Tender Offers versus Block Trades: Empirical Evidence

Transcription

1 Tender Offers versus Block Trades: Empirical Evidence MARTIN HOLMÉN a and EUGENE NIVOROZHKIN b Forthcoming Managerial and Decision Economics Abstract In this paper, we test whether the determinants of block trade and non-partial tender offer probabilities differ and whether the relative magnitude of security and private benefits can explain the choice of transfer mode. We investigate the Swedish market for corporate control. The results emphasize the importance of investigating block trades and tender offers as two competing events. The proxies for private benefits of control, small controling voting blocks and separation of voting rights from cash flow rights are positively related to the likelihood of a block trade but negatively related to the likelihood of a non-partial tender offer. Our results suggest that separation of voting rights from cash flow rights might limit the efficiency of the market for corporate control. The prevalence of block trades in the presence of greater private benefits of control highlights the disadvantages of this control transfer mode in terms of incentive alignment between the buyer and the remaining dispersed shareholders. Keywords: Tender offers; Block trades; Ownership structure; Private benefits of control; Dual class shares; Stock pyramids JEL Classification: G34; G32 The authors would like to thank Tim Broadhurst, Cynthia Campbell, Ettore Croci, Thomas Lagoarde-Segot, Fredrik Lindqvist, and Richard Sweeney for valuable comments. Comments from participants at the EFMA symposium on Corporate Governance in Milan, the FMA meeting in Barcelona, the 2007 SNEE conference, the 2008 INFINITI Conference on International Finance at Trinity College Dublin, and a seminar at the Research Institute of Industrial Economics are also acknowledged. Financial Support from Jan Wallander s and Tom Hedelius Research Foundation and VINNOVA is gratefully acknowledged. a Corresponding author: Centre for Finance and Department of Economics, University of Gothenburg, Box 640, Gothenburg, Sweden, and Department of Finance and Statistics, Hanken School of Economics, Arkadiagatan 22, Helsinki, Finland. Martin.Holmen@cff.gu.se ; Phone: b SSEES, University College London, Gower Street, London WC1E 6BT, UK

2 Tender Offers versus Block Trades: Empirical Evidence Abstract In this paper, we test whether the determinants of block trade and non-partial tender offer probabilities differ and whether the relative magnitude of security and private benefits can explain the choice of transfer mode. We investigate the Swedish market for corporate control. The results emphasize the importance of investigating block trades and tender offers as two competing events. The proxies for private benefits of control, small controling voting blocks and separation of voting rights from cash flow rights, are positively related to the likelihood of a block trade but negatively related to the likelihood of a non-partial tender offer. Our results suggest that separation of voting rights from cash flow rights might limit the efficiency of the market for corporate control. The prevalence of block trades in the presence of greater private benefits of control highlights the disadvantages of this control transfer mode in terms of incentive alignment between the buyer and the remaining dispersed shareholders. Keywords: Tender offers; Block trades; Ownership structure; Private benefits of control; Dual class shares; Stock pyramids JEL Classification: G34; G32

3 1. Introduction This paper addresses two questions. First, do the determinants of block trade and non-partial tender offer probabilities differ? And second, does the controlling shareholder s relative magnitude of security benefits and pecuniary private benefits explain the choice of transfer mode? 1 Security and private benefits should be substitutes. We hypothesize that the relative magnitude of private benefits proportionately increases with a reduction in a controlling ownership stake and with an increase in a wedge between voting rights and cash flow rights. A wedge between voting rights and cash flow rights can be achieved by the use of dual class shares and stock pyramids. A greater proportion of private benefits would increase the incentives for preserving these benefits and would therefore increase the probability of a block trade transfer mode (Burkart, Gromb, and Panunzi, 2000). Given that private benefits tend to be internalised in a non-partial tender offer, we expect that the likelihood of a tender offer would tend to increase with a reduction in the factors used as proxies for private benefits of control. Researchers in corporate finance have given considerable attention to issues affecting transfer of control to the most efficient user of corporate resources. 2 The market for corporate control represents the institutional framework facilitating control changes. The typical modes of control transfers in publicly traded companies are block trades and nonpartial tender offers. 3 Until recently, the finance literature tended to treat the two transfer modes as equivalent as long as a block trade or tender offer resulted in the same party having control (Burkart et al., 2000). Burkart et al. (2000) argue theoretically that the incentive effect of the controling party s final holding has largely been overlooked in the control transfer literature (e.g. Grossman and Hart, 1988; Bebchuk, 1994; Zingales, 1995) and is likely to 1 With security benefits, we mean dividends and capital gains that accrue to all shareholders. 2 See e.g. Grossman and Hart (1980) and Shleifer and Vishny (1986). 3 Control changes through a block trade tend to be more common than tender offers. In 2000, only 27 percent of the changes in control of US firms were created by tender offers (Schmid, 2002). 1

4 have a significant effect on the choice of transfer mode. One suggested factor affecting this choice is interdependence of security benefits and private benefits attached to a controling ownership stake. Bethel, Liebeskind, and Opler (1998) document that the market for partial corporate control (block trades) plays an important role in limiting the agency costs in US corporations. However, little is known about the markets of corporate control outside the Anglo-Saxon countries. 4 Denis and McConnell (2003) point out that takeover activity does not appear to be an important governance mechanism throughout the world. Several papers have documented significant differences between the Anglo-Saxon countries and the rest of the world in terms of the typical corporate ownership structure (see e.g. La Porta et al., 1999). However, Franks and Mayer (2001) argue that the most important differences between Anglo-Saxon countries and continental Europe (Germany) are in terms of dynamic relations involving transfers of control. For example, the gains from German block trades are small and appear to solely accrue to large blockholders. In this paper, we investigate the Swedish market for corporate control. Exploring the effect of private benefits of control on the choice of a control transfer mode in the context of the Swedish equity market offers a couple of advantages. First, Sweden has an extreme separation of ownership and control. It ranks #1 in terms of the use of dual class shares and #2 in terms of the frequency of pyramid structures (after Belgium) (La Porta, Lopez-de-Silanes, and Shleifer, 1999). 5 Thus, we have a simple measure of the incentives to extract private benefits from the corporation, i.e. the separation between ownership and control due to dual class shares and pyramids. This measure varies both between and within firms over time and it has also been used in earlier literature (see e.g. Claessens, Djankov, Fan and Lang, 2002; 4 The value of European mergers and acquisitions exceeded that in America for the first time in 2006 (Economist, Jan ). 5 Typically, Swedish firms issue A and B shares. The A-shares are one-share one-vote while the B shares are one-share 0.1 votes. 2

5 Faccio, Lang and Young, 2001; Giannetti and Simonov, 2006). Second, the high frequency of dual class shares and pyramids allows us to explore the relation between the market for corporate control and these instruments per se. In particular, we are able to disentangle the effect of dual class shares from the pyramid effect on the market for corporate control. Theoretically, dual class shares and pyramids are often considered to be perfect substitutes. Most of the existing financial economics literature on the likelihood of control changes relies on binary choice models. We adopt another approach and estimate the hazard rate of a non-partial tender offer and a block trade, respectively. 6 We use panel data where a majority of the firms are not subject to a non-partial tender offer or involved in a block trade during our sampling period. The hazard function approach allows us to investigate whether, given that a firm has not been subject to a non-partial tender offer (experienced a block trade) up to a certain point, changes in particular characteristics (e.g. ownership) of the firm will lead to a non-partial tender offer (block trade) event. In our main tests, we use an unbalanced panel of 195 large Swedish non-financial firms listed on the Stockholm Stock Exchange The sample contains 1461 firm years. On average, about 70 percent of the Swedish stock market capitalization are included in the sample each year. The sample includes 28 successful non-partial tender offers and 62 block trades. The evidence presented in this paper points out the importance of studying the control transfers through non-partial tender offers and block trades as two distinctive events, conditional on the regulatory environment. In accordance with our expectations, we find some important differences in factors determining the likelihood of the two events. Combining 6 Helwege, Pirinsky, and Stulz (2007) estimate hazard functions when estimating the likelihood of a firm becoming widely held. 7 The sample selection stops in 1998 since Sweden introduced a mandatory bid rule in A mandatory bid rule will trigger a non-partial tender offer if the traded block is sufficiently large. Thus, the analysis of the determinants of block trades and tender offers, respectively, becomes ambiguous with a mandatory bid rule in place. However, as a robustness test, we examine the likelihood of block trades in the period. 3

6 these events in a single model would lead to obscuring the information conveyed by a choice of transfer mode (Burkart et al., 2000). An increase in the largest shareholder s ownership stake tends to increase (decrease) the likelihood of a non-partial tender offer (block trade). Conversely, the use of dual class shares by the largest shareholder decreases (increases) the likelihood of a non-partial tender offer (block trade). Private benefits are proportionately increasing with a decrease in a controlling ownership stake and an increased separation between cash flow and control rights. When the relative magnitude of private benefits of control is large, the bidder and the incumbent in the target negotiate a control block transfer since the private benefits will be internalized in a tender offer. To our knowledge, this paper is the first empirical attempt at illustrating the critical importance of the interdependence between private and security benefits for the choice of transfer mode in an equity transaction involving change of control. Furthermore, while several papers have estimated the likelihood of non-partial takeovers, 8 Rapp et al. (2008) is the only other paper of which we are aware that empirically investigates the likelihood of block trades 9. As far as we know, empirical results on the effect of separation of voting rights from cash flow rights on the market for corporate control are also scarce. Our results suggest that the separation of voting rights from cash flow rights affects the market for corporate control by reducing the likelihood of a tender offer while increasing the likelihood of block trades. Even if block trades on average increase firm value, this does not prove that firm value is maximized at block trades or that it is the best feasible outcome (Burkart et al., 2000). It has been shown that separation of voting rights from cash flow rights is the norm outside the Anglo-Saxon countries (see e.g. La Porta et al., 1999). We argue that this partly explains why there are important differences between Anglo-Saxon countries and 8 See e.g. Hasbrouck (1985), Walkling (1985), Palepu (1986), Ambrose and Megginson (1992) and Powell (1997). 9 Unlike our paper, the paper by Rapp et al. (2008) does not consider tender offers and block trades as alternative control transfer modes. 4

7 the rest of the world in terms of dynamic relations involving control transfers (Franks and Mayer, 2001). While the market for corporate control in the Anglo-Saxon countries is an important governance mechanism, it is not clear that takeover activity plays an important role in limiting agency problems in other parts of the world (Bethel et al., 1998; Denis and McConnell, 2003). The rest of the paper is organized as follows. The next section outlines our hypotheses and describes the data. Section 3 discusses the survival analysis methodology. Section 4 presents the empirical results and the final section summarizes and concludes the paper. 2. Hypotheses and Data 2.1. Hypotheses non-partial tender offers and block trades The hypotheses are outlined for a situation without a mandatory bid rule. In the empirical section, we will discuss how the hypotheses might change with the introduction of a mandatory bid rule. Grossman and Hart (1980) show that due to the free-rider problem, an outsider will never take over a diffusely held firm in order to implement value-improving changes. Shleifer and Vishny (1986) consider a firm with a large minority blockholder and a fringe of small shareholders. When the large shareholder s return on his own shares suffices to cover monitoring and takeover costs, performance improving tender offers will be possible. Either the large shareholder makes a tender offer himself or he makes it easier for a well-informed outsider, who has no initial position in the firm, to make a tender offer. The outsider and the incumbent large shareholder would simply split the gains from the rise in the price of the incumbent s shares. The more shares the incumbent owns, the more likely it is that the return associated with improved performance will cover monitoring and takeover costs. Viewed in another way, the more shares the incumbent owns, the easier it is to convince small 5

8 shareholders that a low bid premium indicates a small performance improvement rather than an attempt to profit at their expense. If the incumbent accepts a bid by an outsider, it is convincing for small shareholders; since the incumbent s stake in firm increases, it is in his own interest to accept tender offers associated with some lower value improvements. This is the basis for our first hypothesis Hypothesis 1: The probability of a non-partial tender offer is increasing in the largest shareholder s ownership in the firm. Hypothesis 1 deals with friendly takeovers, i.e. the bidder has negotiated the terms of the tender offer with the management and large shareholders in the target before the tender offer is made public. Even if the frequency of hostile takeovers has increased in Sweden since the late 1990s, hostile takeovers were very rare in most of our sample period. It could be argued that if hostile takeovers dominated the Swedish market for corporate control, hypothesis 1 should be reversed. The risk of a hostile takeover is decreasing in the largest shareholder s ownership in the firm, since the probability of a control contest is inversely related to the ownership concentration of the firm (Nenova, 2003). In addition, Shleifer and Vishny (1986) do not consider the possibility that the large minority shareholder derives private benefits of control and how this would affect the probability of a successful tender offer. 10 Burkart et al. (2000) also consider control transfers in firms with a dominant incumbent minority shareholder and otherwise dispersed owners. The incumbent s influence over corporate decisions enables him to pursue own goals and extract private benefits. The incumbent can sell the minority block to an outside bidder in a negotiated block trade or as part of a non-partial tender offer. All private benefits are 10 We define a tender offer as successful if the target shareholders accept the offer and sell their shares to the bidder. 6

9 transferable to the bidder and the bidder is therefore willing to compensate the incumbent for the private benefits. The bidder will improve firm performance irrespective of transfer mode. However, as compared to a block trade, a tender offer leads to more concentrated ownership. The controlling shareholder s incentive to extract private benefits decreases with ownership concentration, since he internalizes more of the inefficiencies associated with the extraction of private benefits when the size of the block increases. Indeed, after a successful non-partial tender offer, there are no private benefits of control. Burkart et al. consider the incumbent and the bidder as a coalition which must acquire shares at the post tender offer value, since the dispersed shareholders free-ride in tender offers. Thus, since the coalition is not compensated for the reduction in private benefits associated with more concentrated ownership, only the block is traded. We relax this condition and base our second hypothesis on the argument that the contribution to the value of the controlling block of private benefits relative to security benefits increases when the size of the controlling block decreases. Hypothesis 2: The probability of a block trade is decreasing in the largest shareholder s ownership in the firm. When the large shareholder s control of voting rights is separated from cash flow rights by the use of dual class of shares and/or stock pyramids, the large shareholder s mitigation of the free-rider problem is reduced. Grossman and Hart (1988) also show how deviations from oneshare one-vote can reduce the likelihood of takeovers in the presence of private benefits of control. Hypothesis 3: The probability of a non-partial tender offer is decreasing in the large shareholder s separation of voting rights and cash flow rights. 7

10 Finally, all other things equal, a larger wedge between voting and cash flow rights is likely to increase the relative magnitude of the private benefits for the controlling owner, since it increases his ability/incentive to extract private benefits (Bebchuk, Kraakman, and Triantis, 1999; Claessens et al., 2002; Barak and Lauterbach, 2007). Therefore, a larger wedge between voting rights and cash flow rights should increase the likelihood of a block trade (Burkart et al., 2000). Hypothesis 4: The probability of a block trade is increasing in the large shareholder s separation of voting rights and cash flow rights. The private benefits of control are not easily determined or quantified and thus, it is very difficult to design an accurate proxy (Dyck and Zingales, 2004). Following the existing literature on corporate governance (e.g. Bebchuk et al., 1999; La Porta, Lopez-de-Silanes, and Shleifer, 2002), we use the wedge between voting rights and cash flow rights as a gauge on the ability/incentive to extract private benefits. Claessens et al. (2002) use the wedge between voting rights and cash flow rights as a gauge on entrenchment. There are a number of other methods that have been used to measure private benefits. One proxy is the premium paid for a controlling block of shares (Barclay and Holderness, 1989; Dyck and Zingales, 2004). Needless to say, this proxy can only be constructed for firms that experience a block trade and cannot be used to predict a block trade. A second proxy for private benefits is the premium paid for voting shares when a company has both voting and non-voting shares (Nenova, 2003; Zingales, 1994). Although many Swedish firms have dual class shares, only the low voting B-shares would be traded for a large number of these firms while the controlling owner would often keep all A-shares. 8

11 Hence, in many cases it is impossible to measure the voting premium. Relying on this measure would reduce our sample size by roughly 75 percent. Furthermore, Rydqvist (1996) shows that the voting premium dramatically increases in control contests. It is likely that the voting premium would increase prior to public tender offers and block trades. Thus, the interpretation of this variable in terms of predicting public tender offers and block trades would be somewhat ambiguous. Empirical studies that attempt to measure the private benefits of control generally find that the valuation proxies for private benefits of control in Sweden are similar to, but somewhat lower than, those in the US and the UK (see Nenova, 2003; Dyck and Zingales, 2004) Sample Selection Our main tests are performed on an unbalanced panel dataset containing accounting data for the largest non-financial Swedish firms listed on the Stockholm Stock Exchange The sample stops in 1998 since Sweden introduced a Mandatory Bid Rule (MBR) in The rule requires that the bidder in a block trade offers small shareholders the same price per share as he paid the incumbent in the block. The threshold at which the mandatory bid rule comes into play was first set to 40 percent of the voting rights. This threshold was changed to 30 percent of the voting rights in Thus, the MBR directly affects the ability of large shareholders to transfer private benefits and therefore, we do not expect to find any support for hypotheses 2 and 4 with an MBR. In fact, in the competing risk framework we adopt, our model specification is no longer valid in the post- MBR period, given the restrictions imposed on the control transfer mode. However, as a robustness test, we estimate the determinants of block trades on a sample containing data for The rule was introduced as an amendment of the Swedish self-regulation on takeovers. 9

12 The accounting data is collected from the Findata Trust database. The sample contains the vast majority of the largest non-financial public firms in this time period. Some large firms that were only listed for one or two years before delisting are not included in the sample. The accounting data is combined with ownership data from Sundqvist ( ) and Sundin and Sundqvist ( ). This source reports the 25 largest owners in all listed firms as of January each year. Sundin and Sundqvist provide detailed information on coalition structures in a wide sense. Thus, if two large shareholders are known to cooperate, their shareholdings are aggregated by Sundqvist et al. We have followed their definitions of ownership coalitions. After the collection of ownership data, the sample consists of 195 firms and 1461 firm years. To identify control block trades, we collect information on the identity of the largest voteholder coalition in each firm year. When there is a change in the identity of the largest voteholder, we collect information about the changes from the publications by Sundqvist ( ) and Sundin and Sundqvist ( ) and the Affärsdata database. Affärsdata contains Swedish business articles and news from 1981 until today. It now covers 90 sources. We identify 130 control changes (change in the identity of the largest voteholder) in our sample and classify them as due to block trades (62 obs), reconstructions upon financial distress (12 obs), accumulation of shares by an outsider through the acquisitions of minority blocks and/or open market trade (45 obs) and open market divestment by the controling owner (11 obs). Since we are interested in control changes where the incumbent and an outsider negotiate the control transfer, we focus on the 62 block trades. The other control changes do not involve a negotiated control transfer between the incumbent and the outsider gaining control. 10

13 We also identify 20 indirect control changes due to block trades (14 obs), non-partial takeover (2 obs), and accumulation of shares by an outsider (4 obs). We define a control change in firm B as indirect when there is a control change in firm A and firm A controls firm B. We do not define the indirect ownership change in firm B as a control change (non-partial takeover or block trade), since the ownership in firm B does not change. The takeover activity within Swedish pyramids has been analyzed by Holmen and Knopf (2004) and Doukas, Holmen, and Travlos (2002). A first rough estimate of successful non-partial tender offer activity is also collected from the publications by Sundqvist ( ) and Sundin and Sundqvist ( ) since they report all delistings. However, they do not distinguish among actual non-partial takeovers, minority buyouts, and going private transactions. We are only interested in transactions where there has been a change in control, i.e. not in minority buyouts and going private transactions. We use daily newspapers to separate going private transactions from non-partial takeovers after successful tender offers. 12 Moreover, we examine the ownership structure of the firm in the years preceding the delisting. In Sweden, almost all non-partial takeovers are preceded by a public tender offer (Bergström and Rydqvist, 1989). According to Swedish law, any shareholder or group of shareholders in the target that has 10% of the shares can block a legal merger. Therefore, the terms of the tender offer are often negotiated between the bidder and the large shareholders of the target before the public announcement. When the large blockholders have accepted the terms of the bid, a public tender offer is made for all target shares, including the blockholders shares (Rydqvist, 1993). Most bids are non-partial and contingent on 90% of the shareholders accepting the offer. The fact that we only look at successful non-partial tender offers suggests that all blockholders ultimately accepted the offer, sometimes after a revision of the offer. Some non- 12 Part of this data was provided by Clas Bergstrom and Kristian Rydqvist. 11

14 partial takeovers start with a hostile tender offer, i.e. an offer that has not been discussed with the blockholders in the target. The bidder then negotiates with the blockholders in the target and the offer might be revised. If a rival bidder offers a higher price, the blockholders in the target are not forced to sell to the initial bidder, even if they have agreed on the terms of the initial offer. However, since the terms of most tender offers are typically negotiated between the bidder and the blockholders in the target before they are made public, we do not observe many failed bids. On the other hand, failed negotiations are common, i.e. the bidder and the blockholders in the target start negotiations but do not reach an agreement and we do not observe a tender offer. Furthermore, blockholders in potential target firms may negotiate with several potential bidders before reaching an acceptable offer, or deciding not to sell. Our final takeover sample consists of 35 successful non-partial tender offers. Seven successful tender offers were preceded by a block trade, i.e. an outsider acquired the controling block from the incumbent and then made a tender offer for the remaining shares. We classify these seven observations as block trades since control was transferred at the block trade, not at the tender offer. Thus, our final non-partial tender offer sample includes 28 observations. Together with block trades, control changes due to financial distress, accumulation of shares by an outsider, and divestment in the open market, we have 158 control changes, 22 percent of which (28 of 158) are public tender offers. This is similar to the 27 percent reported by Schmid for the US in the year In our total sample, 24 firms were subjected to minority buyouts and four firms were delisted due to bankruptcy. Table 1 panel A summarizes our sample. On average, our sample roughly contains 100 firms each year and roughly comprises 70 percent of the Swedish stock market capitalization. On average, 2 firms are taken over by non-partial tender offers each year and 4.5 firms were subjected to control block transfers each year. 12

15 2.3. Descriptive Statistics In Table 2, we provide descriptive statistics for the 195 firms and 1461 firm years in our sample. The largest block of equity on average contains 33.5 percent of the firm s cash flow rights (Unadj Equity). When the controlling owners net investment in the firm is adjusted for pyramid structures, the average controlling owner holds 26.5 percent of the firm s cash flow rights. If firm B is controlled by firm A, Equity in firm B is adjusted for pyramid structures by multiplying the ultimate owner s fraction of cash flow rights in firm A with firm A s fraction of cash flow rights in firm B. This is in line with e.g. Claessens et al. (2002). However, we define equity ownership for the largest voteholder in each firm and do not use any cut off levels. Claessens et al. (2002) define firms where no shareholder holds 10 percent of the voting rights as having dispersed ownership. In only 22 observations in our sample (1.5% of the sample) is the largest voting block smaller than 10 percent. We control for holdings through multiple control chains according to Faccio and Lang (2002). The average ratio between Equity and Unadj Equity is The median is since less than 50 percent (32 percent) of the firms are part of a pyramid structure (see panel B). The average controlling owner holds 50.8 percent of the voting rights (Votes). 13 The difference in excess of pyramid structures [förstår inte riktigt detta] is due to the high frequency of dual class shares. More than 80 percent of the firms in our sample have dual class shares (see panel B). The median controlling shareholder has almost 50 percent more voting rights than cash flow rights (Votes/ Unadj Equity) and almost twice as many voting rights as pyramid adjusted equity (Votes/ 13 We define Votes as the ultimate owner s fraction of voting rights in the firm, independent of pyramid structures. Claessens et al. (2002) define Votes as the fraction of voting rights in the weakest link in terms of voting rights in the pyramid. We have run all our tests with the definition of Claessens et al. (2002). Votes become smaller for about half of the pyramidal firms. The block trade results reported below virtually remain unchanged. The tender offer results become somewhat less statistically significant, however. 13

16 Equity). 14 The median controlling owner has been the largest voteholder for 12 years (Tenure). We also define a cross-ownership dummy (unreported). For firm B, this dummy is equal to one if Firm A controls firm B and firm B owns shares in firm A, and zero otherwise. Only 80 observations in our sample (5.5%) were involved in cross-ownership. Including a cross-ownership dummy in the hazard models reported in the next section does not affect any of the other results and the dummy per se is insignificant in all models. The median firm has assets with a book value of 1482 million SEK (Size), is 48 years old (Age), generates a return of 3.4 percent on total assets (Profitability), has financed 23.1 percent of the total assets with long-term debt (Leverage), has 57 percent of its total assets in short-term assets (Liquidity) and has a Tobin s q of Tobin s q is defined as the sum of the market value of equity and the book value of total debt divided by the book value of total assets. The sample is split according to whether the firm was subject to a control change in Successful non-partial tender offers and block trades are defined as control changes. All years within the same ownership spell (N=396) prior to a successful non-partial tender offer or a block trade are classified as belonging to a firm experiencing a control change. An ownership spell is defined as all firm years with the same controlling owner. Thus, new ownership spells begin at control block transfers and other control changes. 15 At a non-partial tender offer, the firm is delisted and no new ownership spell begins. 14 Faccio and Lang (2002) and Giannetti and Simonov (2006) analyze the ratio of voting rights and cash flow rights. Claessens et al. (2002) analyze the difference between voting rights and cash flow rights. In our empirical work, we take the natural logarithm of the ratios in order to reduce the impact of extreme values. Using the differences generates similar but statistically somewhat weaker results. 15 We have 195 firms, 62 control block transfers, and 88 other ownership changes resulting in 345 ownership spells. The 62 block trades are related to 48 firms. Thus, some firms experienced more than one block trade during the sample period. Furthermore, 7 firms experienced a block trade before becoming a non-partial takeover target. Note, however, that if it is the shareholder gaining control by the block transfer that makes the non-partial tender offer for remaining shares, the tender offer is not classified as a takeover but as a going private transaction. 14

17 The mean and median difference tests suggest that the controling owners in firms experiencing a control change have less Votes. The median difference test also suggests that controling owners in firms experiencing a control change have less total excess votes (Votes/ Equity). There is no significant difference for the other ownership variables 16 and the descriptive statistics do not suggest any significant relation between control tenure and control changes. The firms experiencing control changes are smaller but of similar age as those not experiencing control changes. Firms experiencing control changes are less profitable (Profitability) according to the median difference test. The other difference tests are insignificant. In panel B, we report statistics for four binary variables. More than 80 percent of the firms have dual class shares. In 29.4 percent of the firms, the controlling shareholders hold all A-shares and only the B-shares are traded on the Stockholm Stock Exchange. The controlling shareholder holds all A-shares in roughly 40 percent of the dual class firms. Almost one third of the firms are controlled by a pyramid structure. The difference tests suggest that dual class shares increase the probability of a control change while there is no significant difference for pyramids. The relative magnitude of private benefits does not only depend on the size of the controlling block but also on the distribution of the remaining shares. To proxy for the distribution of outside shareholdings, we define a dummy variable (Outside Block) equal to one if the second largest shareholder holds more than 10 percent of the voting rights, and zero otherwise. In 43.7 percent of the firms does an outside shareholder hold 10 percent of the voting rights or more. The descriptive statistics suggest that an outside block holder increases the probability of the firm experiencing a control change. 16 Note that mean differences are tested on the natural logarithm of Unadj. Equity/ Equity, Votes/ Unadj. Equity, Votes/ Equity, Tenure, Firm Size, Firm Age and Tobin s q. 15

18 In panel C, the firms experiencing a control change are split by whether they are nonpartial tender offer targets or subject to control block transfers. All ownership spells prior to the non-partial tender offer and block trades are classified accordingly. The mean difference tests indicate that the ownership spells associated with non-partial tender offers have larger Unadj. Equity, Unadj. Equity/ Equity, Votes, Votes/ Equity and Tenure but less Votes/ Unadj Equity. The median difference test suggests that the controlling owners in tender offer targets have less Equity than owners of block trade targets. Compared to firms experiencing control block transfers, tender offer targets are also larger (Size), older (Age), more profitable (Profitability), and have higher Liquidity but lower Leverage. There are no significant differences in terms of Tobin s Q. Finally, the difference tests in panel D indicate that nonpartial tender offer targets are more likely to have Dual Class Shares and be Pyramid firms. It is also more likely that the controlling owners hold all A-shares in tender offer targets as compared to ownership spells associated with control block transfers (Controlling owner holds all A-shares) Statistics on the bidders in the control changes In table 3 panel A, we present some statistics on bidders in non-partial tender offers and block trades, respectively. Half of the bidders in non-partial tender offers are Swedish entities. The frequency of Swedish bidders in block trades is significantly higher, amounting to almost 76 percent. Among the Swedish bidders, there are no statistically significant differences between bidders in non-partial tender offers and block trade bidders in terms of being a listed firm. Among the listed bidders, it is more likely that the bidder will be a pyramidal firm in block trades, as compared to non-partial tender offers. The bidder is defined as being a pyramidal firm if it is controlled (largest voteholder) by another firm listed on the Swedish stock market. 16

19 Gaining control of a firm by a block trade requires less capital as compared to a non-partial tender offer. As a proxy for the bidder s size, we present the average and median market value of equity for the listed bidders in panel B. It is difficult to define and collect the size or wealth for bidders that are non-listed entities or individuals. The averages and medians do not suggest that bidders in block trades are statistically smaller than bidders in non-partial tender offers Announcement returns For illustrative purposes, we present statistics on the announcement returns at non-partial tender offers and block trades, respectively, in table 4. The announcement returns are estimated as the change in share price from five days before to five days after the first announcement of the control change. 17 The announcements are collected from the Affärsdata database. We are able to collect the necessary data for 25 non-partial tender offers and 56 negotiated block trades. The average and the median tender offer in our sample are associated with announcement returns above 25 percent. These numbers are similar to the abnormal price changes at tender offers in the US reported by e.g. Jensen and Ruback (1983). The average announcement return at a negotiated control block transfer is 5 percent. The median return at block trades is only 1.6 percent, but it is still statistically different from zero. Positive abnormal returns have also been reported for block trades in the US (Holderness, 2003). Even though minority shareholders are harmed in some block trades (large negative announcement returns), the average and the median share price increases suggest that improved management is at least part of the source of the gains in block trades (Berglöf and Burkart, 2003). 17 It would be interesting to also explore the control block premium. However, it is not possible to collect the prices at which the control block is traded since this is done outside the stock-market and the parties are obliged to make this information public. 17

20 3. Survival analysis methodology Survival analysis methodology is well-suited to analyses of factors explaining the likelihood of a change in firm control (e.g. Cameron and Trivedi, 1996). 18 In our context, the changes in control constitute an instance of competing risks. 19 We choose to look at two competing risks the change in the controlling owner of a firm due to a transfer of the controlling block of shares and the change due to a non-partial tender offer. The observation spell for a firm is the time until there is a change in the controlling owner or the end of the observation period has been reached. Therefore, the spells ending up with a control transfer due to other causes than block trade or a non-partial tender offer are treated as censored, resulting in the beginning of a new observation spell for the firm. As a result, at each point in time within a spell, firms can be considered to be in one of three mutually-exclusive states the status-quo state, the state of a non-partial tender offer, and the state of block trade. Over time, all firms are exposed to the chance of transiting from the status-quo state to one of the two alternative states block trade or non-partial tender offer. For the present analysis, we adopt an independent competing risks specification. In an independent competing risks model, the hazard associated with a non-partial tender offer is assumed to be independent from that of a block trade, conditional on the effects of the independent variables. 20 The estimation of an independent competing risks model is thus 18 Survival analysis deals with the modeling of time-to-event data, also known as transition data (or survival time data or duration data). Comprehensive treatments of these models can be found in Cox and Oakes (1984), Lancaster (1990), and Blossfeld and Rohwer (1995). 19 It should be noted that the absence of control change can be considered as a residual risk. 20 An important advantage of independent competing risks models is their computational tractability. Extensions to allow for correlated risks have been introduced in the literature, but have not yet been used to any considerable extent in applied work. In fact, the assumption of conditional independence is not very unrealistic given that if a particular covariate affects the hazard of more than one event, and it is in the model, then its effect is controled for. We use a large number of covariates, many of which vary with or are a function of time, which would be likely to mitigate the potential conditional dependence between the risks involved. In addition, if we assume the dependence to be due to a common unit effect, the inclusion of an unobserved heterogeneity term in our models would also ensure conditional independency of risks. 18

21 exactly equivalent to the estimation of separate models for each risk, while treating control transfers due to the other risk(s) and the situations of no control transfer as censored. 21 The advantage of the adopted methodology over more traditional Ordinary Least Squares or multinomial regressions is the ability of the survival analysis to directly account for the sequential nature of the data, and its ability to handle censoring and incorporate timevarying covariates (Jenkins, 2005). In our approach, the hazard rate captures the chances of a non-partial tender offer or a block trade within a year, conditional on survival (no such event) up to that point. The unconditional probability of not experiencing event k (where k is takeover or block trade) from the beginning of the observation period (t=0) until time t>0 for firm i in our sample is equal to λ ik (t)dt, where λ ik (t) is the hazard function defined by the equation Pr t dt Ti t lim 0 dt dt T t i ( t), i k and dt is an infinitesimal interval of time. Alternatively, λ ik (t)dt can be interpreted as an unconditional probability of a firm i experiencing an event in a tiny interval of time [t, t+dt]. The (instantaneous) hazard rate function for firm i at time t>0 is assumed to take the proportional hazards form ( t ) 0 ( t)exp( X ), i k k it where λ 0k (t) is the unknown baseline hazard at time t which may take a parametric or nonparametric form, and X it is a vector of covariates summarizing observed differences between firms at time t and β is a vector of parameters to be estimated. 21 For continuous time models, the log-likelihood for a model with multiple destinations has a separability property meaning that a multiple-destination survival model can be estimated by estimating a number of singledestination models separately, one for each destination (competing risk). With the discrete data, as in our case, an independent competing risk model can still be estimated by setting up the likelihood function as the likelihood for a standard multinomial logit model applied to re-organized data (Allison, 1982). For intervalcensored data with a relatively small interval hazard, it is common to assume that the continuous time hazard rate was constant within each interval. 19

22 Note that the probability density function in our context is a time to failure function that gives the instantaneous probability of the event. That is, in a survival experiment where the event is a non-partial tender offer, the value of the density function at time T is the probability of a firm subject to a tender offer precisely at time T. This differs from the hazard function which gives the probability conditional on a firm having survived to time T. 22 Although the survival of firms occurs in continuous time, our data calls for the discrete time specification of the model given that the spell length is observed only in oneyear intervals. In other words, the underlying continuous durations are only observed in disjoint time intervals [0 = a 0, a 1 ), [a 1, a 2 ), [a 2, a 3 ),, [a k-1, a k = ). Our covariates (e.g. firms characteristics) may vary between time intervals but are assumed to be constant within each of these. In the discrete case, hazard of exit to a particular destination k in the jth interval is given by j k a, a T a } h. ( X it ) Pr{ T j1 j j1 In our case, all intervals have the length of one year, so that the recorded duration for each firm i corresponds to the interval [t i-1, t i ). Firms are recorded as either experiencing a tender offer or a block trade during the interval, or as still remaining in status-quo. The former group, contributing completed spell data, is identified using censoring indicator c i =1. For the latter group, contributing right-censored spell data, c i =0. The number of intervals comprising a censored spell is here defined to include the last interval within which the firm is observed. The log-likelihood for each event k can be written in terms of the hazard function as: LogL k n c ti 1 ti log h ( X ) 1 hsk ( X is ) (1 ci )log1 hsk ( X is ), i ikt ikti i1 s1 s1 22 In this paper, we sometimes use the terms hazard rate of tender offer/block trade or tender offer/block trade risk instead of tender offer/block trade probability since what we model is the probability per time unit that a firm that has survived to the beginning of the respective interval will be taken over or that the controling shareholder will sell her block of shares in that interval. 20

23 where the discrete time hazard in the jth interval is h jk j ( X ij ) 1 exp exp( X ij jk ) with jk log 0k ( ) d. This specification allows for a fully non-parametric baseline hazard with a separate parameter for each duration interval. Alternatively, the γ jk may be described by some semiparametric or parametric function, e.g. θ k (j). If we define an indicator variable y ikt =1 if firm i experienced an event during the interval [t-1,t], and y ikt =0 otherwise, the log-likelihood can be rewritten in sequential binary response form: a a j 1 n ti yikj log h jk ( X ij ) (1 yikj )log1 h jk ( X ij ) log L. k i1 j1 This is one specification of log-likelihood which we estimate for each event. Our second specification incorporates a Gamma distributed random variable to describe unobserved (or omitted) heterogeneity between individuals. The instantaneous hazard rate is now specified as t) ( t) exp( X ) ( t)exp( X log( )), ik ( 0k i it k 0k it k i where ε i is a Gamma distributed random variable with unit mean and variance σ 2 v, and the discrete-time hazard function is now h jk ( X ij ) 1 exp exp( X ij log( )) k jk i. The likelihood functions of the second model are: LogL k where n i1 log 1 c i A i k c B i ik A ik t i 1 v exp X ij k k ( j) j1 (1/ v) and 21

24 B ik 1 v ti 1 j1 expx ij (1/ v) ( ) k k j Ai, if ti 1, or 1 Ai, if ti 1, where θ k (j) is a function describing duration dependence in the hazard rate. The loglikelihood function of the first model is the limiting case as v0. 4. Empirical Results In this section, we report the results of estimations of discrete time proportional hazards regression models with the hazard rate of control changes as the dependent variable. 23 First, we estimate competing risk models for non partial tender offers and block trades, respectively, for the period. A mandatory bid rule was introduced in 1999 with direct implications for the choice of control transfer mode and the ability of controlling shareholders to transfer private benefits of control. Despite the fact that our model specification is no longer valid in the post-mbr period, we do expect to find weaker relationships between ownership proxies and the probability of block trade. Therefore, we reestimate the block trade models for the period The probability of a non-partial tender offer In this section, we estimate the risk of a non-partial tender offer using the hazard regressions models. We include several independent variables. Our first hypothesis deals with the ownership fraction of the largest shareholder and, therefore, we include the natural logarithm of the largest shareholder s voting fraction (Ln(Votes)). To capture the effect of separation of voting rights from cash flow rights, we include two variables. The effect of dual class shares is captured by the natural logarithm of the ratio of the largest shareholder s vote fraction and 23 The models are estimated using the pgmhaz8 procedure of STATA 8.2. For each specification, two models are estimated by maximum likelihood methods: (1) the Prentice-Gloeckler (1978) model; and (2) the Prentice- Gloeckler (1978) model incorporating a gamma mixture distribution to summarize unobserved individual heterogeneity, as proposed by Meyer (1990). We estimate a fully non-parametric specification for the baseline hazard with four interval-specific baseline hazards. 22