The Changing Nature of Investment Banking Relationships

Transcription

1 The Changing Nature of Investment Banking Relationships Shane A. Corwin * Mendoza College of Business University of Notre Dame Notre Dame, IN scorwin@nd.edu Mike Stegemoller Hankamer School of Business Baylor University Waco, TX michael_stegemoller@baylor.edu April 2014 Abstract We examine relationships between firms and their investment banks (IBs), both within and across different IB functions. For all transaction types, we find an increase in the number of relationships and a decrease in relationship exclusivity over time. We identify significant transaction-type specific relationship components, with relationships being strongest for equity and weakest for M&A. Relationships are less exclusive for firms with more bargaining power and more exclusive for younger, smaller firms, and firms with more growth options. Most importantly, we find strong evidence of a firmwide relationship component that spans the functional areas of I-banking and increases over time. JEL classification: G10, G24, G34, L14 Keywords: Investment Banking, Underwriting, M&A Advising, Underwriter Relationships * We thank Robert Conway, Jack Cooney, Paul Schultz, Sophie Shive, and seminar participants at the University of Notre Dame, Baylor University, and Babson College for comments. Steven Carroll, Brian Ford, and Travis Johnson provided excellent research assistance. Any remaining errors are the responsibility of the authors.

2 1. Introduction Relationships between firms and their investment banks are generally managed centrally within the investment bank (IB). However, much of the prior literature analyzes only a specific area of investment banking, such as equity or debt underwriting, or M&A advising. 1 In addition, much of the previous research focuses on factors that influence the decision either to hire a specific IB or to switch lead underwriters on consecutive deals. In this paper, we provide a more comprehensive analysis of IB relationships both within and across the various IB functions by analyzing over 20,000 M&A transactions and public and private security issues by U.S. exchange-listed firms between 1996 and These data allow us to address three broad research questions. First, what are the principal characteristics and determinants of IB relationships? Second, how have IB relationships changed over time? And, third, to what extent do relationships carry over across the various aspects of the IB business? An IB relationship can constitute a wide range of functions, including securities underwriting, M&A advising, lending, derivatives contracts, transaction services such as cash management, and other services and advice that may not be observable to the researcher. As a proxy for the overall IB relationship, we focus on securities underwriting and M&A advisory mandates, as well as ties between these functions and lending. We expect these observable relationship components to capture the most important aspects of IB-firm relationships. While we separate some of our analysis by transaction type, we emphasize the firm-wide nature of IB relationships and the carryover of relationships across IB functions. We believe this firm-level analysis is appropriate for two reasons. First, while the responsibility for hiring decisions on various types of deals may fall to different people within a client-firm, top level executives are generally involved in the selection of the underwriters and M&A advisors that are the focus of our analysis. Second, from the perspective of the IB, these relationships tend to be handled on a firm-wide basis with one banker acting as the point of contact for the client-firm and bringing in personnel from specific product teams as appropriate. The extent to which relationships encompass multiple IB functions will depend largely on what 1 Exceptions include Burch, Nanda, and Warther (2005), Ljungqvist, Marston, and Wilhelm (2006), and Bharath, Dahiya, Saunders, and Srinivasan (2007). A more detailed discussion of the prior literature is provided in Section 2. 1

3 drives the relationship from the perspectives of both the client-firm and the IB. The literature suggests that the value of IB relationships derives primarily from information production. For example, James (1992) provides evidence that set-up costs in the due diligence process lead to reduced underwriting costs for IPO firms that are expected to issue additional securities in the future. These types of information effects lead to switching costs in IB hiring decisions and may also cause firms to avoid using the same IB as their primary product market rivals (Asker and Ljungqvist (2010)). Further, prior research suggests that information plays a more important role in equity relationships than in other IB relationships (see, Burch, Nanda, and Warther (2005) and Fernando, May, and Megginson et al. (2012)). Thus, if information effects dominate, we expect stronger relationship effects in equity than in debt transactions and we expect equity relationships to be a primary determinant of firm-wide IB relationships. 2 While information is also likely to play a predominant role in M&A transactions, these transactions rely first and foremost on ideas, which may not come from the primary relationship bank. We expect this difference in transaction initiation to lead to weaker relationships in M&A advisory services than in underwriting services and a weaker link between M&A advisory relationships and firm-wide IB relationships. In contrast to these information-based stories, Fernando, Gatchev, and Spindt (2005, 2013) argue for transaction rather than relationship-based associations between firms and underwriters, with the two sides choosing to part ways when their relative quality no longer aligns. One possible interpretation of their results is that firms will select the IB most suited to a particular transaction type, rather than rely on a firm-wide relationship. In this paper, we examine directly the extent to which relationships carry over across different IB functions. Our sample period encompasses a number of important changes to the structure and economics of the IB industry. The passage of the Gramm-Leach-Bliley Act in 1999 led to a substantial increase in the role of commercial banks in investment banking and more direct ties between lending and underwriting relationships. The link between lending and IB relationships was further magnified by the recent financial crisis, which led many firms to seek out additional sources of capital. We expect this shift to increase the 2 Throughout the paper, we use the term firm-wide relationship to denote the component of IB relationships that spans the functional areas within an investment bank. IB relationships that are related only to a specific transaction type, such as equity, debt, or M&A, are referred to as transaction-type specific relationships. 2

4 importance of debt and lending in IB relationships. The dot-com bubble and the subsequent Global Settlement in 2003 also caused considerable changes in the business models of IBs and in the related determination of underwriting mandates. Against this backdrop, there have been significant changes in the structure of underwriting syndicates. For example, Hu and Ritter (2007) document a rise in the use of multiple bookrunners in IPOs, and Shivdasani and Song (2011) find a similar increase in the frequency of joint leads in the debt underwriting market. Corwin and Schultz (2005) also document a significant increase in the number of comanagers in IPOs and a simultaneous decrease in the number of non-managing syndicate members. Our analysis allows us to examine the impact of these changes on the characteristics and determinants of IB relationships, both within and across transaction types. We begin our analysis with a comprehensive examination of the characteristics and determinants of IB relationships. We find that both the total number of IBs that a firm works with and the number of IBs involved in at least 50% of the firm s deals increase over time for all types of transactions, but these trends are most significant for debt transactions. Assigning each lead underwriter/advisor full credit for each deal, we find weak evidence of an increase over time in market share for the top relationship bank. However, after adjusting for joint appointments, we find a significant decrease in market share for the top bank, in combined market share for the top three banks, and in a market share Herfindahl index. These results suggest that the changing structure of investment banking has led to an increase in the number of IB relationships per firm, but these relationships are less exclusive, on average. Using panel regressions, we find that relationships are more exclusive in cases where information is likely to be more important, especially for equity issues. For example, relationships tend to be more concentrated for firms with high market-to-book ratios and high idiosyncratic volatility. We also find that relationships are less exclusive for firms with more bargaining power, as measured by firm size or the firm s fraction of total equity, debt, and M&A deal value. These results are consistent with the informational role of IB relationships and with the existence of informational hold-up problems (see Rajan (1992) and Huston and James (1996)). Debt relationships are also less exclusive for firms with higher leverage ratios, consistent with high debt firms maintaining multiple sources of capital. Even after controlling for other factors, we find 3

5 strong evidence that the exclusivity of both firm-wide and transaction-type specific IB relationships decreases significantly during our sample period. This decrease in exclusivity contrasts with the IB industry overall, for which concentration has remained relatively flat or even increased over time. Next, we extend existing research that examines why and how often firms switch underwriters. We find over our entire sample period that firms retain at least one lead underwriter (advisor) in 48.4% of consecutive M&A deals, 55.2% of debt deals, 69.7% of SEOs following the IPO, and 53.3% of SEOs following other SEOs. These results are generally consistent with prior studies. However, our data also allow us to analyze IB retention across transactions of different types. For example, at least one IPO lead is retained in 49.2% of subsequent M&A deals and 60.5% of subsequent debt deals. 3 A probit model of lead banker/advisor retention shows that the likelihood of retention for equity and debt deals increases significantly if the prior deal was of the same type. Thus, while we find evidence of substantial carry-over in relationships across IB functions, there also appears to be a significant transaction-type specific component, with retention being strongest for equity deals and weakest for M&A deals. While prior studies report a decrease in underwriter retention over the 1980s and 1990s, we find that the likelihood of IB retention has increased during our sample period, especially for debt issues. These results are, in part, a reflection of the increase over our sample period in the use of joint lead managers and larger groups of comanagers. The results may also reflect an increased link between IB relationships and the provision of lending services. Prior research suggests that past relationships with firms have a significant effect on the selection of underwriters and M&A advisors, with some limited evidence that relationships may carry over across IB functions. In our final set of tests, we extend these prior studies by analyzing the firm-wide nature of IB relationships and the extent to which relationships in one functional area carry over to other functional areas. In a frequency analysis for firms that have at least two deals of each type, we find that the top equity IB is also the top debt IB in 62.4% of cases and the top M&A advisor in 56.9% of cases. Similarly, the top debt IB 3 The conclusions are generally similar if comanager and non-advisory roles are included, with 67.0% of M&A deals, 84.9% of debt deals, and 88.2% of equity deals retaining at least one lead or co-manager from the IPO. Krigman, Shaw, and Womack (2001) find that 48.9% of firms that switch underwriters retain the previous lead underwriter as a comanager, and 25.6% of switchers use a previous comanager as the new lead. 4

6 is also the top M&A advisor in 50.3% of cases. Like the IB retention results, these findings suggest that relationships span the functional areas within the IB. However, we also find numerous cases where the top bank in one area plays no role in other areas. These results point to a significant transaction-type specific component in IB relationships, with the carry-over in relationships being strongest between equity and debt underwriting and weakest for M&A advisory services. We also find that the firm-wide nature of relationships is stronger when the IB also lends to the firm, especially in the latter part of the sample period. Probit models for IB selection support the importance of both firm-wide and transaction-type specific relationships, with the firm-wide component increasing over time. For example, from , the probit model coefficients suggest that the likelihood of being hired as lead equity underwriter for an IB that was on the prior deal and has a 50% equity relationship is 15.8%. Adding a 50% relationship in debt, M&A, or lending increases the likelihood to between 21.2% and 30.6%, and adding 50% relationships in all three areas increases the likelihood to 48.1%. From , the likelihood of being hired as lead equity underwriter for an IB that was on the prior deal and has a 50% equity relationship is 16.4%. Adding a 50% relationship in debt, M&A, or lending increases the likelihood to between 29.8% and 33.3%, and adding 50% relationships in all three areas increases the likelihood to 71.0%. Overall, our results are consistent with important relationship effects in IB hiring decisions. More importantly, these relationships incorporate both firm-wide and transaction-type specific components, with both the firm-wide nature of relationships and the link between relationships and lending increasing over time. 4 The remainder of the paper is organized as follows. In Section 2, we provide a brief overview of prior research related to investment banking relationships. Section 3 describes the data and sample characteristics, including the time trends in underwriting syndicate and M&A advisory group structure. In Section 4, we analyze the characteristics and determinants of IB relationships. Section 5 presents our results related to IB retention and Section 6 examines the extent to which IB relationships carry over across deal types and the impact of these relationships on IB selection decisions. Section 7 concludes. 4 Consistent with Asker and Ljungqvist (2010), we also find that the effects of past relationships are weakened when the investment bank in question is the primary relationship bank of the firm s main product market rival. 5

7 2. Background The importance of ongoing relationships in investment banking has been well documented going back at least to Eccles and Crane (1988). Existing research has addressed a number of related issues, including the impact of relationships on IB hiring and retention decisions, the impact of relationships on the pricing of IB services, and the link between lending and IB relationships. The importance of IB-firm relationships is highlighted by the significant role that past relationships play in the selection of underwriters and M&A advisors. For example, Ljungqvist et al. (2006) and Bharath et al. (2007) find that the likelihood of winning a debt or equity underwriting mandate is positively related to past underwriting relationships with the firm. 5 Similarly, for M&A transactions, Bao and Edmans (2011) find that past M&A market share is the primary determinant of advisor selection and Forte, Iannotta, and Navone (2010) find that firms are more likely to select as advisor the IB with which they have the strongest past relationship. 6 Many of the previous studies on IB relationships focus on the firm s decision to retain the lead underwriter across consecutive deals, with a particular emphasis on IB retention following the IPO. Taken together, these studies suggest that the frequency of underwriter switching has increased over time, from approximately 25-30% in the 1980s, to roughly 60% by the early 2000s (see, for example, Beatty and Ritter (1986), Welch (1989), James (1992), Dunbar (2000), Krigman, Shaw, and Womack (2001), Ljungqvist and Wilhelm (2005), and Burch, Nanda, and Warther (2005)). Evidence on the determinants of underwriting switching is more mixed. Beatty and Ritter (1986), Dunbar (2000), and Ljungqvist and Wilhelm (2005) present evidence that IB switches are related to IPO pricing or the perception of poor IB performance at the IPO. In contrast, Krigman et al. (2001) find that switches are not related to IPO performance, but instead involve firms switching toward higher-quality IBs or IBs offering better analyst coverage. Consistent with the role of non-price factors in IB hiring and 5 Relationships among investment banks are also important. For example, Corwin and Schultz (2005) find that an underwriter s chances of being included in an IPO syndicate increase if the underwriter has a previous relationship with the IPO lead underwriter. 6 Prior research also examines the decision to hire investment banks in M&A advisory roles and the relation between advisors and M&A performance. Servaes and Zenner (1996) find that investment banks are more likely to be employed as advisors by acquirers with less prior acquisition experience and when transaction complexity is high. While Bowers and Miller (1990) and Rau (2000) find little evidence that top-tier investment banks provide superior acquisition performance, Kale, Kini, and Ryan (2003) and Boone and Mulherin (2008) find evidence that prestigious advisors provide benefits to the firms that hire them. Bao and Edmans (2011) find a persistent advisor-specific component in M&A performance, but find that this component has little impact on M&A market share. 6

8 retention, Ljungqvist et al. (2006, 2009) find that analyst optimism affects underwriter selection. Although Ljungqvist et al. (2006) find no evidence that optimistic analyst coverage improves an IB s chances of earning a lead underwriting mandate, Ljungqvist et al. (2009) find that optimistic analyst coverage does improve an IB s chances of earning co-managing appointments, which in turn increase the bank s chances of winning future lead underwriting mandates. To capture the multidimensional nature of IB services, Ellis, Michaely, and O Hara (2011) develop a measure of IB competitiveness across multiple functions, including fees, pricing, analyst coverage, market making, debt capacity, and reputation. They then test the relation between IB competitiveness and underwriter switching in seasoned equity offers. Consistent with prior studies, they show that underwriters providing poor service are less likely to be retained. Further, they find that firms making lateral IB switches see more competition for their business, while poorly-performing firms and firms upgrading to higher quality IBs see less IB competition and face higher switching costs. Prior research also suggests that lending plays an important role in IB relationships. In particular, the passage of the Gramm-Leach-Bliley Act in 1999 led to a substantial increase in the role of commercial banks in investment banking and more direct ties between lending and underwriting. 7 For example, Ljungqvist et al. (2006), Drucker and Puri (2005), Yasuda (2005), and Bharath et al. (2007) find that lending relationships increase the likelihood of a bank being awarded debt and equity underwriting business. Existing studies also suggest links between lending relationships and IPO underpricing (James and Wier (1990) and Schenone (2004)), the long-run performance of IPOs (Benzoni and Schenone (2010)), SEO pricing (Narayanan, Rangan, and Rangan (2004)), and SEO underwriting fees (Drucker and Puri (2005)). 8 While our paper contributes to this literature along a number of dimensions, it is most closely related to Burch et al. (2005), Ljungqvist et al. (2006), and Bharath et al. (2007). Using a sample of securities issues 7 Numerous studies address the entry of commercial banks into investment banking and the resulting competition between commercial banks and traditional I-banks. See, for example, Ang and Richardson (1994), Puri (1996, 1999), Saunders (1999), Gande, Puri, Saunders, and Walter (1997), Gande, Puri, and Saunders (1999), Roten and Mullineaux (2002, 2005), Fields, Fraser, Bhargava (2003), Song (2004), Drucker (2005), Kim, Palia, Saunders (2008), Chaplinsky and Erwin (2009), and Papaioannou (2011). 8 Drucker and Puri (2005) find that concurrent lending at the time of an SEO leads to discounted yields on loans and discounted underwriting fees on the SEO, with commercial banks primarily offering yield discounts and IBs primarily offering fee discounts. In contrast, Calomiris and Pornrojnangkool (2009) find that the effects of bundled services depend on the sequence of transactions. In general, when banks provide both lending and underwriting to the same client, they charge premiums on both products. However, when the loan follows the equity issue, it is IBs rather than commercial banks that provide discounted loan pricing. 7

9 from , Burch et al. (2005) find that the strength of IB relationships decreased during the 1980s and 1990s for both equity and debt issues. They also find that stronger relationships are associated with reduced underwriting fees for equity issues but higher fees for debt issues. Ljungqvist et al. (2006) focus on the effects of analyst coverage on the selection of lead underwriters in equity and debt offers from 1993 through 2002, but also control for the effects of previous lending and underwriting relationships. Their results suggest that both past underwriting and lending relationships have significant effects on underwriter selection, with same-transaction-type relationships having a larger impact than cross-market relationships. Finally, Bharath et al. (2007) use a sample of borrowers from 1986 through 2002 to analyze the effects of lending relationships on a bank s ability to win future lending and underwriting business. Their results support the importance of past lending and underwriting relationships in winning future business, but suggest that the economic impact of cross-market relationships is small. Our analysis differs from these prior studies in several important respects. First, we examine securities issues, M&A activity, and lending for a broad sample of firms and IBs, allowing us to provide a more comprehensive analysis of IB relationships and to address whether these relationships carry over across transaction types. Second, we incorporate additional analysis of comanaging and non-advisor roles, and the link between lending and investment banking, which have become increasingly important over time. Finally, our sample period includes the period from the end of the dot-com bubble through the financial crisis, during which substantial changes occurred in the investment banking industry. This sample period allows us to examine the changing economics of the investment banking industry and the associated effects on IB-firm relationships. Our results point to a strong and growing firm-wide component in IB relationships, which encompasses equity and debt underwriting, M&A advising, and lending activities. 3. Data and Sample Characteristics A. Sample Firms and Transactions In constructing the sample, our goal is to identify all security issue and M&A events for which a firm might hire an investment banker. To identify firms, we use CRSP to collect the sample of U.S. firms with listed common stock (CRSP share codes 10 or 11) on any dates between 1996 and Because our 8

10 analysis is at the firm level, we identify CRSP firms based on the permanent company number (PERMCO). This search produces an initial sample of 11,502 firms. The security issue and M&A events for these firms are identified using data from SDC. To begin, we use the SDC new issues database to collect all public and private security issues by U.S. firms between 1996 and Next, we use the SDC M&A database to collect all transactions with effective dates between 1996 and 2009 for which either the target or acquirer is a U.S. firm (public or private). The resulting sample includes 140,677 transactions, 18,229 equity issues, 268,784 debt issues, and 4,536 issues of other securities types, such as preferred stock, units, etc. This initial sample and the subsequent restrictions applied to obtain the final sample, as described below, are shown in Table A1 in the appendix. Because we are working at the firm level with multiple types of securities, we cannot match CRSP and SDC data based simply on Ticker symbol or CUSIP. We therefore define firms based on the CRSP PERMCO and match to SDC using a firm-level identifier called CIDGEN. 9 To begin the matching process, we identify the date range associated with every unique PERMCO-Ticker-CUSIP combination in the CRSP data. An analogous process is performed with the SDC data using CIDGEN. Specifically, we identify the earliest and latest event dates for each unique CIDGEN-Ticker-CUSIP combination in the equity, debt, and M&A event samples. To create PERMCO-CIDGEN matches, we begin by matching the equity, debt, and M&A samples to the CRSP sample based on CUSIP and Ticker. For unmatched CIDGENs, we then attempt to match based on CUSIP only, and then based on Ticker only. For each PERMCO, we keep the highestlevel CIDGEN match identified across the three event samples. Finally, we manually check all CUSIP-only and Ticker-only matches, as well as matches for which the CRSP date range does not correspond to the SDC date range for that firm. After making corrections, this process results in 14,480 PERMCO-CIDGEN matches, of which 13,018 (89.9%) are matched based on CUSIP and Ticker, 1,274 (8.8%) based on CUSIP only, and 188 (1.3%) based on Ticker only. 9 CIDGEN is a 10-digit firm-level identifier provided for every event in SDC. While the last three digits can change across events, the first seven digits are unique to the firm and appear to change only as the result of major firm-level changes, such as reorganizations or mergers. While changes in 7-digit CIDGENs tend to correspond with changes in CRSP PERMCOs, this is not always the case. In those cases where the CIDGEN number changes and the PERMCO does not, we match both CIDGENs to the same PERMCO. In the small number of cases where the PERMCO changes and the CIDGEN does not, we redefine the CIDGENs to create separate SDC identifiers for events involving the two PERMCOs. 9

11 The PERMCO-CIDGEN matches described above allow us to identify all SDC events associated with each firm (or each unique CRSP PERMCO). Using this process, we are able to assign a PERMCO to 80.9% of equity issues, 39.6% of debt issues, 64.4% of other security issues, 40.9% of M&A acquirers, and 22.0% of M&A targets. Once SDC events are matched to firms, we eliminate duplicate events and apply several additional restrictions. We exclude financials, utilities, and government agencies, as well as firms that list during the last year of the sample or delist during the first year of the sample. These restrictions reduce the number of firms in our final sample to 8,322. We further restrict the set of SDC events by applying the following criteria: (1) aggregate same-day debt offers by the same firm using the same lead IBs, (2) exclude purely secondary equity issues, (3) exclude M&A events for which the acquirer owns more than 50% of the target prior to the event or less than 50% after the event and restrict M&A form to acquisition, merger, acquisition of majority interest, or acquisition of assets 10, (4) exclude all events associated with captive finance subsidiaries 11, and (5) exclude all events that occur more than five days outside the CRSP date range associated with that PERMCO. These additional restrictions result in a sample of 60,458 events, including 7,316 equity issues, 7,614 debt issues, 650 securities issues of other types, 32,451 M&A acquirer events, and 12,427 M&A target events. To analyze lending activity for our sample firms, we match the firms identified above to DealScan using the link table provided by Michael Roberts and Wharton Research Data Services (see Chava and Roberts (2008)). 12 Our initial DealScan sample includes all loan facilities by U.S. firms from 1996 through 2009, excluding loans by financial firms, utilities, and government agencies. After matching to the sample firms, the final loan sample includes 29,469 loans by 4,964 firms. The sample is described in more detail in the first three columns of Table I. In total, the sample accounts for more than $17 trillion in aggregate transaction value. Public and private security issues 10 We exclude M&A events classified as acquisitions of certain assets, partial interest, or remaining interest. 11 We view captive finance subsidiaries as distinct entities that are similar to financial firms and may have their own IB relationships. As a result, we separate them from their parent firms and exclude them from the sample. The most active finance subsidiaries in our sample are GE Capital, GMAC, Caterpillar Financial Services, and John Deere Capital. For a discussion of captive finance, see Bodnaruk, O Brien, and Simonov (2012). We thank Andriy Bodnaruk for providing data identifying these subsidiaries. 12 DealScan consists primarily of large syndicated loans by medium and large firms. According to Carey and Hrycray (1999), DealScan covered 50-75% of U.S. commercial loans in the early 1990s, with coverage increasing from 1995 onwards. For a more complete description of the DealScan data, see Ivashina (2009) and Chava and Roberts (2008). 10

12 represent an aggregate transaction value of $3.54 trillion. Debt transactions dominate the security issue sample, accounting for 49% of all issues and 77% of total transaction value. Within the equity sample, public seasoned offerings dominate, accounting for 43% of issues and over 60% of aggregate transaction value. IPOs account for about one third of the equity issues by both frequency and aggregate transaction value. While private equity offers account for about 20% of total equity issues, these offers tend to be small, representing only 6% of aggregate equity transaction value. The debt sample is concentrated in nonconvertible offers, with public nonconvertible debt accounting for 45% of transactions and 59% of total transaction value and private nonconvertible debt accounting for 35% of transactions and 27% of aggregate transaction value. Aggregate transaction value in the M&A sample is significantly larger than that in the security issue sample. Of the 32,451 acquirer events in the sample, transaction values are available for 17,547 (54%) with an aggregate value of $5.50 trillion. 13 Of the 12,427 sample target events, transaction values are available for 8,172 (66%), reflecting an aggregate value of $6.35 trillion. For comparison, lending activity is described in the last row of Table I. The aggregate loan value of $8.07 trillion is more than double the aggregate security issue value, with an average loan size of $273.8 million. The time series of aggregate transaction values for each transaction type are plotted in Figure 1. While security issuance frequencies tend to decrease over time, aggregate transaction values follow more variable patterns, with equity issuance peaking in and debt issues being higher in the early 2000s and toward the end of the sample period. M&A transaction values increase sharply in the late 1990s, decrease through the mid-2000s, and peak again in Table II describes firm-level activity for the sample of 8,322 firms. Of these firms, all but 768 have at least one transaction during the sample period and the mean (median) number of transactions is 7.3 (4.0). The firm at the 99 th percentile is involved in 57 combined security issue and M&A transactions, with the most active firm accounting for 225 combined transactions. Of the sample firms, 64.3% have at least one security issue, 67.1% are acquirers in at least one M&A event, and 58.9% are targets or target parents in at 13 In general, transaction values must be reported for M&A transactions that are considered material to the firm. See Rodrigues and Stegemoller (2007) for a discussion of these reporting requirements. In addition, Netter, Stegemoller, and Wintoki (2011) provide a comprehensive discussion of data screens related to the analysis of M&A transactions. 11

13 least one M&A event. Examining securities issues by type, 50.9% of the sample firms have at least one equity issue and 27.5% have at least one debt issue. The last row in Table II describes lending activity. Of the 8,322 firms in our sample, 4,964 have at least one loan during the sample period and the mean (median) number of loans is 3.5 (1.0) per firm. Event frequencies for the most active firms in the sample are described in more detail in Table A2 in the appendix. The most frequent security issuers in the sample are Union Pacific, United Parcel Service, and IBM, with 71, 70, and 68 securities issues, respectively. The most active acquirers in the sample are Cisco Systems with 133 acquisitions, IBM with 102 acquisitions, and Microsoft with 101 acquisitions. General Electric is the most frequent target (as target parent) with 126 target events, followed by Alderwoods Group with 69 target events, and AT&T with 62 target events. Sample firms are described in more detail in Table III. The average market value for these firms is $1.47 billion, the average stock price is $15.06, and the average daily trading volume is 444 thousand shares. 14 The number of years from a firm s first listing in CRSP to the end of our sample period (or the firm s delisting date, if earlier) averages 18.9 years. The number of years in the sample averages 6.89, and ranges from 0.23 to Because firms enter and leave the sample at different times, it is useful to examine transaction frequencies and deal values per year. On average, sample firms experience 0.18 equity issues per year, 0.14 debt issues per year, and 0.82 M&A transactions per year (as either acquirer or target). The average transaction value per year equals $16.8 million for equity, $40.0 million for debt, and $241.6 million for M&A transactions. In addition, as shown in the last two rows of Table III, the average firm in the sample is involved in 0.57 loans per year with an average loan value per year of $119.5 million. B. Investment Bank Identification and Roles Using SDC, we collect data on all underwriting syndicate members for equity, debt, and other security offerings, as well as all advisors used by targets and acquirers in M&A transactions. Across the full sample of 60,458 transactions, we identify over 90,000 individual underwriter and advisor observations. As 14 For the purposes of this table, market value, stock price, and trading volume are averaged for each firm across all trading days in the sample period. The table then reports the cross-sectional average of these sample period averages. Market values are not adjusted for inflation. Again, we exclude firms that delist during the first year or list during the last year of the sample period. 12

14 shown in rows (7) and (8) of Table A1, we identify underwriters for 89.0% of equity issues, 97.8% of debt issues, and 74.8% of other security issues. In addition, we identify advisors for 17.3% of M&A acquirer events and 44.8% of M&A target events. To clean the underwriter/advisor data, we first combine all variations of the same underwriter/advisor name into a single name. We then adjust the underwriter names to account for mergers and acquisitions among underwriting/advising firms. We identify underwriter/advisor combinations using several different sources. We begin with the merger samples identified in Corwin and Schultz (2005), Asker and Ljungqvist (2010), and Bao and Edmans (2011). We then verify these combinations and identify additional combination events using SDC s parent information for each IB, news stories, and the SDC mergers and acquisitions database. In total, we identify 170 IB combination events during our sample period. In some cases, such as the acquisition of Hampshire Securities by Gruntal & Company in 1997, our interpretation is that one firm is subsumed by another. In these cases, the acquired underwriter/advisor exits our sample on the event date and the acquiring firm is assumed to absorb all market share and relationships of the acquired firm. In other cases, such as the acquisition of Dillon Read by SBC Warburg in 1997, our interpretation is that two firms combine to form a new firm. In this example, the new firm is Warburg Dillon Read. In these cases, a new firm is created starting on the event date and this firm absorbs all market share and relationships from both predecessor firms. 15 To illustrate this process, Figure A1 in the appendix shows all of the predecessor firms and combination events associated with what ultimately become Bank of America Merrill Lynch and UBS Paine Webber. After cleaning, the data include 1,436 unique underwriter/advisor names. Of these, 492 are involved in equity issues in some capacity, but play no role in debt or M&A events during the sample period. Another 117 are involved only in debt events and 417 are involved only in M&A events. Ninety-eight of the sample underwriters/advisors are involved in securities issues but not M&A events, and 195 participate in at least one transaction of each type. This sample of 195 underwriters/advisors is what we refer to as full-service 15 Because the post-merger IB in both types of cases is assumed to absorb the pre-merger relationships and market share of the two pre-merger entities, the naming convention used to define post-merger IBs has no impact on the results. 13

15 IBs. The ten most active banks for each transaction type are listed in Table A4 of the appendix. Using SDC s role descriptions, we classify underwriters as leads, comanagers, or non-managing syndicate members and we classify M&A advisory group members as either advisor or other. 16,17 Throughout the paper, we combine these roles into two broad categories. The first category, lead roles, includes all lead underwriters in security issues and all advisor roles in M&A transactions. The second category, important roles, includes lead roles plus all comanagers and all other M&A advisory roles. To identify lending relationships between firms and their IBs, we hand match all IB names that appear in the SDC sample to the DealScan lender IDs associated with our sample loans. Following Gopalan, Nanda, and Yerramilli (2011), we focus on the subset of lenders who are assigned lead arranger credit and divide the credit for each loan equally in cases with multiple lead arrangers. 18 Using this method, at least one SDC-identified IB is assigned lead arranger credit in 77.3% of the sample loans, accounting for 89.6% of the aggregate lending in our sample. In most subsequent tests, we impose the additional restrictions that a transaction must have a value of at least $1 million and employ in some role on the deal either a full-service bank or a bank in the top 25 of market share for the particular transaction type involved, where market shares and full-service designations are based on transactions during the preceding twelve months. These restrictions are intended to capture the set of transactions in which a full-service IB would be both considered and willing to participate, allowing us 16 We identify an underwriter as a lead if that manager is assigned an SDC underwriter role of global coordinator, book runner, joint book runner, lead manager, joint lead manager, lead placement agent, or joint lead placement agent. We identify an underwriter as a comanager if that manager is assigned an SDC underwriter role of co-manager or co-placement agent. We identify non-managing syndicate members using the SDC role syndicate member. Because Corwin and Schultz (2005) find that these non-managing syndicate members play only a minor role in the underwriting process and the use of these syndicate members has decreased substantially over time, we exclude them from our analysis of IB relationships. Hu and Ritter (2007) distinguish between lead managers and book running managers, noting that all book runners are leads, but not all leads are book runners. Following much of the prior research, we do not distinguish between book running and non-book running lead managers. 17 For M&A target (acquirer) events, 89.3% (92.2%) of advisory group members in the data have advisor listed as one of their roles. Among those not listed as advisor, the vast majority are identified as providing a fairness opinion. For target advisors, 32.9% of those listed as advisor have at least one other listed role, with over 90% of these providing a fairness opinion. Of those not listed with an advisor role, 58.4% provide fairness opinions and the rest are primarily listed as representing the seller or a majority shareholder. For acquirer advisors, 12.6% of those listed as advisor have at least one other listed role, with 100% of these providing a fairness opinion. Of those not listed with an advisor role, 77.6% provide fairness opinions and the rest have no listed role. 18 Based on DealScan s lead arranger credit variable, 80.0% of the sample loans have one lead and 19.3% have two. The maximum number of leads is 11, but only 26 loans (0.09%) have more than five leads. Of the lenders with lead arranger credit, DealScan identifies 62% as Administrative Agent, 16% as Syndication Agent, 11% as Agent, and 9% as Arranger. The remaining 2% are spread across 42 different role descriptions. Ivashina (2009) describes the difficulty in interpreting lender roles by title, noting that different titles don t necessarily correspond to different roles. 14

16 to analyze IB relationships that span the functional areas within the IB and to account for the two-sided matching problem described in Fernando et al. (2005). Summary statistics for this restricted sample are provided in the last three columns of Table I. As the table shows, the additional restrictions produce a sample of 20,190 transactions, including 5,285 equity issues, 7,228 debt issues, 394 security issues of other types, 3,759 M&A acquirer events, and 3,524 M&A target events. The restricted sample accounts for 98.1% of aggregate security issue value, 81.3% of reported M&A acquirer transaction value, and 91.2% of reported M&A target transaction value. C. The Characteristics of Underwriting Syndicates and M&A Advisory Groups As noted above, previous research identifies several important changes in equity syndicates over time. Before turning to the question of what impact these patterns have on IB relationships, we first examine the patterns during our sample period and compare the equity and debt syndicate patterns to those for M&A advising groups. Panels A and B of Figure 2 plot the average number of lead, comanaging, and nonmanaging syndicate members by year for the full sample of equity issues and the subsample of IPOs, respectively. The patterns are consistent with those reported in prior research and show that the previouslydocumented trends have continued in the late 2000s. For all equity issues, the average number of nonmanaging syndicate members decreased from approximately 9 in 1997 to 1 in 2003, and to essentially zero by the end of the sample period. At the same time, the average number of leads increased from 1 to 1.6 and the average number of comanagers increased from 1.5 to over 3. The patterns for IPOs are similar, with a slightly larger increase in leads and comanagers and a sharper decrease in non-managing syndicate members. Underwriting syndicates for debt issues are characterized in Panel C of Figure 2. While debt issues tend not to include non-managing syndicate members, the trends in leads and comanagers are even more striking than those for equity issues. During the sample period, the average number of leads increases from approximately 1 to over 3, and the average number of comanagers increases from just under 1 to over Syndicate characteristics for other types of issues are difficult to interpret due to the small number of issues per year, especially toward the end of the sample period. For example, the number of sample issues in this category drops to 9 in 2008 and 14 in In general, however, the patterns are similar to those for equity and debt. The average numbers of leads and comanagers increase through 2007 and there is little evidence of non-managing syndicate member participation after the middle of the sample period. 15

17 The characteristics of M&A advisory groups are illustrated in Panel D of Figure 2. From 1996 through 2004, the number of advisors in M&A transactions averaged close to 1.1, and was slightly higher for acquirers than for targets. This number increased through the end of the sample period, reaching nearly 1.25 for targets and 1.35 for acquirers. While these trends are small relative to those in the equity and debt syndicates, they do suggest a notable shift in the composition of M&A advisory groups over time. To provide additional detail, Figure 3 plots the proportion of security issues with multiple lead underwriters (panel A), the proportion of M&A deals with multiple advisors (panel B), and the proportion of security issues with zero non-managing syndicate members (panel C). For both IPOs and debt issues, the proportion of deals using multiple leads increases steadily from close to zero at the beginning of the sample period to over 80% in For all equity deals, the proportion levels off at just over 40% after The increased numbers of leads and comanagers in equity syndicates is accompanied by a significant decrease in the use of non-managing syndicate members. For both IPOs and total equity offerings, the proportion of issues with zero non-managing syndicate members increased from 20%-30% in the late 1990s to nearly 100% by Debt underwriting syndicates do not tend to include non-managing syndicate members. These results suggest that the trends reported in Corwin and Schultz (2005) have continued through the mid and late 2000s. The proportion of M&A acquirers with multiple advisors increases from just under 10% to over 15%. In comparison, the proportion of M&A targets with multiple advisors increases through the mid- 2000s before slightly dropping off again in the late 2000s. The results in this section point to several important time trends in the characteristics of underwriting syndicates and M&A advisory groups. In particular, there is a significant increase over time in the number of leads and comanagers in underwriting syndicates, with a corresponding trend toward zero non-managing syndicate members. M&A advisory groups have also grown over time, but to a much smaller extent than underwriting syndicates. In the next three sections, we examine various aspects of IB relationships, including the extent to which these trends in syndicate structure have affected the characteristics and determinants of IB relationships. We begin with a comprehensive analysis of the characteristics and determinants of IB relationships both across firms and across time. We then analyze underwriter retention on consecutive deals 16

18 and extend previous research by examining retention across deal types. Finally, we examine the extent to which IB relationships carry over across different investment banking functions and the role these relationships have on IB hiring decisions. 4. The Characteristics and Determinants of Investment Banking Relationships We begin our analysis with a detailed characterization of firm-ib relationships and their determinants. To examine possible changes in relationships over time, we measure relationships using threeyear rolling windows. Because there may be many different aspects to these relationships, we define several alternative relationship measures. In addition, for each measure, we create four alternative versions based on lead roles and four alternative versions based on important roles. The four versions include a firm-wide measure based on all transactions by the firm, and three transaction-type specific measures based on equity, debt, and M&A transactions, respectively. To define the relationship measures, we first create two measures of market share for each IB-firm pair based on the dollar value of transactions done by the firm during the three-year window. The first measure of IB-firm market share is defined as the proportion of the firm s aggregate transaction value for which the IB acted in a lead (or important) role, where each IB in a lead (or important) role is given full credit for the deal. Adjusted IB-firm market share is defined analogously, but with each IB in a lead (or important) role receiving credit for a proportion 1/N of deal value. To capture the number of IB relationships maintained by the firm, we calculate the total number of IB relationships and the number of strong IB relationships, where we define a strong relationship as any IB that earns at least a 50% IB-firm market share. To focus on the top IB relationships for each firm, we define the market share of the firm s top IB, the adjusted market share of the firm s top IB, and the aggregate adjusted market share of the firm s top three IBs. Finally, to capture the exclusivity of the firm s IB relationships, we define a Herfindahl index based on the adjusted market shares of all IBs that work with the firm. For the analysis to follow, we analyze IB relationships only for those firms with at least two relevant deals during a three-year window. To illustrate the time-series patterns in these relationship measures, Figure 4 plots the cross-sectional 17

19 average of each relationship measure during each three-year window ending in 1998 through Figures A through D show that firms have increased both the total number of IB relationships they maintain and the number of strong relationships they maintain, whether measured based on lead roles only or based on all important roles. While these patterns are evident across all types of security issues, there is little if any pattern for M&A relationships. For example, the number of different IBs hired in lead roles on debt transactions increases from an average of 2.3 per firm in to 4.9 per firm in Likewise, the average number of IBs hired in important roles on debt transactions increases over this period from 4.9 to 12.0 per firm. The patterns for equity issues are similar, though less dramatic, with the average number of lead IB relationships increasing from 1.4 to 2.4 and the average number of important relationships increasing from 3.9 to 5.8. These results show that the changes in equity and debt syndicate structure documented in Section 2 led to a related increase in the number of separate IBs hired by firms in both lead and important roles. While a similar pattern exists for M&A advisory relationships, the average number of important relationships in this category increases only from 2.1 to 2.4 per firm. Panels E through H of Figure 4 provide results related to the strength and exclusivity of IB relationships. When each lead IB is assigned full credit for each deal, Panel E shows that there is very little change over time in the market share of the top relationship bank for equity and M&A transactions, and a modest increase over time for debt transactions. However, when market shares are adjusted to account for multiple leads, the conclusions change dramatically. As Panel F shows, the average adjusted market share of the top relationship bank drops significantly over time for all types of deals. On average, the top IB s adjusted market share drops from 73.7% to 40.8% for debt transactions and from 86.5% to 64.4% for equity transactions. Although more modest, a decrease is also evident for M&A transactions, for which adjusted market share decreases from an average of 78.2% to 72.3%. While the unadjusted market shares illustrated in Panel E may closely track league table results, the adjusted market shares shown in Panel F may be a better representation of the division of fees across IBs. 20 Table A3 in the Appendix provides a more detailed description of the cross-sectional distribution of various relationship measures for the three-year windows ending in 1998 and

20 Thus, our results are consistent with a change over time in the economics of the investment banking business, in which IBs earn fees from more firms but earn a smaller portion of fees from each firm. Unless specifically stated, the analyses throughout the rest of the paper utilize adjusted market share and relationship measures. Although the results in Panel F suggest that the adjusted market shares of the top relationship IB have decreased over time, Panel G shows that IB relationships continue to be dominated by a small number of banks. At the beginning of our sample period, the top three relationship IBs in all transaction types earn an aggregate adjusted market share of at least 98%. This aggregate market share decreases only slightly over time for equity and M&A transactions. Debt transactions are significantly less concentrated by the end of the sample period than the other transaction types. Even for debt transactions, however, the top three relationship IBs earn an aggregate adjusted market share of 80.1% in These results suggest that IB relationships have evolved over time from a single primary bank toward a small group of primary relationship banks. 21 Results based on the market share Herfindahl index are consistent with the adjusted market share results and suggest that the exclusivity of IB relationships has decreased significantly over time, especially for relationships involving equity and debt transactions. Notably, this trend contrasts with that for the overall industry, where IB concentration has remained relatively flat or even increased over time. 22 To better understand the determinants of IB relationships and the related time series patterns, we provide pooled time-series and cross-sectional regressions of IB relationship measures on several firm and IB characteristics. We note that exclusive IB relationships may reflect an information monopoly or hold-up problem, which makes it difficult for firms to switch banks (see, for example, Rajan (1992) and Huston and James (1996)). 23 If hold-up problems are most severe in cases where private information is more valuable, we expect fewer and more exclusive relationships for small firms, young firms, and firms with high marketto-book ratios and idiosyncratic volatility. In addition, these effects may be strongest for relationships 21 If we separate the adjusted market shares of the top three relationship banks in debt and equity transactions, we find that while the top IB s adjusted market share decreases over time (Figure 4 Panel F), the adjusted market shares of the second- and third-ranked IBs increase over time. This pattern is much weaker for M&A transactions. 22 For all transaction types, Herfindahl indices calculated from market-wide (rather than firm-specific) market shares increase sharply from , drift downward slightly from , and increase from Comparing results in 1996 and 2009 suggests a slight upward trend overall. The sharp increase from reflects the large number of IB mergers during this period. 23 These models are developed in the context of lending and the choice of bank vs. public debt. However, Rajan (1992) notes that the analysis would also apply in the context of investment banking, where IBs acquire information and negotiate with firms. 19

21 involving equity issues, which are particularly sensitive to information effects. As described in Rajan (1992) and Huston and James (1996), the use of multiple banks can mitigate the types of hold-up problems discussed above. However, the ability to utilize multiple banks may be limited by the firm s bargaining power relative to the IB. We expect larger firms and more active firms to have more bargaining power, allowing them to maintain a higher number of less exclusive IB relationships. To identify more active firms, we define each firm s relative transaction value for each relevant transaction type (i.e., all transactions, equity, debt, or M&A) as the aggregate transaction value for the firm during the three-year window divided by the aggregate transaction value of all firms during the same period. To the extent that firms with higher leverage ratios access debt markets more frequently, leverage ratios may serve as an additional proxy for bargaining power. We therefore expect firms with high leverage ratios to have a higher number of less exclusive relationships, especially with respect to debt relationships. 24 To capture the bargaining power of IBs, we include the bank s market share in the relevant transaction type across all issuers during the most recent twelve months. We expect IBs with higher market share to command more exclusive relationships. Finally, we include year fixed effects to capture any time trends unrelated to variation in the other explanatory variables. The results from the panel regression are described in Table IV. As expected, we find that large firms and firms that are more active participants in capital markets have a higher number of less exclusive relationships across all IB functions. For most specifications, we also find more exclusive relationships for firms with higher market-to-book ratios and higher idiosyncratic volatility. Taken together, these results are consistent with the informational effects of IB relationships and with bargaining power being used to mitigate informational hold-up problems. However, the results based on firm age are mixed and the findings are not consistent across all relationship measures. 25 We find that firms with higher leverage ratios have less exclusive relationships. These findings are consistent with high debt firms having more bargaining power 24 Large relative transaction values and high leverage ratios may also reflect a higher demand for sources of capital. Like the bargaining power explanation above, this would suggest less exclusive relationships for more active firms and firms with higher leverage ratios. 25 We define firm age in this regression as years since the initial listing date in CRSP. The mixed results on this variable may reflect the poor quality of this proxy. 20

22 and searching out more sources of capital. After controlling for other factors, we find little evidence of a consistent relationship between IB market share and IB relationships. Coefficients on year fixed effects (not shown) suggest that, even after controlling for other factors, both firm-wide and transaction-type specific relationships have become less exclusive over time, with the patterns being strongest for equity and debt. 5. Investment Bank Retention The strength of IB relationships is revealed, in part, through a firm s decision to retain an IB from one deal to the next. As noted above, the decision to switch lead underwriters has been studied extensively in the prior literature, especially with respect to the debt and equity issues that follow a firm s IPO. In this section, we extend this literature by analyzing IB retention during our more recent sample period and across transactions of different types. While most prior studies define retention as any case where the lead on one deal is also the lead on the next deal, the increase over time in jointly-led deals suggests the need for a modified definition. We begin by defining IB retention as any case where at least one IB in a lead (or important) role on the prior transaction is hired in a lead (or important) role on the current transaction. We then provide a Probit analysis of retention for each individual IB in a lead (or important) role. Retention rates in sequential transactions over the full sample period are described in Panel A of Table V. For each transaction type pair, the top line lists the average retention rate based on lead roles and the second line lists average retention based on important roles (in brackets). We present results both for the full sample of events that have an identifiable prior transaction by the same firm and for the subsample of paired events that are no more than three years apart. We also provide separate results categorizing both the current and prior events by transaction type. Ignoring transaction type, we find that at least one lead IB is retained in 50.7% of consecutive deals and at least one important IB is retained in 65.1% of consecutive deals. These rates increase to 53.4% and 67.7% if the deals are within three years of each other. Focusing on consecutive deals of the same type, we find that at least one lead IB is retained in 48.4% of consecutive M&A deals, 55.2% of consecutive debt deals, 69.7% of SEOs following an IPO, and 53.3% of SEOs following other SEOs. For equity and debt deals, retention is particularly strong if non-lead roles are considered and if a 3-year restriction is imposed. 21

23 For example, for deals no more than three years apart, at least one important IB is retained in 74% of consecutive debt deals, 94.1% of SEOs following the IPO, and 79.9% of SEOs following other SEOs. The results suggest that ongoing relationships are particularly important for equity and debt issues. Our data also allow us to analyze IB retention across transactions of different types. For example, following IPOs, at least one lead IB is retained in 49.2% of subsequent M&A deals, 60.5% of subsequent debt issues, and 69.7% of subsequent equity issues. More generally, retention rates tend to be higher for equity and debt transactions than for M&A deals, and retention tends to be highest following the IPO, regardless of the subsequent transaction type. In all cases, differences across transaction types are statistically significant. However, even for deals of different types, at least one lead IB is generally retained in over 40% of deals. These results are consistent with the importance of both firm-wide and transaction-type specific components in IB relationships. Further, retention rates increase if we impose a maximum of three years between deals, suggesting that IB relationships weaken as time passes between deals. Panel B of Table V provides separate results for deals during and Across all transactions, retention rates for both lead roles and important roles increase over time. However, these differences appear to be driven primarily by retention rates related to debt transactions. For example, at least one lead (important) IB is retained in 50.2% (68.6%) of consecutive debt transactions during , compared to 65.4% (84.7%) during Similarly, at least one lead IB is retained in 53.3% of equity transactions that follow debt transactions during , compared to 64.3% during These results suggest that changes in the economics of investment banking, such as the increased use of multiple leads and larger groups of comanagers, have increased the likelihood that a particular IB is retained from one deal to the next, especially for debt deals. The results may also reflect an increased link between IB relationships and the provision of lending services. We address this possibility below. To control for other factors that may affect the decision to retain an IB on consecutive deals, we estimate a probit model for retention. The model includes one observation for each IB on the prior transaction and the dependent variable equals one if the IB is retained on the current transaction. Because the model includes multiple observations per transaction, the standard errors are clustered by transaction. We 22

24 estimate the model separately for equity, debt, and M&A deals both using the full sample of prior deals and eliminating all deals of different types. We also provide separate results based on lead (panel A) and important (panel B) roles. The independent variables include several firm, IB, and transaction characteristics, as well as time period effects, where firm and transaction characteristics are based on the current deal. Results for lead roles are presented in Panel A of Table VI. The coefficient on firm age is negative and statistically significant in all but the equity-only model, and retention is higher for transactions immediately following the IPO. These results are consistent with the retention rates in Table V and suggest that relationships are strongest for young firms. In some cases, multiple deals of different types or multiple M&A events for the same firm occur on the same date. The coefficient on a dummy variable identifying these cases is positive and significant, showing that firms tend to use the same IB for transactions that occur on the same day. 26 Similarly, the coefficient on years since prior transaction is always negative and significant, confirming our earlier finding that the likelihood of retention decreases as time passes between transactions. Transaction value is positive and marginally significant in some models, suggesting that IBs are more likely to be retained if the current deal is large. More interestingly, however, the interaction terms suggest that retention likelihood decreases when the current deal is substantially larger or substantially smaller than the prior deal. These results are consistent with Fernando et al. (2005) who find evidence in support of transaction-based IB selection decisions. Finally, the results from the time period dummies are consistent with an increase in retention rates over time, as suggested by our univariate results. To examine the importance of transaction type, we include two additional variables. First, in the transaction-type specific models, we include a dummy variable to identify those cases where a transaction of another type occurs between the two same-type deals. The coefficient on this variable is negative and significant in the merger-merger model, suggesting that retention likelihood decreases if an equity or debt deal occurs between the two merger deals. However, the variable is insignificant in the debt and equity 26 Transactions that occur close together in time may be negotiated together rather than as separate transactions. Consistent with this, we find that such transactions often use the same IB. However, there are many cases where deals occurring in close proximity do not use the same IB. For example, only 68.5% (41.2%) of transactions occurring on the same day (within one week) of a merger use at least one IB from among the merger advisors. 23

25 models. In the combined deal models, we instead include a variable to identify whether the current transaction and the prior transaction are of the same type. The results show that for all transactions, retention is significantly higher when the prior transaction is of the same type. In addition, this transaction-type specific effect has decreased over time for debt issues and increased for mergers. We examine the importance of IB characteristics using four additional variables. The first three are the IB s market shares in equity, debt, and M&A transactions over the 12 months prior to the current deal. The fourth is a dummy variable identifying IBs who have done at least some lending during the prior 12 months. The results suggest that equity market share improves retention in both equity and M&A transactions and M&A market share improves retention in both M&A and equity transactions. In contrast, debt market share has a positive impact on retention in debt models, but a negative effect on retention in equity and M&A transactions. Finally, there is some evidence that IBs with lending capacity experience higher retention rates in debt transactions, but lending capacity appears to have little impact on retention in equity and M&A deals. Results based on important roles are provided in Panel B. In general, these results are similar to those for lead roles and suggest that the lead vs. important designation does not affect the conclusions. To better understand the economic significance of these results, we estimate the implied probability from each model based on the mean values of the explanatory variables and assuming a deal size of $250 million. The results suggest that the during the first subperiod, the estimated likelihood of being retained on consecutive deals is 62% when the current deal is debt, 51% when the current deal is equity, and 52% when the current deal is a merger. However, the difference between deal types increases over time. During the third subperiod, the estimated likelihood of retention is 73%, 61%, and 48% for debt, equity, and mergers, respectively. During the first two subperiods, a prior deal of the same type increases retention likelihood by between 3% and 9%. In contrast, during the most recent subperiod, this same-type effect is 5% for equity, insignificant for debt, and jumps to 19% for mergers. For the most recent subperiod, these results suggest that IBs from all types of prior transactions tend to be retained in equity and debt deals, while M&A advisor retention tends to be more type-specific. 24

26 Overall, the retention analysis in this section points to significant differences across deal types and over time. As expected, relationships appear to be stronger for equity issues where information effects are likely to be most pronounced than for M&A transactions which rely heavily on idea generation. In addition, the retention of debt underwriters has increased significantly over time, whether or not the paired transaction is also a debt issue. Despite the existence of transaction-type specific effects, the retention analysis points to a strong firm-wide component to IB relationships. The importance of this firm-wide component and the possible ties between this component and lending are analyzed in more detail in the following section. 6. Investment Bank Relationships across Deal Types One of the advantages of our data is the ability to examine the extent to which IB relationships carry over across transaction types. The results in the previous sections provide evidence of both firm-wide and transaction-type specific components in IB relationships. In this section, we provide a more detailed analysis of the links between relationships in the different functional areas of investment banking. We begin by analyzing the frequencies with which the top bank in one transaction type is one of the top three banks in another transaction type. For this test, we use a variation of the top bank market share measure described in Section 4, where we define market shares based on the full sample period (or subperiod) rather than with three-year windows and we require a firm to have at least two transactions of each relevant type. Using this measure, we identify the top three relationship IBs for each firm for each transaction type, based on lead roles, and compare the top relationship IBs in different functional areas. The results are provided in Table VII. As an illustration, there are 311 firms that have at least two debt transactions and two equity transactions during the sample period. For these firms, we find that the top debt relationship bank is also the top equity relationship bank in 62.4% of cases. The results point to substantial crossover of relationships from one transaction type to another. As noted above, the top debt relationship bank is also the top equity relationship bank in 62.4% of cases, and this bank is one of the top three equity relationship banks in 74.0% of cases. Similarly, the top equity relationship bank falls in the top three for debt in 78.8% of cases. While these results are consistent with a substantial crossover in relationships between functional areas, there are also a surprising number of cases 25

27 where the top bank in one area plays no role in other areas. For example, for firms with at least two deals of each type, the top debt bank fails to lead an equity issue in 25.7% of cases and the top equity bank fails to lead a debt issue in 17.7% of cases. 27 Results involving M&A relationships reflect a weaker crossover in relationships between functional areas. For example, the top M&A bank fails to lead an equity issue in 31.6% of cases and fails to lead a debt issue in 24.2% of cases. Even in these cases, however, the top M&A bank is also the top equity bank in 56.9% of eligible cases and is the top debt bank in 50.3% of eligible cases. Thus, firm-wide IB relationships appear to have an important impact on the selection of M&A advisors, though the impact is notably weaker than that for debt and equity underwriter selection. The M&A results may reflect the decision by some firms to select advisors that are M&A specialists, rather than full-service IBs. Panels B and C of Table VII provide separate results based on transactions from and , respectively. To analyze the impact of lending relationships on the firm-wide nature of IB relationships, we also separate cases where the firm has a lending relationship with their top IB at some point during the sample period from those cases where no lending relationship exists. For these tests, we require that a firm have at least two transactions during the subperiod in each of the relevant transaction types. Ignoring lending relationships, there are several notable patterns in the subperiod results. First, the likelihood that the top equity bank and top debt bank are the same increases over time. This likelihood is 63.6% in the first subperiod and 72.0% in the second subperiod. At the same time, the likelihood that the top M&A bank is also the top equity or debt bank decreases. For M&A and equity lead roles, this likelihood decreases from 68.7% to 60.0%. For M&A and debt roles, this likelihood decreases from 57.3% to 47.2%. When we separate results based on whether a lending relationship exists between the firm and their IB, we find that the presence of a lending relationship leads to a much stronger cross-over in relationships, especially during the second half of the sample period. For example, for cases in the second subperiod where 27 The incorporation of important roles has little impact on the results involving M&A relationships. However, for equity and debt relationships, the incorporation of important roles decreases the likelihood that the top IB in one area does not appear in the other area. Results are also generally consistent if we require four deals of each type rather than two. However, among these more frequent issuers, there is a slight decrease in the likelihood that the top bank in one area is also the top bank in another area, accompanied by an increase in the likelihood that the bank is ranked second in the other transaction type. 26

28 a lending relationship exists, the top equity bank was also the top debt bank in 81.6% of cases and appears on a debt deal in all but 5.3% of cases. The comparable numbers in the absence of a lending relationship are 62.2% and 16.2%. The results across other transaction types are similar. Taken together, the results in Table VII suggest that there is a strong tendency for IB relationships to carry over across transaction types, but these patterns differ by transaction type and across time. The crossover of equity and debt relationships has increased over time, while the crossover between M&A relationships and underwriting relationships has decreased over time. In addition, the presence of a lending relationship is associated with a stronger firm-wide component in the IB relationship, especially in the second half of the sample period. These results are consistent with increasing ties between lending and the other functional areas within the bank. As a final test, we provide a probit model for underwriter/advisor selection in equity, debt, and M&A transactions. For each transaction in the model, the set of eligible IBs includes all full-service IBs plus any additional IBs ranked in the top 25 by market share in the current type of transaction, where market shares and full-service designations are defined over the previous twelve months. Because the model includes multiple observations per transaction, the standard errors are clustered by transaction. We include the log of transaction size to allow for differences between large and small deals. To account for changes in syndicate structure and other time trends, we include year fixed effects. We also include several IB characteristics, including IB market share in aggregate equity, debt, and M&A transactions over the year prior to the current transaction, and a dummy variable identifying those IBs providing lending services during this period. These variables are intended to capture the IB s reputation or expertise in a particular area. We also include IB fixed effects to capture any unidentified IB characteristics that are not captured by our other variables. The primary variables of interest are related to relationships between firms and their IBs. The first three variables are the IB s share of the firm s equity, debt, and M&A transactions over the three years immediately prior to the current transaction. To examine the impact of lending relationships on IB selection, we include an analogous measure of the IB s lending relationship with the firm over the prior three years. To capture the importance of more recent transactions, we also include dummy variables identifying IBs that 27

29 had lead roles (or advising roles for M&A deals) and comanaging roles (or non-advising roles for M&A deals) on the prior transaction. We expect both market share in the firm s prior deals and involvement in the preceding deal to improve an IB s chances of being hired on the current deal. We also separate the relationship variables by type in order to test whether IB selection in one functional area is affected by past relationships in other functional areas. Asker and Ljungqvist (2010) find that firms are reluctant to hire IBs that are utilized by their product market rivals. To capture this effect, we first rank firms by sales in each 4-digit SIC industry. We then identify the top relationship bank for the top-ranked and second-ranked firm in each industry based on the proportion of that firm s combined equity, debt, and M&A transactions over the prior three years for which the bank held a lead role. For sample firms ranked second or lower in an industry, we define the competitor bank as the top bank of the number one firm in the same industry. For firms ranked first in an industry, we define the competitor bank as the top bank of the number two firm. To test whether the effects of IB relationships are impacted when an IB is used by a firm s competitors, we interact the top competitor bank dummy with the lead on prior deal and comanager on prior deal dummies. We expect any effects of past relationships on hiring decisions to be reduced in those cases where the IB has a strong relationship with the firm s primary product market rival. The results for equity, debt, and M&A deals are provided in Panels A through C of Table VIII, respectively. Models (1) and (2) provide results based on lead roles and model (3) provides comparable results based on important roles. In models (2) and (3), all coefficients are interacted with subperiod dummies representing transactions from , , and Examining the IB market share variables, we find that equity market share improves a bank s chances of being included on an equity deal, debt market share improves the likelihood of inclusion on debt deals, and M&A market share improves the likelihood of inclusion on M&A deals. After controlling for type-specific market share, however, there is little evidence that market share in unrelated transaction types improves an IB s chances of being hired. This suggests that IB expertise is transaction-type specific. Examining changes across time, we find that the coefficient on debt market share in the debt model increases significantly over time and the coefficient on 28

30 M&A market share in the M&A model is significantly lower in the middle subperiod than in the earlier or later periods. For equity deals, the coefficient on equity market share also drops during the middle subperiod, but this time trend is statistically significant only when important IB roles are considered. Finally, the lending dummy is generally positive and significant in the middle subperiod, but is insignificant elsewhere. Turning to the relationship variables, we find evidence of substantial cross-over in relationships across functional areas. This is consistent with the results in Table VII and contrasts sharply with the results for overall market share. In all of the models, we find that all three types of IB relationships matter. However, equity relationships are more important for inclusion in equity deals and debt relationships are more important for debt deals. In addition, holding a lead or important role on the prior deal has a positive and significant effect on the likelihood of inclusion, even after controlling for the continuous relationship variables. We also find that prior lending relationships have a significant positive effect on the likelihood of earning future equity and debt underwriting mandates, as well as future M&A advising roles. This finding is consistent with the results of Ljungqvist et al. (2006) and Bharath et al. (2007) for equity and debt offers and suggests that the effects of lending relationships also carry over to the selection of M&A advisors. Examining changes over time, we find that firm-wide relationships increase in importance over time. For example, the importance of prior debt, M&A, and lending relationships in the equity underwriting model are strongest during the third subperiod. Similarly, the importance of equity and lending relationships increase over time in the debt model, and the importance of debt and lending relationships increase over time in the M&A model. The strongest time trends are evident with respect to lending relationships, which have increased in importance over time for all transaction types. Together, this evidence suggests that both the firm-wide component of IB relationships and the link between this component and lending have increased over time. In contrast to Bharath et al. (2007), who find that the effects of cross-market relationships on underwriter choice are economically small, we find that these effects are economically important and have increased substantially over time, especially with respect to lending relationships. The economic significance of our results is discussed in more detail below. Consistent with Asker and Ljungqvist (2010), we find for both equity and debt deals that the effects 29

31 of past relationships are weaker in cases where the bank under consideration is the utilized by the firm s primary product market rival. Subperiod results show that this effect increases over time in the equity model, with the coefficient being statistically significant only in the third subperiod for lead roles and in the second and third subperiods for important roles. The pattern is the opposite for the debt and M&A models, where the competing firm effect decreases over time. In the debt model, the effect is significant during the first two subperiods for lead roles and during all three subperiods for important roles. In the M&A model, the effect is significant only during the first subperiod. As a final test, we run a joint model including all transaction types. This model is presented in Panel D. The advantage of this model is that it allows us to examine the interactions of different deal types and to test the importance of the sequence of transaction types. To analyze the relative importance of the various IB market share measures, we interact each transaction-type specific market share measure with a dummy variable to identify transactions that are not of the same type. As in Panels A through C, the results suggest that IB market share has a significant positive effect on the likelihood of being included in an underwriting syndicate or M&A advising group. However, this effect is completely negated or even reversed for transactions that are not of the same type. Lending capacity tends to have a positive effect on IB selection, but this effect is concentrated in the middle subperiod. To examine the separate importance of past equity, debt, and M&A relationships, we focus on the lead on prior deal variable and we break this variable down by transaction type in two ways. First, we separately define dummy variables for whether an IB held a lead role on a prior deal that was an equity, debt, or M&A deal. Second, we interact these dummy variables with indicators for whether the current deal is of the same type. As controls, we also include the overall IB relationship measure, based on all transaction types, and the IB lending relationship measure. As expected, the coefficients on these relationship measures are positive and significant. In addition, as in Panels A through C, the importance of lending relationships increases over time. The results show that being a lead on a prior equity deal has a positive and significant effect on lead choice in the current deal. While this effect is reduced if the current deal is not equity, it is not eliminated. 30

32 This suggests that having an equity relationship with a firm helps the IB earn lead roles in all types of future deals, though the effect is greatest for future equity deals. The results for leads on prior debt and M&A deals are similar. The effect is positive and significant, but is reduced if the current deal is not of the same type. These results are consistent with those from the type-specific models in Panels A through C. To highlight the economic significance of the results, we estimate the implied probability of IB selection from each model in Table VIII, based on the mean values of all variables. These results confirm the importance of both firm-wide and transaction-type specific relationships, with the firm-wide component increasing over time. For example, from , the probit model coefficients suggest that the likelihood of being hired as lead equity underwriter for an IB that was on the prior deal and has a 50% equity relationship is 15.8%. Adding a 50% relationship in debt, M&A, or lending increases the likelihood to between 21.2% and 30.6%, and adding 50% relationships in all three areas increases the likelihood to 48.1%. From , the likelihood of being hired as lead equity underwriter for an IB that was on the prior deal and has a 50% equity relationship is 16.4%. Adding a 50% relationship in debt, M&A, or lending increases the likelihood to between 29.8% and 33.3%, and adding 50% relationships in all three areas increases the likelihood to 71.0%. Overall, the results confirm our earlier findings that relationship effects are strongest for equity and weakest for M&A, with the importance of debt relationships increasing significantly over time. The results in Tables VII and VIII suggest that there is substantial crossover in relationships across transactions of different types. While prior equity, debt, and M&A relationships have a larger impact on future roles of the same type, these relationships also carry over to transactions of different types. Together with the findings in the previous sections, these results suggest that IB relationships include a strong firmwide component that encompasses equity and debt underwriting as well as M&A advising. In addition, our results suggest that both the firm-wide component of IB relationships and the link between IB relationships and lending have become more important over time. 7. Summary and Conclusions Although relationships between firms and their IBs are generally managed centrally within the investment bank, academic research tends to focus on only a specific area of investment banking, such as 31

33 equity or debt underwriting, or M&A advising, and on the factors that influence the decision to hire a specific IB. As a result, very little is known about the extent to which relationships span the functional areas of investment banking. Using a sample of over 20,000 M&A transactions and public and private security issues by U.S. exchange-listed firms between 1996 and 2009, we provide a comprehensive analysis of IB relationships both within and across functional areas. We address three broad research questions. First, what are the characteristics and determinants of IB relationships? Second, how have relationships between firms and their IBs changed over time? And, third, to what extent do relationships carry over across the various aspects of the investment banking business? Our sample period encompasses a number of important changes to the economics of the investment banking business. Most notably, there have been significant changes over time in the structure of underwriting syndicates, with both debt and equity syndicates being increasingly characterized by multiple lead underwriters and large numbers of comanagers. In addition, the repeal of Glass-Steagall restrictions has led to an increased role for commercial banks in investment banking. Our evidence suggests that these changes have led to substantial changes in IB relationships. For all types of transactions, we find that firms have increased the number and decreased the exclusivity of their IB relationships. In part, these changes appear to reflect the increasing link between lending and more traditional investment banking functions, as firms seek to maintain multiple sources of capital. We provide evidence of a strong firm-wide component in IB relationships. For example, while retention of lead underwriters and advisors tends to be strongest on consecutive deals of the same type, at least one IB in a lead role is retained in 40% or more of consecutive deals, regardless of transaction type. Despite this firm-wide component of relationships, we also find evidence of significant transaction-type specific relationship components, with relationships being strongest for equity deals and weakest for M&A deals. Consistent with information-based explanations, we find that IB relationships tend to be more exclusive for small firms, young firms, and firms with high market-to-book ratios and idiosyncratic volatility. Overall, our evidence suggests that the firm-wide nature of IB relationships and the link between these relationships and lending have increased significantly over time. 32

34 References Ang, J. S., and T. Richardson, 1994, The underwriting experience of commercial bank affiliates prior to Glass-Steagall Act: A reexamination of evidence for passage of the act, Journal of Banking and Finance 18, Asker, J., and A. Ljungqvist, 2010, Competition and the structure of vertical relationships in capital markets, Journal of Political Economy 188, Bao, J., and A. Edmans, 2011, Do investment banks matter for M&A returns?, Review of Financial Studies 24, Bharath, S., S. Dahiya, A. Saunders, and A. Srinivasan, 2007, So what do I get? The bank s view of lending relationships, Journal of Financial Economics 85, Beatty, R., and J. Ritter, 1986, Investment banking, reputation, and the underpricing of initial public offerings, Journal of Financial Economics 15, Benzoni, L., and C. Schenone, 2010, Conflict of interest and certification in the U.S. IPO market, Journal of Financial Intermediation 19, Bodnaruk, A., W. O Brien, and A. Simonov, 2012, Captive finance and firm s competitiveness, working paper, University of Notre Dame. Boone, A. and H. Mulherin, 2008, Do auctions induce a winner s curse? New evidence from the corporate takeover market, Journal of Financial Economics 89, Bowers, H. and R. Miller, 1990, Choice of investment banker and shareholders wealth of firms involved in acquisitions. Financial Management 19, Burch, T. R., V. Nanda, and V. Warther, 2005, Does it pay to be loyal? An empirical analysis of underwriting relationships and fees, Journal of Financial Economics 77, Calomiris, C. W., and T. Pornrojnangkool, 2009, Relationship banking and the pricing of financial services, Journal of Financial Services Research 35, Carey, M., and M. Hrycray, 1999, Credit flow, risk, and the role of private debt in capital structure, Unpublished Working Paper, Board of Governors of the Federal Reserve System, Washington, DC. Chaplinsky, S., and G. R. Erwin, 2009, Great expectations: Banks as equity underwriters, Journal of Banking and Finance 33, Chava, S., and M. Roberts, 2008, How does financing impact investment? The role of debt covenants, Journal of Finance 63, Corwin, S. A., and P. Schultz, 2005, The role of IPO underwriting syndicates: Pricing, information production, and underwriter competition, Journal of Finance 60, Dunbar, C., 2000, Factors affecting investment bank initial public offering market share, Journal of Financial Economics 55,

35 Drucker, S., 2005, Information asymmetries, cross-product banking mergers, and the effects on corporate borrowers, Working Paper, Columbia Business School. Drucker, S., and M. Puri, 2005, On the benefits of concurrent lending and underwriting, Journal of Finance 60, Eccles, R. G., and D. B. Crane, 1998, Doing Deals: Investment Banks at Work, Boston, MA, Harvard Business School Press. Ellis, K., R. Michaely, and M. O Hara, 2011, Competition in investment banking, Review of Development Finance 1, Fernando, C. S., V. A. Gatchev, and P. A. Spindt, 2005, Wanna dance? How firms and underwriters choose each other, Journal of Finance 60, Fernando, C. S., V. A. Gatchev, and P. A. Spindt, 2005, Two-sided matching: How corporate issuers and their underwriters choose each other, Journal of Applied Corporate Finance 25, Fernando, C. S., A. D. May, and W. L. Megginson, 2012, The value of investment banking relationships: Evidence from the collapse of Lehman Brothers, Journal of Finance 67, Fields, P., D. Fraser, and R. Bhargava, 2003, A comparison of underwriting costs of initial public offerings by investment and commercial banks, Journal of Financial Research 26, Forte, Iannotta, and Navone, 2010, The banking relationship s role in the choice of the target s advisor in mergers and acquisitions, European Financial Management. Gande, A., M. Puri, A. Saunders, and I. Walter, 1997, Bank underwriting of debt securities: Modern evidence, Review of Financial Studies 10, Gande, A., M. Puri, and A. Saunders, 1999, Bank entry, competition, and the market for corporate securities underwriting, Journal of Financial Economics 54, Hu, W. Y., and J. R. Ritter, 2007, Multiple bookrunners in IPOs, working paper, University of Florida. Huston, J., and C. James, 1996, Bank information monopolies and the mix of private and public debt claims, Journal of Finance 51, Ivashina, V., 2009, Asymmetric information effects on loan spreads, Journal of Financial Economics 92, James, C., 1992, Relationship-specific assets and the pricing of underwriter services, Journal of Finance 48, James, C., and P. Wier, 1990, Borrower relationships, intermediation, and the cost of issuing public securities, Journal of Financial Economics 28, Kale, J., O. Kini, and H. Ryan, Financial advisors and shareholder wealth gains in corporate takeovers. Journal of Financial and Quantitative Analysis 38, Kim, D., D. Palia, and A. Saunders, 2008, The impact of commercial banks on underwriting spreads: Evidence from three decades, Journal of Financial and Quantitative Analysis 43,

36 Krigman, L., W. H. Shaw, and K. L. Womack, 2001, Why do firms switch underwriters, Journal of Financial Economics 60, Krishnan, K., 2013, Commercial banks getting underwriting business: Tying or business building?, Journal of Economics and Business 66, Ljungqvist, A., and W. J. Wilhelm, Jr., 2005, Does prospect theory explain IPO market behavior, Journal of Finance 40, Ljungqvist, A., F. Marston, and W. J. Wilhelm, Jr., 2006, Competing for securities underwriting mandates: Banking relationships and analyst recommendations, Journal of Finance 41, Ljungqvist, A., F. Marston, and W. J. Wilhelm, Jr., 2009, Scaling the hierarchy: How and why investment banks compete for syndicate co-management appointments, Review of Financial Studies 22, Narayanan, R. P., K. P. Rangan, and N. K. Rangan, 2004, The role of syndicate structure in bank underwriting, Journal of Financial Economics 72, Netter, J., M. Stegemoller, and J. Wintoki, 2011, Implications of data screens on merger and acquisition analysis: A large sample study of mergers and acquisitions from 1992 to 2009, Review of Financial Studies 24, Papaioannou, G. J., 2011, Competing for underwriting market share: The case of commercial banks and securities firms, Journal of Financial Services Marketing 16, Puri, M., 1996, Commercial banks in investment banking: Conflict of interest or certification role?, Journal of Financial Economics 40, Puri, M., 1999, Commercial banks as underwriters: Implications for the going public process, Journal of Financial Economics 54, Rajan, R. G., 1992, Insiders and outsiders: The choice between informed and arm s-length debt, Journal of Finance 47, Rau, R., 2000, Investment bank market share, contingent fee payments, and the performance of acquiring Roten, I. C., and D. J. Mullineaux, 2002, Debt underwriting by commercial bank-affiliated firms and investment banks: More evidence, Journal of Banking & Finance 26, Roten, I. C., and D. J. Mullineaux, 2005, Equity underwriting spreads at commercial bank holding companies and investment banks, Journal of Financial Services Research 27, Saunders, A., 1999, Consolidation and universal banking, Journal of Banking & Finance 23, Schenone, C., 2004, The effect of banking relationships on the firm s IPO underpricing, Journal of Finance 59, Servaes, H. and M. Zenner, 1996, The role of investment banks in acquisitions, Review of Financial Studies 9, Shivdasani, A., and W. Song, 2011, Breaking down the barriers, Competition, syndicate structure, and underwriting incentives, Journal of Financial Economics 99,

37 Song, W., 2004, Competition and coalition among underwriters: The decision to join a syndicate, Journal of Finance 59, Welch, I., 1989, Seasoned offerings and the pricing of new issues, Journal of Finance 44, Yasuda, A., 2005, Do bank relationships affect the firm s underwriter choice in the corporate bond underwriting market?, Journal of Finance 60,

38 Table I Transaction Summary Statistics The table lists number of deals, average deal size, and aggregate deal value by transaction type. The sample includes 8,322 firms, as identified by CRSP Permco, and related security issue and M&A events from 1996 through 2009 that meet the sample restrictions described in Section 2. M&A transactions that do not report deal size are excluded. Results are provided for the full sample of deals and for the restricted sample of deals larger than $1 million that involved at least one full-service investment bank or top 25 investment bank. A full service investment bank is defined as a bank that acted as lead underwriter on at least one equity and one debt transaction and also advised one merger deal during the previous 12 months. A top 25 investment bank is defined as a bank with a top 25 market share in the transaction type being considered, based on lead underwriting and advising roles during the previous 12 months. The table also provides summary statistics for loan facilities by the sample firms from 1996 through 2009, as identified by Dealscan. Number of Deals Full Transaction Sample Mean Deal Size ($mil) Aggregate Deal Value ($mil) Restricted Transaction Sample Mean Deal Size ($mil) Number of Deals Aggregate Deal Value ($mil) Equity: Public IPO 2, ,584 2, ,797 Public SEO 3, ,555 2, ,494 Private 1, , ,116 All Equity Issues 7, ,296 5, ,407 Debt: Public Nonconvertible 3, ,609,012 3, ,608,006 Public Convertible , ,030 Private Nonconvertible 2, ,519 2, ,694 Private Convertible 1, , ,159 All Debt Issues 7, ,723,901 7, ,704,889 Other Security Issues: Public , ,695 Private , ,343 All Other Security Issues , ,038 Total Security Issues 15, ,541,163 12, ,474,334 M&A Transactions: Acquirer Events 17, ,502,599 3,759 1, ,475,169 Target Events 8, ,345,968 3,524 1, ,784,417 Dealscan Loan Facilities 29, ,068,

39 Table II Summary of Firm-Level Activity The table describes total and per firm event frequencies across all types of sample events. The sample includes 8,322 firms, as identified by CRSP Permco, and all security issue and M&A events from 1996 through 2009 that meet the sample restrictions described in Section 2. The table also describes loan facilities by the sample firms from 1996 through 2009, as identified by Dealscan. Total Sample Events N Zero Activity % Zero Activity Activity Per Firm (N=8,322 firms) Mean Median Min 99 th Perc. Max All Event Types 60, % Security Issue Events 15,580 2, % M&A Events 44,878 1, % Acquirer Events 32,451 2, % Target Events 12,427 3, % Equity Issues 7,316 4, % IPOs 2,628 5, % SEOs 3,178 6, % Private Equity Issues 1,510 7, % Debt Issues 7,614 6, % Convertible Debt Issues 1,492 7, % Non-convertible Debt Issues 6,122 6, % Other Security Issues 650 7, % Loan Facilities 29,469 3, %

40 Table III Firm Summary Statistics The table provides summary statistics for sample firms, as identified by CRSP Permco. Events include all security issue and M&A events from 1996 through 2009 that meet the sample restrictions described in Section 2. Firm characteristics are taken from CRSP and are measured across all days during the sample period on which the firm was listed on CRSP. Market Value and Volume for each firm are defined as the average across all CRSP days. For firms with multiple classes of shares, both market value and volume are summed across the share classes on each trading day. Price is the average across all CRSP days of the closing stock price for the firm s most active share class. Firm Age is defined as the number of years between the firm s initial listing on CRSP and the end of our sample period in Years in Sample is defined as the total number of days between the first CRSP date and the last CRSP date, divided by 365. The table also describes loan facilities by the sample firms from 1996 through 2009, as identified by Dealscan. N Mean Median Min 10 th Perc. 90 th Perc. Max Market Value ($mil) 8,322 1, , , Price 8, Volume (000) 8, , Firm Age 8, Years in Sample 8, Total Deals/year 8, Equity Deals/year 8, Debt Deals/year 8, M&A Deals/year 8, Total Deal Value/year ($mil) 8, , Equity Deal Value/year ($mil) 8, , Debt Deal Value/year ($mil) 8, , M&A Deal Value/year ($mil) 8, , Loan Facilities/year 8, Loan Value/year ($mil) 8, ,

41 Table IV The Determinants of Investment Banking Relationships This table describes panel regressions in which the dependent variable is one of three measures of underwriter-firm relationships. The first measure is a Herfindahl index (HH) based on the adjusted market shares of all investment banks (IBs) who led at least one deal for the firm. The second measure, Top, is the adjusted market share of the firm s top IB. The third measure is the number of IBs that are involved in 50% or more of the firm s transactions (N50), based on unadjusted market shares. Adjusted and unadjusted market shares are calculated for each category of transactions and across all transactions based on lead roles over three-year rolling windows, where adjusted market shares give each co-lead/co-advisor credit for a proportion 1/N of deal value. The independent variables are computed at the end of the last year in the three-year window. Ln MV is the natural log of the firm s equity value. Relative transaction value is the ratio of the dollar value of the firm s transactions scaled by the total value of transactions for the entire market. M/B is the firm s market value to book value of equity. Ln Age is the natural log of the difference between the last year of the three-year window and the initial CRSP listing date for the firm. Idiosyncratic volatility is the standard deviation of the market-adjusted residuals of daily stock returns from 205 days to 6 days prior to the announcement. Leverage is debt (defined as the sum of long-term debt, the current portion of long-term debt, and notes payable) scaled by debt plus the book value of equity. For each transaction type, IB market share is defined as the adjusted market share of the firm s top IB across all sample transactions of that type during the most-recent twelve months. Each model contains year fixed effects. The p-values based on robust standard errors are presented in parentheses below the coefficients. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. All Transactions Equity Transactions Debt Transactions Mergers HH Top N50 HH Top N50 HH Top N50 HH Top N50 Intercept *** *** *** *** *** *** *** *** *** *** *** *** Ln MV *** *** *** (.648) (.905) *** *** *** *** (.007) *** *** (.135) Rel. Transaction Value *** *** *** *** *** *** *** *** *** (.007) *** *** *** M/B *** *** *** (.410) (.726) *** *** *** * (.061) *** *** *** Ln Age (.782) (.299) *** * (.054) (.238) *** (.636) (.342) ** (.017) * (.086) (.128) *** Idiosyncratic Volatility (.469) (.407) ** (.042) *** (.001) *** (.001) *** *** *** ** (.012) ** (.016) ** (.036) (.458) Leverage *** *** *** (.409) (.191) *** (.007) *** *** *** *** *** *** IB Market Share *** *** (.211) (.718) (.873) (.106) * (.082) (.204) (.485) ** (.022) *** (.004) (.181) N 10,568 10,568 10,568 2,107 2,107 2,107 3,148 3,148 3,148 3,602 3,602 3,602 Adjusted R

42 Table V Investment Bank Retention Frequency This table shows the frequency with which any investment bank from transaction t-1 appears in transaction t. The top number in each cell represents the frequency when we examine only the lead investment bank role, and the bracketed number is the frequency when we examine important investment bank roles. Rows represent the type of transaction at time t-1 and columns represent the type of transaction at time t. In Panel A we report frequencies from the entire sample and from the entire sample when we restrict the time between t and t-1 to three years. In Panel B, we report frequencies for sequential transactions in and In the last column of each sample we report the p-value from a difference in means test that tests whether the frequencies in each of the previous columns are equal. In Panel B, we report the significance of the difference in frequencies between and by asterisks: ***, **, * representing statistical significance at the 1%, 5%, and 10% levels, respectively. First Transaction All 50.7% [65.1%] 14,846 Panel A Full Sample Period Results Second Transaction All Transactions All M&A Debt SEO p-value for difference across types 46.2% [55.2%] 5, % [69.6%] 6, % [76.0%] 2, [.0001] 3-Year Restriction All M&A Debt SEO p-value for difference across types 53.4% [67.7%] 13, % [58.9%] 4, % [70.4%] 5, % [80.3%] 2, [.0001] M&A 45.0% [52.5%] 4, % [48.9%] 2, % [54.0%] 1, % [63.6%] [.0001] 47.1% [54.9%] 3, % [52.1%] 1, % [55.4%] 1, % [66.4%] [.0001] Debt 51.1% [68.2%] 6, % [55.1%] 2, % [74.1%] 3, % [75.8%] [.0001] 52.3% [69.5%] 5, % [57.6%] 1, % [74.0%] 3, % [80.0%] [.0000] Equity IPO 59.4% [78.7%] 1, % [67.0%] % [84.9%] % [88.2%] [.0001] 68.1% [86.7%] 1, % [76.7%] % [91.7%] % [94.1%] [.0001] SEO 55.4% [72.5%] 2, % [64.0%] % [78.8%] % [75.0%] [.0001] 61.5% [77.6%] 1, % [71.2%] % [81.0%] % [79.9%] [.0001] 41

43 First Transaction All 49.6% *** [65.5%] *** 8,459 Table V Continued Panel B Subperiod Results Second Transaction All M&A Debt Equity p-value for difference across types 47.6% ** [58.1%] 3, % *** [66.6%] *** 3, % * [80.5%] * 1, [.0001] All M&A Debt Equity p-value for difference across types 57.6% [69.4%] 4, % [57.1%] 1, % [76.8%] 1, % [77.2%] [.0001] M&A 44.4% *** [52.8%] *** 2, % [50.8%] 1, % *** [51.8%] *** % [66.9%] [.0001] 50.2% [57.1%] 1, % [52.5%] % [59.9%] % [67.3%] [.0018] Debt 47.8% *** [65.4%] *** 3, % *** [55.3%] 1, % *** [68.6%] *** 2, % ** [77.0%] [.0001] 58.5% [74.5%] 2, % [58.2%] % [84.7%] 1, % [79.1%] [.0001] Equity IPO 62.2% *** [83.5%] % * [73.9%] % *** [88.4%] * % [92.0%] [.0001] 72.4% [85.3%] % [71.9%] % [96.4%] % [91.4%] [.0001] SEO 56.1% *** [77.2%] ** 1, % [70.3%] ** % *** [81.0%] % ** [82.8%] *** [.0001] 62.6% [72.9%] % [62.9%] % [79.8%] % [74.1%] [.0004] 42

44 Table VI Probit Analysis of Investment Bank Retention This table provides results from a Probit model of the decision to retain an investment bank on consecutive transactions, where the current transaction is denoted as t and the previous transaction is t-1. The model includes one observation for each investment bank on deal t-1 and the dependent variable equals one if the investment bank is also retained on deal t. Panels A presents results based on lead roles only and Panel B presents results based on all important roles. The model in column (1) includes all transaction types. Columns (2) (4) present results for paired transactions in which transaction t is equity, debt, or merger, respectively. Columns (5) (7) present results for paired transactions of the same transaction type. Firm age is the year in which t occurs minus the start date of the firm s CRSP Permco. Last transaction was an IPO is equal to one if t-1 is an IPO and is zero otherwise. Same day is equal to one if the paired transactions occur on the same day and is zero otherwise. Years since t-1 is the date of t minus the date of t-1. Transaction value is the issue size or deal value reported in SDC for transaction t. The variables Value t ½Value t-1 and Value t 2Value t-1 are dummy variables taking on the value of one if the current transaction value is less than half or more than twice the size of the prior transaction, respectively. Dummy variables are used to identify transactions during the first third and the last third of the sample period. Interim transaction is a dummy variable equal to one if there is a transaction between the pair analyzed in the model. Same transaction type is a dummy equal to one if the current and previous transactions are of the same type (i.e., equity, debt, or M&A). IB market share is the adjusted market share of an IB across all transactions of a particular type (i.e., equity transactions, debt transactions, or mergers) during the previous twelve months, where adjusted market shares give each colead/co-advisor credit for a proportion 1/N of deal value. The lending dummy equals one if the IB provided lending services during the previous twelve months. Each model contains both year and investment bank specific fixed effects. p-values based on robust standard errors are presented in parentheses below the coefficients, where standard errors are clustered by transaction. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. 43

45 (1) All Table VI - Continued Panel A. Lead Models (3) t=equity (2) t=debt (4) t=merger (5) t=debt t-1=debt *** (.006) *** (6) t=equity t-1=equity (.274) (.823) (7) t=merger t-1=merger ** (.022) *** (.002) - Intercept *** *** (.165) *** Firm age (years) *** *** * *** (.054) Last transaction was IPO *** ** *** *** *** (.020) (.002) (.001) Same day *** *** *** *** *** Years since t *** *** *** *** *** *** *** Transaction value ** ** ** (.759) (.922) (.026) (.272) (.627) (.037) (.011) Value t ½Value t *** *** *** *** *** *** *** Value t 2Value t *** *** *** (.325) (.193) (.005) (.203) (.817) ** (.305) (.643) (.513) (.023) (.288) (.909) (.413) *** *** * *** ** *** (.004) (.054) (.912) (.003) (.012) Interim transaction *** (.323) (.350) (.005) Same transaction type *** *** * *** (.003) (.083) Same transaction type (.182) (.364) (.525) (.945) Same transaction type ** *** (.135) (.040) (.821) (.001) IB s Equity Market Share *** * *** *** *** (.092) (.002) (.327) (.511) IB s Debt Market Share *** *** *** *** *** (.191) (.008) (.803) IB s M&A Market Share *** * *** * * *** (.176) (.055) (.065) (.081) IB Lending Dummy *** *** (.008) (.343) (.542) (.630) (.336) (.260) # of Observations 20,711 9,268 3,085 7,895 8,308 2,443 4,428 # of Transactions 14,070 6,328 2,413 5,010 5,147 1,940 3,412 Pseudo R

46 (1) All Table VI - Continued Panel B. Important Models (2) (3) (4) t=debt t=equity t=merger (5) t=debt t-1=debt *** (.388) (6) t=equity t-1=equity *** (.603) (7) t=merger t-1=merger *** (.003) *** (.002) - Intercept *** *** ** (.018) *** Firm age (years) *** *** (.225) (.310) Last transaction was IPO *** *** *** *** (.371) (.002) (.006) (.001) Same day *** *** *** *** *** (.001) Years since t *** *** *** *** *** *** *** Transaction value ** *** *** *** ** (.013) (.184) (.602) (.005) (.012) Value t ½Value t *** *** *** *** *** *** *** Value t 2Value t *** *** ** *** (.005) (.817) (.038) (.003) (.917) (.637) *** *** (.351) (.251) (.688) (.005) (.006) (.622) (.451) *** *** *** *** *** (.153) (.272) (.003) Interim transaction *** (.614) (.515) (.004) Same transaction type *** *** ** *** (.002) (.028) Same transaction type * *** (.065) (.001) (.921) (.139) Same transaction type *** ** *** (.028) (.514) (.001) IB s Equity Market Share *** *** *** *** (.376) (.645) (.381) IB s Debt Market Share *** *** *** *** *** (.530) (.775) IB s M&A Market Share *** *** *** *** *** (.293) (.001) (.270) (.001) IB Lending Dummy *** *** *** * *** *** (.092) (.003) (.699) # of Observations 44,398 20,150 6,695 16,440 20,617 6,330 4,559 # of Transactions 14,070 6,328 2,413 5,010 5,147 1,947 3,412 Pseudo R

47 Table VII Top Investment Banks Across Functional Areas The table examines the frequencies with which the firm s top investment bank in one transaction type is among the firm s top investment banks in other transaction types. For each transaction type, IB ranks are determined based on adjusted market shares on the firm s deals during the specified time period, where adjusted market shares give each co-lead/co-advisor credit for a proportion 1/N of deal value. For each transaction type comparison, firms are required to have at least two transactions of each type during the sample period or during the specified subperiod, with the last column in each grouping listing the number of firms with the required number of transactions. Results in Panel A are based on transactions during the entire sample period from 1996 through Results in Panels B and C are based on transactions during and , respectively. The column labeled Did Not Appear shows the frequency with which the firm s top investment bank in one transaction type did not lead a deal in the other transaction type. Equity IB Rank Debt IB Rank Merger IB Rank Did Not Did Not Did Not N N Appear Appear Appear N Panel A: Firms with at least two deals of each type from Equity Top IB % 12.9% 3.5% 17.7% % 9.7% 2.0% 30.4% 392 Debt Top IB 62.4% 10.3% 1.3% 25.7% % 13.1% 3.1% 31.4% 682 Merger Top IB 56.9% 10.2% 1.3% 31.6% % 13.5% 5.9% 24.2% Panel B: Firms with at least two deals of each type from All Firms: Equity Top IB % 13.2% 4.1% 15.7% % 6.8% 0.7% 23.1% 147 Debt Top IB 63.6% 9.9% 0.8% 25.6% % 11.0% 2.4% 28.1% 335 Merger Top IB 68.7% 6.8% 0.0% 24.5% % 9.9% 4.2% 24.2% With Lending Relationship: Equity Top IB % 11.1% 2.8% 2.8% % 3.1% 3.1% 12.5% 32 Debt Top IB 52.8% 11.3% 0.0% 35.9% % 6.4% 2.7% 33.6% 110 Merger Top IB 81.3% 9.4% 0.0% 9.4% % 12.8% 4.7% 11.6% Without Lending Relationship: Equity Top IB % 14.1% 4.7% 21.2% % 7.8% 0.9% 26.1% 115 Debt Top IB 72.1% 8.8% 1.5% 17.7% % 13.3% 2.2% 25.3% 225 Merger Top IB 65.2% 6.1% 0.0% 28.7% % 8.8% 4.0% 28.5%

48 Table VII Continued Equity IB Rank Debt IB Rank Merger IB Rank Did Not Did Not Did Not N N Appear Appear Appear N Panel C: Firms with at least two deals of each type from All Firms: Equity Top IB % 10.7% 4.0% 10.7% % 14.3% 2.9% 22.9% 70 Debt Top IB 72.0% 8.0% 1.3% 17.3% % 14.9% 3.0% 33.6% 235 Merger Top IB 60.0% 11.4% 0.0% 27.1% % 15.7% 7.2% 22.6% With Lending Relationship: Equity Top IB % 10.5% 0.0% 5.3% % 19.2% 7.7% 3.9% 26 Debt Top IB 81.6% 2.6% 0.0% 13.2% % 17.4% 4.2% 27.1% 144 Merger Top IB 72.0% 8.0% 0.0% 20.0% % 21.1% 5.5% 1.8% Without Lending Relationship: Equity Top IB % 10.8% 8.1% 16.2% % 11.4% 0.0% 34.1% 44 Debt Top IB 62.2% 13.5% 2.7% 21.6% % 11.0% 1.1% 44.0% 91 Merger Top IB 53.3% 13.3% 0.0% 31.1% % 11.1% 8.7% 40.5%

49 Table VIII Probit Analysis of Investment Bank Selection This table presents probit models in which the dependent variable is equal to one if a particular investment bank is the lead bank on a particular transaction and is zero otherwise. In Panels A, B, C, and D we examine equity transactions, debt transaction, mergers, and all transaction types, respectively. Models (2) and (3) in every panel interact all independent variables with sub-period dummies, which represent (Subperiod 1), (Subperiod 2), and (Subperiod 3). We throw out all transaction from 1996 so that the market share and relationship variables contain at least one year of observations. Ln(Transaction size) is the natural logarithm of issue size or deal value reported in SDC. IB market share is the adjusted market share of an IB across all transactions of a particular type (i.e., equity transactions, debt transactions, mergers, or lending) during the previous twelve months, where adjusted market shares give each co-lead/co-advisor credit for a proportion 1/N of deal value. In Panels A, B, and C, IB s share of a firm s deals is the adjusted market share of an IB across all of the firm s transactions of a particular type (i.e., equity, debt, mergers, or lending) during the previous three years, where adjusted market shares give each co-lead/co-advisor credit for a proportion 1/N of deal value. In Panel D IB s Share of Firm s deals is defined similarly based on all equity, debt, and merger transactions. Lead on prior deal is a dummy variable equal to one if the IB was a lead bank on the firm s prior transaction or security issue. Competitor s Top IB is a dummy variable equal to one if the IB is the most used bank by the firm with the highest revenue amount for a particular four digit SIC code. For the top firm in the industry Competitor s Top IB is equal to one for the top bank of the firm with second-highest revenue. Comanager on Prior Deal is equal to one if the investment bank is a comanager on the firm s prior transaction. In Panel D we include interactions with dummies for whether the transaction is not a specific transaction type. Each model contains both year and investment bank specific fixed effects. p-values based on robust standard errors are presented in parentheses below the coefficients, where standard errors are clustered by transaction. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. 48

50 Table VIII Continued (1) Lead Intercept *** Ln(Transaction Size) *** IB s Equity Market Share *** IB s Debt Market Share ** (.037) IB s M&A Market Share (.709) IB Lender (.195) IB s Share of Firm s Equity Deals *** IB s Share of Firm s Debt Deals *** IB s Share of Firm s M&A Deals *** IB s Share of Firm s Loans *** Lead on Prior Deal *** Lead on Prior Deal Competitor s Top IB *** (.005) Comanager on Prior Deal *** Comanager on Prior Deal Competitor s Top IB (.888) Panel A Equity Deals (2) Lead Variable Variable Subperiod 1 Subperiod *** (.367) *** (.001) ** (.025) (.761) (.462) *** ** (.012) *** (.221) *** (.414) *** (.252) (.204) *** (.403) (.746) ** (.042) *** *** *** *** *** (.137) *** (.326) Variable Subperiod ***a *** c (.860) (.874) (.480) *** *** ***c ***a ***b ** (.014) *** (.782) Variable Subperiod *** *** (.216) (.937) (.853) *** *** *** ** (.043) *** (.623) (.203) (.932) (3) Important Variable Subperiod *** *** *** (.226) (.319) * (.095) *** *** *** *** *** * (.098) ** (.020) (.846) Variable Subperiod ***b ***a (.684) (.518) (.937) ***b ***c ***c *** ***c ** (.044) **b (.045) (.326) Year Fixed Effects Yes Yes Yes IB Fixed Effects Yes Yes Yes # of Transactions (not unique) 1,895 1,895 1,895 # of Observations 53,316 53,316 80,197 Pseudo R

51 Table VIII Continued (1) Lead Intercept *** Ln(Transaction Size) *** IB s Equity Market Share (.362) IB s Debt Market Share *** IB s M&A Market Share (.993) IB Lender (.148) IB s Share of Firm s Equity Deals *** IB s Share of Firm s Debt Deals *** IB s Share of Firm s M&A Deals *** IB s Share of Firm s Loans *** Lead on Prior Deal *** Lead on Prior Deal *** Competitor s Top IB (.005) Comanager on Prior Deal *** Comanager on Prior Deal Competitor s Top IB (.975) Panel B Debt Deals (2) Lead Variable Subperiod *** Variable Subperiod (.455) (.928) *** (.758) (.916) *** *** *** *** *** ** (.021) *** (.213) *** (.238) *** (.696) *** (.004) *** *** *** *** *** * (.056) *** (.170) Variable Subperiod ***a (.267) ***a * (.065) b (.477) ***a ***a *** ***a *** (.219) *** * (.093) Variable Subperiod *** (.834) *** (.641) (.359) *** *** *** *** *** ** (.012) *** (.641) (3) Important Variable Subperiod *** *** (.843) *** (.487) *** *** *** *** *** *** ** (.003) (.743) (.702) Variable Subperiod ***a ***a ***a *b (.058) a (.841) ***a ***a *** ***a ***a * (.062) **a (.015) (.535) Year Fixed Effects Yes Yes Yes IB Fixed Effects Yes Yes Yes # of Transactions (not unique) 4,875 4,875 4,875 # of Observations 143, , ,865 Pseudo R

52 Table VIII Continued (1) Lead Intercept *** Ln(Transaction Size) *** IB s Equity Market Share (.883) IB s Debt Market Share ** (.049) IB s M&A Market Share *** IB Lender (.125) IB s Share of Firm s Equity Deals *** IB s Share of Firm s Debt Deals *** IB s Share of Firm s M&A Deals *** IB s Share of Firm s Loans *** Adviser on Prior Deal *** Adviser on Prior Deal Competitor s Top IB (.110) Panel C Mergers (2) Lead Variable Subperiod *** Variable Subperiod *** (.265) (.584) *** (.225) *** *** *** *** *** ** (.011) *** (.175) *** *** (.103) *** *** *** *** *** (.546) Variable Subperiod *** (.611) a (.352) ***b (.545) *** ***c *** ***b ***b c (.604) Variable Subperiod *** (.129) (.406) *** (.370) *** *** *** *** *** ** (.012) (3) Important Variable Subperiod *** *** (.804) ** (.031) *** * (.090) *** *** *** *** *** (.902) Variable Subperiod *** (.719) b (.189) *** (.606) ***a *** ***c ***a ***c (.350) Year Fixed Effects Yes Yes Yes IB Fixed Effects Yes Yes Yes # of Transactions (not unique) 3,288 3,288 3,288 # of Observations 112, , ,959 Pseudo R

53 Table VIII Continued (1) Lead Intercept *** Ln(Transaction Size) *** IB s Equity Market Share *** IB s Equity Market Share Current Deal is Not Equity *** IB s Debt Market Share *** IB s Debt Market Share Current Deal is Not Debt *** IB s M&A Market Share *** IB s M&A Market Share Current Deal is Not M&A *** IB Lender *** (.008) IB s Share of Firm s Transactions *** IB s Share of Firm s Loans *** Lead on Prior Deal that is Equity *** Lead on Prior Deal that is Equity Current Deal is Not Equity *** Lead on Prior Deal that is Debt *** Lead on Prior Deal that is Debt Current Deal is Not Debt *** (.002) Lead on Prior Deal that is M&A *** Lead on Prior Deal that is M&A Current Deal is Not M&A *** Panel D All transaction types (2) Lead Variable Subperiod *** Variable Subperiod *** *** *** *** *** *** *** (.325) *** *** *** *** (.001) *** (.163) *** *** *** *** *** *** *** *** *** *** (.001) *** *** *** ** (.016) *** *** (.004) *** *** Variable Subperiod ***a ***b ***a ***a ***a ***a ***c * (.096) ***a ***a *** *** (.010) ***a (.834) *** *** Variable Subperiod *** *** *** *** *** *** *** (.637) *** *** *** *** *** ** (.020) *** *** (3) Important Variable Subperiod *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** Variable Subperiod ***a ***a ***a ***a ***a ***a ***a a (.348) ***a ***a ***a ***a (.006) ***a ***a ***c *** Year Fixed Effects Yes Yes Yes IB Fixed Effects Yes Yes Yes # of Transactions (not unique) 13,352 13,352 13,352 # of Observations 448, , ,149 Pseudo R

54 5,000 Panel A Aggregate Deal Frequency by Year 4,500 4,000 3,500 Number of Deals 3,000 2,500 2,000 1,500 1, Equity Debt Other Issues M&A Acquirer M&A Target Panel B Aggregate Deal Value by Year ($bil) 1,100 1, Aggregate Deal Value ($ bil) Equity Debt Other Issues M&A Acquirer M&A Target Figure 1 Total Event Activity by Year 53

55 16.0 Panel A Equity Syndicate Characteristics by Year 16.0 Panel B IPO Syndicate Characteristics by Year NLEAD NCOMGR NSYND NLEAD NCOMGR NSYND Panel C Debt Syndicate Characteristics by Year Panel D M&A Advisory Group Characteristics by Year Figure 2 Syndicate Characteristics by Year 54