Multiple Controlling Shareholders and Firm Value **

Transcription

1 Multiple Controlling Shareholders and Firm Value ** C. Benjamin Maury a, Anete Pajuste b, * a Department of Finance and Statistics, Swedish School of Economics and Business Administration, P.O. Box 479, Helsinki, Finland b Department of Finance, Stockholm School of Economics, P.O. Box 6501, Stockholm S-11383, Sweden Abstract We show that firm value depends on the interaction between the blocks of large shareholders. The paper analyses the trade-off facing controlling shareholders of whether to monitor or collude with the management in the expropriation of minority shareholders. Firm value is positively related to the presence of a third strong blockholder, particularly when the other two large shareholders have blocks of comparable size, reflecting a monitoring incentive. Firm value is negatively related to the presence of a second large shareholder, especially when the largest owners combined own the majority of voting rights, reflecting a collusion incentive. JEL classification: G3, G32 Keywords: Corporate governance, Ownership structure, Multiple owners * Corresponding author. Tel: ; fax: address: Anete.Pajuste@hhs.se (A. Pajuste). ** We are grateful to Henrik Cronqvist, Guido Friebel, Mariassunta Giannetti, Peter Högfeldt, Matti Keloharju, Yrjö Koskinen, Eva Liljeblom, Anders Löflund, Andrei Shleifer, Jeremy Stein, Tuomo Vuolteenaho and seminar participants at Harvard University and Toulouse University for valuable comments. Financial support from the Finnish Academy of Sciences, the Hanken Foundation, OKOBANK Group Research Foundation, Stiftelsen för främjande av värdepappersmarknaden i Finland (Maury), and Handelsbankens Forskningsstiftelse (Pajuste) is gratefully acknowledged.

2 1. Introduction Since the seminal work by Berle and Means (1932), the conflict of interest between managers and dispersed shareholders has attracted a lot of attention. Recently, empirical work has shown that ownership is typically concentrated in the hands of a few large shareholders, giving rise to an equally important agency conflict between controlling shareholders and minority shareholders (e.g., La Porta et al., 1999; Barca and Becht, 2001). This agency problem may be particularly severe if cash flow rights differ from control rights. In practice, this separation occurs through the use of high voting shares, cross-holdings or through a pyramid ownership structure, i.e. when there is a departure from one share, one vote (Grossman and Hart, 1988; and Harris and Raviv, 1988). Outside the United States, the presence of several large shareholders with substantial blocks of shares is common (Barca and Becht, 2001; Becht and Boehmer, 2002; Claessens et al., 2002). Data on 5232 European companies collected by Faccio and Lang (2002) show that 39 percent of firms have at least two controlling shareholders that hold at least 10 percent of voting rights, and 16 percent of firms have three controlling shareholders 1. Therefore, it is important to study the allocation of control between controlling shareholders, as well as its impact on firm performance. So far, very few papers have addressed this issue. Theoretical models of the interaction between multiple large shareholders are presented by Gomes and Novaes (2001), Bennedsen and Wolfenzon (2000), and Block and Hege (2001). Gomes and Novaes (2001) focus on how ex-post bargaining problems among large shareholders protect 1 Throughout the paper, the controlling shareholder is defined as a shareholder having at least 10 percent of the voting power. 1

3 minority shareholders by preventing large shareholders from undertaking actions that would otherwise harm minority shareholders. Bennedsen and Wolfenzon (2000) explore the formation of coalitions between large shareholders. They suggest that the best ownership structure is one with either a single large shareholder or shareholders of roughly the same size. Block and Hege (2001) introduce a model where two blockholders compete for effective control in a company. They claim that the relevant concept of control power is how contestable the leading shareholder s position is, and not just ownership concentration. According to their model, control becomes more contestable if, for example, the controlling shareholder s block decreases in size relative to the competitor s. Empirical evidence on the role of multiple owners in firm performance has been limited. Lehman and Weigand (2000) report that the presence of a strong second largest shareholder enhances profitability in German listed companies. Faccio et al. (2001) test the multiple large shareholder effect on dividends. They find that the presence of multiple large shareholders dampens expropriation in Europe (due to monitoring), but exacerbates it in Asia (due to collusion). In Italy, Volpin (2002) provides evidence that valuation is higher when a voting syndicate, as compared to a single shareholder, controls the firm. All these studies focus only on the presence of multiple controlling shareholders and not on the characteristics of blocks or contestability of power. To our knowledge, this paper is among the first attempts to empirically explore the effects of multiple blockholders on firm value by taking into account not only the presence of blockholders but also the contestability of their control, and blockholders type. This paper uses detailed ownership data from Finland to study the role of multiple controlling shareholders. The sample consists of ownership data collected from 174 Finnish 2

4 firms during the period 1993 to The total number of firm-year observations with ownership data is For each firm, we collect data on cash flow rights and votes for the largest three owners 3. Even in large firms both directly and through their private companies private persons hold significant proportions of the control rights. The data give a comprehensive picture of ultimate ownership of Finnish listed companies. Ownership structures are rather stable over time. Concentration of power and the separation of ownership and control, mostly achieved by the use of dual class shares, are set historically and usually remain intact unless specific external factors arise, for example merger with a U.S. company. The stability of ownership structures thus provides a motivation to focus on cross-sectional differences in firm valuation. The Finnish data set is representative in several respects. First, the presence of large blockholders is common outside the United States. Second, in Finland, just like in many other countries worldwide, corporate law gives controlling shareholders significant powers through the use of multiple voting class shares. Finally, Finnish ownership data allows us an in-depth study of multiple blocks providing information about cash flow rights and voting power of several largest shareholders, as well as their type (e.g., family, financial institution, and state). In addition, a one-country study allows us to focus on the effect of multiple owners by avoiding the country-level institutional differences. 2 Missing values in some of the explanatory variables reduce our sample to 712 in the subsequent analysis. 3 The owners are ranked according to their voting stakes, e.g. the largest owner is defined as the shareholder with the highest voting power. 3

5 The data is clearly clustered into two characteristic groups, defined simply by the voting power of the single largest shareholder. The first group consists of small size firms, with one majority block, a high use of the ownership and control separation, high tangibility of assets and low Tobin s Q. The second group includes large firms with dispersed ownership, a high presence of foreign ownership, alignment of ownership and control, and high Tobin s Q. This paper, however, focuses on firms that fall in the middle range those which are neither dispersed nor highly concentrated. How does ownership affect value between these firms? We propose that in the middle range firm value depends on the interaction between several blocks, particularly on the level of control contestability. After controlling for other factors that might affect firm value, we find that voting power contestability (using different measures) has a positive effect on firm value, as measured by Tobin s Q. We suggest that the contestability increases with the number of controlling shareholders that have relatively equal voting power, and with the ability to challenge the voting power of the largest block. For example, firms with three controlling shareholders, but for which the two largest owners hold less than 50 percent, have higher value than firms that are controlled by a single majority owner or than firms controlled by two shareholders forming a majority. Interestingly, the presence of a strong second largest shareholder does not have a positive effect on firm value. Our model explains these results in the following way. A controlling shareholder faces a trade-off between monitoring other large shareholders (and the manager), and colluding with them to extract private benefits. Collusion may be more optimal if her cash flow stake is low. The data show that the largest and second largest shareholders separate ownership and control they have more votes than equity more often than the third 4

6 shareholder does. Hence the incentives to collude are higher for the first and second largest shareholder, but not for the third one. Finally, we emphasize that we cannot make strong claims about the direction of causality between different ownership structures and firm value. Instead, we try to explain the cross-sectional variation in firm value not captured by other firm characteristics, and find that the interaction between multiple controlling shareholders has economically significant explanatory power. We present our interpretation of the causal links behind this cross-sectional variation. The paper proceeds as follows. Section 2 presents a simple model that analyzes the controlling shareholders incentives to collude or monitor. Section 3 describes the data set. Section 4 discusses the determinants of ownership structure. Section 5 presents the results. Section 6 offers robustness checks and alternative interpretations of results. Section 7 concludes. 2. A simple model We present a model of a firm with one, two or three controlling shareholders (blockholders). Controlling shareholder is defined as an owner with at least 10 percent of votes. The motivation for using this threshold is that outside shareholders holding at least 10 percent of voting power are given certain important rights against abuse by insiders (see Appendix A). The motivation behind looking at the three largest shareholders comes from prior empirical studies (at least in the continental Europe) reporting that usually no more than two to three largest shareholders exercise their power at general meetings (e.g. on Finland, see Karhu et al., 1998). Previous literature has focused predominantly on the effects of control concentration on 5

7 firm performance, assuming that there is only one controlling shareholder. Many of these studies find a non-linear relationship between ownership concentration and firm value 4. This suggests that there might be other things going on beyond a pure one-to-one relation between control concentration and firm value, motivating our choice of focus on the characteristics of multiple blocks rather than their total share. The presence of multiple large shareholders rather than a single controlling shareholder could affect the extraction of private benefits at the cost of minority shareholders. We propose that the agency problems between several controlling shareholders are at least as important as between controlling shareholders and outside investors. The model presented in this section is borrowed from La Porta et al. (2002), with some added features to illustrate the effect of interaction between multiple blocks. We assume that the controlling shareholders have cash flow or equity ownership α 1, α 2, α 3, respectively. As in La Porta et al., we assume that the cash flow stakes are exogeneously determined. In particular, we do not consider the choice of optimal ownership structure, but rather the effect of the chosen ownership structure on firm performance. This assumption is consistent with our findings that the ownership structures of Finnish firms do not change that often. The firm s profits are RI, where I is the amount of cash, which is invested in a project with the gross rate of return R; the firm has no costs. There is no manager, other than representation of the largest shareholder. One may regard the largest shareholder as equivalent to the manager or insider. This is consistent with Pagano and Röell (1988) who claim that the largest shareholder generally takes an active interest in running the company, by choosing the management and/ or directly taking executive positions. The controlling shareholder(-s) can 4 See for example Morck, Shleifer and Vishny (1988), Stulz (1988) and McConnel and Servaes (1990). 6

8 enjoy private benefits that are not shared with minority shareholders. In particular, a share s of the profits can be diverted from the firm. We assume that the largest shareholder (= manager) is the one that chooses s, the level of private benefit extraction. The private benefits can take the form of excess salaries, beneficial transfer pricing to controlling shareholders privately held firms, subsidized personal loans, etc. Multiple controlling shareholders may collude in private benefit extraction. This assumption follows Zwiebel (1995), who argues that many investors choose to hold less than majority blocks of equity because blocks are generally considered necessary for active participation in controlling coalitions. The diversion of the profits is inefficient the controlling shareholders receive sri c(s,n)ri, where c(s,n) is the share of profits that is wasted when s is diverted. Here n denotes the number of shareholders in the controlling coalition 5. We have added this parameter to capture the intuition that the total cost of diversion is different depending on the number of shareholders participating in a coalition. Like in LLSV, we assume that c s >0, c ss >0, and add the assumption that c n <0 and c nn >0. These assumptions mean that the marginal cost of stealing is positive; the marginal cost of stealing rises as more is stolen; the total cost of stealing decreases with the number of shareholders in a controlling coalition; and the marginal cost of stealing increases with the number of shareholders in the coalition. The last two assumptions capture the idea of economies of scale in stealing: more partners in a scam, other things equal, add the knowledge and resources to hide the diversion of profits. However, the marginal gain in cost reduction decreases with every new partner. 5 LLSV (2002) assume c(s,k), where k is the quality of shareholder protection that varies between countries. Since we study only one country, we drop this parameter. 7

9 The model introduces a trade-off that the controlling shareholders face, namely to collude, to monitor, or to remain passive (i.e. to be in the same position as minority shareholders). We assume that the largest shareholder never remains passive, and further that she has some positive private benefits of control. The trade-off between colluding and monitoring stems from the following agency problem. Burkart et al. (1997) argue that the controlling shareholder may be value enhancing because she monitors the manager (here the largest shareholder), but at the same time it may also lead to opportunistic behavior by the delegated monitor. After verification, the monitor observes the actual cash flow that is larger than the reported cash flow, which she should seize and share with other investors. Instead she can do better by colluding with the manager and sharing the difference only with the manager. When there are multiple controlling shareholders, the same agency problem arises. If any of the controlling shareholders delegates the verification of cash flows to one shareholder, there is a risk that the delegated monitor will collude with the manager. As we will see, delegated monitoring may nevertheless persist when private benefits to be shared by the monitor are low, and if moreover the monitor holds a large fraction of cash flow rights (not only voting rights). First, let us assume that there is only one controlling block. The largest shareholder maximizes: α 1 (1 s) RI + sri - c(s,1)ri (1) The first order condition gives the LLSV (2002) result, namely that: c s (s,1) = 1 - α 1 (2) We denote the s that solves this expression as s 1 *. 8

10 Now, let us assume that there are two controlling blocks. In this case, the second largest shareholder has to decide whether to monitor or to collude with the largest block or to remain passive. Collusion case If the two blocks are colluding, the largest shareholder maximizes: α 1 (1 s) RI + sλri - λc(s,2)ri (3) where λ is the relative bargaining power of the largest shareholder; λ (0,1]. We assume that the bargaining power can depend on the shareholder s relative voting power, managerial representation, and type. The total cost, c(s,2), is shared by the two parties proportional to their bargaining power, i.e. proportional to their share in private benefits. The first order condition for problem (3) is given by: U s =[-α 1 + λ - λc s (s,2)] RI = 0 (4) which gives the expression for s: c s (s,2) = 1 - α 1 /λ (5) We denote the s that solves this expression as s 2c *. From (5), we get the very intuitive result that higher bargaining power in hands of the largest shareholder (λ) increases the private benefit extraction. In the extreme case, if λ=1, the largest owner has the whole bargaining power in a coalition, and expression (5) collapses to c s (s,2) = 1 - α 1 (6) Given the assumption that the total cost of stealing decreases with the number of shareholders in a coalition, and here λ=1, then s 2c * > s 1 *. This result suggests that the largest shareholder will divert more if he can benefit from the presence of another conspirator, and 9

11 enjoy all the private benefits. When λ<1, s 2c * goes down, and at some point will reach s 1 *. If λ is very low, then the extraction of private benefits in a coalition of two (s 2c *) will be below that of a single blockholder (s 1 *). Comparing expressions (2) and (5) we also see that if the coalition of two owners reduces the extraction costs considerably (i.e. the cost of stealing is much lower for two than for one owner), then the private benefit extraction in a coalition of two blocks can actually be higher than with a single owner. The second blockholder s utility function is given by: α 2 (1 s 2c *) RI + s 2c *(1-λ)RI (1-λ)c(s 2c *,2)RI (7) Monitoring case The second largest shareholder can also take the monitoring role. We assume that by monitoring the largest shareholder, the monitor can recover share γ [0,1] of the diverted profits, i.e. γsri. The recovered profits are distributed among all shareholders on a pro rata basis. The monitoring requires a fixed cost m, that is incurred by the monitor. The second blockholder thus has the following payoffs depending on monitoring or not monitoring, respectively: (monitoring) α 2 (1 s 2m *) RI m (8) (not monitoring) α 2 (1 s 1 *) RI (9) where s 2m * = γ s 1 * s 1 *. With no monitoring, we are back to the single shareholder case with s 1 *. With monitoring, share γ of s 1 * is recovered, and s 2m * < s 1 *, i.e. the diversion is lower than in the single shareholder case. If α 2 (s 1 * s 2m *)RI m, i.e. the monitoring is optimal from the perspective of the second blockholder, then the largest shareholder s utility is: α 1 (1 s 2m *) RI + s 2m *RI - c(s 2m *,1)RI (10) The different levels of private benefit extraction (s) are shown in Figure 1. 10

12 c s (s,1) 1-α 1 c s (s,2) Figure 1 s 2m * s 1 * s 2c * (λ=1) s With one controlling shareholder, the share of private benefits extracted is s 1 *. When there are two controlling shareholders, the private benefit extraction depends on the relative bargaining power of the two parties. In the extreme case, when the controlling shareholder has full bargaining power (λ=1), but the second shareholder still joins the coalition thus reducing the total cost of private benefit extraction, the profit diversion is at maximum (s 2c *). As the largest shareholder s bargaining power decreases, the private benefit extraction decreases moving towards s 1 * or even below it. The lowest level of profit diversion is when the second largest shareholder finds it optimal to monitor. Then the level of private benefit extraction is s 2m *. Finally, we can compare the payoffs of the second largest shareholder in case of collusion and monitoring: (collusion) α 2 (1 s 2c *) RI + s 2c * (1-λ)RI (1-λ)c(s 2c *,2)RI (11) (monitoring) α 2 (1 s 2m *) RI m (12) These expressions allow us to do some simple comparative static analysis. If λ=1 (full bargaining power for the largest shareholder), then the second blockholder optimally chooses to monitor, because s 2m * < s 1 *, and thus he can have higher security benefits. Higher cash flow participation and lower bargaining power for the largest shareholder decreases the private benefit extraction. This implication is in line with numerous studies that find that the cash flow 11

13 stake and the voting stake (here, the bargaining power) has an opposing effect on private benefit extraction and hence firm value (e.g. Claessens et al. (2002), La Porta et al. (2002)). The model thus far suggests that a simple presence of the second shareholder does not tell us much about it s effect on private benefit extraction. We observe that what matters is the relative cash flow stake of the largest and second blockholder, the relative bargaining power, and the monitoring cost. Depending on different parameter values, we may observe either monitoring or collusion, which has a very different effect on private benefit extraction. Now let s add a third controlling shareholder. The analysis follow directly from the two controlling shareholder case. There are several potential coalitions to be formed. If the two largest blocks have formed a coalition, the third block evaluates the costs and benefits of joining the coalition versus monitoring it (exactly as in the two shareholder case). It can be seen that low monitoring costs, low bargaining power in private benefit extraction, and higher cash flow stake of the third block all work in one direction, namely towards high incentives to monitor the largest block(s). If the second block is monitoring the largest shareholder or is passive, then the third block evaluates the costs and benefits of monitoring or colluding with the largest shareholder. 6 We can again regard this as a two shareholder case with the largest and third largest blockholder involved. There can be another coalition, namely, between the second and third shareholder. If this coalition has higher voting power than the largest shareholder, it is fare to assume that it would gain the power to choose the level of private benefits. The analysis then extends from the two shareholder case presented above. 6 We assume that the monitoring costs are duplicated if both the second and the third blockholders monitor the largest shareholder (i.e. each of them individually incurs m). Thus we rule out the potential collusion in monitoring. 12

14 We can now turn to the implications of this model for firm valuation. As in La Porta et al. (2002), we measure valuation with Tobin s q, which is given by q=(1-s)r. This measure is taken from the perspective of a minority outside shareholder who only receives the security benefits. The level of private benefits has a direct implications on Tobin s q, namely that higher s has a negative effect on firm valuation. The model suggested different levels of s depending on the number and the nature of firm s blockholders. Lets reconcile these results with firm valuation. Firm value is expected to be the lowest when the largest blockholder (or a coalition) holds more than 50 percent of votes (high bargaining power), and at the same time uses separation of ownership and control (i.e. less equity than votes). High voting power gives discretion in private benefit extraction while lower cash flow stakes underweight the importance of security benefits 7. A coalition formed by two controlling shareholder may have a more negative effect than just one blockholder if the second shareholder can substantially reduce the cost of private benefit extraction. On the other hand, if the cost of private benefit extraction is not changed, a strong second blockholder may improve firm value, because it reduces the bargaining power of the largest shareholder, thus reducing the private benefit extraction. Intuitively, it is more difficult to form a coalition of three (or more) partners than just two. Moreover, the marginal gain of profit diversion cost reduces with the number of partners. Three strong blockholders mean that the control power of the largest shareholder is very 7 This is consistent with both theoretical papers (Grossman and Hart, 1988; Harris and Raviv, 1988; and empirical work (La Porta et al., 2002; Claessens et al., 2002, Cronqvist and Nilsson, 2003) that show that the negative effect of large shareholders is magnified if there is a substantial departure from one-share-one-vote. 13

15 contestable. The contestability of power may increase when there is a small difference in block size between the leading shareholders. Higher contestability means lower bargaining power in hands of the largest shareholder, and hence lower private benefit extraction. Therefore, we expect firm value to be positively related to control contestability. To summarize, the presence of multiple blockholders introduces the following trade-off. More than one controlling shareholder may increase firm value, because it reduces the largest shareholder s bargaining power and thus the level of private benefit extraction. On the other hand, it may reduce firm value, because with more partners in profit diversion the private benefit extraction can become more efficient, and thus increase the level of private benefit extraction. The true effect on firm value depends on the relative size of the blocks, as well as the blockholders characteristics (type). In Section 5, we will try to disentangle these effects empirically. 3. Data and methodology 3.1. The sample and variable descriptions This paper is based on a new database of ownership structures in Finnish listed companies during The total number of firm-year observations with ownership data is 804. For several explanatory variables the number of observations is lower (see Table 1 for number of observations for each variable), therefore the majority of regression results are based on a sample of 712 firm-year observations. There are 174 different firms in the sample. We thus have an unbalanced panel that permits firms to enter and leave the market freely. Our main source for ownership data is the yearbook Pörssitieto (Kock, ) that relies on the data from Finnish Central Securities Depository or the firm s own ownership 14

16 register. The book reports the cash flow ownership and votes of the 20 largest shareholders ranked by ownership. Where the data is inadequate in the book we use firms annual reports. We are mainly interested in the controlling shareholders and how they use control. Therefore, we collect data on equity and votes and the identity (type) of the three largest owners in each firm 8. We classify the shareholders into the main types, family, corporation, financial institution, state and other. Ownership by families is aggregated to include family members with the same surname. Families are assumed to own and vote collectively. Pörssitieto sums up the ownership of financial firms belonging to various insurance groups, although these do not legally form a group. As these firms probably act in concert, we use the same group classification. We have tried to identify the ultimate owners in Finnish listed companies. We include indirect holdings through private firms by private persons when they are reported among the 20 largest shareholders. If a corporation or financial institution owns a company in our sample, we check further to see if it has a majority owner and report the ultimate owner s type, if there is one. If an owner is a private corporation and none of the insiders (board members and managers) have a controlling stake in it, we report this owner type as a corporation. In most cases, the data are at the end of the financial year. The fact that all ownership data is not from exactly the same date does not cause a problem, because the ownership structures were quite stable over the studied period. 8 One motivation for studying the control of the three largest shareholders in a firm is the finding by Karhu et al. (1998) who report that in 80 percent of the general meetings the majority of votes and shares were in the hands of one, two, or three large shareholders. 15

17 The accounting data for the control variables comes from Datastream. The missing values were added from the annual reports. We followed the Helsinki Stock Exchange classification of industry groups (15 categories) for Main List (first tier market) companies. For I List and NM List (second tier market) companies we extracted the industry classification from information given in the annual reports. Market value of equity (used in the Tobin s Q ratio) was collected from the annual reports. The market value of unlisted shares (if any) is assumed to equal to the market price of the listed shares. Tobin s Q is measured as the sum of market value of listed shares, the implied value of unlisted shares (by extrapolating the market value of listed shares), and the book value of debt all divided by book value of total assets. As a result, low values of Tobin s Q might be biased since the extrapolated market value does not take into account the value of private benefits accruing to owners of unlisted shares. Since we are interested in the value effects to outside (=minority) shareholders, this alternative interpretation of low Tobin s Q values works in our favor. It again demonstrates that they value different characteristics than do large blockholders. We will return to this discussion in Section 6. We control for the impact of foreign ownership by collecting the aggregate number of voting shares of registered as well as nominee registered foreign owners for each company. The data comes from Helsinki Stock Exchange or companies annual reports. Lastly, to have a more in-depth understanding of reasons behind particular ownership structures and transfers of control, we use the Lexis-Nexis database that compiles past news on acquisitions, transfers of control and other ownership related company information. 16

18 3.2. Descriptive statistics Table 1 shows summary statistics for variables used in this study. The variables are defined in Appendix B. The distribution of votes and cash-flow ownership together with the control-to-ownership ratio reveal that the largest owners generally use control rights in excess of their cash-flow rights. The wedge between votes and equity is largest for the largest controlling shareholder, having a mean control-to-capital ratio of In other words, the largest shareholder has 1.39 percent of voting rights per 1 percent of cash flow participation. The use of dual-class shares is common among the largest shareholders, especially privately controlled firms. Interestingly, the largest and second largest shareholder s wedge between the voting and cash flow rights are highly correlated (i.e. they do not differ much), but the wedge of the third largest owner is lower 9. [INSERT TABLE 1 HERE] The distribution of votes and cash flow ownership by the largest shareholder in Finnish listed companies is displayed in Table 2, Panel A. Private persons or families control 32 percent of the firms. Private control is important because it often comes with managerial and board representation. The concentration of votes and equity, amounting to 48 and 35 percent respectively, is high in privately controlled firms. The second largest ownership category is corporations, controlling 25 percent of the firms, followed by financial institutions that control 19 percent of the firms but with smaller average stakes. The state, cities, and municipalities are the largest shareholders in 12 percent of the firms. Other categories, including associations, control 10 percent of the firms. 9 See also Maury and Pajuste (2002) for a discussion on this issue. 17

19 Panel B in Table 2 describes the type of controlling shareholder, who is defined as an owner with at least 10 percent of votes. We classify blockholders according to firm size measured by sales 10. In the lowest 20 percentile, the controlling shareholder is typically a private individual or a family. In the mid-sized firms, private owners constitute the largest group, controlling 31 percent of firms, followed by corporations controlling 27 percent of firms. [INSERT TABLE 2 HERE] Table 3 describes firms having an unlisted share class a class of shares not traded on the Finnish Stock Exchange. We observe unlisted shares more often among firms where the largest owner has more votes than cash flow rights (Panel A), and secondly among firms that have a single majority shareholder (Panel B). Panel A shows that unlisted firms as percent of all firms that separate ownership and control amount to 55 percent (201 out of 366), while unlisted firms as percent of total firms is only 29 percent (230 out of 804). Panel A also summarizes the number of firms in which the largest shareholder has votes in excess of cashflow rights. While there is no substantial increase in the number of companies for which ownership differs from control, the total number of companies listed increases during our sample (newly listed companies choose a one-share-one-vote mechanism). Panel B shows that among majority controlled firms 46.5 percent have unlisted shares, while among non-majority firms the respective number is only 19.5 percent. Panel C shows that Tobin s Q is higher for firms without an unlisted share class. [INSERT TABLE 3 HERE] 10 Results are similar when firms are classified by assets. 18

20 Table 4 provides a more detailed description of ownership blocks. In 45.7 percent of firm-year observations (335 out of 733) there is only one blockholder with at least 10 percent of votes. Two blockholders are present in 31 percent of the firm-year observations, while in 14.6 percent of the cases there are three blockholders. In 8.7 percent of the observations no shareholder has an ownership block of ten percent; we consider these firms as having a dispersed ownership structure. We are particularly interested whether the largest (or two largest) controlling shareholder(-s) have a 50 percent majority or not. When the largest shareholder holds a majority, we might expect the presence of a second (or third) block to be less important, simply because it is very hard to contest actions taken by the majority block. [INSERT TABLE 4 HERE] Table 4 suggests that the valuation consequences of control differ depending on the number of blocks and on the aggregate voting power of the largest shareholders. Median Tobin s Q is clearly higher under dispersed ownership structure (no block), 1.45, than any other ownership structure. However, if we simply compare the median Tobin s Q under one, two or three blocks, there is not very much of a variation ( ). A very different picture arises when we differentiate between aggregate voting power above 50 percent and below 50 percent. Panel B of Table 4 shows that the median Tobin s Qs are significantly higher if the largest (or two largest) blocks do not have majority. 4. What determines the ownership structure? Before analyzing determinants of firm value, we address the potential endogeneity problems related to different ownership structures. We focus on three ownership variables ownership concentration, the separation of ownership and control, and foreign ownership. The 19

21 ownership structures in Finnish listed companies are stable in the sense that ownership concentration and (to a lesser extent) foreign ownership do not show substantial variation over time. Therefore, to estimate the effect of various explanatory variables on ownership concentration and foreign ownership we use group means estimator (e.g. Green, 2000). To estimate the effect on separation of ownership control, we use a probit specification, because the dependent variable a dummy that takes a value of 1 if the largest owner has more votes than cash flow rights is discrete. As a robustness check we re-run the regressions with an ordinary-least-squares (OLS) and probit specification, respectively, in each separate year subsample; the results (not reported) do not change. We note that ownership structures in Finnish listed companies are stable. What are then the cross-sectional differences between firms with various ownership structures? The results are presented in Table 5. We find that the control concentration measured by sum of votes held by three largest shareholders is positively related to asset tangibility, measured by tangible assets divided by total assets, and negatively related to firm size; Model 1. The positive relationship between asset tangibility and concentration holds also for equity concentration (not reported), i.e. the sum of cash flow rights held by three largest shareholders. This result might arise for two reasons. First, high tangibility may simply be a feature of low growth companies. Second, high tangibility implies a high collateral value of assets, which may make it easier to access bank financing, thereby reducing the need to give up control when the firm needs capital. Therefore, we propose that ownership concentration is endogenously chosen by firms that have lower demand for equity financing. [INSERT TABLE 5 HERE] 20

22 There is evidence that the use of dual class shares is a heritage that older firms carry on, and that does not change unless certain external factors affect the firm. Model 2 reports the probit regression showing that older firms are more likely to choose the ownership and control separation (wedge dummy). The table also shows that family owners are more likely to separate ownership from control. Multiple class shares are less apparent in larger size firms. Again, this is consistent with the argument that large firms have higher needs for capital and thus they have to diminish potential sources of agency problems that would scare away equity investors. We have claimed that separation of ownership and control by the use of dual-class shares is a historical pattern that firms keep unless some external pressure arises. Two examples of why firms may want to abandon the dual-class shares follow. In 1997, Amer Group Ltd, producer of sports equipment including Wilson brand, abolished its vote-heavy K shares by converting them into ordinary A shares. The move came after UK investor Lord Moyne tried to become a core owner of the company by buying K shares. The company redeemed all the shares to be sold to Lord Moyne, and at the same time decided to reactivate the earlier planned conversion of K into A shares at a ratio of 1:1.5. According to analysts, this move ended the uncertainty surrounding the future of its powerful K shares and was positive to the company s long term success. It was believed that the conversion of the shares would increase interest in Amer shares in the U.S. In 2001, Eimo, an international manufacturer of precision plastic components primarily selling to the mobile communications industry, entered into a formal merger agreement with U.S. company Triple S. As part of the agreement, all Eimo Series K shares (20 votes each) were converted one-to-one into Series A ordinary shares (1 vote each), resulting in a single series of common stock. 21

23 These two examples are rather representative, and show that the most common reason for abandoning dual class shares is in case of an external pressure, such as merger or cooperation agreement with a foreign company or willingness to appeal to foreign, particularly, the U.S. investors. This brings us to the next discussion on role of foreign owners. We find that foreign owners choose better performing firms (see Table 5, Model 3). We test the contemporaneous relations for each year, but the results remain significant for lagged values, i.e. one-year or even two-year lagged values of Tobin s Q are economically determinants of the level of foreign ownership. The relationship is significant for each and every year in the sample. Moreover, foreign owners seem to pick the firms that do not have dual-class share structures. The causality from Tobin s Q to foreign ownership runs in the direction we specify because the foreign owners are mostly nominee-registered accounts, which are very unlikely to exercise control and influence company performance. Nevertheless, one could argue that the foreign ownership is a self-propagating factor, which may attract more investors, give higher credibility and thus further increase firm value. It seems plausible to assume that foreigners choose better performing (higher Tobin s Q) and more transparent firms (lower control-to-ownership ratio). We hope that these findings, for the purposes of subsequent discussion on the role of multiple blocks, allow us to treat the ownership structures as contemporaneously exogenous and focus on cross-sectional variation. Although, tangibility and age affect the ownership structure, they are factors that do not change rapidly. Therefore we propose to use the wedge between votes and cash flow rights as the proxy for a package of firm characteristics that potentially contain more agency problems than in other firms. 22

24 5. Results This section presents the cross-sectional analysis of differences in firm value. We try to explain differences in value by taking separation of ownership and control (the wedge) as a proxy for certain ownership related agency problems. We analyze the value effects in different ownership ranges and test the role of multiple blocks and their type. As a general rule, we use pooled ordinary-least-squares (OLS) specification with industry and year dummies wherever the ownership effect has substantial variation over years. Group means estimators are used when there is little variation in the main explanatory variables. Since most of the explanatory variables such as separation of ownership and control and relative size of multiple blocks, do vary over years, we use OLS. Repeating the analysis on a year-by-year basis leads to similar inferences (some years may loose significance due to low number of observations, but the signs of the parameters remain intact). The results are also corroborated using a balanced panel of firms (78 firms, 4 years ( )). Ownership and control variables are highly correlated. Therefore, to control for potential multicollinearity problems we estimate variance inflation factors (VIFs) for the independent variables in all the regressions we perform. The variance inflation factors are normally below 2, with a maximum of 2.8 (below the critical level of 5). We control for heteroskedasticity by using robust standard errors. Throughout this section, the dependent variable is ln(q), the logarithmic transformation of Tobin s Q. The simple model presented in Section 2 suggested that private benefit extraction is higher (i.e. value is lower) when the controlling shareholder s bargaining power is high and the cash flow rights are low. Presumably, this happens when the largest shareholder holds a majority stake (more than 50 per cent of votes), but less equity. This hypothesis is tested in 23

25 Table 6. Regression Model 1 uses the control-to-ownership ratio of the largest shareholder as an explanatory variable. The relationship between the wedge between ownership and control and firm value is significantly negative. Moreover, the negative effect is intensified if the largest owner holds a majority stake the interactive term between ln(co1) and Majority dummy is significantly negative. This result gives some support to the model s prediction that the value of a firm is lower if there is high control concentration but low equity stakes. We also study the effect of the wedge between ownership and control of the second and the third shareholders separately (not reported). The effect of the second largest shareholder s wedge is significantly negative, but that of the third largest shareholder is not significant, close to zero. This result once again indicates that there might be a significant role for the second and third largest owners. To compare our results with previous empirical findings, we test the relation between control concentration and firm value in Finland. We corroborate the previous findings that the relation between largest owner s stake and firm value is non-linear (not reported). In particular, there is a positive effect of dispersed ownership structures, a negative effect of very concentrated ownership, and a mixed relationship in the middle range the range where ownership is neither dispersed nor highly concentrated. The results thus far suggest that we cannot tell much about cross-sectional differences in firm value simply by analyzing the control stake held by the single largest owner. Simple concentration (combined voting power of the three largest owners) does not explain the cross-sectional variation in Tobin s Q within the middle range of firms the firms that are neither majority controlled nor widely held again suggesting that interaction between various ownership blocks and their relative size matters. Moreover we note that this middle range is quite large both in Finland (459 observations) and 24

26 Western Europe in general. Of the 5232 European companies in Faccio and Lang (2002) sample, in 57 percent of firms the largest shareholder has voting power in the range of 10 to 50 percent. Moreover 39 percent firms have at least two controlling shareholders (at 10 percent cut-off), and 16 percent firms have three controlling shareholders. Many earlier studies do not extend the discussion beyond the single largest owner; therefore we attempt to fill this gap. We propose that the cross-sectional differences within this range be driven by the interaction of multiple blocks. Table 6 regression Models 2 and 3 analyze the effect of multiple blocks when the ownership structure is neither dispersed nor is there a single majority owner, i.e. a sub-sample of firm-years where voting power of the largest owner is between 10 and 50 percent. Model 3 of Table 6 shows that there is a negative effect of two controlling blocks, when they jointly hold a majority (i.e. combined voting power exceeds 50%). Moreover, the simple presence of two blocks (Model (2)) does not have a statistically significant negative impact on value; what matters is the relative size of the blocks, i.e. whether they have a motivation to collude or not. The simple model presented in Section 2 showed that profit diversion may be higher when two shareholders form a coalition, because they can minimize costs of extracting private benefits. What are the second largest shareholder s incentives to collude? If the largest owner lacks majority, but can reach it together with the second owner, there are strong incentives to cooperate to expropriate minority shareholders and share the private benefits of control 11. This, 11 This hypothesis is supported also in regression (not reported here) where we test the effect of combined voting power of the largest and the second largest owners. The firms with combined voting power of the two largest owners above 50% have significantly lower Tobin s Q values. 25

27 consistent with our model, suggests a collusion when agent s bargaining power is high. The second shareholder s participation is detrimental to the coalition; therefore, it is plausible to assume that her bargaining power if high. [INSERT TABLE 6 HERE] We also observe a positive effect of the presence of the third block. Model 2 shows that the presence of three blocks has a significantly positive effect on firm value. The relationship holds also on the whole sample (Model 5). This result supports the positive effect of contestability of voting power, which is generally in line with theoretical predictions by Block and Hege (2001), and our model. Model 3 in Table 6 confirms the positive effect of three blocks, particularly when the largest two shareholders do not jointly hold majority. This structure implies that the three block owners are of rather equal size and that control is more easily contestable. Thus, the results suggest that three control blocks of comparatively equal size enhance the value due to mutual monitoring. This result remains robust in the group means specification; see Model 4. We also find (not reported) that the voting power (and equity stake) of the third largest shareholder is positively related to firm value. A big question arises: Why two shareholders are tempted to collude while the third one introduces a monitoring role? According to the model, it should be that the second blockholder has either higher relative bargaining power and/ or lower cash flow stake than the third blockholder. Another reasoning is that the marginal gains of cost reduction are going down with more partners in the coalition (recall that c nn > 0), while the decrease in relative bargaining power is more substantial. Therefore, the benefits of adding another coalition partner may be lower than the costs. How to test this prediction? One attempt to do it is to see what are the 26