Shareholdings Structure and Separation between Ownership and Control in Belgium

Transcription

1 Chapter 2: Shareholdings Structure and Separation between Ownership and Control in Belgium 1. Introduction The following chapter tends to describe, in a comprehensive manner, the full ownership structure of the 140 Belgian listed companies by the end of December 1995, as well as the portfolio of their shareholdings. The paper details the methodology used, based on the contribution of Brioschi et al. (1989, 1997) and the database built to this end. As far as we now, this has never been done on that scale. Available data on ownership were dispersed and fragmented. Summary descriptive statistics are followed by international comparisons between European countries. Finally, we develop our own methodology to estimate the extent of the separation between ownership and control among ultimate shareholders of listed companies, in Belgium especially, but also abroad. This paper is part of a much wider project initiated by the European Community aimed at defining the characteristics of the Corporate Governance systems in the European countries and, in particular the ownership of listed - and non listed if possible - companies in these countries. This project gave birth to the ECGN (European Corporate Governance Network), composed of several researchers in each of the European countries. The standardised methods used by the Network allow us to present comparable data of ownership across countries. For Belgium, data coming from the Brussels Stock Exchange and from specific CD-Rom have been gathered and computed to build a comprehensive database. The database includes direct and indirect ownership of the 140 listed Belgian firms, the identity of the owners both in Belgium and abroad. It includes also the cross-shareholdings between companies, the own-shares held via indirect participations, and finally the direct and indirect shareholdings that listed firms themselves have in other listed and non-listed companies. The methodology uses an inverted matrix to compute pyramidal links and indirect ownership characteristics from extensive direct ownership data. The main point of interest of this work is made possible by the great precision of the database and will be detailed in section 6. It is to define - using some simplifying hypothesis - the level of separation between ownership and control in Belgium, for each large shareholder, for each listed company. The results can then be confronted to the situation in countries belonging to other Corporate Governance systems, like the United States or the Netherlands. The chapter structures as follows. Section 2 presents a synthetic review of the abundant literature on the Corporate Governance subject. Section 3 explains the methodology used to build the database. Section 4 displays the main descriptive figures of the ownership and the portfolio of shareholdings of the listed Belgian companies. Section 5 presents some international comparisons of ownership data in European countries. Section 6 estimates the magnitude of separation between ownership and control in Belgium. Section 7 applies the same method to estimate the separation between ownership and control in other countries. Section 8 concludes

2 2. Review of the Literature 2.1. Theory on Corporate Governance The first to have focused on the question of Corporate Governance and, in particular, on the question of separation between ownership and control are Berle and Means (1932). Observing the evolution of US Corporations in the industrial development preceding the thirties in the United States, Berle and Means predicted a great success to the Modern Corporation. The Modern Corporation grows and concentrates economic power. Growth is only possible with ownership dispersion. It results in separation between ownership and control, since wealth is becoming unnecessary to exercise control. To illustrate their point, Berle and Means calculated that the 200 largest corporations in the US controlled almost half (49.2%) of the total corporate wealth. And they choose these 200 corporations as a sample for further investigations on the separation between ownership and control. They divided the 200 corporations between ownership categories 27. The authors provided this way five control definitions that are still valid today: 1. Private Ownership: 80% of voting stock owned by a compact group or individual (12 companies) 2. Majority Control: public interest larger than 20% but less than 50% (10 companies) 3. Control Through a Legal Device: considerable separation of ownership and voting power through issuing non-voting stock and/or pyramiding and/or voting trust (21 companies) 4. Minority Control: single block between 20% and 50% but sometimes smaller (44 companies) 5. Management Control: not identified as minority control (residual category), but none of the known blocks larger than 5% (21 companies) Doubtful, presumed minority (29 companies) Doubtful, presumed management ( 44 companies) Joint Minority-Management Control: several blocks between 5% and 50% (16 companies) Leech (1987) re-examines the evidence used by Berle and Means, contesting the ownership level of 20% as a threshold to avoid management control. With such a threshold, Berle and Means (1932) end to the conclusion that 44 percent of the largest US corporations were management controlled ( residual cases). Leech judges the proportion far too high compared to the reality. To illustrate his point, the author uses a probabilistic model of random voting by shareholders. The model is adapted to British data of 1970/1971. Leech shows that, given the highly dispersed ownership of British corporations, 6 to 15% of the votes are enough to control the firm without the influence of the management, showing thus that the limit of 20% set by Berle and Means was too restrictive. Among the various possible ways to approach the question of corporate control, the most popular one is the Principal - Agent approach. The Principal - Agent theory has first been defined by Jensen and Meckling (1976) and finds multiples applications. For corporate governance questions, it is the theoretical approach used by Shleifer and Vishny. Shleifer and Vishny (1995) have probably made the most relevant review of the theoretical literature on the corporate governance. In their opinion, the essential problem of corporate 27 Berle and Means (1932), p

3 governance is to find the ways by which investors can get return on their investments from the managers. The source of the problem is the agency costs: since the shareholder (principal) and the manager (agent) cannot sign complete contracts, there remain some residual control rights on the firm. The distribution of those rights among the different parties will significantly influence the balance of powers. Since investors bring funds to the managers, they could ask for the residual control rights. However, two facts make the situation hard to realise: (1) the contracts signed between managers and shareholders are subject to too broad interpretation to be enforced by outside courts, (2) investors are often too small and too poorly informed to face alone the costs of monitoring (free-rider problem). As a result, managers often end up with the totality of residual control rights, which gives them in practice much more power than they would normally be entitled to have. These extensive powers may give managers incentives to extract some private benefits from their control situation, acting more in their own interest than in the interest of the firm and of the shareholders. This is the core of the agency problem. Many empirical studies have confirmed the existence of agency costs in practice (Shleifer and Vishny (1995), Morck, Shleifer and Vishny (1990), among others). There is, however, two ways to mitigate this problem. The first one is legal protection: the legal system might place some restrictions on manager s self-dealing, including excessive compensation, or issue of additional securities to the management and its relatives. Next to this type of solution, there is a whole part of the literature dedicated to the analysis of incentive contracts that can be designed between owners and managers. Such contracts are aimed to give a manager long term incentive ex ante to align his interests with those of investors. The second alternative to the agency problem is concentrated ownership. Indeed, large shareholdings mitigate the free-rider problem since the investor has reached the critical size such that its monitoring costs are outweighed by the benefits of value increase of the firm. Concentrated ownership can be achieved in three ways: (1) large shareholdings, (2) takeovers, and (3) large creditors. One can thus foresee that concentrated ownership will naturally lead to increased profits of the firm compared to dispersed ownership. Concentrated ownership seems to have been a universal response to agency problem (Shleifer and Vishny (1986), Franks and Mayer (1994)). However, the relation between ownership concentration and performance is likely not to be linear. In fact, when a shareholder gets too large, he may tend to expropriate minority shareholders, leading then to a decrease in the firm value. Expropriation can take the form of special dividends, superior voting rights, exploitation of other business relationship with other controlled companies, among any other examples. Theoretically, one should then expect a non-linear relation between concentrated ownership and performance, with an upward slope at the beginning and a downward slope once concentration reach some critical level. Taking a slightly different perspective, Brukart et al. (1995) develop a model where the ownership structure of a firm acts as a commitment device to delegate some degree of authority from the shareholders to the management. Dispersed ownership supposedly leaves more initiative power to the management while concentrated ownership induces close monitoring. The optimal trade-off between the gains from monitoring and those from managerial activism will depend on the congruence of interests of both parties. The optimal

4 trade-off will also vary with the life cycle of the firm, or with its decision to go public or to remain private. The authors show that the ratio between short term profits and long term profits is a measure of congruence between managers and shareholders. If this ratio is low, congruence is low, giving room for conflict of interest. Concentrated ownership is then advisable. Similarly, concentrated ownership is advisable for younger firms in which monitoring is essential, while dispersed ownership is preferable for more mature activities in which managerial initiatives are beneficial. From a completely different point of view, Zwiebel (1995) examines the dynamics of coalitions of shareholders. The author elaborates a model of co-operative game where small shareholders join together and form controlling coalitions, conferring to their members partial benefits of control. The model has three implications : (1) the largest shareholders tend to create their own space, their presence dissuade other large shareholders from investing in the firm, (2) in the same spirit, there is a threshold above which a large shareholder in not challenged within the firm, (3) the shareholder structure in corporations will exhibit a clientele effect among block investors : the larger the first investor, the smaller the other shareholders will be. Empirical data for the US reasonably match the theory. This contribution and its application to Belgium will be extensively treated in chapter 3. The main corporate governance issue can then be summarised by an agency problem between shareholders and managers. Empirically, ownership concentration seems to be the privileged answer to agency costs. Theory suspects a non-linear relation between ownership concentration and firm performance. The choice of ownership structure has to solve the trade-off between monitoring costs and managerial initiative and the answers may differ widely according to the firms characteristics. Sections 5 will present the main figures of ownership concentration in various European countries Empirical Studies on Corporate Control Among empirical studies on corporate control, only a few concern continental European countries. Several were initiated by J. Franks and C. Mayer. They examine the empirical relation between ownership structure and board turnover. They apply to the following countries: Germany (Franks and Mayer, 1995), UK (Franks, Mayer and Renneboog, 1996) and Belgium (Renneboog, 1996). Franks and Mayer (1995) report four types of corporate control in Germany: shareholder stakes, complex pattern of ownership, bank control in widely held companies, and a market in sales of share stakes. They find that the first three forms tend to be dominated by private benefits of the various stakeholders rather than disciplining the management. Only sales of share stakes are directly related to board turnover and performance. The authors suggest that where there is poor performance, large shareholders often exercise corporate control by selling their stakes. Franks, Mayer and Renneboog (1996), using the same methodology, apply the previous study to the cases of UK and Belgium. For UK, they find little relation between hostile takeovers and poor performance. On the other hand, the authors find a strong relation between the composition of the board, share ownership, and the exercise of corporate control. Indeed, like in Germany, there is a significant positive correlation between board turnover and ownership concentration for poorly performing firms. Moreover, the correlation is stronger in case of large outside shareholders. Like in Germany again, the trade of large stakes is

5 another way to exercise control in poorly performing firms in the United Kingdom. One can thus note that, both for German and for UK, take-overs are not the prevalent method for corporate control. For Belgium, Renneboog (1996) finds rather similar results as for the other countries. When performance is low, ownership concentration is positively correlated to board turnover. Moreover, management is more frequently disciplined in firms with a high non-executive board representation. Amongst shareholders, institutional shareholders are the least active in sanctioning the management, whereas holding companies are the most active, families and industrial companies are also active investors and controllers. Correlation is found for ultimate controlling shareholders but not for direct controlling shareholders. Like in the other countries studied, a market for stakes arises when performance in poor. Speaking of take-overs, for the Netherlands, Kabir, Cantrijn and Jeuninck (1997) examine the relationship between a firm s take-over defences and its ownership structure for Dutch listed companies. The authors find that multiple firms take-over defences is increasingly adopted as ownership gets more dispersed. Ninety percent of Dutch listed companies have at least one take-over defence. They state that Managers in the Netherlands seem to be immune from the disciplinary threat of the market for corporate control. However, there is a lack of evidence about the price reaction to the adoption of take-over defence measure and the question remains on the effect of concentrated ownership and corporate performance. Among the impacts of concentrated ownership for the US, Demstez and Lehn (1985) found a negative correlation between the firm s size and ownership concentration. The authors interpret their results, as the higher price of a given fraction of the firm should reduce the degree to which ownership is concentrated. Another result is a positive correlation between ownership concentration and the instability of the firm s profit. This second result is explained by the fact that the noisier the firm s environment, the greater the payoff to owners in maintaining tighter control. Hence, noisier environment may favour concentrated ownership. But the authors did not find any significant linear relation between performance and ownership concentration. However, their study did not explore the eventuality of a nonlinear relationship between ownership structure and performance. Talking about international comparisons, one should mention the contribution of La Porta et al. (1996). To compare laws across countries, the authors have assembled a data set covering governance-related legal rules and their enforcement in 49 countries that have publicly traded companies. They found the most striking differences between common law countries (like US, UK) and French Civil Code countries (including Belgium). While common law countries give both shareholders and creditors strongest protections, the second group of countries are those who protect investors the least. Starting from this statement, the authors looked for some possible adaptations to the lack of investors protection in civil law countries compared to common law countries. Their most interesting result concerns ownership concentration. The authors found a strong negative correlation between concentration of ownership and the quality of legal protection of investors in a country. Everything is as if, when law does not protect them, shareholders have to be large and powerful to balance the power of the management and extract payments from them. Data thus confirm the idea that legal systems matter for corporate governance Studies Specific to Belgium

6 The closest paper related to ours is the descriptive study by Renneboog (1996) on the listed Belgian firms in 1989 and in The author presents ownership of listed firms per investor class: institutional, families, holding and companies, and foreign vs. domestic investors. He gives some specific examples of pyramidal ownership and shows how this type of structure can lead to the violation of the one share one vote rule, that is, the separation between ownership and control for ultimate shareholders. The author then synthesises the influence of pyramidal ownership by calculating both the direct stakes held in a firm by immediate shareholders, and the ultimate controlling stakes held through pyramids by the different classes of investors. Ultimate control arises for a continuous chain of holdings of at least 50% at every tier, or of 25% at least in the absence of any larger shareholder. Indirect control is significant essentially for holding companies and families. Although the work of Renneboog is an interesting attempt to describe and quantify the ownership structure of the Belgian Stock Exchange, our work is made in a different perspective and uses a more detailed database. We work with individual data (each firm, each shareholder is identified in the matrix) without grouping them by investor classes. Our objective is to use the most possible detailed description of ownership structure of Belgian listed firms, for both domestic and foreign investors. Working with individual data rather than summary data allows us to assess more precisely to degree of separation between ownership and control in Belgium. Moreover, our methodology using the Brioschi matrix 28 makes possible to identify the other structures common in the ownership of Belgian firms but that are not pyramidal, such as rings, cross-shareholdings, and own shares. To estimate the separation between ownership and control, Renneboog identifies by hand the 155 ultimate controlling shareholders of the 155 listed companies of his sample. In contrast, our methodology, using automatic computation, allows us to estimate separation on a much larger scale. We are able to identify all the controlling shareholders of the listed companies, but also of the non-listed companies of the matrix (several hundreds elements) and the ratio of separation between ownership and control yielding in each of them. Besides controlling owners, we can estimate separation for every shareholder, even the small ones. Besides the work of Renneboog (1995, 1996) very few publications exist on corporate governance specific to Belgium. We can mention however, the paper of C. Van Hulle (1991) that analyses take-overs involving a Belgian firm form 1970 to 1985, thus before the introduction of disclosure rules. She finds 37% on average of abnormal return for targets, due to a change in control and no abnormal return for the bidder. Since these figures are quite similar to those for US, UK, and for France, the author claims that targets are not worse off despite the lack of disclosure requirements. Moreover, she finds no confirmation of the Shleifer and Vishny s result (1986) of lower take-over premium for large shareholders, since blocking minorities have negotiating power in many cases. 28 See next section for details

7 3. Methodology 3.1. The Model The methodology used to compute direct and indirect ownership of the listed companies is based on the model developed by Baldone, Brioschi and Paleari (1997) improving on Brioschi et al. (1989). The models uses the input-output matrix methodology to determine the integrated (direct + indirect) ownership shares in a set of firms connected by crossshareholdings, on the basis of all the direct ownership links existing between those firms. The authors motivate their research by the importance of cross-shareholdings and ownership links between large and listed firms that characterises most countries in Continental Europe and in Japan: With the relevant exception of the UK and US, in most industrialised countries large firms and in particular listed firms are connected by ownership links or cross-shareholdings. Crossshareholdings are associated with the phenomenon of business groups* 29. ( ) In addition to ownership links within each single group also share interlocks among firms belonging to different groups do exist. Probably the main effect of share interlocks is to alter the control structure of the firm: interlocking shareholdings, in fact, appear to be important means for firms to insulate themselves from the external market for corporate control. ( ) Furthermore a new definition of ownership, integrated ownership is needed which takes into account both the direct ownership of an outside stockholder and the indirect ownership through the network of cross-shareholdings. 30 The integrated ownership of a firm is computed the following way: Let A= [a ij ] be the matrix of direct ownership where a ij is the percentage share of total equity shareholder i holds in firm j directly. What we want is to compute the matrix of integrated ownership Y = [y ij ] where y ij is the sum of percentage shares of total equity shareholder i holds in firm j directly, indirectly, and through cross-participations. Let D(Y) be a diagonal matrix in which the k-th element is y kk. D(Y) can be interpreted as the matrix of the own shares of each firm of the set. So we have : meaning : Y = A D(Y)A + YA (1) Integrated stake of i in k = Direct stake of i in k Reciprocal stake of k in i +Indirect stake of i in k One can show 31 that the solution to this equation is: Y = [D(I-A) -1 ] -1 A(I-A) -1 (2) 29 All words followed by an «*» in the text are defined in appendix 5 of the chapter. 30 Baldone, Brioschi, and Paleari, 1997, p3. 31 The proof is provided in Appendix

8 We used the Matlab software to compute Y in this paper. The following examples of A and Y in simple structures are aimed at showing how A changes into Y according to the original structure of direct ownership Examples of Matrixes A and Y Example One: Direct Ownership Only The economy is made of three ultimate shareholders. An ultimate shareholder is an investor that has no shareholder. The typical example is an individual, or the state. However, in practice, an ultimate shareholder can be a foreign firm or a foundation for which we cannot trace back the identity of the owners. Since the shareholders have only direct stakes, integrated ownership equals direct ownership and Y equals A % 33% 33% 4 A = Y Example Two: Direct and Indirect Ownership Consider an economy made of three ultimate shareholders (Box 1,2,3) one intermediate shareholder (Box 4) and one bottom firm (Box 5). A bottom firm is a firm that has no shareholdings. Here, since there are two indirect holdings, Y differs from A % 5 60% 40% 4 50% A Y

9 Example Three: Cross-Shareholdings in Rings There is no ultimate shareholder or bottom firm in this structure. Shareholdings are in ring meaning that in a group of companies, each one is at the same time owner and owned. We see here that A and Y differ completely. 1 10% 100% % A Y Y can be interpreted as follows: * the ring creates own shares for each firm that equates : 100% * 20% * 10% = 2% * and indirect stakes due to the ring are : 100% * 20% = 20% for firm 1 in firm 2; 20% * 10% = 2% for firm 3 in firm 1; 10% * 100% = 10% for firm 2 in firm 3. In this chapter, we will tend to evaluate the type of shareholdings structure that characterises the most the Belgian listed companies and their linked corporations. Is it the rings, the hierarchical pyramids, a network of cross-shareholdings or isolated firms? 3.3. Data Collection We use two main sources of ownership information for listed companies: one for direct ownership data, one for indirect ownership data. For direct ownership, data are found in the printouts of the on-line Database of the Brussels Stock Exchange (DB Part). It contains continuously updated records of all notifications by shareholders of listed firms published according to the Transparency Law ( Loi sur la transparence des participations importantes dans les sociétés cotées ) of March 2, This law is the transposition in National Law of the European Transparency Directive. It stipulates that each time a shareholder crosses, upwards or downwards, a multiple of five percent ownership in a listed company (5, 10, 15, 20%, etc.), he or she has to notify both the listed firm and the market authorities within 24 hours. The information is made public within 48 hours via the financial press and on DB Part. The first ownership declarations were made by December of 1989 when the law became effective. At that time, all shareholders owning more than 5% of the votes in a listed company had to fill a declaration form and send it both to the Market Authorities and to the listed company owned. For business groups*, the ultimate controlling agent in the group can make declarations. In such case the control structure has to be explained, preferably in the form of an organisational chart. Each member of the group has to make a separate declaration. In the case of voting pacts*, the same rules as for business groups* apply

10 The number of ownership notifications varies widely from year to year. The maximum registered was in 1991 (due to legal obligations) with 350 notifications, and the minimum was in 1995 with 140 notifications. On average, there are about 200 notifications registered per year. And it often happens that stakes below 5% are notified to the Market Authorities. There are three main reasons for that. First, some listed companies have lowered the notification threshold to 3% of the voting capital, as they are allowed to do so by law. Second, shareholders are free to notify as soon as they own shares, such that some owners declare very small stakes. And third, voting pacts exceeding 5% of the shares are to be notified and each member of the pact has to make a separate declaration. For indirect ownership data, our main source is the CD-ROM BNB. This is produced and distributed by Bureau van Dijk. It is put together from magnetic tapes that are supplied by the Central Bank. These tapes gather all the annual accounts that each Belgian firm has to provide once a year to the Central Bank (Centrale des Bilans). Smaller firms only provide abridged accounts (less detailed, the abridged accounts do not contain, for example, the name of the administrators), bigger firms have to provide full accounts. Due to various delays, information is not always up-to-date, but most accounts are on the CD-ROM within one year from their official publication. In the annual accounts, Belgian firms are required to mention the list of their shareholdings in other firms exceeding 10% of the voting capital. Therefore, with these published shareholdings portfolios, we are able to identify the firms that hold stakes in the direct shareholders of listed Belgian firms, to reconstitute the ownership pyramid. Here, although firms are required to publish shareholdings exceeding 10% only, most of them publish their whole portfolio, even with much smaller stakes. The drawback of this source, although very convenient, is that these indirect shareholders are only identified if they are Belgian firms. Individuals and shareholders of non-listed firms cannot be identified nor the foreign shareholders of non-listed firms. Imagine, for example, a listed firm L having 2 declared shareholders A and B, (published on DB Part) having each 30% of the votes. A is a Belgian company that is owned by C and D having each 25% of the votes. We can identify C and D because C and D have to publish, in their annual report, the list of their shareholdings exceeding 10% of the capital. C and D have thus each declared a shareholding of 25% in A. But, if B is owned by a foreign company, or by individuals, we are not able to identify its owners. Indeed, individuals do not publish annual accounts, and foreign companies may not be obliged to publish their shareholdings, and they are not included in the CD-Rom we use anyway Building up A According to the methodology described, we have to fill in the complete matrix of direct ownership links around listed Belgian firms as they were in December 31, It means, - Upward: the direct shareholders of every listed firms, and the shareholders of these shareholders, up to three levels of ownership and, - Downward: the direct shareholdings of every listed firm and the shareholdings of these shareholdings, down to three levels of shareholdings. The total number of firms and individuals included in such a set is This matrix A - filled in by hand - is thus a square matrix 1125 X 1125 were the value of each cell (a ij ) is the percentage of shares shareholder i owns in firm j

11 Ownership is defined here as the percentage of equity capital held in a firm and not as the percentage of votes. So we talk about ownership rights and not about control rights. In direct ownership links, discrepancies between ownership and control rights can come either from non-voting shares or from different voting rights attached to certain categories of shares. In Belgium however, according to the legal principle of shareholder equality, all shares must carry the same rights and they must have the same value. Each share gives right to one vote. Non-voting shares are admitted but only a few listed Belgian companies had issued non-voting stocks by the end of 1995 and those shares were in all cases held at 100% by a single owner 32. The difference between ownership and control rights induced by the existence of these non-voting shares is thus totally insignificant statistically. So, ownership rights are equivalent to control rights in this analysis. In the matrix, the firms owned are in columns and the shareholders are in lines. The matrix must be square to be inverted when computing Y, the matrix of integrated ownership. Therefore, each time we add a shareholder (a line) to A, we add its corresponding column. The construction of A is performed sequentially, adding step by step a category of shareholders and a category of shareholdings. We start by creating 140 column for the 140 listed Belgian firms. Then we add to the matrix the lines for: - the direct shareholders of every listed firms (578 lines), - the shareholders of these shareholders, up to the third level, in order to have two levels of indirect ownership of the listed companies ( lines). To leave the matrix square, we add the corresponding columns. Then, to include the portfolio of shareholdings of the listed companies in the analysis, we add columns for: - the direct shareholdings of the listed companies (207 columns), - the shareholdings of the firms in which listed companies have participations, to make the second and third level of shareholdings ( columns). To leave the matrix square, we add the corresponding lines. Finally, to take into account the rest of the environment of the Belgian listed companies, we fill the remaining of the matrix with the stakes that direct and indirect shareholders of listed firms hold in all companies included in the matrix. To make things clearer, we illustrate the procedure by the real example of the organisational chart of the Générale de Banque (GBK), first bank in the country. Direct Shareholders of Listed Firms: We add a line to the matrix each time we find another shareholder to a Belgian listed company. If the shareholder identified is already included in the matrix, we just fill in the corresponding cell with the percentage of shares he owns in the listed company. The different types of shareholders are classified in the following categories: Listed Belgian firms Non listed Belgian firms Individuals 32 Mercantile Beliard : 100% non-voting shares held by GIMV ; Clabecq : 100% non-voting shares held by SWS; La Hersautoise : 100% non-voting shares held by SOWAGEP. Soucre : DBPart - Brussels Stock Exchange

12 Foreign companies Matrix A counts 578 direct shareholders (other than the listed Belgian companies) among which, 299 non listed Belgian firms, 77 individuals, and 150 foreign companies. And 56 listed Belgian firms are shareholders of other listed Belgian companies. In the case of GBK, we have: GBK GBK Listed firm FIN (Finoutremer) 0.4% Listed Firm SGB (Générale de Belgique) 27.2% Listed Firm Bancinvest (Fortis group) 1.6% Non Listed Firm Mutuelle Solvay 2.9% Non Listed Firm Sika (SGB group) 9.0 % Non Listed Firm Second and Third Layers of Shareholders of Listed Firms: We fill in here the shareholders of the shareholders identified on the last step. For the GBK case, we add the shareholders of Finoutremer and Générale de Belgique (SGB). These are SGB, BBLfonds for Finoutremer, and Cofiblan, Frabepar, AG 1824 and Suez for Générale de Belgique (SGB). There is no shareholder registered for Bancinvest and Mutuelle Solvay. Sika is controlled at 54.5% by the listed company Union Minière. The next step would be to include the shareholders of these indirect owners, it is done in the database but it makes the matrix too large to be included here. Its gives: FIN GBK SGB Sika GBK FIN (Finoutremer) 0.4% Listed firms SGB (Générale de 55.2% 27.2% Belgique) UM (Union Minière) AG 1824 (Fortis group) 10.7% Bancinvest (Fortis group) 1.6% Non listed firms BBL fonds 1.5% Cofiblan (Suez group) 3.8% Frabepar (Suez group) 9.0% Mutuelle Solvay 2.9% Sika (SGB group) 1.5% 9.0 % 54.5% Foreign firms Suez (France) 61.0% The second layer of shareholders in the complete matrix A counts only 42 new non listed Belgian firms; other indirect shareholders are firms that were already put in the matrix as direct shareholders of one or several listed Belgian firms. And perhaps more significantly,

13 the third layer of shareholders of listed Belgian firms counts only 10 new non-listed Belgian firms, others are either individuals, or foreign companies, or other investors already in the matrix. These figures are a strong first indication that most ownership pyramid* of listed firms in Belgium are organised in network*: there is a limited group of companies owning one another at various levels. This network type of organisation reflects in some sense the culture of a small country. There must be also a large presence of individuals at low levels of the pyramid and/or foreign companies, although it is not very likely that foreign investors hold stakes in small non-listed Belgian firms. Statistical analyses will quantify these characteristics. Direct and Indirect Shareholdings of Listed Belgian Firms: For the shareholdings of the listed companies, we have used some selection criteria in order to include only the main, significant shareholdings. We kept all the shareholdings in firms that were already included in the matrix and the stakes held in large non listed Belgian corporations (where the capital exceeds BEF 50,000,000). We excluded the shareholdings in subsidiaries and branches and the other minor shareholdings (below 2% of the capital). Being a bank, GBK has no shareholdings in industrial companies, but she controls the listed bank Belgolaise and she has a shareholding in the listed financial company Definance. There is no indirect shareholding for GBK. It gives: BELS DFIN FIN GBK SGB Sika BELS (Belgolaise) DFIN (Définance) GBK (Générale de Banque) 53.4% 4.8% FIN (Finoutremer) 0.4% SGB (Générale de Belgique) 55.2% 27.2% UM (Union Minière) AG 1824 (Fortis group) 10.7% Bancinvest (Fortis group) 1.6% BBL fonds 1.5% Cofiblan (Suez group) 3.8% Frabepar (Suez group) 9.0% Mutuelle Solvay 2.9% Sika (SGB group) 1.5% 9.0 % 54.5% Suez (France) 61.0%

14 4. Belgium 4.1. Specific Features about Belgium Belgium counts 220,000 firms approximately. Many of them are small. Half of the firms count less than five employees or less than BF 10 million of total assets. Two legal forms dominate: the Société Privée à Responsabilité Limitée (SPRL, a limited liability partnership) and the Société Anonyme (SA, a stock corporation). There are about 90,000 companies of each type. SPRLs are the most numerous among small firms (99% of SPRLs are firms under BF 100 millions of total assets). Their ownership certificates are nominative and the transferability of the certificates is subject to restrictions, such as the agreement of the other partners. Most large firms are SAs (84% of firms over 100 millions of total assets are SAs). Their distinguishing feature is the possibility of issuing anonymous shares with no restriction on their transferability. There are 140 Belgian registered firms that are listed on the official market of the Brussels Stock Exchange in Their market capitalisation accounts for about 38% of the GDP. Listed firms are of various sizes and belong to all sectors of the economy. Holding companies account for 23% of the market capitalisation, while electricity and gas companies represent 20% of the capitalisation on the Brussels Stock Exchange. These are followed by banks and financial services companies (14%), chemical companies (9%) and insurance companies (8%). Market capitalisation is highly concentrated among a few large firms: the Top 10 account for 50% of the total market capitalisation, while the Top 50 represent 95% of the market capitalisation. Turnover is low for smaller listed firms: the BEL20 market index, which includes 20 firms, accounts for 83% of the total market turnover. In terms of owners, Belgium is characterised by the dominance of large holding companies since the banking reform of After the big crisis of the thirties, a reform of the banking law forbade banks to hold shares in non-financial companies and to invest in those by any means except credit allocation. Banks then split into investment branches that became independent holding companies and banking branches that became the banks of today. The law changed in 1992 with the second European banking directive allowing some shareholdings of banks in industrial companies as investments. However, for various reasons like the loss of expertise and the role of holding companies, Belgian bankers did not take advantage of this possibility and remained out of the ownership of industrial firms. That is why one never finds a bank as a shareholder of a firm other than another bank or insurance company. For Belgian holding companies however, the situation started to change recently toward a diminution of their influence. In that matter, the most striking example is the buying in the beginning of 1998, by the French holding Suez-Lyonnaise des Eaux, of the totality of the shares of the biggest Belgian holding Générale de Belgique. This trade increased even further the influence of the French investors on the Belgian economy, growing in all sectors since the last decade. It led to the delisting of the Belgian holding, but, in counterpart, the French holding Suez-Lyonnaise became listed on the Brussels Stock Exchange. On the other hand, over the last few years, the holding Générale de Belgique tends to concentrate its activities toward a core-business in the energy sector by increasing its shareholdings in firms like Electrabel, Petrofina and Tractebel. The holding bought 25% of

15 Tractebel in September 1996 acquiring then a control position with 52% of the votes, and sold in 1998 its 27% shareholding in the Générale de Banque (largest bank in the country). However, these elements are not taken into account in this analysis, since the database ends in December Family ownership has traditionally been very important in the country and still is. Family groups control 27 blocks of shares in the 140 listed firms. Some family empires are very old, dating back to the industrial revolution, others are more recent but the influence of individuals in the economic life has always been a characteristic of the country. In terms of activities, the trading volume of the Brussels Stock Exchange has considerably augmented over the last years. In 1997, the trading volume of BEF 1230 billion represented 52% more than 1996 and more than the double compared to The turnover has doubled also between 1995 and 1997, reaching 21% of the market capitalisation. For the first time, the market capitalisation has exceeded the half of the GDP (58%), compared to 38% in Internationally, the Brussels Stock Exchange ranks now in the median, ahead of Paris (50%) and Frankfurt (41%) Quantitative Analysis The goal of this section is to highlight and to comment the main descriptive statistics about direct and ultimate ownership of Belgian listed firms. Subsection A details the number and size distribution of direct stakes and voting blocks of listed Belgian companies. Subsection B presents the direct shareholders and the ultimate blockholders, their type and the composition of their portfolio of investments. Subsection C gives summary statistics of ultimate ownership, compared to direct ownership. Subsection D, at last, gives an overview of the portfolios of shareholdings of the listed Belgian firms themselves. Except for the first table, tables and figures are displayed in appendix 1 of the chapter 33. Table 1 summarises the number of companies, stakes, blocks, shareholders and blockholders we analyse. A direct stake* is a stake held directly in a one of those firms by a shareholder considered individually. A group block* is the sum of the stakes held by shareholders belonging to the same business group and who vote jointly. A voting block* is the sum of the stakes held by shareholders linked by a voting agreement*. TABLE 1. NUMBER OF LISTED COMPANIES, STAKES, BLOCKS, SHARE- AND BLOCKHOLDERS Number of Listed Companies (31 Dec. 1995), Brussels Official Market 140 Number of Listed Companies with at least one ownership notification 135 Number of companies with no notification 5 No. of Notified Voting Blocks 269 No. of Group Blocks notified with Voting Blocks 489 No. of Direct Stakes notified with Voting Blocks 750 No. of Voting Block Holders who filed the notification 195 No. of Group Block Holders mentioned in the notifications 328 No. of Shareholders mentioned in the notifications Tables and figures presented in subsections A and B (plus Table 1) come from Becht and Chapelle (1997), Preliminary Report of the ECGN to the European Commission

16 Table 1 shows that the average listed firm has notified 5 direct shareholders, 3 group blocks* and 2 voting blocks*. The 750 direct stakes notified gather in 269 voting blocks (note that, in many cases, a voting block counts only one stake, as we will see further). Thus, we can consider that voting agreements concentrate on average voting power almost by a factor 3. There are 195 voting blockholders who filed 269 notifications, which makes an average of 1.4 voting blocks ultimately controlled by a blockholder. There are 562 shareholders who filed 750 notifications, which makes an average of 1.3 direct stakes per shareholder. These averages close to 1 let suppose that most of the investors in listed companies have very small portfolios, including one listed firm in most cases. We will examine in more details the portfolios of shareholders and blockholders in subsection B. A. Stakes and Blocks per Listed Firm Number of stakes and blocks and their stability Looking at the number of direct stakes per company (table 2), it appears that three quarters of the firms count 7 or less notified direct stakes, and half of the listed firms count 3 or less notified direct skates. Looking at voting blocks, three quarters of the firms count only two voting blocks, and one voting block only in 45% of the cases. (Table 2). Extreme cases are one company with 58 direct stakes, 50 group blocks and 3 voting blocks (Almanij) and one company with 8 voting blocks (Definance). The five companies without notification could be considered to have zero stakes, group and voting blocks. However, some voting pacts, resulting from simple contractual agreements, might be temporary of fragile, especially in bad times, so that they do not always reflect the real voting structure of a General Assembly. It is the reason why voting blocks are not always taken into account by certain studies, or by the firm itself, and by the other shareholders. In order to assess the stability of the voting blocks, table 3 shows the number of stakes in a voting block. In more than 50 percent of the cases (150 elements), the voting blocks is only composed of one single shareholder. If we consider the 150 voting blocks remaining and made of several shareholders, 50 voting blocks are coalitions between two shareholders and 50 others are coalitions between three to five shareholders. The maximum value is 49 shareholders in one block; it is a coalition of individuals and non-listed firms voting in the holding company Almanij. Table 3b shows the number of group blocks in a voting block. In 75% of the cases, a group block corresponds to a voting block: shareholders belonging to the same group of firms form a coalition of voters with no external members. In 25% of the cases however, one or several voters join an existing group block to vote, or individuals and stand-alone firms decide to vote jointly, without necessarily being part of an industrial group. So, in most cases, voting blocks are either made of one single owner or they correspond to a voting agreement between firms belonging to a same industrial group ( group block ), being this way quite stable. In other cases, voting blocks are often agreements between few shareholders, reducing the co-ordination problems. However, it must be noted that voting blocks are not always agreements on all types of votes. For example, the voting pact between the holding companies Générale de Belgique (27.35%) and Groupe Bruxelles Lambert (24.52%) shareholders of the industrial company Tractebel does not include joint voting on strategic decisions concerning the management of the firm. So that, the voting block formed by the two holding companies has not been

17 considered as a joint control 34. So, a case-to-case analysis is necessary before considering voting blocks as single, unified, entities of control. Stakes and blocks by rank and concentration measures To give a synthetic view of ownership concentration, Figure 1 displays the direct stakes by rank. For each rank, the minimum, median, mean and maximum are computed in % of the votes. The median and the mean of the largest direct stake in a listed Belgian firm is around 45%. This confirms that Belgian direct ownership is highly concentrated. In order to get majority control of 50%, the largest shareholder ought to form a coalition with the second with the third largest shareholder. Still, in practice, atomistic shareholders do not usually exercise their voting rights on the annual meetings such that the largest average shareholder might have de facto absolute control. A remarkable point (Figure 1) is that the second largest shareholder lags neatly behind the first one, with a median stake and a mean stake around 11%, which is two times less than a blocking minority*. The third rank of stakes is even much smaller, around five percent, and the rest is negligible. It follows that direct shareholdings are characterised by a small number of significant shareholders - one to three - with a largest investor approaching the majority level and that is not challenged by the second largest owner, lying below the blocking minority* level. To compare ownership characteristics both in terms of direct stakes and voting blocks, the same tables and figures are displayed for both variables. The median and the mean of the largest voting block in a listed Belgian firm is around 56%, which is already above the simple majority level (Figure 2). And, like in the case of direct stakes, the second largest blockholder lags far behind the first one, with a median stake around 10%, and a mean stake even smaller. Further ranks do not attain 5% of the votes on average. We see here that direct ownership of listed companies are characterised by the control exercised by a single voting block, when other minor blockholders stay around 10% and 5% of the votes. Concentration measures (C 1, C 3, C 5, C all ) in table 5 and 6 give the same trend of results as the ranks of stakes and blocks. There is a large gap between the largest shareholder and the top three shareholders, but the distance gets reduced as the ranks of shareholders get bigger, both for direct stakes and voting blocks. The top three of direct shareholders own on average a cumulated percentage of 59.3% of the votes (62.6% with voting blocks), and the total of shareholders declared own together 63.8% of the votes of the firm (same with voting blocks). Distribution size of stakes and blocks Figure 3 presents a histogram with the maximum direct stake for 135 notified companies. The two main peaks are at 25%-30% and 50%-60%, corresponding respectively to the blocking minority and the single majority levels of ownership. There is an additional peak for the 15%-20%. The same histogram for voting blocks is more revealing since any of the smaller direct stakes could actually come from a larger group or voting block. The histogram of maximum voting blocks indicates three clear peaks: 50%-55%, 55%-60%, and 65%-70% (Figure 4). The first peak corresponds to the simple majority level while the peak 65%-70% corresponds to the qualified majority (66%) required for certain decisions at the General Meeting. A rather surprising result, compared for instance to a country like 34 Commission Bancaire et Financière, Rapport annuel, 1996/1997, pp