Transaction costs, behavioral uncertainty and the formation of interfirm cooperations: Syndication in the UK private equity market

Transcription

1 FACULTEIT ECONOMIE EN BEDRIJFSKUNDE HOVENIERSBERG 24 B-9000 GENT Tel. : 32 - (0) Fax. : 32 - (0) WORKING PAPER Transaction costs, behavioral uncertainty and the formation of interfirm cooperations: Syndication in the UK private equity market Miguel Meuleman Ghent University Sophie Manigart Vlerick Leuven Ghent Management School and Ghent University Andy Lockett Nottingham University Business School Mike Wright Nottingham University Business School JANUARY /359 We acknowledge financial support from the Interuniversity College of Management Sciences (I.C.M.), Ghent University and Gate2Growth Academic Network in Entrepreneurship, Innovation and Finance. The Centre for Management Buyout Research (CMBOR) at the University of Nottingham generously provided the data. D/2006/7012/04

2 Transaction costs, behavioral uncertainty and the formation of interfirm cooperations: Syndication in the UK private equity market Summary Empirical results on the relationship between uncertainty and interfirm cooperation are conflicting. To address this puzzle, we distinguish between two different sources of uncertainty: primary and behavioral uncertainty. We hypothesize that the effect of primary uncertainty on interfirm cooperation is moderated by the behavioral uncertainty associated with an exchange partner. Evidence from the formation of investment syndicates in the management buyout market shows that for high levels of behavioral uncertainty, there is a negative effect of primary uncertainty on interfirm cooperation, whereas for low levels of behavioral uncertainty there is a positive effect. Implications for theory and research are suggested. Keywords: transaction cost economics, uncertainty, interorganizational strategy, social capital, reputation 2

3 Transaction costs, behavioral uncertainty and the formation of interfirm cooperations: : Syndication in the UK private equity market There has been an increased interest in different forms of interfirm collaborations in recent years. Using diverse theoretical perspectives, researchers have studied various aspects of joint ventures, licensing agreements, distribution and supply agreements, research and development partnerships, syndications and other forms of interfirm collaborations. One major difference between single-firm strategies and collaborative strategies is the presence of behavioral uncertainty or uncertainty regarding the behavior of exchange partners (Williamson, 1985). Recent studies have increasingly started to focus on informal mechanisms such as trust and trustworthiness that reduce the extent of behavioral uncertainty associated with an exchange partner (Barney and Hansen, 1994; Gulati, 1995; Uzzi, 1997; Dyer and Chu, 2003). Firms, it has been argued, can even gain a competitive advantage by having a reputation of being trustworthy as it lowers transaction costs (Barney and Hansen, 1994). Few studies, however, have actually examined the extent to which the trustworthiness associated with an economic actor pays off in economic benefits. The main research question in this study is, therefore, to what extent do differences in behavioral uncertainty enable some firms to adopt governance structures that would not be available to others under similar circumstances? This paper builds on transaction cost economics and looks at the impact that the wider context has for the behavioral uncertainty associated with an exchange partner. Transaction cost theory relies on two basic assumptions of human behavior namely bounded rationality and opportunistic behavior. Starting from these assumptions, the 3

4 effect of uncertainty, frequency and asset specificity are examined in order to explain the extent to which firms rely on hierarchies, hybrids or markets to execute transactions. According to transaction cost theory, asset specificity in combination with uncertainty can potentially lead to costly haggling and maladaptation costs when actors behave opportunistically (Williamson, 1985). Previous studies examining the relationship between uncertainty and firm boundary decisions have produced a conflicting set of results (David and Han, 2004). One major critique associated with transaction cost economics has been the assumption of opportunism. Structural sociologists and behavioral researchers have emphasized the role of embeddedness and trust in alleviating the risk of opportunism (e.g. Granovetter, 1985; Podolny, 1994; Mayer et al., 1995, Zaheer and Venkatraman, 1995). Others have also questioned the assumption of opportunism by exploring the role of reputation in mitigating transaction costs (Hill, 1990; Parkhe, 1993). To clarify the empirical puzzle currently present in the transaction cost literature, we relax the assumption of opportunism by focusing on the interrelation between two different sources of uncertainty associated with an economic exchange: primary uncertainty related to the transaction and behavioral uncertainty related to the partner initiating the transaction (Williamson, 1985). As such, we respond to the suggestions made by David and Han (2004) to consider the conditions under which transaction cost theory is appropriate. Further, we extend previous research indicating that firm specific attributes influence organizational governance (Leiblein and Miller, 2003). The few studies that have examined the interrelation between primary uncertainty and behavioral uncertainty have considered the role of previous experience between exchange partners in reducing the extent of behavioral uncertainty (Sutcliffe and Zaheer, 1998; Artz and Brush, 2000). Our study focuses on the role of the reputation and social embeddedness 4

5 of an exchange partner in reducing the degree of behavioral uncertainty. Behavioral uncertainty in this study is, therefore, closely related to the concept of trustworthiness as defined by Barney and Hansen (1994) as it is influenced by the attributes of an individual actor rather than by the attributes of an exchange relation between two actors. As such, we investigate whether reputable and socially embedded firms have an advantage in the formation of interfirm collaborations. We examine the interrelation between primary and behavioral uncertainty by studying the formation of one particular form of interfirm collaboration, namely equity investment syndicates in the management buyout market. 1 Syndicated investments are a common feature of venture capital and later stage buyout investments both in the US and in Europe (Sorenson and Stuart, 2001; Wright and Lockett, 2003). An equity investor leading a buyout investment may either act as sole equity provider or invite other investors to join in a syndicate. From a transaction cost perspective, a sole investment corresponds to a more hierarchical form of governance whereas a syndicated investment corresponds to a hybrid form of governance. For each separate investment, we examine the probability that an investment syndicate will form. The hypotheses are tested on UK management buyout deals from 1993 to Studying the formation of investment syndicates extends research on transaction cost economics and interfirm cooperation in particular. First, our empirical setting enables us to observe on a project by project basis which governance structure is adopted by investors using ex-ante measures of uncertainty associated with the transactions and the partner initiating the transaction. This helps to avoid the confounding effect of endogeneity frequently found in cross-sectional studies testing transaction cost theory. Second, previous studies have stressed the role of reputation and 5

6 social embeddedness in the private equity market (Black and Gilson, 1998). There is a considerable degree of heterogeneity among the actors in this market which enables us to empirically identify the effect of reputation and social embeddedness. Third, compared to other forms of interfirm collaborations such as joint ventures, syndicates are a relatively homogenous type of transaction which avoids problems of unobserved heterogeneity. The rest of this paper is organized as follows. First, we introduce the research setting of our study. Second, we discuss the theory and provide our theoretical framework and related hypotheses. The following section outlines the data and method used in the analyses. Next, we present the findings from the empirical analyses. Finally, we discuss our findings, conclude and outline potential avenues for future research. Research setting: The UK private equity market for management buyouts We investigate the foregoing issues by looking at the formation of investment syndicates in the buyout segment of the private equity market in the UK. The UK buyout market has been established since the middle of 1980s and during the 1990s on average about 300 transactions took place each year (Wright, et al., 2000). Equity investment syndicates are a form of interfirm alliance in which two or more firms coinvest in an investee firm (the buyout target) and share a joint pay-off. As noted above, a sole investment in which the lead investor holds all the outside equity corresponds to a hierarchical form of governance whereas a syndicated investment in which the equity is distributed among several partners corresponds to a hybrid form of governance as investment partners are mutually dependent on each other in a nontrivial way. A syndicate is not a hierarchy as decisions are usually taken in consensus (Wright and 1 We use the term buyout to refer to management buyouts as well as related transactions such as buy-ins, 6

7 Lockett, 2003). Investment syndicates are similar to strategic alliances as they involve long term resource pooling by two or more partner firms to achieve shared strategic goals (Chung et al., 2000; Wright and Lockett, 2003). The buyout setting is a particularly fruitful arena for studying interfirm collaborations as buyout investing is characterized by a considerable degree of uncertainty and therefore issues of interfirm governance are especially important. There are several rationales for syndication: portfolio diversification (Manigart et al., 2005), window dressing (Lerner, 1994), deal flow generation (Sorenson and Stuart, 2001), improved investment selection (Lerner, 1994), value adding (Brander et al., 2002) and certification (Stuart et al., 1999). 2 Each syndicate typically contains a lead firm, in general the investor initiating the transaction, and one or more non-lead firms, with an individual firm taking on both roles depending on the particular deal. In syndicated investments, non-lead investors rely on the lead investor to perform the task of managing and monitoring the underlying investment. Gorman and Sahlman (1989) report that in venture capital investing, the lead investor spends roughly ten times as much time monitoring and managing the company as non-lead investors. As a result, non-lead members of a syndicate may suffer a severe informational disadvantage compared to the lead investor (Admati and Pfleiderer, 1994; Wright and Lockett, 2003). Even though the investment agreement between the syndicate members will stipulate the kind of information that should be reported and the frequency of reporting by the lead investor to the syndicate members, problems of asymmetric information are unlikely to be resolved completely. Further, lead investors generally hold a larger equity stake as compared to non-lead investors in order to compensate them for their effort buy-in and buyout combinations (so-called BIMBOs), or investor-led buyouts. 2 Even though most of the studies on private equity investment syndicates have focused on the venture capital industry, most of the arguments also apply to later stage buyout investing. 7

8 (Wright and Lockett, 2003). Overall, this means that lead investors have more informal control through their privileged access to information and more formal control through their residual rights of control. Given the long term and uncertain nature of private equity investments, the interests of the lead investor may not always coincide with the interests of non-lead investors. The uncertainty associated with the underlying investment combined with more formal and informal control by the lead investor could potentially lead to opportunism by the lead investor and hence create transaction costs for non-lead investors (Pichler and Wilhelm, 2001; Piskorski, 2004). The lead investor might monitor the investee poorly or could take decisions to its own advantage at the expense of non-lead investors. Overall, as non-lead investors are vulnerable to the actions taken by the lead investor, non-lead investors will only join when they have confidence in the lead investor. Potential hazards of interfirm cooperation in investment syndicates are intensified as the illiquid nature of buyout investments creates a lock-up for the nonlead investors of a syndicate for a considerable period of time. This corresponds to a situation of asset specificity. If syndicate partners seek an early exit from the investment, serious losses may result because of the illiquid nature of the investment. Further, an early exit may convey a negative signal to potential outside investors (Lerner and Schoar, 2004). As a consequence, syndicate partners are mutually dependent for a considerable period of time and hence are locked into dealing with the lead investor. 3 Given that each investments is characterized by a nontrivial degree of asset specificity, we do not explicitly examine the effect of asset specificity in the remainder of this paper. 8

9 THEORETICAL FRAMEWORK AND HYPOTHESES Our basic theoretical premise is that in order to understand the governance of economic transactions one should look simultaneously at the uncertainty associated with the underlying transaction and the uncertainty associated with the economic actors involved in the transaction. Early transaction cost literature did not explicitly make a distinction between these two different forms of uncertainty (Williamson, 1975). More recent literature with respect to transaction cost economics has disaggregated the construct of uncertainty (Williamson, 1985). For example, Sutcliffe and Zaheer (1998) distinguished between primary, competitive and supplier uncertainty. In the present study, we focus on two different sources of uncertainty associated with an economic exchange: primary and behavioral uncertainty (Williamson, 1985: 56-59). Primary uncertainty can be defined as uncertainty surrounding the underlying transaction which arises from changing conditions in the economic environment or random acts of nature and will tend to be higher as the number of changes and their unpredictability increase. Primary uncertainty may cause problems of communication, technological difficulties, coordination problems and, as a consequence, adversely impact the final execution of transactions. Behavioral uncertainty as defined by Williamson (1985) arises from human action and refers to the effects of opportunism on transactions that are executed through incomplete contracts. Following Koopman (1957), behavioral uncertainty also results from a lack of knowledge about the actions or decision rules of other economic actors. 4 In this study, behavioral uncertainty refers both to uncertainty with respect to the risk of opportunism by exchange partners (Williamson, 1985) and uncertainty with 3 As lead investors have more formal and informal control, it will be much more easy to force an overall exit. 9

10 respect to the decision rules used by exchange partners (Koopman, 1957). Behavioral uncertainty as such has a close relationship with trust as trust depends both on the perceived risk of opportunism and the perceived competence of an exchange partner (Das and Teng, 2001). Thus, the absence of behavioral uncertainty will lead to a higher level of trust between exchange partners. Primary and behavioral uncertainty are interrelated in Williamson s framework. Both sources of uncertainty are necessary to generate transaction costs. Williamson (1985:63) notes that absent the hazard of opportunism, costly haggling and maladaptation costs would vanish as partners can rely on a general clause device to solve disputes. An important question then is to what extent are potential transaction costs caused by a combination of primary uncertainty and asset specificity mitigated by the extent of behavioral uncertainty associated with an exchange partner? Most empirical studies testing transaction cost theory have focused on primary uncertainty while assuming explicitly or implicitly a nontrivial degree of behavioral uncertainty. More recently, behavioral uncertainty has received more attention in the academic literature by studies focusing on the role of reputation and trust in economic exchanges (Hill, 1990; Parkhe, 1993; Barney and Hansen, 1994; Mayer et al., 1995; Zaheer and Venkatraman, 1995; Dyer and Chu, 2003). Few studies, however, have focused explicitly on the interrelation between these two sources of uncertainty and how they impact the governance of economic transactions. Sutcliffe and Zaheer (1998) argue that primary uncertainty moderates the relation between supplier uncertainty and the decision to vertically integrate, but failed to find any significant relationship in their experimental study. Further, Artz and Brush (2000) show how the extent of cooperation 4 Koopman (1957) used the term secondary uncertainty to describe this type of uncertainty. Even though secondary uncertainty is related to behavioral uncertainty as defined by Williamson (1985), it differs from the latter by its innocent and non-strategic nature. 10

11 in buyer supplier relationships moderates the relation between primary uncertainty and negotiation costs. In contrast with their study, our study analyses outcomes by looking which governance structures are actually adopted rather than measuring transaction costs directly within one governance structure. One potential problem with the latter approach is that it suffers from problems of endogeneity, which is not present in our setting. Further, we measure the extent of behavioural uncertainty by looking at attributes of individual exchange partners rather than attributes of a relationship between two partners. Figure 1 shows our conceptual model. Based on transaction cost theory, we hypothesize that the primary uncertainty underlying a transaction decreases the probability of interfirm cooperation. We further hypothesize that this relation is moderated by the behavioral uncertainty associated with an exchange partner. We elaborate on these arguments below. INSERT FIGURE 1 ABOUT HERE Transaction costs and primary uncertainty In syndicate arrangements, primary uncertainty is the possibility that different eventualities influence the economic outcome of the underlying buyout investment. There are several reasons why transaction costs associated with a cooperative arrangement tend to increase when the underlying transaction is characterized by higher levels of primary uncertainty. First, the higher the level of primary uncertainty associated with a transaction, the more likely firms will have to adapt to unforeseen environmental conditions and the higher the transaction costs will be. As a result, the more likely is it that cooperation 11

12 problems or problems arising from opportunism will emerge. After all, as the number of possible changes increases, the number of possible unexpected contingencies that may affect contracts between firms increases too. Because of bounded rationality, future contingencies are hard to stipulate and contracts will be incomplete (Williamson, 1985). Even if complete contracts were possible, their enforcement would be prohibitively expensive and time consuming (Hart, 1995). As unforeseen contingencies emerge, partners are vulnerable to opportunism. Further, the higher the level of primary uncertainty, the more likely is it that opportunism by one of the partners involved in a transaction may go undetected (Hill, 1990). When there is more uncertainty regarding the outcome of the underlying transaction, it is more likely that negative outcomes will be attributed to external factors rather than to opportunism by one of the partners. Each partner can protect itself from opportunism by monitoring the collaborating partners directly. In most situations, however, partners cannot costlessly observe the actions by other partners. The cost of monitoring also increases the cost of cooperation. Further, monitoring might be difficult and could even have adverse consequence on the incentives of the partners involved in a collaboration (Ghoshal and Moran, 1996). Therefore, interfirm cooperation is less likely to occur when the underlying transaction is characterized by higher levels of primary uncertainty. Second, most studies testing transaction cost economics have focused on cooperation problems. Another class of problems described within transaction cost theory is involved with problems of coordination. Coordination problems refer to the cost of aligning the actions of different partners to a transaction. Even when interests are aligned, coordination problems can arise due to the lack of shared and accurate knowledge about the decision rules that partners within a cooperative agreement will use (Malmgren, 1961; Williamson, 1991; Milgrom and Roberts, 1992; Gulati et al., 12

13 2005). As decisions in a syndicate are usually taken by consensus, the lead investor has to consult non-lead investors in order to take decisions. Therefore, a syndicate will involve more complicated and time consuming decision making, and hence higher transaction costs, than non-syndicated deals (Fried and Hisrich, 1995; Wright and Lockett, 2003). The costs are likely to be higher when primary uncertainty increases, as unexpected future contingencies will put greater demand on the joint decision making capability of the syndicate. Our discussion above suggests that transaction costs associated with a cooperative arrangement will be higher when there is more primary uncertainty with respect to the underlying transaction, making syndication less attractive. Hence: H1: Interfirm cooperation will be negatively related to the primary uncertainty associated with the underlying transaction. Transaction costs and behavioral uncertainty The identity of exchange partners is important in alleviating potential hazards associated with interorganisational exchange as this will influence the extent of behavioral uncertainty (Granovetter, 1985; Parkhe, 1993; Podolny, 1994; Chiles and McMackin, 1996; Dollinger et al., 1997). Behavioral uncertainty is one of the major assumptions underlying transaction cost theory. Different mechanisms, however, have been described to deal with opportunism or cooperation problems in interfirm relationships. One set of mechanisms relies on contracts and third party enforcement through courts or other legitimate authorities. Under conditions of bounded rationality, comprehensive contracting is not feasible (Williamson, 1991). Another set of mechanisms relies on self-enforcing agreements, in which no third party intervenes to determine whether a 13

14 violation has taken place (Telser, 1980:27). Self-enforcing agreements involve safeguards that allow for self-enforcement. Within this class of self-enforcing mechanisms, a distinction can be made between formal safeguards, such as financial and investment hostages, and informal safeguards, such as embeddedness and reputation (Dyer and Singh, 1998). The risk of opportunism associated with an exchange partner will be lower as the reliance on self-enforcing mechanism increases. In this study, we focus on the role of informal safeguards i.e. reputation and social embeddedness in reducing the extent of behavioral uncertainty as previous research has indicated that formal safeguards are less effective for the governance of private equity syndicates (Wright and Lockett, 2003). Both the reputation and social embeddedness of an exchange partner will influence its trustworthiness and hence have an impact on the extent of behavioral uncertainty. In the next paragraphs, we argue that the negative relation between primary uncertainty and interfirm cooperation posited in hypothesis 1, is moderated by the extent of behavioral uncertainty associated with the exchange partner i.e. the lead investor in a syndicate. We do not formulate hypotheses with respect to the main effect of behavioral uncertainty as we are mainly interested under which conditions transaction theory is appropriate. Transaction cost theory primarily focuses on the effect of transaction specific characteristics whereas it is less clear about the main effects of behavioral uncertainty on the governance of economic transaction. We include the main effect of our behavioral uncertainty variables as control variables in our model however. Reputation and behavioral uncertainty The reputation associated with an exchange partner helps to alleviate problems of cooperation and problems of coordination. First, the reputation associated with an 14

15 exchange partner reduces the perceived risk of opportunism and hence problems of cooperation (Axelrod, 1984; Kreps, 1990; Hill, 1990; Parkhe, 1993; Mayer et al., 1995). After all, the costs of losing a reputation by behaving improperly will be higher for high reputable investors as compared to investors who have yet to establish a reputation. Reputable firms will therefore have less incentives to behave opportunistically and hence will be more trustworthy exchange partners. These reputation effects are especially effective when firms have long term interests in an industry. One motive for syndication is the expectation of reciprocity by the exchange partner in the future, thereby assuring future deal flow (Sorenson and Stuart, 2001). Through such expectations of reciprocity, the shadow of the future helps to govern interfirm relationships (Axelrod, 1984). Further, by collaborating with reputable partners, firms can reduce transaction costs involved with screening potential partners, writing contracts and monitoring partners ex-post (Chiles and McMackin, 1996). The transaction costs of interfirm collaboration will therefore be lower if exchange partners have established a reputation. Second, the reputation of a firm helps to reduce potential coordination problems. Since a firm s reputation acts as an important signal of its overall effectiveness (Fombrun and Shanley, 1990), there will be less uncertainty regarding the decision rules that reputable exchange partners use when unexpected contingencies should arise. In the private equity market, firms mainly gain a reputation by building up a track record of previous investments. Since it is mainly the lead investor who manages the day-to-day relations with the investee (Gorman and Sahlman, 1989; Wright and Lockett, 2003), potential syndicate partners will prefer to join a syndicate when the lead investor has build up a reputation for managing and monitoring investees. In general, if the lead investor has established a reputation for managing and monitoring investments, the 15

16 willingness of syndicate members to rely on the lead investor to oversee the investee will increase and the perceived risk of interorganisational collaboration will be lower. As a result, transaction costs resulting from coordination problems will be lower. The above arguments indicate that, all other things equal, the reputation of an exchange partner will help to offset transaction costs associated with interfirm cooperation. Therefore, we argue that the negative effect of primary uncertainty on interfirm cooperation is positively moderated by the reputation of an exchange partner. Hence: H2a: The reputation of an exchange partner will positively moderate the relationship between primary uncertainty and interfirm cooperation. Social embeddedness and behavioral uncertainty Economic exchange is embedded in a network of interfirm relations through previously established social relations between firms (Granovetter, 1985). The repeated nature of syndication in the private equity market may be a response to the aforementioned transaction costs. There are two different mechanisms by which the social embeddedness of an exchange partner is likely to protect against transaction costs associated with interorganisational exchange. First, social networks play an important role in communicating information about an actor s behavior and hence help to reduce ex-ante transaction costs of locating and screening partners (Granovetter, 1985; Raub and Weesie, 1990; Robinson and Stuart, 2005). The information provided by networks serves as an important basis for assessing the reliability and specific capabilities of partners (Gulati, 1995). Firms that are more socially embedded will therefore benefit as they are more visible and information on 16

17 their behavior will be more widely available. This in turn reduces the uncertainty that potential exchange partners may have when joining a syndicate arrangement. Further, due to the inclination to limit the search for exchange partners to a circumscribed social circle, more socially embedded firms are likely to have more potential partners that are willing to collaborate (Piskorski, 2004). Second, the ex-post transaction costs of collaborating with highly embedded firms will be lower as firms that are deeply embedded in social networks put those networks of relations at risk by behaving opportunistically. The imposition of these social costs acts to reduce the threat of opportunistic behavior (Granovetter, 1985). The network of a firm s relationships therefore has a disciplining effect on a firm s behavior and reduces ex-post transaction costs (Robinson and Stuart, 2005). Social embeddedness is therefore likely to reduce the chances of opportunism and hence the transaction costs associated with syndication will decrease. Potential collaborative partners are therefore more likely to engage in interfirm cooperation where exchange partners are embedded in a broader social network of relations. Therefore, it is argued that the negative effect of primary uncertainty on interfirm cooperation is positively moderated by the social embeddedness of an exchange partner. Hence: H2b: The social embeddedness of an exchange partner will positively moderate the relationship between primary uncertainty and syndication. DATA AND METHODS Data 17

18 One of the major constraints in research on buyout investments is gaining access to the often highly confidential and sensitive data concerning deal structures, given the private nature of the transaction. The buyout deals for this study are identified through a hand collected database maintained by the Centre for Management Buyout Research (CMBOR). This database covers the entire population of buyouts in the UK from the beginning of the 1980s onwards. In order to enhance reliability, the data are drawn from a variety of sources so as to reduce common method bias. 5 The sample consists of 2738 buyouts that occurred between 1993 and 2001 in the UK and that were backed by at least one private equity investor. 6 For each transaction, data on the investment was collected by CMBOR. Characteristics of buyout investors were primarily found in directories issued by the British Venture Capital Association (BVCA) and the European Venture Capital Association (EVCA). 7 Additional information sources were used to collect information when individual investor characteristics were missing, notably the Guide to Venture Capital in the UK and Europe and Venture Economics. The data with individual deal characteristics and the data with characteristics of private equity firms were then merged into one dataset with separate records for each investment, which is the unit of analysis. These separate records contain information on individual deal characteristics and characteristics of the lead investor in the year the deal was completed. 8 5 Most importantly, a semi-annual survey is conducted with organizations such as banks and private equity companies investing in buyouts. All participants in the market respond to this survey, enabling a comprehensive picture of the population to be obtained. This data collection method enables private information on full details of individual financing structures to be obtained. This data is further completed with data from the business press. 6 We thus exclude buyout transactions that did not involve a private equity investor, but were solely financed with debt. 7 The directories issued by the BVCA cover two yearly periods for , , and The directories of EVCA are yearly. 8 Matching deal characteristics with characteristics of the lead investor assumes that the lead investor initiates the deal and decides whether or not the deal will be syndicated. This is in line with observed practice in the private equity industry. If the lead investor was not explicitly mentioned, the investor with the largest equity investment was assigned the lead role. Previous research has generally found that lead 18

19 Measures The dependent variable in our analysis is binary and indicates whether or not the deal is syndicated. A value of 1 is attributed if the deal was syndicated and 0 if not. The independent variables measure primary uncertainty, reputation and social embeddedness. Control variables are added. Primary uncertainty We use three measures that capture different dimensions of the primary uncertainty associated with the underlying transaction i.e. the buyout investment. A main advantage of the variables employed is that they represent ex-ante indicators of the expected uncertainty associated with the transaction. Leverage. A higher leverage of a deal is associated with lower overall uncertainty. First, high leverage signals the potential for future cash flow generation to service high debt levels. Highly levered transactions are therefore characterized by less operational or market uncertainty. Second, leverage is positively associated with the liquidation value of assets, lowering the downside risk associated with failure (Williamson, 1988). Finally, the disciplining effect created by high debt levels in a buyout reduces agency risk (Jensen, 1986). Our first measure of uncertainty is therefore the total amount of senior secured debt, high yield debt and subordinated (mezzanine) debt as a percentage of total financing used to structure the buyout transaction. The negative of this variable is taken for ease of interpretation. A higher value points to more primary uncertainty. investors on average hold larger equity stakes compared to non-lead investors (Barry et al., 1990; Wright and Lockett, 2003). 19

20 Logarithm of absolute amount invested by management. One source of uncertainty in buyout investments is the extent of agency risk posed by the management of the buyout company. Agency risk attributable to the separation of full ownership and control varies inversely with the management s ownership stake (Jensen and Meckling, 1976). In the context of buyouts, increased management stockholdings have usually been associated with performance improvements in the years following the buyout (Phan and Hill, 1995). Traditionally, the percentage of equity held by management has been employed as a proxy for agency risk. With respect to buyouts, however, it is more difficult to get a substantial stake of the equity in very large transactions compared to small transactions, given that managers are wealth constrained. Further, Sapienza and Gupta (1994) and Kaplan and Stein (1993) have suggested that the absolute investment made by management may be a better predictor for agency risk compared to the equity percentage held. Therefore, the absolute amount invested by management is used as an indicator of the uncertainty posed by the management of the company. We log transformed this variable and multiplied the value by -1 for ease of interpretation. A higher value indicates a higher level of primary uncertainty. Ratchet dummy. Contingent contractual provisions such as equity ratchets are another mechanism to cope with the uncertainty associated with the investment. An equity ratchet is an antidilution provision that helps to protect the investor in bad states by increasing the number of shares the investor has. It also addresses the problem of agency risk by tying the management s equity share to the performance of the firm (Thompson and Wright, 1991). Kaplan and Strömberg (2004) empirically show that equity ratchets are more often used in venture capital deals when the internal risk associated with the venture is high. Internal risk reflects issues such as uncertainty with respect to the capabilities of the management team or uncertainty with respect to the 20

21 financial projections of the company. Therefore, our third uncertainty measure is a dummy variable where a value of 1 indicates the use of an equity ratchet in the deal structure and 0 otherwise. Again, a higher value indicates a higher level of primary uncertainty. Reputation Two different measures are used to capture the reputation of the investor initiating the deal. Logarithm of number of previous investments. Reputation is measured by counting the number of investments by the lead investor from the first observations in the dataset until the year prior to the investment. This measure includes investments from the early eighties onwards. This measure also captures the ability of private equity firms to structure and manage deals and their ability to evaluate general aspects of the companies business plan and management team which are not specific to the industry. 9 The logarithm of this measure is used. Logarithm of number of times lead. The number of times a private equity firm has acted as lead arranger in an investment syndicate with other private equity companies is an indicator of its reputation for being a lead investor. This measure is related to the reputation measure used for investment banks in underwriting syndicates (Megginson and Weiss, 1991) and the reputation of banks in lending syndicates (Lee and Mullineaux, 2004). This measure was calculated from the start of the observation period onwards i.e. the beginning of the 1980s. 9 In order to check the robustness of the results, we also used an industry specific reputation variable which measures the number of investments the private equity company had in the same industry of the target company prior to the year of the investment as recorded in the complete CMBOR dataset. A distinction is made between 35 different industries. The logarithm of this number is used. The results remain qualitatively the same. 21

22 Social embeddedness The embeddedness of a firm in the network of syndicate relationships is measured by the firm s degree centrality (Freeman, 1979), i.e. the number of relations a firm has with different private equity firms. This measure is a good indicator of the amount of information that is available on a particular firm and is often used in studies on director interlocks (e.g. Beckman and Haunschild, 2002). A five-year moving window is used to calculate this variable. The length of this window is chosen based on the average lifespan of a syndicate relationship. 10 To compute this measure, we construct adjacency matrices representing the syndication networks in each year based on the syndicate relationships in the previous five years. We only count the number of relations the lead investor has with different non-lead members of a syndicate. The relations between nonlead investors are not included as non-lead investors mainly interact with the lead investor (Wright and Lockett, 2003). All matrices serve as input into UCINET 6, a network software package, to calculate a firm s degree centrality (Borgatti, Everett and Freeman, 2002). This variable is lagged one year. We use the logarithm as this variable is highly skewed. Control variables Several control variables are included in the multivariate analyses. First, as we develop hypotheses related to the interaction effect of behavioral uncertainty with primary uncertainty, and not with respect to the main effect of behavioral uncertainty, our reputation and social embeddedness variables are included as controls to test their main 10 Our data indicates that the average time to exit for a syndicated investment lies between 3 and 4 years. This figure, however, does not take into account right censoring as we do not know all the realized exits. 22

23 effect on syndication. For example, private equity companies that have no established reputation may want to be associated with more reputable investors through syndication (Lerner, 1994). Further, because of resource constraints and portfolio risk considerations, it is more likely that larger deals will be syndicated in order to allow for portfolio diversification (Manigart et al., 2005). Larger deals are further expected to get more media attention and therefore attract more potential investors increasing the likelihood of syndication. The size of the deal is measured by the total value of the deal including both debt and equity investments. We take the logarithm of this measure. We also include a relative measure for the size of the deal that takes into account the size of the funds under management of a private equity firm. This measure is calculated by taking the total amount of equity used to finance the deal minus the amount invested by management, and dividing it by the total amount of funds available to a private equity company. This variable represents the potential concentration of the deal within the portfolio when there is no syndication. The higher this figure, the more likely a deal will be syndicated as not syndicating it would decrease the diversification of the portfolio. We also include a measure that captures the size of the private equity company namely the number of investment executives. In order to capture repetitive momentum or unobserved heterogeneity in the proclivity to syndicate, the number of deals done as a sole investor relative to the total number of deals done in the two years prior to the year of the investment is used (Heckman and Borjas, 1980). This variable is expected to have a negative sign, i.e., the more a private equity firm invests on a standalone basis, the less likely it will syndicate future deals. Therefore, we used an average of five years. We tested the sensitivity of this assumption by using a 23

24 To control for unobserved temporal factors that may influence syndication activity, dummy variables were included for each year in the observation period. Moreover, industry differences are controlled for by introducing separate dummy variables for each of the 35 activity sectors. Several control variables are included to account for the heterogeneity observed among buyouts. First, a distinction is made between management buyouts, management buy-ins, a combination of a buy-in and a buyout, investor-led buyouts and others such as public buy-ins since there may be different informational asymmetries according to whether the transaction involves insider or outsider management (Wright and Robbie, 1998). Additionally, dummy variables are introduced to identify the vendor source of the buyout since these too may involve different informational asymmetries (Wright, et al., 2000). The following categories are identified: foreign divestments, local divestments, private buyouts, public to private buyouts, secondary buyouts, receiverships, privatizations and a category for which the source is unknown. A dummy is also included to measure if management has more than 50% of the total equity. More management control could decrease the willingness of private equity investors to syndicate as this would further reduce their control. Further, if management has a majority stake, private equity firms will have less power to take decisions. As a result, syndicate partners may be less willing to join a syndicate as they are more dependent on the management of the buyout company. 11 Lastly, separate dummy variables are included for the most active players in the buyout market. Each of these investors was involved in more than 30 buyouts in our sample. 12 moving window of 4 years and 6 years. The results stay the same. 11 The robustness of this variable is tested by using cut-off points at both 40% and 60%. The results are qualitatively the same. 24

25 Sample description For each of the private equity backed buyouts identified, it is known whether the deal is syndicated or not. There are, however, missing values for some of the explanatory variables, particularly those relating to financial measures which are not always available for reasons of confidentiality. The traditional method of dealing with missing values, used here, is to apply complete case analysis, i.e. to exclude all cases for which one or more of the variables is missing, thereby assuming the missing values being missing completely at random (Little and Rubin, 1987). In the present case, the remaining sample is not a completely random sub-set of the total sample. In particular, the subsample with complete cases has a greater proportion of syndicated deals (28.6%) than the total sample (27.2%). A t-test indicates, however, that this difference is not statistically significant. Further, as we employ logistic regression, listwise deletion is problematic only when the probability of any missing data depends on both the dependent and independent variables (Allison, 2002). There are no reasons to suspect that missing values in this dataset are related to both the dependent and the independent variables. Note that in order to compute the reputation and social embeddedness variable, the full dataset is used. INSERT TABLE 1 ABOUT HERE The sample consists of 732 private equity backed buyout transactions in which 64 private equity firms participated either as sole investor or as lead investor. The summary statistics for the characteristics of the private equity firms in our sample are shown in 12 In total there are 13 private equity firms with more than 30 buyouts in our sample. 25

26 table The cumulative number of previous investments by private equity firms in our sample is on average 66. As the standard deviation indicates, there is a huge difference between the private equity firms in the sample. One investor, namely 3i, was on average involved in more than 1574 investments in the year prior to its investments. The median cumulative number of investments is 23 for each private equity company. The number of times a firm has acted as lead investor in the years preceding the investment is approximately 13 with a minimum of 0 and a maximum of 227. The average number of different partners a private equity firm has invested with in the 5 years preceding the year of the buyout, the degree centrality, is slightly more than 1.5, ranging between 0 and 55. The average number of investment executives is a little over 11. The percentage of deals done as sole investor in the two years preceding the deal is on average about 50%. The number of investments per private equity firm included in the sample is slightly more than 12 with a minimum of 1 and a maximum of 172. Overall, the standard deviations indicate that there is considerable heterogeneity among the private equity firms in our sample. INSERT TABLE 2 ABOUT HERE The summary statistics for the investments included in the analyses are shown in table 2. On average, almost 29% of the deals in our dataset are syndicated. The average absolute amount invested by the management totals 533,000 with a minimum of 10,000 and a maximum of almost 35 million. The average leverage, i.e. debt to total 13 In order to calculate these statistics, for each private equity company the mean was calculated over the years in which it made investments as recorded in the sample. Table 1 then presents the mean of these firms means. 26

27 enterprise value, of the deals is 59%. Further, an equity ratchet is used in slightly more than 37% of the buyouts. The mean value of a deal is a little higher than 27 million with a huge diversity of values as indicated by the standard deviation. The largest deal is worth 860 million, while the smallest deal is worth only 100,000. The relative size measure indicates that the total amount of equity financing needed was slightly more than 4% of the funds under management. In general, these figures clearly illustrate the degree of variation among buyouts which has often been neglected in US studies (see e.g. Kaplan and Stein, 1993). Table 3 presents bivariate analyses comparing syndicated with non syndicated deals. The absolute amount invested by management is higher for deals that are not syndicated compared to deals that are syndicated. The difference is not statistically significant. Leverage is almost identical for syndicated versus non-syndicated deals. Ratchets are used significantly more frequently in syndicated versus non-syndicated deals which contrasts with hypothesis 1. None of the reputation measures is significantly different for syndicated versus non-syndicated deals. The social embeddedness variable as measured by the degree centrality of the lead investor is significantly higher for syndicated versus non syndicated deals. All the control variables are significantly different for syndicated versus non-syndicated deals. Syndicated deals are on average larger than non-syndicated deals as expected. The relative size of the deal is higher for syndicated versus not syndicated investments. The number of investment executives is larger for private equity companies backing deals that were not syndicated. Not surprisingly, private equity firms backing non-syndicated deals act more as sole investors compared to those who back syndicated deals. INSERT TABLE 3 ABOUT HERE 27