VALUE EFFECTS OF INVESTMENT BANKING RELATIONSHIPS. Alexander Borisov University of Cincinnati. Ya Gao University of Manitoba

Transcription

1 VALUE EFFECTS OF INVESTMENT BANKING RELATIONSHIPS Alexander Borisov University of Cincinnati Ya Gao University of Manitoba This Version: January 2018 Abstract This paper examines the firm value effects of the business relationship between investment banks and their corporate clients. Using an event study approach and a sample of consolidation activities in the investment banking industry, we show that firms that rely more on the services provided by their investment banks experience significant negative abnormal returns upon the announcement of a merger involving their bank as target. The adverse effect is more pronounced for firms that are more likely to be financially constrained. By contrast, the negative price impact is attenuated when the acquiring bank is more reputable or has larger underwriting capacity. We conduct several tests to establish that the observed effect is indeed relationship-specific and not a reflection of unobservable deal or client characteristics. JEL Classification: G24, G14, G34 Keywords: Investment Banking, Bank-firm Relationship, Mergers and Acquisitions, Event Study 1

2 VALUE EFFECTS OF INVESTMENT BANKING RELATIONSHIPS 1. Introduction The interaction and business relationships between investment banks (IBs) and their client firms have been of significant academic and policy interest, especially in the aftermath of the global financial crisis of that was accompanied by the failure of major investment banks such as Lehman Brothers and an almost unprecedented collapse of the investment banking industry. While existing studies advance various arguments, ranging from acquisition of valuable private information and monitoring to switching and transactions costs, for the importance of long-lasting relationships between IBs and their client firms, quantifying or even identifying the firm value implications of such relationships is challenging. From an empirical point of view the main difficulty rests in the endogenous nature of the decision to sustain the business relationship. In this paper we propose an empirical strategy that could allow us to offer insights into the question whether maintaining investment banking relationships creates value for the client firms. To this end, we use the incidence of mergers and acquisitions (M&As) in the investment banking industry as an experimental setting to identify the value effect of investment banking relationships and examine its determinants. 1 Specifically, we propose that a merger between two IBs could serve as a shock to the established relationship between these banks and their client firms. Consequently, we use an event study approach to test if such shocks are reflected in the returns experienced by client firms at the announcement of the business combination. In efficient markets, such (abnormal) announcement returns should capture the value created or destroyed by the shock. We also investigate what factors affect the magnitude of any value effects. Based on a large sample of mergers and acquisitions between IBs that took place between 1980 and 2014, we first document that client firms of the target bank experience a cumulative 1 In what follows we use the terms mergers, acquisitions, and business combinations interchangeably. 2

3 abnormal return (CAR) of -.2% in the 3-day window centered at the announcement of the business combination. While this average magnitude is modest, we demonstrate that it masks important heterogeneities across client firms. Specifically, we find that firms rely on their IBs more extensively for equity underwriting and/or M&A advising services prior to the event have a CAR of -1.6% in the 3-day window. To capture reliance, we develop two metrics: The first measure is the number of equity offerings by the firm conducted with the IB during the past five years, while the second measure is the number of instances of M&A advising of the firm by the IB during the same period. We also construct the value-based (in dollar terms) counterparts of these measures. Overall, the preliminary univariate analysis suggests that firms that depend more on their IBs derive more value from these relationships and an adverse shock could result in value destruction. We next proceed to examine the value of investment banking relationships in a series of multivariate regressions that control for firm characteristics, aggregate economic conditions and industry factors, and even unobservable heterogeneity at the business combination level. This analysis further allows us to generate insights into what drives the value implications of these relationships. We show that the loss of value is more pronounced for firms that appear to be more financially constrained. Existing research suggests that IB reputation is an important driver of firms decision to switch IBs or maintain the relationship (Krigman, Shaw, and Womack (2001), Fernando, Gatchev, and Spindt (2005)). Hence, we examine the extent to which IB s reputation influences the value derived by the client firms. Consistently, we find that if an IB is acquired by another IB with a higher reputation, or with a larger capacity to provide investment banking services, client 3

4 firms of the target bank experience less negative abnormal returns. These findings further highlight the importance of IBs reputational capital. We perform several tests to verify the robustness of our results. First, while the main analysis uses market-adjusted CARs calculated over a 3-day window centered at the business combination announcement, we confirm that our insights continue to hold if we use other windows (e.g., 11-day window) or different procedures for calculating CARs (e.g., marketmodel or Fama-French 3-factor model). Second, our sample period is from 1980 to 2014, thus including the burst of the dot-com bubble and the Great Recession of the financial crisis that witnessed the demise of the investment banking industry and the collapse of Lehman Brothers. Therefore, we perform our analysis using only pre-2000 events and find qualitatively similar effects. Last, in order to establish that the observed market reactions are indeed due to the relationship between the IB and its client, i.e. the pair-wise IB-firm combination, and not client characteristics that make it sensitive to factors pertaining to aggregate equity issuance or M&A activity, we confirm that the CARs are not related to the total amount of investment banking services received by the client firm but only to the activity attributable to the specific IB relationship. Our paper contributes to the existing literature that advances various arguments in favor or against the value of investment banking relationships. For instance, Hansen and Torregrosa (1992) argue that in the process of providing services, IBs exert monitoring over the manager and thus could increase the value of the client firms. Benveniste and Spindt (1989), Cornelli and Goldreich (2001), Ritter and Welch (2002), and Ljungqvist, Jenkinson, and Wilhelm Jr. (2003) suggest that by soliciting bids in their established institutional networks, IBs reduce underpricing of equity offerings. Another source of value arises from the ability of IBs to generate and reuse 4

5 valuable private information (James (1992), Schenone (2004), and Benveniste et al. (2003)). In addition to the arguments above, the value loss associated with the disruption of the relationship might also arise from presence of switching and transactions costs (e.g., Burch, Nanda, and Warther (2005)) or as in Fernando, Gatchev, and Spindt (2005), disruption of an optimal firm-ib matching. By examining the value implications of shocks to IB relationships generated through business combinations between banks, and drivers of the value effects, our paper speaks directly to this line of research. The papers closest to ours are Fernando, May, and Megginson (2012) and Kovner (2012). Both examine the value implications of IB relationships focusing on the unique instance in time resulting from the global financial crisis and the failure or near failure of major IBs. Consistent with our results, both studies observe significant negative market reaction for client firms at the announcement of an adverse shock to the IB. Our study builds upon these works along two dimensions. First, we use a more comprehensive time period that covers rather than a specific point in time that is associated with large simultaneous shocks to market and economy as the financial crisis. While the fall of Lehman Brothers, or other IBs, undoubtedly provides a strong natural experiment that can be seen as exogenous to established relationships, the unusual nature of the event could create a system-wide impact that may bias the economic magnitudes of the observed effects. Second, in our setting we can further examine how the relative capacity for provision of services and reputation of the two IBs can affect the value implications. The cross-sectional variation of IB-client pairs allows us to study the effect of additional characteristics on the process of value creation. 5

6 This rest of the paper is organized as follows. The next section describes the data and sample construction. Section 3 presents our main findings. Section 4 offers robustness tests, while Section 5 concludes. 2. Sample Construction and Summary Statistics We construct our sample using data from three different sources: data on business combinations between IBs; data on services received by client firms of these IBs; and data on client firms performance around the merger announcement dates. For the sample of business combinations between IBs, we start with all mergers and acquisitions announced between 1980 and 2014 where both the bidder and target are financial institutions (SIC codes between 6000 and 7000) from the Securities Data Corporation (SDC) Mergers and Acquisitions (M&A) Database. We do not explicitly require both bidder and target to be IBs in order to generate the largest possible initial sample and to include deals involving institutions that provide investment banking services but do not identify themselves as IBs based on SIC code. 2 The resultant dataset contains 53,384 transactions over the period of 35 years. To identify the client firms of the IBs involved in mergers, we match the investment bank merger dataset with SDC M&A Database and SDC Global New Issuance Database, respectively, for records of advisory and underwriting services received by the client firms. With the absence of a unique identifier across these databases, we first perform a fuzzy matching process between the full name of the IBs and their names as advisors and/or underwriters that appear in the SDC M&A Database and SDC Global New Issuance Database. To minimize incorrect matches, we discard observations with a generalized edit distance matching score greater than 500 and 2 We use the term investment bank loosely to refer to financial institutions that provide either underwriting or advisory services. 6

7 manually verify the remaining matches. 3 After matching, in order to avoid issues created by potential variations in IB s full names, we use IB s short name in the SDC M&A Database and SDC Global New Issuance Database as unique identifiers of IBs. The resultant list of IBs consists of 134 unique institutions involved in a total of 900 mergers. With the short name as unique identifier for each IB, we match the IBs with their active client firms prior to the merger announcement date. We consider a firm as an active client of the IB if the bank has been listed either as M&A advisor or leading manager of an equity offering by the client firm within the 5 years prior to the merger announcement. If the client firm switched to another IB for the same type of service (advisory or underwriting) prior to the merger, the client firm will no longer be considered as a client of the previous IB. If an IB is involved in multiple mergers within the 5-year window, client firms experience multiple shocks to their existing relationships. However, the impact of later shocks could be influenced by earlier shocks and market expectation of future M&A activity of the IB. Therefore, we only consider the first shock after the client s transaction with its IB. After imposing these criteria, the resultant dataset consists of 2,549 investment bank-client firm pairs. With these pair-wise relationships, we calculate the reliance of client firms on their IBs, the length of their relationship, and some client-firm characteristics as controls. We present summary statistics for these variables in Table 1 and detailed definitions in the Appendix. [Insert Table 1 Here] 3 Generalized edit distance (GED) measures the dissimilarity between two strings. Each editing operation is assigned a unit cost (e.g. additional blank at 10, appending characters at 10, deletion at 100), and the generalized edit distance is calculated as the minimum-cost sequence of operations for constructing one string from another. For example, the GED from Dean Witter to Dean Witter & Co is 170, calculated as three instances of characters appending at 50 each and two blanks at 10 each. 7

8 We use the equity offerings and M&A advising services provided by IBs to capture the reliance of client firms on their IBs. Specifically, Equity Count is the number of offerings completed by the client firm with the IB in the 5 years prior to the merger. We note from Panel A in Table 1 that the average Equity Count is.93, while the median is 1. The corresponding average dollar value, captured through Equity Value, is about $118 million. We use a similar set of measures for M&A advisory services. M&A Count is the number of advisory services a client firm receives from the IB in the 5-year period prior to the merger. As reported in Table 1 Panel A, the average number of advisory services a client firm receives from its IB is 0.40, and the 75 th percentile is 1. Hence, only about a quarter of the IB relationships involve advisory services. Statistics on M&A Value show that the average dollar amount of these deals is $328 million. As some of the variables are highly skewed, especially the value-based measures, we take natural logarithm after adding 1 to preserve observations without either equity underwriting or M&A advising. Another factor that may affect the value of the relationship is its length. We consider two measures of relationship length. Days since First Deal counts the number of days between the first transaction a client firm had with its IB and the announcement date of the business combination involving the IB. In our sample, client firms on average have 972 days history with their IBs before the shock, with a median of 808 days. 4 The second measure for relationship length is Days since Last Deal, which is the number of days between the most recent transaction a client firm had with its IB and the business combination announcement date. The average gap between the most recent transaction and the IB s merger is 543 days, with a median of 528 days. 4 The minimum Days since First Deal is 1, when Enron Global Power & Pipelines used Dillon Read & Co as their M&A advisor on May 14 th, 1997 and Dillon Read was announced to be purchased by Swiss Bank Corporation the next day on May 15 th. The longest Days since First Deal is between Goldman Sachs and Circuit City, for almost 24 years (8,754 days). 8

9 We obtain client firms stock prices and return information from the Center for Research in Security Prices (CRSP) and market-adjusted cumulative abnormal returns (CAR) for (-1,+1) and (-5,+5) windows centered at the IB merger announcement date from EVENTUS. 5 We note that the average 3-day CAR is -.2%, while the median is -.4%. While these figures are modest, we show that they mask important heterogeneity across firms. Last, we use accounting data from COMPUSTAT to construct several variables related to firm characteristics that might influence the value implications of IB relationships. Specifically, we construct measures that capture size, growth opportunities, capital structure, and financial constraints. In Panel B of Table 1, we show CARs for three levels (high, medium, and low) of client dependence on services provided by the IB. A client s dependence is considered as high if both its Equity Count and M&A Count are above their respective cross-sectional medians. If a firm has either Equity Count or M&A Count above the median, then the firm is assigned to the medium level of dependence group. The rest of the firms are in the low dependence category. We show that while the 3-day and 5-day CARs of client firms in the medium and low dependence groups are not statistically distinguishable, the CARs for the client firms in the high dependence group are negative and significantly lower than those of firms in the median group. This is consistent with the idea that firms with more pronounced reliance on the services of their IBs and stronger relationships tend to lose more value around the IB mergers. 5 We use the CRSP value-weighted index to capture market returns. 9

10 3. Empirical Results 3.1. Main Result The preliminary analysis in the previous section suggests that firms that rely more on their IBs for services experience more negative abnormal returns at the announcement of a merger involving their IB. We next proceed to verify this result in a multivariate setting. Specifically, we estimate a series of ordinary least squares (OLS) regressions with the 3-day CAR as dependent variable and our measures of reliance on IB services as key independent variables of interest, while controlling for various firm-specific characteristics and aggregate economic and industry conditions. Our main results are presented in Table 2. [Insert Table 2 Here] In column (1) of Table 2, we regress the 3-day CAR on ln(1+equity Count) and ln(1+m&a Count). The coefficient on the former is negative and statistically significant at the 5% level, while the coefficient on the latter is negative and statistically significant at the 10% level. This is consistent with the idea that firms that rely more on the IBs experience more negative market reaction at the announcement of the merger involving their IB. Using the estimates in column (1) we compute that an increase of 1 in Equity Count, i.e. an additional equity offering, is associated with a drop in the 3-day CAR of.5%. For the median firm, this corresponds to a $1.5 million loss of value in the 3-day window around the merger event. We estimate an effect of similar magnitude for an increase of 1 in M&A Count. Thus, we document not only the statistical significance of the effects we study but also their economic importance. 10

11 In column (2), we replace the count measures with their value-based counterparts. We note that while both ln(1+equity Value) and ln(1+m&a Value) are negative, only the former is significant at the 5% level. This result is in line with the empirical findings in Krigman, Shaw, and Womack (2001) and Fernando, May, and Megginson (2012). In columns (3) and (4), we augment the baseline specifications from columns (1) and (2) with controls for firm size, leverage, and market-to-book. We also control for aggregate economic and market conditions through year fixed effects and time-invariant industry characteristics through industry (2-digit SIC code level) fixed effects. The importance of the value effect of client firms reliance on their IB for equity underwriting continues to hold. The coefficients on both the count-based measure in column (3) and the dollar-based measure in column (4) are negative and statistically significant at the 5% level. Last, we confirm that the observed value effects are not driven by certain deals. To this end, we estimate a model with deal fixed effects in columns (5) and (6). The coefficients on both count-based and dollar-based measures continue to be negative and significant at the 10% level or better Financial Constraints and IB Reputation Our results so far are consistent with the notion that IB relationships are valuable for client firms and shocks to these relationships lead to destruction of firm value. Another finding from our analysis is that the effects appear to be stronger for relationships built on underwriting services, than on M&A advising. To further corroborate this point, and to offer an insight into possible channels underlying the value effects, we examine the role of firms financial constraints. On one hand, the IB relationship would be most valuable for financially constrained 11

12 firms that depend on their banks. On the other, these might be firms that are more likely to be captured in the relationship. Thus, we hypothesize that clients with less financial slack will be more adversely affected by a shock to the established relationship. We adopt two approaches to examine this argument. First, we follow Kaplan and Zingales (1997) and construct an index (KZ Index) intended to reflect financial constraints at the firm level. 6 Using the index, we rank all firms into three categories (from 1 to 3, where 3 is most constrained) based on the magnitude of their index and assign to each firm the rank of its category (KZ-rank). We then add the KZ-rank as an additional explanatory variable. We note from columns (1) and (2) of Table 3 that the coefficient of this variable is negative and statistically significant at the 10% level. This result is consistent with the idea that more financially constrained firms lose more value upon the shock to their IB relationship [Insert Table 3 Here] Our second approach to capturing the role of financial constraints follows Benmelech, Bergman, and Seru (2011). Specifically, we construct the measure Maturing Debt as the amount of long-term debt maturing at year t+1 that was issued at least one year prior to the merger announcement date of the client firm s IB. In columns (3) and (4) of the table we show that the coefficient on Maturing Debt is negative and statistically significant at 5% level. This is consistent with our hypothesis about the role of financial constraints. Last, we add in columns (5) and (6) both measures and note that both measures are negative and statistically significant. Existing research (e.g., Krigman, Shaw, and Womack (2001)) documents that client firms tend to switch to higher reputation underwriters in their SEO. This suggests that the reputation of 6 The index specification and coefficients are based on Lamont, Polk, and Saá-Requejo (2001). 12

13 an IB may affect the value created by the relationship. To test for the existence of such an effect, we consider the underwriting reputation of the IBs involved in the mergers using the underwriter ranking provided by Loughran and Ritter (2004). The ranking is on a scale from 0 to 9, with 9 being the highest. Since each shock, i.e. business combination between IBs, has two parties (target and acquirer), we create a variable Diff Reputation. It is calculated as the acquiring IB s underwriter prestige rank minus the target IB s rank. [Insert Table 4 Here] In columns (1) and (2) of Table 4 we observe that the average effect of the difference in the reputations of the IBs is positive bot not statistically significant. Thus, we cannot establish robust evidence of an effect of IB reputation on the value derived from IB relationships. The prestige ranking is based on the underwriting activities of IBs and some IBs in our sample might focus on M&A advising rather than underwriting. Therefore, we also use the IB s service capacity measured as the total number of underwriting and/or advisory services provided by the IB within the last 5 years. Specifically, we create two indicators. The variable D(A Eq Ct>T Eq Ct) takes the value of 1 if the acquiring IB has underwritten more offerings in the past 5 years relative to the target IB, and 0 otherwise. The variable D(A M&A Ct>T M&A Ct) takes the value of 1 if the acquiring IB has advised more M&As in the past 5 years relative to the target IB, and 0 otherwise. The idea is to reflect the relative difference in the capacity for service provision between the IBs. We also construct similar indicators using the dollar-based capacity of the IB instead of the count-based one. 13

14 In columns (3) and (4) of Table 4 we show that the coefficient estimates of the indicators based on equity underwriting are positive and statistically significant at the 1% and 5% levels, respectively. While the indicators based on M&A advising are negative, they are not robustly significant. Thus, we find some evidence consistent with the argument that the mergers might have a positive value effect if the acquiring IB has a larger underwriting capacity. Overall, this analysis suggests that IB reputation and service provision capacity might have some effects on the value of IB relationships. However, these effects are not robustly significant and thus we could only offer limited evidence. 4. Robustness Tests Our main analysis uses 3-day market adjusted CARs to capture value implications. In this section, we verify that our results are robust to alternative event windows and approaches to estimation of abnormal returns. First, we use 11-day window centered at the announcement of the business combination between IBs to construct CAR(-5,+5). In columns (1) and (2) of Table 5 we show that our main results continue to hold. The coefficient on ln(1+equity Count) is negative and statistically significant at 10% level. The coefficients on ln(1+equity Value) and ln(1+m&a Value) in column (2) are negative and statistically significant at 1% and 5% levels, respectively. [Insert Table 5 Here] Our sample covers 1980 to 2014, a period that witnessed substantial changes and shocks to the investment banking industry such as the burst of the dot-com bubble and the Great 14

15 Recession period. Therefore, we repeat our analysis using deals that took place prior to From columns (3) and (4) of the table we note that our results continue to hold since the coefficients of the equity underwriting measures are negative and significant, albeit only at 10% level. One may also argue that the number of transactions between an IB and its client might reflect the firm s sensitivity to aggregate equity or M&A market conditions. Hence, in columns (5) and (6) we add as additional explanatory variables measures of the underwriting and advising services received by the firm by all IBs and not only those involved in the business combination. We note that the aggregate measures are not statistically significant. By contrast, the measures of the firm s reliance on the IB involved in the merger for equity underwriting continue to have negative and significant coefficients. This finding provides further support for the idea that the value effects we capture are indeed attributable to the shock to the IB-client relationship. Last, we verify the robustness of our main results to alternative procedures for calculating CARs. In columns (1) to (4) of Table 6, we calculate abnormal returns using the Fama-French 3- factor model and use the 3-day CAR as dependent variable. In columns (5) to (8), the abnormal returns are calculated using the market model as reference for return generating process. For both models, factor loadings are calculated from 255 days of return observations 46 days prior to the announcement of the IBs business combination. We note that our main results continue to hold and are in line with the idea that firms that rely more on their IBs experience more negative market reaction at the announcement of the merger involving their IB. [Insert Table 6 Here] 15

16 5. Conclusion In this paper we examine the firm value effect of business relationships between IBs and their clients firms utilizing instances of business combinations between IBs as external shocks to these relationships. Using an event-study approach, we document that firms that rely more on their IBs for provision of underwriting and advising services experience significantly more negative abnormal returns upon announcement of a merger involving their bank as a target. We also show that the effect is more pronounced for firms that are more likely to be financially constrained but the negative price impact could be attenuated when the acquiring IB has a larger capacity for advising and/or underwriting. Overall, we document the value effects of investment banking relationship and identify financial constraints as their main driver. 16

17 Appendix See list below for definition of variables: Variable Definition Book-to-Market Ratio: We follow the definition in Fama and French (2008). Book value is lagged total assets (Compustat Data Item 6) for year t-1, plus balance sheet deferred taxes and investment credit (35) if available, minus liabilities (181), preferred stock liquidating value (10) if available, or redemption value (56) if available, or carrying value (130). Market equity is the product of price and shares outstanding as of December of last year. CAR (-n, +n): Cumulative abnormal return of the client firm n days before and after the IB merger announcement date. Days since First Deal: Number of days between the IB merger announcement date and the date when the IB announced its first service (advisory or underwriting) to the client. Days since Last Deal: Number of days between the IB merger announcement date and the date when the IB announced its last service (advisory or underwriting) to the client. Diff Reputation: The difference between acquiring IB s underwriting rank and the target IB s underwriting rank one year prior to the IB merger announcement date. We use the underwriter ranking provided by Loughran and Ritter (2004). D(A Eq Ct>T Eq Ct) (D(A MA Ct>T MA Ct)): An indicator variable that equals 1 if the acquiring IB has higher sum of Equity Count (M&A Count) from all client firms than the target IB, and 0 otherwise. 17

18 D(A Eq Val>T Eq Val) (D(A MA Val>T MA Val)) An indicator variable that equals 1 if the acquiring IB has higher sum of Equity Value (M&A Value) from all client firms than the target IB, and 0 otherwise. Equity (M&A) Count: Number of underwriting (advisory) services an IB provided to the client firm during the 5 years prior to the IB merger. Equity (M&A) Value: Cumulative total dollar value (in millions) of transactions of all the underwriting (advisory) services an IB provided to the client firm during the 5 years prior to the IB merger. KZ Index: We follow Lamont, Polk, and Saá-Requejo (2001) to construct the Kaplan- Zingales Index. It is calculated as *[(Compustat Data Item 18 + Item 14) / Item 8] *[(Item 6 + CRSP December Market Equity-Item 60 Item 74) / Item 6] *[(Item 9 + Item 34) / (Item 9 + Item 34 + Item 216)] *[(Item 21 + Item 19) / Item 8] *[Item 1 / Item 8]. Item numbers are COMPUSTAT annual data items and Item 8 is lagged. KZ-rank: Tercile rank of the firm s Kaplan-Zingales Index, with highest index value (most financially constrained) ranked at 3 and lowest index value ranked at 1. Leverage: The ratio between book value of debt and total assets. Book value of debt is the sum of total long-term debt (Compustat Data Item 9) and total debt in current liabilities (34). ln(1+equity Count, Tot5) (ln(1+m&a Count, Tot5)): Natural logarithm of one plus the total number of underwriting services (M&A advising) a client firm received from all IBs during the 5 years prior to the IB merger announcement date. 18

19 ln(1+equity Value, Tot5) and (ln(1+m&a Value, Tot5)): Natural logarithm of one plus the total value of underwriting (M&A advising) transactions (in millions) a client firm received during the 5 years prior to the IB merger announcement date. Maturing Debt: We follow Benmelech, Bergman, and Seru (2011) and construct maturing debt as the amount of long-term debt maturing at year t+1 that was issued at least one year prior to the IB merger announcement date. It is the sum of one-year lagged debt maturing in two years (Compustat Data Item 91), two-year lagged debt maturing in three years (Item 92), three-year lagged debt maturing in four years (Item 93) and four year lagged debt maturing in five years (Item 94). Size (Assets): Natural logarithm of total assets (Compustat Data Item 6) of the client firm one year prior to the IB merger announcement date. Total Assets: Compustat Data Item 6 of the client firm one year prior to the IB merger announcement date. 19

20 Table 1 Summary Statistics The sample consists of 2,549 client firms that have business relationship with 134 IBs involved in M&A events from 1980 to The CAR variables are market adjusted cumulative abnormal returns for the client firms over 3-day and 11-day windows centered at the merger announcement date. Equity (M&A) Count is the number of underwriting (M&A advisory) services provided by the IB to the client firm during the 5 years prior to the IB merger. Equity (M&A) Value is the cumulative dollar value (in millions) of the underwriting (M&A advisory) services provided by the IB to the client firm during the 5 years prior to the IB merger. Size(Assets) is the natural logarithm of the book value of the client firm s total assets. Book-to-Market Ratio is the natural log of the ratio of the book value of equity to the market value of equity. Leverage is the ratio of total debt to total assets. KZ Index is the Kaplan-Zingales Index and Maturing Debt is the amount of long-term debt maturing in the upcoming year. Panel B reports the average CAR for three levels of firm dependence on IB services. A client firm is considered as high dependence if both of its Equity Count and M&A Count are above their respective cross-sectional medians. If a firm has either Equity Count or M&A Count above the median but not both, its dependence is medium. The rest of the firms are considered as low dependence. The p-values are for t-tests of difference in average CAR between dependence groups. Detailed definitions and construction of the variables are provided in the Appendix. Panel A Mean St. Dev. P25 Median P75 N IB-Client Firm Transactions Equity Count ,549 ln(1+ Equity Count) ,549 Equity Value ,549 ln(1+ Equity Value) ,549 M&A Count ,549 ln(1+ M&A Count) ,549 M&A Value , ,549 ln(1+ M&A Value) ,549 Days since Last Deal ,007 2,549 ln(1+days since Last Deal) ,264 Days since First Deal ,294 2,549 ln(1+days since First Deal) ,549 Client Firm CAR(-1,+1) ,390 CAR(-5,+5) ,390 Size(Assets) ,549 Total Assets 1, , , ,549 Book-to-Market Ratio ,529 Leverage ,534 KZ Index ,221 KZ-rank ,221 Maturing Debt ,516 20

21 Panel B Low N = 586 Medium N = 1,664 High N = 140 CAR(-1,+1) p-value p-value CAR(-5,+5) p-value p-value

22 Table 2 Determinants of the Value of IB Relationships The table presents OLS estimation results of analysis of the determinants of 3-day CAR of the client firms around the IB merger announcement date. The sample consists of all active clients of IBs involved in business combinations, i.e. mergers and acquisitions. A client firm is considered an active client if the firm used the IB s services within 5 years prior to the merger and did not switch to another IB for the same type (e.g., underwriting) of service prior to merger. The table reports coefficient estimates followed in square brackets by standard errors clustered at the merger deal level. Detailed definitions and construction of the variables are provided in the Appendix. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. CAR (-1, +1) (1) (2) (3) (4) (5) (6) ln(1+equity Count) ** ** ** [0.004] [0.005] [0.005] ln(1+ M&A Count) * * [0.005] [0.006] [0.006] ln(1+equity Value) ** ** ** [0.001] [0.001] [0.001] ln(1+ M&A Value) * [0.001] [0.001] [0.001] Size(Assets) 0.002* 0.002** * [0.001] [0.001] [0.001] [0.001] Leverage ** ** ** ** [0.011] [0.011] [0.009] [0.009] Book-to-Market Ratio *** *** ** *** [0.002] [0.002] [0.003] [0.003] Constant ** ** [0.005] [0.004] [0.016] [0.017] [0.012] [0.011] Year FE No No Yes Yes No No SIC 2-digit FE No No Yes Yes No No Deal FE No No No No Yes Yes Observations 2,390 2,390 2,375 2,375 2,375 2,375 Adjusted R-squared

23 Table 3 IB Relationships and Financial Constraints The table presents OLS estimation results of the analysis of determinants of 3-day CAR of the client firms around the IB merger announcement date. The sample consists of all active clients of IBs involved in business combinations, i.e. mergers and acquisitions. A client firm is considered an active client if the firm used the IB s services within 5 years prior to the merger and did not switch to another IB for the same type (e.g., underwriting) of service prior to merger. The table reports coefficient estimates followed in square brackets by standard errors clustered at the merger deal level. Detailed definitions and construction of the variables are provided in the Appendix. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. CAR (-1, +1) (1) (2) (3) (4) (5) (6) ln(1+equity Count) ** ** * [0.006] [0.006] [0.006] ln(1+ M&A Count) [0.007] [0.006] [0.007] ln(1+equity Value) ** ** ** [0.001] [0.001] [0.001] ln(1+ M&A Value) [0.001] [0.001] [0.001] KZ-rank * * * ** [0.002] [0.002] [0.002] [0.002] Maturing Debt ** ** ** ** [0.025] [0.025] [0.024] [0.023] Size(Assets) 0.002** 0.003** * 0.002* 0.002* [0.001] [0.001] [0.001] [0.001] [0.001] [0.001] Leverage [0.011] [0.011] [0.01] [0.01] [0.01] [0.01] Book-to-Market Ratio *** *** ** ** ** ** [0.002] [0.002] [0.001] [0.001] [0.001] [0.001] Constant ** ** 0.040*** 0.039*** * [0.015] [0.016] [0.012] [0.012] [0.014] [0.014] Year FE Yes Yes Yes Yes Yes Yes SIC 2-digit FE Yes Yes Yes Yes Yes Yes Observations 2,097 2,097 2,353 2,353 2,076 2,076 Adjusted R-squared

24 Table 4 Investment Bank Reputation and Capacity The table presents OLS estimation results of analysis of the determinants of 3-day CAR of client firms around the IB merger announcement date. The sample consists of all active clients of IBs involved in business combinations, i.e. mergers and acquisitions. A client firm is considered an active client if the firm used the IB s services within 5 years prior to the merger and did not switch to another IB for the same type (e.g., underwriting) of service prior to merger. Diff Reputation is the difference between acquiring IB s underwriter rank and target IB s rank. D(A Eq Ct>T Eq Ct) is a dummy variable that equals 1 if the acquiring IB had underwritten more equity in the past 5 years than the target IB. D(A M&A Ct>T M&A Ct) is a dummy variable that equals 1 if the acquiring IB had advised more M&As in the past 5 years than the target IB. D(A Eq Val>T Eq Val) and D(A M&A Val>T M&A Val) are dummy variables based on the dollar counterparts to the count measures. The table reports coefficient estimates followed in square brackets by standard errors clustered at the merger deal level. Detailed definitions and construction of the variables are provided in the Appendix.*, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. CAR (-1, +1) (1) (2) (3) (4) ln(1+equity Count) ** ** [0.005] [0.006] ln(1+ M&A Count) [0.006] [0.006] ln(1+equity Value) ** ** [0.001] [0.001] ln(1+ M&A Value) [0.001] [0.001] Diff Reputation [0.008] [0.007] D(A Eq Ct>T Eq Ct) 0.023*** [0.007] D(A M&A Ct>T M&A Ct) ** [0.007] D(A Eq Val>T Eq Val) 0.016** [0.008] D(A M&A Val>T M&A Val) [0.009] Size(Assets) 0.002** 0.002** 0.002* 0.002** [0.001] [0.001] [0.001] [0.001] Leverage ** * ** ** [0.011] [0.011] [0.010] [0.011] Book-to-Market Ratio *** *** *** *** [0.002] [0.002] [0.002] [0.002] Constant 0.035** 0.026* [0.016] [0.015] [0.016] [0.017] Year FE Yes Yes Yes Yes SIC 2-digit FE Yes Yes Yes Yes Observations 2,375 2,375 2,375 2,375 Adjusted R-squared

25 Table 5 Robustness Tests The table presents results of various robustness tests. The sample consists of all active clients of IBs involved in business combinations, i.e. mergers and acquisitions. A client firm is considered an active client if the firm used the IB s services within 5 years prior to the merger and did not switch to another IB for the same type (e.g., underwriting) of service prior to merger. The dependent variable in columns (1) and (2) is the 11-day (Day -5 to Day +5) CAR of the client firm around the merger announcement date. The dependent variable in columns (3) to (6) is the 3-day CAR of the client firm around the merger announcement date. The table reports coefficient estimates followed in square brackets by standard errors clustered at the merger deal level. Detailed definitions and construction of the variables are provided in the Appendix. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. CAR(-5,+5) CAR(-1,+1) Pre-2000 Aggregate Activity (1) (2) (3) (4) (5) (6) ln(1+equity Count) * * ** [0.011] [0.008] [0.006] ln(1+m&a Count) [0.011] [0.010] [0.006] ln(1+equity Value) *** * ** [0.002] [0.002] [0.001] ln(1+m&a Value) ** [0.002] [0.002] [0.001] ln(1+equity Count, Tot5) [0.003] ln(1+m&a Count, Tot5) [0.003] ln(1+equity Value, Tot5) [0.001] ln(1+m&a Value, Tot5) [0.001] Size(Assets) ** 0.005** 0.006** * [0.002] [0.002] [0.002] [0.003] [0.001] [0.001] Leverage *** *** ** ** [0.014] [0.014] [0.020] [0.020] [0.011] [0.011] Book-to-Market Ratio * * *** *** *** *** [0.004] [0.004] [0.002] [0.002] [0.002] [0.002] Constant * * [0.026] [0.026] [0.015] [0.017] [0.017] [0.017] Year FE Yes Yes Yes Yes Yes Yes SIC 2-digit FE Yes Yes Yes Yes Yes Yes Observations 2,375 2,375 1,169 1,169 2,375 2,375 Adjusted R-squared

26 Table 6 Alternative CAR Estimations The table presents robustness tests based on alternative definitions of abnormal returns in the 3-day window around the IB merger announcement date. Columns (1) to (4) report results with 3-day CARs adjusted by Fama-French three factor model. Columns (5) to (8) report results with 3-day CARs adjusted using the market model. The sample consists of all active clients of IBs involved in business combinations, i.e. mergers and acquisitions. A client firm is considered an active client if the firm used the IB s services within 5 years prior to the merger and did not switch to another IB for the same type (e.g., underwriting) of service prior to merger. Controls for Size(Assets), Leverage, and Book-to-Market Ratio are included in the models but their coefficients are not tabulated for brevity. The table reports coefficient estimates followed in square brackets by standard errors clustered at the merger deal level. Detailed definitions and construction of the variables are provided in the Appendix. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. Fama-French 3-factor Model Market Model (1) (2) (3) (4) (5) (6) (7) (8) ln(1+equity Count) ** ** ** ** [0.006] [0.005] [0.006] [0.005] ln(1+m&a Count) * * [0.007] [0.007] [0.007] [0.006] ln(1+equity Value) ** ** ** ** [0.001] [0.001] [0.001] [0.001] ln(1+m&a Value) * * [0.001] [0.001] [0.001] [0.001] Controls Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes No No Yes Yes No No SIC 2-digit FE Yes Yes No No Yes Yes No No Deal FE No No Yes Yes No No Yes Yes Observations 2,375 2,375 2,375 2,375 2,375 2,375 2,375 2,375 Adjusted R-squared

27 References Benmelech, Efraim, Nittai Bergman, and Amit Seru, 2011, Financing Labor National Bureau of Economic Research (NBER). Benveniste, Lawrence M., Alexander Ljungqvist, William J. Wilhelm, and Xiaoyun Yu, 2003, Evidence of Information Spillovers in the Production of Investment Banking Services, Journal of Finance 58, Benveniste, Lawrence M., and Paul A. Spindt, 1989, How Investment Bankers Determine the Offer Price and Allocation of New Issues, Journal of Financial Economics 24, Burch, Timothy R., Vikram Nanda, and Vincent Warther, 2005, Does It Pay to Be Loyal? An Empirical Analysis of Underwriting Relationships and Fees, Journal of Financial Economics 77, Cornelli, Francesca, and David Goldreich, 2001, Bookbuilding and Strategic Allocation, Journal of Finance 56, Fama, Eugene F., and Kenneth R. French, 2008, Dissecting Anomalies, Journal of Finance 63, Fernando, Chitru S., Vladimir A. Gatchev, and Paul A. Spindt, 2005, Wanna Dance? How Firms and Underwriters Choose Each Other, Journal of Finance 60, Fernando, Chitru S., Anthony D. May, and William L. Megginson, 2012, The Value of Investment Banking Relationships: Evidence from the Collapse of Lehman brothers, Journal of Finance 67, Hansen, Robert S., and Paul Torregrosa, 1992, Underwriter Compensation and Corporate Monitoring, Journal of Finance 47, James, Christopher, 1992, Relationship-Specific Assets and the Pricing of Underwriter Services, Journal of Finance 47, Kaplan, Steven N, and Luigi Zingales, 1997, Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financing Constraints?, The Quarterly Journal of Economics 112, Kovner, Anna, 2012, Do Underwriters Matter? The Impact of the Near Failure of an Equity Underwriter, Journal of Financial Intermediation 21, Krigman, Laurie, Wayne H Shaw, and Kent L Womack, 2001, Why Do Firms Switch Underwriters?, Journal of Financial Economics 60, Lamont, Owen, Christopher Polk, and Jesús Saá-Requejo, 2001, Financial Constraints and Stock Returns, Review of Financial Studies 14, Ljungqvist, Alexander P, Tim Jenkinson, and Jr. Wilhelm, William J., 2003, Global Integration in Primary Equity Markets : The Role of U.S. Banks and U.S. Investors, Review of Financial Studies 16, Loughran, Tim, and Jay Ritter, 2004, Why Has IPO Underpricing Changed Over Time?, Financial Management 33, Ritter, Jay R., and Ivo Welch, 2002, A Review of IPO Activity, Pricing, and Allocations, Journal of Finance 57, Schenone, Carola, 2004, The Effect of Banking Relationships on the Firm s IPO Underpricing, Journal of Finance 59,