daily abnormal returns over the firm's fiscal year is used by Coughlan and Schmidt (1985) and they also conclude that low performance leads to increased turnover.

Longer term performance measures are used by Warner, Watts and Wruck (1988), but they do not find a consistent pattern in the parameter estimates of three year lagged stock returns. Yungsan (1993) uses both short and long term performance measures: the average of current and one year lagged stock returns, changes in EBIT/TA, and a stock return measure that incorporates the current year stock return and up to nine years of lagged stock returns whereby the more recent stock returns are given more weight. He finds that both short and long term performance measures are important in predicting CEO turnover.

11

TURN stands for CEO turnover. SIZE represents the logarithm of the total assets. PERFORM stands for performance variables : share price, accounting returns and changes in dividends per share. OLS regressions are estimated. The market adjusted returns are calculated over a periods of 1, 2, 3, 5 and 10 years before the year of turnover (T). T stand for the year of turnover in the period 1989-1992. (T-1) and (T-2) represent respectively 1 and 2 years before the year of turnover. EBIT/TA : earnings before financial and extraordinary results and taxes / total assets, EBT/TA : earnings before extraordinary results and taxes / total assets, EAT/TA : earnings after taxes / total assets. EBIT/TA (T-1,T), EBT/TA (T-1,T), EAT/TA (T-1,T), EBIT/TA (T-2,T-1), EBT/TA (T-2,T-1) and EAT/TA (T-2,T-1) are dummy variables indicating whether the respective earnings were negative (dummy equals 1) in at least one of the years of period (T-1,T) or (T-2,T-1).

DVD/SH (T-1,T) is dummy variable indicating whether there was a reduction in dividends per share of 25% or more over the period (T-1,T) or whether dividends remained at zero over this period (dummy equals 1).

Indep var.->	MARKET ADJUSTED RETURNS					OPERATING INCOME		EARN. BEFORE TAX		EARN. AFTER TAX		DIV/SH
	1 year period	2 year period	3 year period	5 year period	10 year period	EBIT/TA (T-1,T)	EBIT/TA (T-2,T-1)	EBT/TA (T-1,T)	EBT/TA (T-2,T-1)	EAT/TA (T-1,T)	EAT/TA (T-2,T-1)	DIV/SH (T-1,T)
PANEL A : ALL SAMPLE CO'S
sample size	623	613	599	564	534	590	574	602	586	647	633	561
-2 log L	0.18	0.02	0.01	0.05	0.05	0.01	0.01	0.00	0.00	0.00	0.00	0.00
param. est.	-0.333	-0.359	-0.301	-0.122	-0.032	0.021	0.054	0.382	0.554	0.395	0.534	0.382
Wald chi-sq	1.360	3.504	4.483	2.998	2.918	0.010	0.071	2.821	5.390	2.964	5.066	2.919
p-value	0.24	0.06	0.03	0.08	0.08	0.91	0.79	0.09	0.02	0.08	0.02	0.08
PANEL B : ALL HOLDING COMPANIES
sample size	260	256	253	250	241	228	219	240	231	250	243	222
-2 log L	0.26	0.42	0.33	0.36	0.38	0.10	0.14	0.13	0.02	0.04	0.01	0.05
param. est.	-0.177	-0.158	-0.158	-0.088	-0.025	0.611	0.407	0.265	0.861	0.468	0.833	0.444
Wald chi-sq	0.199	0.411	0.622	0.648	0.636	1.330	0.74	0.415	4.145	1.322	4.081	1.435
p-value	0.65	0.52	0.43	0.42	0.42	0.24	0.38	0.51	0.04	0.25	0.04	0.23
PANEL C : INDUSTRIAL AND COMMERCIAL COMPANIES
sample size	284	279	273	252	234	328	323	328	323	328	323	276
-2 log L	0.24	0.03	0.01	0.11	0.18	0.13	0.08	0.05	0.04	0.06	0.03	0.07
param. est.	-0.440	-0.575	-0.493	-0.160	-0.030	0.187	0.354	0.482	0.569	0.473	0.591	0.499
Wald chi-sq	1.197	3.779	5.354	2.149	1.476	0.502	1.805	2.741	3.478	2.673	3.731	2.569
p-value	0.27	0.05	0.02	0.14	0.22	0.47	0.17	0.09	0.06	0.10	0.05	0.10