On ownership structure, investor protection, and company value in the Italian financial market

On ownership structure, investor protection, and company value in the Italian financial market Emilio Barucci Dipartimento di Matemnatica Politecnico di Milano Via Bonardi 9, 20133, Milano, ITALY barucci@mate.polimi.it Silvia Ceccacci Dipartimento di Economia e Istituzioni Università di Roma TorVergata Via Columbia 2-00183, Roma, ITALY silvia.ceccacci@uniroma2.it Corresponding author 1

On ownership structure, investor protection, and company value in the Italian financial market Abstract We evaluate the relationship among ownership structure, corporate governanceinvestor protection and company value in the Italian stock exchange. We show that company value is negatively related to the stake of the largest shareholder and that controlled companies have valuation ratios lower than non controlled-public companies. Companies controlled by another company (a family) are negatively (positively) valued by the market. Recent innovations on regulation of financial markets aiming to improve investor protection are associated with a higher company value and a weaker negative effect of ownership concentration on company value. JEL keywords: G32, G34, G38. Keywords: Ownership Structure, Corporate value, corporate governance, investor protection. 2

1 Introduction We analyze the relationship among ownership structure, corporate governance-investor protection and company value in the Italian stock exchange. We consider a dataset made up of all companies listed in the period 1990 2002. During this period, the Italian financial market went though a deep process of privatizations and modernization of its regulation with an increase of investor protection. On certain aspects, reforms have been effective, i.e., ownership is less concentrated, pyramidal groups are shrinking, but some features are not changing, e.g., there are still very few hostile takeovers, there are few public companies, and institutional investors weakly exert their voice. In our analysis we would like to address two major topics posed by the literature on corporate governance: How does the ownership structure-company value relation work in a financial market with weak investor protection and concentrated ownership? Does the new regulation has increased company value and affected the ownership structure-company value relation? Handling the first topic we relate company value to ownership concentration, voting-cash flow rights wedge, family-non family control, corporate governance. The literature on these topics mainly concerns the US financial markets and the relationship between management s ownership and company value. In a financial market with concentrated ownership, the key issue is not management-ownership separation but the role of large shareholders (insidersoutsiders), their identity, the relationship with the management and the control-ownership wedge. In this setting, as the largest-controlling shareholder s stake increases we observe an incentive effect, i.e., either the largest shareholder strictly monitors the management or they coincide and therefore the management is more focused on the shareholder s value (incentive hypothesis), and two negative effects: managers and/or the largest shareholder may be entrenched, they may adopt conservative policies, they may look for perquisites (entrenchment hypothesis); controlling shareholders may exploit minority shareholders (shareholders exploitation hypothesis). So the ownership concentration-company value relation is ambiguous. The empirical literature on emerging countries and continental Europe has shown that the 3

incentive effect prevails with a positive relation between the stake of the largest shareholder and company value, see [Claessens, et al., 2002, Gorton and Schmid, 2000, Lemmon and Lins, 2002, La Porta et al., 2002]; improper mechanisms to separate ownership from control such as dual shares, pyramidal groups, cross holding negatively affect the company value, see [Claessens, et al., 2002, Gompers, et al., 2004, Lemmon and Lins, 2002]. The literature on these topics for the Italian financial market is quite limited, see [Volpin, 2002, Barontini and Siciliano, 2003, Bianco and Casavo [Volpin, 2002] addresses the topic considering all non financial companies listed in the Italian financial market during the period 1986-1997. He addresses the agency costs reduction and exploitation of minority shareholders tradeoff associated with a large stake by the controlling shareholder showing three main results: Tobin s Q is significantly smaller in firms where the controlling shareholders are among top executives, is significantly larger when control is partially contestable, e.g. company controlled by a shareholders syndicate, is high when the fraction of cash flow rights owned by the controlling shareholder is higher than 50 %. In our analysis we use a richer set of information to investigate the phenomenon and we consider more recent data that allow us to analyze the effect of recent reforms. It is difficult to assess the effect of the regulation on the ownership structure-company value relation. [La Porta et al., 2002, Shleifer and Wolfenzon, 2000] claim (and find some evidence) that an increase of investor protection causes a higher company value, less concentrated ownership, and a weaker incentive effect associated with entrepreneur s cash flow ownership. When investor protection is weak, a positive effect of ownership concentration on company value is also suggested implicitly by the so called political economy theory, see [Roe, 2000]: in a country with weak investor protection it is too costly to separate ownership from control and therefore company value should be increasing in ownership concentration. However, in a financial market with poor investor protection also the entrenchment effect and the risk of exploitation of minority shareholders are strong. Therefore, it is difficult to predict the effect of a new regulation increasing investor protection on the ownership structure-company value relation. Our analysis sheds light on the two topics. Company value decreases in the stake of the largest shareholder, and controlled companies have valuation ratios lower than non controlled-public companies. While outside blockholders do not monitor-improve company 4

value, the institutional investors stake positively affects company value. Dual class shares do not affect the company value. Risk of expropriation inside pyramidal groups is relevant: a company controlled by another limited company is valued negatively by the market. Instead, family firms are valued positively. This dichotomy is also reflected on the effect associated with a coincidence between management and largest shareholder: an executive controlling the company directly (through another limited company) positively (negatively) affects its value. To evaluate the effect of the new regulation, we consider two subsamples (1990-1996 and 1997-2003). We show that the negative effect associated with the stake of the largest shareholder and of blockholders is confirmed only in the first subsample so it seems that the recent regulation has contributed to eliminate the negative relation between ownership concentration and company value. The paper is organized as follows. In Section 2 we present the dataset and some descriptive statistics. In Section 3 we outline the experimental design of our analysis. In Section 4 we provide our empirical results on ownership structure and company value. In Section 5 we evaluate the effect of the new regulation. In Section 6 we assess robustness of the results by providing a fixed effect analysis, an analysis of the full dataset (including financial companies), and a two stages ordinary least squares regression to account for endogeneity between performance and ownership structure. 2 The institutional setting and the dataset The dataset includes all firms listed in the Italian stock exchange during the period 1990-2002. We exclude firms belonging to the Mercato Ristretto (a market for companies with a very small market value) and companies based as exchange market in other countries. As a result, the final sample contains 425 firms (346 non financial companies) and 2.909 yearly-firm observations (2.321 for non financial companies). The sample allows us to analyze the phenomenon in a period characterized by a process of modernization of the regulation of Italian financial markets. Up to 90s, Italian financial markets were characterized by concentrated ownership, weak investor protection and State 5

ownership. Financial intermediaries and institutional investors did not play an active role on monitoring companies and on their corporate governance; companies were mainly under control of the State or of families. The system was quite opaque and the stock market was underdeveloped. The start of the privatization process in 1992 was joined by a process of modernization of the regulation of financial markets. The major goal was to increase investor protection favoring dispersed ownership and to make corporate governance effective. There are three main regulatory pieces that have affected financial markets and corporate governance in the last ten years: the Testo Unico Bancario (TUB) in 1993, the Testo Unico della Finanza (TUF) in 1998, the Preda code of best practice for listed companies in 1999. The privatization process is relevant in our research for two main reasons. First of all, it has reduced the presence of the State in the economy, and in particular banks are not anymore under control of the State; second it has contributed to raise the debate on corporate governance and financial markets in Italy, as a matter of fact political parties discussed for a long time on the best way to privatize companies. The law on privatization in 1994 has introduced some norms to protect shareholders in privatized companies. The TUB allowed commercial banks to detain stakes of non financial companies and vice-versa, under some constraints. The TUF has established a new regulation on takeovers, has introduced a strict regulation on the flow of information to the market and has reinforced shareholders rights, i.e., lower quorums to convene a shareholders meeting, to appeal to the board of auditors, to appeal to the court against CEOs and auditors, to promote a law suit against directors, higher quorums for extraordinary shareholders meetings decisions, minority shareholders can be represented inside the board of auditors. According to the index on investor protection computed by [La Porta et al., 1998], thanks to the TUF Italy has gone from 1 to 5 over 6. Finally, the code of best practice of the Italian stock exchange clearly stresses that the management of the company should be focused on the shareholder s value, recommends executive remuneration be market value sensitive and introduces committees inside the board of directors as in the AngloSaxon tradition. As determinants of company value, we consider three sets of data: balance sheet data, ownership structure data, and corporate governance features. Our main data sources are Il Taccuino dell Azionista, an annual publication edited by Il Sole 24 Ore, Il Calepino and 6

Indici e Dati, which are published yearly by Mediobanca. Accounting data are from Il Calepino. Information on ownership structure is obtained from Il Taccuino dell Azionista and from CONSOB official releases. Information on market capitalization and gross dividends is drawn from Indici e Dati. In Table 1 we provide a brief description of the variables employed in our analysis. In Table 7 we provide descriptive statistics of the dataset (including financial companies). The table reports the mean, median, standard deviation, minimum and maximum, the number of observations and the mean for the non financial companies subsample. For variables involving debt we restrict our attention to non financial companies. In Table 3 we provide descriptive statistics for variables used in our analysis by identity of the largest shareholder (limited company, family, bank, insurance company, foundation, State) and by control type (absolute majority, de facto, shareholders coalition, non controlled). In Table 4 we provide descriptive statistics for two subsamples (1990-1996) and (1997-2002) and a test on mean and median difference. Considering the full sample we would like to point out some facts on a pure descriptive basis, see Table 7. Ownership is highly concentrated: the average stake detained by the largest shareholder is above 50%, 64% of companies are controlled by a shareholder with a stake above 50% and only 9% are non controlled, i.e., public companies. Non financial companies have an ownership structure more concentrated than financial companies. Note that the management is strictly related to the largest-controlling shareholder: in 36-40% of companies the ceo or the chairman coincide with the largest shareholder and the ceo (chairman) owns on average 18% (20%) of the shares of the company. The phenomenon is stronger when only non financial companies are considered. Notice the weak role played by institutional investors, as a whole they own on average a 13% stake, but restricting our attention to non financial companies the average stake is only 6.6%, moreover observe that investment funds (mutual funds, private equity funds) as a whole on average hold less than 2%. Obviously these figures are biased by the fact that shareholders with a small stake are not reported by our sources. The Italian financial market is not bank based as the German one, i.e., only 8% of listed companies are controlled by banks but excluding financial companies the fraction decreases to 2%. 7

As in other countries with concentrated ownership, separation of ownership from control is obtained via non voting shares, cross shareholding, pyramidal groups. Many companies are characterized by non voting shares (35%), but the fraction of non voting (or with limited rights) shares on average is small (6%). Pyramidal groups are relevant, [Bianchi et al., 2001] analyzing listed companies in 1996 obtain a voting rights over cash flow rights ratio of 2.4, on the full sample we obtain a similar figure: 2.2 (2.3 for non financial companies). 52% of companies are controlled by another limited company and considering also limited liability companies we have that the fraction goes up to 61%. It is difficult to evaluate the degree of cross holding among listed companies, it is easier to observe the degree of interlocking of directors. Interlocking directorates are widely diffused: 80% of companies have a director sitting on the board of another company, 57% of companies have an executive sitting on the board of another listed company, restricting our attention to non financial companies we observe that 1/3 of the companies are interlocked with banks. In Table 3 we provide descriptive statistics discriminating for the identity of the largest shareholder (Spa, Family, Bank, Insurance, Foundation, State) and control type (cont > 50, cont < 50, coalition, noncontr). Largest shareholder s identity and control type affect valuation ratios. Mvbv and qtobin are the highest in companies controlled by a family, low valuation ratios are associated with companies controlled by the State and by another limited company. As far as control type is concerned, we observe that company value decreases as control of the company becomes stricter; control types are ranked as follows: noncontr, coalition, cont < 50, cont > 50. Financial market regulation has changed in the 90s in the direction of an increase of investor protection and of a more effective corporate governance. Some features of ownership structure and corporate governance are changing, see Table 4. First of all we notice that ownership is less concentrated in recent years: the average stake of the largest shareholder has gone from 57% to 49%, while the stake of outside relevant shareholders is remaining almost constant, the average stake of small shareholders has gone up from 30% to 37% on average. The difference in the mean and in the median of the stake of the largest shareholder and of small shareholders is statistically significant. As a consequence, the fraction of companies controlled by absolute majority has decreased, and the fraction of non 8

controlled companies and of companies controlled by a shareholders agreement has gone up. Improper forms of separation between ownership and control (pyramidal groups, interlocked directors, dual shares) are less frequent in the 1997-2002 subsample than in the 1990-1996 subsample. However, cash flow rights of the largest shareholder are constant in the two subsamples. Firm value has gone up: mvbv and qtobin are higher in the 1997-2002 sample than in the 1990-1996 sample. The difference is statistically significant. This evidence confirms the main claims of the law and finance thesis (law matters) and of the financial system convergence hypothesis, see [La Porta et al., 1998, Shleifer and Wolfenzon, 2000, La Porta et al., 2002, Bebchuk, 1999]. As a matter of fact, increasing investor protection, ownership becomes less concentrated, improper mechanisms to separate ownership from control are less frequent and firm value goes up. This evidence confirms [Doidge, et al., 2004] showing that foreign companies listing also in US are valued more by the market, and [Doidge, 2001] showing that foreign companies listed in US are characterized by a decrease in ownership concentration. However, we interpret these results with caution, overvaluation of financial markets during the internet bubble may distort the result on company value, and the above facts can be due to other structural breaks occurred during the last ten years in Italy: globalization, increase of competitive pressure, adhesion to the European Union, see [Rajan and Zingales, 2002, Bris et al., 2003] and [Morck et al., 2000] for this type of interpretation on Canada experience. Moreover, ownership concentration decreased until 1998, but then in recent years no further decrease is observed and we are far away from a public company system. 3 Review of the literature: variables and hypotheses To shed light on the relation among ownership structure, corporate governance, investor protection and corporate value, we consider ownership structure, corporate governance features, and balance sheet data. The relation between company value and ownership structure is complex because more than one dimension is relevant. While in a public company system, the major point is provided by the classical agency conflict between managers and shareholders and therefore 9

the key determinant is the stake of the company detained by managers, in a system with concentrated ownership and weak investor protection we have to consider not only the stake detained by managers but also the shareholder s identity, the stake of insiders-controlling shareholders and the stake of outside shareholders, management entrenchment, cash flowvoting rights wedge. The key element is provided by the stake of the largest-controlling shareholder. In what follows, we concentrate on four different features of ownership structure-corporate governance: ownership concentration, cash flow-voting rights wedge, family-non family firms, corporate governance. 1) Ownership concentration Ownership structure is introduced considering the stake of the largest shareholder or of shareholders participating to the coalition controlling the company (lrgshrs), and the stake of outside blockholders (outsideshr), i.e., the stake detained by shareholders with a relevant stake unrelated to the largest-controlling shareholder. We also consider the stake detained by institutional investors (totinst). When the largest shareholder and/or executives detain a large stake, there is a weak ownership-control separation and therefore the management of the company should be more focused on the shareholder s value with a positive effect on corporate value (incentive hypothesis). However, a large stake by the largest shareholder and or executives is not purely beneficial, it may cause negative effects on company value. The literature has identified two main (potential) negative effects. A risk averse shareholder-executive detaining a large stake may be inclined to look for perquisites, to adopt a conservative policy and therefore there is a potential loss for all shareholders due to non optimal management (entrenchment hypothesis), see [Huddart, 1993, Admati et al., 1994]. Moreover, a large shareholder may exploit minority shareholders (shareholders exploitation hypothesis), e.g. see [Burkart and Panunzi, 2001]. Combination of these hypotheses may lead to a nonlinear relation between firm performance and ownership of the largest shareholder-executives. The above hypotheses assume the ownership structure as an exogenous datum, however [Demsetz, 1983, Demsetz and Lehn, 1985] point out that the ownership structure and firm performance are endogenously determined, each firm s ownership structure will be optimal 10

for that firm. According to this view, there will be no relation between ownership structure and firm value. Empirical evidence on the relation between ownership structure and firm value is driven by the literature on the US experience and therefore is concentrated on the relation between managers ownership and firm value. [Morck, et al., 1988, McConnell and Servaes, 1990, Holderness, et al., 1999, Hermalin and Weisbach, 1991] find evidence of a nonlinear relation with respect to managers ownership: firm performance is positively related to managerial ownership at low levels, above a certain percentage of shares the relation turns out to be negative. So according to this analysis, increasing managers shareholding up to a certain threshold, the incentive hypothesis dominates the entrenchment hypothesis, the opposite effect holds above the threshold. [Demsetz and Lehn, 1985, Agrawal and Knoeber, 1996, Cho, 1998, Himmelberg, et al., 1999, Demsetz and Villalonga, 2001, Loderer and Martin, 1997] confirm the above curvilinear relation through simple ordinary least squares regressions, but show that company value is not related to managerial ownership when we allow for unobserved firm heterogeneity and endogeneity of managerial ownership and firm performance. There is a small literature on other countries. As far as ownership of the largest shareholder is concerned, [Claessens, et al., 2002, Lemmon and Lins, 2002, La Porta et al., 2002] find a positive effect in emerging markets. [Gorton and Schmid, 2000] show that in Germany firm performance improves in ownership concentration. Large outside blockholders have enough incentives to monitor the management-performance of the company and therefore a positive effect on company value is expected to be associated with a large stake of outside shareholders (outsideshr) and of institutional investors (totinst) in non financial companies, i.e., when they are not the controlling shareholder (monitoring hypothesis), see [Maug, 1998, Kahn and Winton, 1998]. Empirical evidence on this point is weak. [McConnell and Servaes, 1990] find a positive relation between firm value and institutional investors stake but no significant relation with the stake detained by outside blockholders. [Mehran, 1995, Agrawal and Knoeber, 1996] find no effect for both classes of shareholders. [Gorton and Schmid, 2000], addressing the role of German banks on firm performance, evaluate that bank s monitoring significantly improves firm performance. [Volpin, 2002], analyzing the Italian financial market, shows that a large minority 11

shareholder has no effect on company value. Entrenchment of executives is further evaluated considering the dummy variable (ownceochair) assuming value equal to one if a ceo or the chairman of the board of directors is the largest/controlling shareholder of the company. According to the incentive hypothesis, the coefficient associated with this variable should be positive and statistically significant; on the other hand if executives are entrenched then the effect on company value should be negative (entrenchment hypothesis). 2) Cash flow-voting rights wedge Misalignment between cash flow and voting rights and business group affiliation may cause large agency costs with a negative effect on company value, see [Bebchuk, 1999, Bebchuk, et al., 2000, Morck, et al., 2004]. In what follows, cash flow rights and voting rights misalignment is analyzed through several indicators. First of all we consider a dummy variable assuming value equal to one when the company is controlled by another listed/unlisted limited company (1 spa ). This dummy variable represents affiliation to a business group; when it is zero the ratio voting rights/cash flow rights of the controlling shareholder is equal to one, when the company is controlled by another company, the ratio can be higher than one and the company belongs to a pyramidal group. To quantify the wedge between voting rights and cash flow rights, we consider the ratio voting rights/cash flow rights of the largest shareholder (vrgt/cfrgt), where cash flow rights are determined by multiplying voting rights along the controlling path. Business group affiliation and a wedge between voting and cash flow rights represent improper mechanisms to separate ownership and control; as a matter of fact, the largest shareholder appreciates only a small fraction of the costs associated with choices not directed towards the shareholders value and therefore he may be tempted to exploit minority shareholders. A company affiliated with a business group and/or with a vrgt/cfrgt higher than one is less likely to be concentrated on the shareholders value. We consider cash flow rights detained by the largest shareholder (lrgshcf ) and the difference between voting rights and cash flow rights (diffcash=lrgshrs-lrgshcf). lrgshcf is defined as the product of lrgshrs for the fraction of voting rights along the controlling path, the number has been corrected for shares with non voting rights (azioni risparmio) assuming as in [Volpin, 2002] that the controlling shareholder owns none of them: lrgshcf=lrgshrs/(vrgt/cfrgt)*fracord. Accord- 12

ing to our interpretation, we expect 1 spa and diffcash to negatively affect company value. Negative effects associated with ownership concentration should be weaker when lrgshcf is considered instead of lrgshrs. To retain control of a company, the largest shareholder may decide to issue shares with limited voting rights and/or to form a coalition with other shareholders. These routes provide improper forms of separation between ownership and control with a wedge between voting and cash flow rights, i.e., as far as control is concerned it is the composition inside the coalition to be relevant and the fraction of shares with voting rights. To take into account this feature, we consider a dummy variable assuming value equal to one when the company is controlled by a shareholders coalition (coalition), and a dummy variable for companies with limited voting rights shares (risp). Companies controlled by a shareholders agreement, issuing shares with limited voting rights are less likely to be concentrated on the shareholders value and therefore we expect a negative effect on company value be associated with coalition, risp. As far as voting rights-cash flow rights misalignment is concerned, the literature on company value and ownership structure has shown that when a management s group voting rights exceed its cash flow rights firm values are lower, see [Lemmon and Lins, 2002, Lins, 2002, Baek, et al., 2004], and that firm value is increasing in cash flow rights and decreasing in voting rights or in the control-cash flow rights wedge, see [Claessens, et al., 2002, Gompers, et al., 2004]. [Claessens, et al., 2002] show that classical tools used to separate control rights from cash flow rights, i.e., pyramidal group, cross holding and dual class shares, negatively affect the firm value. As far as business group affiliation is concerned, a positive effect on company value has been detected in emerging markets, see [Khanna and Palepu, 2000, Khanna and Rivkin, 2001]. 3) Family firms Family firms are characterized by some peculiarities that may affect company valueperformance. Companies held directly by an individual or a family (not through a limited company) are characterized by no wedge between voting rights and cash flow rights and therefore agency costs are low, moreover in many cases the founder (with his skills as entrepreneur) either is the chairman or the ceo of the company. These features should positively 13

affect company value. On the other hand, in many cases heirs are not as good as the founder and therefore there is a succession problem inside the family and a family holding a large stake may be inclined to adopt conservative policies. Evidence is mixed: while [Morck et al., 2000] observe that heir controlled companies in Canada have a poor performance, [Anderson and Reeb, 2003, Sraer and Thesmar, 2004] show that companies held by an individual or a family are characterized by a higher market value/performance. 4) Corporate governance We take into account some governance features introducing the board size (boardzise), interlocking of an executive of the company, i.e., an executive (ceo, chairman, vice-chairman) of the company sits on the board of another listed company (interexec). [Barucci, et al., 2005, Fich and White, 2004] have shown that interlocking of directors is not directed towards shareholder s value maximization so we expect a negative relation between a company having executives sitting on the boards of other companies and firm value. As far the relation between board size and company value is concerned, there is a wide literature showing that a company with a large board is characterized by a low market value, see [Yermack, 1996], the rationale for this result is that a large board exerts a weak monitoring role on the management of the company. In our analysis we control for balance sheet data. As in large part of the literature we control for company size measured by the logarithm of total assets (lntotass), for cash flow over total assets ratio (cashta) and leverage (leverage). 4 Ownership structure and company value To investigate the phenomenon we proceed in two steps. First we consider a reference model made up only of ownership structure variables as exogenous variables, see Tables 5 and 6, then we consider more refined models including variables addressing the largest shareholder s identity, cash flow rights and voting rights wedge and corporate governance features, see Tables 7 and 8. In the reference model, company value ratio for non financial companies measured as qtobin is explained through ownership structure data: the stake of the largest shareholder- 14

coalition of shareholders controlling the company (lrgshrs), the stake of outside blockholders (outsideshr), and the stake of institutional investors (totinst). We control for the size of the company (lntotass) and the size of the board of directors (boardsize). As pointed out above, the relation between the stake of the largest shareholder and company value may be nonlinear, to address this point we consider the square of the stake of the largest shareholdercontrolling coalition (lrgshrs 2 ). Columns (1)-(4) of Tables 5 and 6 only consider lrgshrs, columns (5)-(8) also consider lrgshrs 2. As pointed out in many papers, e.g. see [Demsetz and Lehn, 1985, Himmelberg, et al., 1999], the relation between ownership structure and firm value may be affected by unobserved firm heterogeneity, i.e., if some determinants of company value are also determinants of ownership structure then results showing existence of a relation between ownership and company value may be spurious. To address this point, we provide five different econometric specifications of the reference model: pooled regression (no firm, industry, year, and panel effects), pooled regression with industry effect, panel analysis with fixed effects (firm effects), panel analysis with random effects and year and industry effects. The reference model is also estimated with industry and year effects and results are presented in column (1) and (6) of Table 8. As far as choice between random effects and fixed effect is concerned, we have performed an Hausman test that has rejected the null hypothesis of no correlation of random effects with regressors, therefore in the analysis below we do not provide random effect regressions. If unobserved firm heterogeneity affects the ownership structure-company value relationship, then we should obtain different results depending on the econometric specification, and the relation should become unsignificant when firm or industry effects are inserted as it is shown in [Himmelberg, et al., 1999]. However, it has been observed that as ownership structure changes slowly from year to year within a company, a fixed effect panel analysis may not detect the effect of ownership on company value, i.e., ownership structure effects may be fully captured by firm specific variables, see [Zhou, 2001]. To address these drawbacks, in this section we concentrate our attention on regressions with industry and year effects, then in Section 6 we will provide fixed effect regressions. In Table 5 we show that there is a negative relation between the stake of the largest shareholder and company value. The relation is statistically significant in all the four econometric 15

specifications. There is no evidence of nonlinearity, i.e., the linear-quadratic specification for lrgshrs is not statistically significant. As far as monitoring by outside-institutional investors hypothesis is concerned, the evidence is mixed. While the coefficient associated with the stake detained by institutional investors is positive, the relation with the stake of outside blockholders is negative. As in large part of the literature and in [Volpin, 2002] for the Italian financial market, it turns out that large companies are characterized by a lower valuation ratio. There is also some evidence that companies with a large board are characterized by a low market value. As expected, significance of all variables is weak when a fixed effect model is estimated. In Table 3 we have shown that controlled companies (by absolute majority or de facto) are characterized by valuation ratios lower than those of non controlled companies and of companies controlled by a shareholders coalition. To check for the statistical significance of this relation, in Table 6 we estimate the reference model eliminating the stake of the largest shareholder and introducing dummy variables for different types of control. We drop the dummy variable for non controlled companies and we consider dummy variables for companies controlled by a shareholders coalition (coalition), de facto (cont < 50) and by absolute majority (cont > 50). In these models, the constant stands for non controlled companies. The analysis confirms results established considering the stake of the largest shareholder: non controlled companies have a valuation ratio higher than controlled companies (de facto or by absolute majority). Valuation ratios of companies controlled by a coalition of shareholders are not different in a statistically significant measure to those of non controlled companies. This result confirms the positive effect associated with a shareholder agreement to control the company as detected in [Volpin, 2002]. Note that control dummy variables are not statistically significant when fixed effects are considered, this evidence was widely expected: control type does not change over time a lot and therefore the control type is almost fully reflected by variables associated with companies. Results on the effects due to outside shareholders and institutional investors confirm those obtained in the previous analysis. As a second step we introduce governance features considering regressions with year and industry effects. In Table 7 we insert variables addressing the largest shareholder identity: 16

the largest shareholder is another limited company, i.e., the company belongs to a pyramidal group (1 spa ), a family (1 family ), the State (1 State ), a foreign shareholder (1 foreign ). Risk of exploitation is high when the largest shareholder is a company compared to the case of a company controlled by a family, therefore we expect a positive coefficient for 1 family and a negative coefficient for 1 spa. To evaluate the entrenchment effect, we introduce the dummy variable ownceochair which assumes value equal to one if an executive of the company is also the largest shareholder. The entrenchment effect hypothesis suggests a negative coefficient for this variable. However, the literature on family firms and the discussion provided in the last section suggest that coincidence of executives with the largest shareholder may be an important resource in case of a family firm, while the effect is likely to be negative in case of an executive who owns the company through another limited company, i.e., risk of exploitation of minority shareholders is high. To address this point, we also consider the case of an executive belonging to the controlling family (1 family ownceochair) and of an executive controlling the company through a limited company (1 spa ownceochair). To cope with the evidence documented in Table 6 on the effect associated with non controlled companies-companies controlled by a shareholders agreement we insert dummy variables for companies controlled by a coalition of shareholders (coalition) and non controlled companies (noncontr). For both variables we expect a positive coefficient. Governance features and dual class shares are introduced considering a dummy variable assuming value equal to 1 if executives of the company sit on the board of another listed company (interexec) and a dummy variable for a company that has issued shares with limited voting rights (risp). In specifications not shown below we have considered interlocking of a generic director, interlocking with a bank, the voting rights/cash flow rights ratio, the number of layers in the business group. Results are quite similar to those presented below. Results on the stake of the largest shareholder-coalition controlling the company confirm the negative relation established above. Statistical significance depends on the presence of other variables, when corporate governance variables and variables addressing the type of control are inserted, the coefficient of lrgshrs is not statistically significant. It is confirmed that the stake of institutional investors plays a positive effect and that the stake of outside blockholders negatively affects company value (with weak statistical significance). 17

Company value depends on the identity of the largest shareholder: companies belonging to a business group, i.e., the largest shareholder is another limited company, have a lower valuation ratio, see column (1). As in [Anderson and Reeb, 2003] for the US financial market, companies controlled directly by a family are valued positively by the market, see column (2). There is no sign of executive entrenchment, i.e., the dummy variable assuming value equal to one if an executive of the company is also the largest-controlling shareholder is not statistically significant, see column (3). A similar result is obtained when we insert a dummy variable for coincidence of the ceo of the company and of the largest shareholder. However the dichotomy between companies controlled by another limited company and companies controlled directly by a family influences the effect of executive entrenchment. While the effect of executive entrenchment (ownceochair) is not statistically significant, the effect of an executive belonging to the controlling family is positive, see column (6), and the effect of an executive controlling the company through another listed company is negative, see column (5). Both variables are statistically significant. A negative effect is associated with a foreign controlling shareholder and with a company controlled by the State. Dummy variables for non controlled companies or controlled by a shareholders coalition are not statistically significant. While there is no statistically significant effect associated with dual class shares, companies with an executive sitting on a board of directors of another listed company are negatively valued by the market. This result confirms that interlocking of directors is not beneficial to company value. As in large part of the literature we observe that company value is negatively related to leverage. In Table 8 we investigate how the wedge between cash-flow rights and voting rights affects the relationship between the stake of the largest shareholder and company value, see column (1)-(7) and we further investigate the role played by different forms of control by interacting the stake of the largest shareholder with its identity: a limited company (1 spa ), a family (1 family ) and with ownceochair, see column (8)-(10). On the wedge between cash flow and voting rights of the largest shareholder, the literature has established that company value is increasing in cash flow rights, see [Claessens, et al., 2002, Gompers, et al., 2004, Volpin, 2002], decreasing in the voting rights/cash flow rights wedge, see [Claessens, et al., 2002], increasing in cash flow rights and decreasing in voting rights 18

when both variables are inserted, see [Gompers, et al., 2004]. We do not provide evidence confirming these results. We confirm that company value is decreasing in the stake of the largest shareholder/coalition controlling the company, also the coefficient for cash flow rights is negative although non statistically significant, see column (2). When both variables are inserted, coefficients are negative but are not statistically significant, see column (3). There is no evidence of a negative effect associated with a wedge between voting rights and cash flow rights and of nonlinearity, see column (4), (5) and (6). Compared with the negative effect associated with the stake of the largest shareholder, non significance of cash flow rights of the largest shareholder means that the incentive effect is more relevant when cash flow rights increase than in the case of voting rights. Confirming the evidence shown in Table 7, we observe that the stake of the largest shareholder interacted with the dummy variable representing a family as controlling shareholder positively affects company value, see column (8), instead the coefficient associated with the stake of the largest shareholder interacted with control by another limited company is negative with a weak statistical significance, see column (10). Therefore, it seems that the incentive effect works only when there is not a business group affiliation, i.e., the controlling shareholder directly detains the shares. No effect is associated with the stake of the largest shareholder when he is also an executive of the company. On the whole our analysis has shown that the company price incorporates the minority shareholders risk of exploitation associated with business group affiliation, but no effect is associated with dual shares. Risk of exploitation of minority shareholders prevails over the incentive effect for the largest shareholder, a company controlled with a large stake is valued negatively by the market. Outside monitoring is ineffective with a significant exception provided by institutional investors. Entrenchment of executives negatively affects company value only if the company is controlled by another company, if the largest shareholder directly detains shares of the company the effect is positive. 19

5 Investor protection and company value As pointed out in Section 2, during the period considered in our analysis, we have assisted to changes in the regulation of financial markets and corporate governance in the direction of improving investor protection. Although the TUF in 1998 contains the main innovations, during the 90s we have assisted to a continuum of innovations going in the direction of an increase of investor protection and of a more effective corporate governance. To check the effectiveness of the regulation for financial markets put forward during the 90s we provide an analysis dividing the sample in two subsamples (1990-1996 and 1997-2002). We provide regressions for the two subsamples separately and regressions of the main models with variables interacted with 1 97, a variable assuming value equal to 1 when data refer to the 1997-2002 period. In Table 9 we show the results of the regressions for the two subsamples. Column (1)-(5) refer to the 1990-1996 subsample, column (6)-(10) refer to the 1997-2002. Results provide evidence in favor of the hypothesis that regulation has become more effective during the sample period and that ownership structure as a corporate governance mechanism works more properly. As a matter of fact, the stake of the largest shareholder-coalition controlling the company and the stake of outside blockholders negatively affect company value only in the 1990-1996 subsample. The positive effect associated with the stake of institutional investors is detected in both subsamples. Results on the effects associated with control by another limited company and by a family are confirmed in both subsamples. Although the results on the stake of the largest shareholder do not agree with the analysis provided in [La Porta et al., 2002] on the relation between investor protection and ownership concentration-company value, results agree with the centrality of investor protection and the convergence hypothesis: increasing investor protection, risk of exploitation/entrenchment effects are weaker and the incentive effect stronger as the stake of the largest shareholder increases. 20

6 Robustness analysis To evaluate robustness of our results, we investigate three directions. First of all we provide results of models analyzed in Table 5 for the dataset including all companies and using as valuation ratio the market value over the book value (mvbv), results are shown in Table 10. Then in Table 11 for the non financial companies sample we provide fixed effect regressions of models presented in Table 7. Finally we address endogeneity between mvbv and lrgshrs by providing two stage ordinary least regressions considering the number of shareholders as instrumental variable, see Table 12. As far as the analysis on mvbv for the full dataset is concerned, company value is decreasing in the stake of the largest shareholder and in the stake of outside blockholders (statistical significance is weak). No statistically significant effect is associated with the stake of institutional investors. The no effect associated with the stake of institutional investors is probably due to the fact that in many financial companies institutional investors are not outside blockholders but are the largest/controlling shareholders, and therefore they do not monitor the company performance-management with a positive effect on company value. When fixed effect regressions are carried out, it is only confirmed that the stake of the largest shareholder negatively affects the company value; the stake of outside blockholders and of institutional investors are not statistically significant. Governance features are not statistically significant. Notice that dual class shares (azioni risparmio) negatively affect company value. As stressed by [Demsetz and Lehn, 1985, Demsetz and Villalonga, 2001, Himmelberg, et al., 1999], the relationship between ownership and company value suffers of endogeneity problems, i.e. they may be endogenously determined. In particular, company value and the stake of the largest shareholder may be endogenously determined. To address this problem, we provide a two stage ordinary least squares regression of some models analyzed above considering the number of shareholders with a relevant stake as instrumental variable, as a matter of fact this variable is negatively correlated with the stake of the largest shareholder and is uncorrelated with company value. To check for endogeneity, we have performed an Hausman test 21

on the endogeneity of lrgshrs and qtobin; considering models with dummy variables for years and industries we do not detect endogeneity, instead when we remove dummy for years and industry sectors we observe that there is some evidence of endogeneity. In Table 12 we provide estimates of various models. The estimate of the reference model, column (1), confirms results established above: company value is negatively affected by the stake of the largest shareholder, by the stake of outside blockholders and is positively affected by the stake of institutional investors. However, evidence on the effect of institutional investors is weak. It is confirmed that affiliation to a business group as well as executive-largest shareholder entrenchment negatively affect company value. Surprisingly, family control turns out to have a negative effect. 7 Conclusions Italian financial markets in the 90s provide an interesting case study to understand the company value-ownership structure relation: investor protection standards are low, ownership is concentrated, and a deep process of modernization of the financial markets regulation has been carried out to improve investor protection. As far as ownership concentration is concerned, we have shown that there is no monitoring by outside shareholders, entrenchment-risk of exploitation of minority shareholders prevail over the incentive effect for the largest/controlling shareholder. As a consequence, company value is decreasing in the stake of the largest shareholder and of outside blockholders; only institutional investors enhance company value. Risk of exploitation of minority shareholders is high in a company controlled by another company but not in family companies, i.e., companies where the largest shareholder directly detains the shares. While affiliation to a business group negatively affects company value, family firms are valued positively. The dichotomy also influences the relation between executive entrenchment and company value. Small boards of directors and boards with non interlocked executives positively affect company value. Dual class shares do not affect company performance. This evidence suggests that in the Italian financial market minority shareholders have been mainly exploited inside a business group and not through shares with limited voting rights. 22