Working for the enemy? The impact of investment banker job changes ondeal flow
Daniel Bradley a,1, Hyung-Suk Choi b,2, Jonathan Clarke c,⁎a University of South Florida, United Statesb Hongik University, South Koreac Georgia Institute of Technology, United States
a r t i c l e i n f o a b s t r a c t
Article history:Received 2 August 2010Received in revised form 11 January 2011Accepted 16 May 2011Available online 24 May 2011
This paper examines the impact of job changes by prominent investment bankers on theM&Aandequitymarket shares of investment banks. Using a hand-collected sample of job changes between1998 and 2006, we find that after controlling for deal and bank-level characteristics, hiring abanker from an investment bankwith a more prominent industry presence has a positive impacton both equity and M&A market share for the gaining bank and a negative impact on the losingbank's M&A market share. After the banker switches firms, we find a significant amount ofbusiness following the banker from the losing bank to the gaining bank, particularly when therelationship is strong between the client firm and the banker. Abnormal returns around theannouncement of a banker changing employers are positive and significant for the gaining bank,suggesting that themarket views banker additions as value increasing. Overall, our results suggesthuman capital is a critical component of investment banking deal flow.
“Merrill Lynch & Co. spent decades building one of Wall Street's premier investment banks. Undoing that work has taken justmonths. Merrill has lost at least 18 veteran investment bankers since the firm agreed to sell itself to Bank of America Corp. In abusiness where human capital is often a firm's most valuable asset, such an exodus can prove fatal.”—Wall Street Journal, July8th, 2009
1. Introduction
There is a wide body of academic evidence suggesting that the investment banking industry is driven by relationships. Anand andGaletovic (2000) suggest that long-term relationships between firms and investment banks are the norm in the United States and inmuchof the developedworld. Yasuda (2005)finds that bank relationships are important in determining afirm's underwriter choice inthe corporate bond market. Fernando et al. (2010) highlight the value of investment banking relationships by showing that equityunderwriting clients of Lehman Brothers experienced a −5% abnormal return around Lehman's bankruptcy.
The extent towhich relationships are embodiedwithin the key personnel of a bank has receivedmuch less attention. Anecdotally,the above quote suggests that the loss of a key banker can have a profound impact on a bank's market share. Similarly, Morrison andWilhelm (2007) suggest that the high degree of labor mobility observed in the investment banking sector has the potential to createsignificant problems for client–bank relationships.
Currently, there is little direct evidence as to whether the addition or loss of a key banker has the potential to create arelationship shock. In the context of the IPO market, Dunbar (2000) finds that for established banks, an increase in the quality ofthe bank's analysts has a positive impact on an investment bank's share of new issues. More directly, Clarke et al. (2007)investigate the impact of Institutional Investor all-star analyst job switches on market share and investment banking deal flow.They find that the bank gaining the all-star analyst increases its market share for equity transactions, but find no such evidence forM&A or bond transactions. Finally, Ljungqvist et al. (2006) control for the arrival and departure of investment bankers in theiranalysis of investment bankingmandates. They findmixed evidence as to whether themovement of key bankers has an impact onthe probability of winning a mandate.
In this paper, we focus on the relationship shock caused when a prominent banker defects from his or her current employer towork for a rival bank. We compile a hand-collected dataset of 339 investment banker job changes over the August 1998 toDecember 2006 period and adopt an approach similar to Clarke et al. (2007). We examinemarket share changes to the investmentbanks affected by the switch. We then examine if client firms switch their business to the banker's new employer. Finally, weexamine the market reaction to news of the switch to determine if these events are economically meaningful.
We document three main empirical findings. First, we find that, on average, the bank gaining the banker experiences aninsignificant increase in its industry-level market share. The losing bank however, experiences an economically and statisticallysignificant decrease in its industry-level market share of approximately −0.63%. This loss is driven primarily by a decline in M&Abusiness. In cross-sectional regressions, we find evidence that the ability to hire an investment banker from a rival bank that has agreater industry presence than the hiring firm is positively related to the gaining bank's market share and negatively related to thelosing bank's market share. For instance, depending on the model specification and type of transaction (equity or M&A), hiring aprominent rival increases industry market share anywhere from 1% to 2%. It corresponds to an approximately 3% decline in theindustry M&A market share for the bank losing the investment banker. This effect is distinct from the overall reputation of theinvestment bank.
Second, and perhaps more importantly, we find evidence that a significant amount of business follows the investment bankerfrom their old bank to their new bank. This result holds for both equity and M&A transactions. Further, we examine the likelihoodof a firm following the banker to her new employer and find that it is positively related to the strength of their past relationships.
Finally, we find significant wealth effects upon the announcement of banker job switches. The difference in market-adjustedreturns between the gaining bank and the losing bank is greater than 1%, which is both statistically and economically meaningful,indicating themarket reacts tonewsof investmentbanker job changes. Collectively, ourevidence indicates thathumancapital is a vitalcomponent of an investment banks' ability to generate business.
These results complement and expand on those presented in Ljungqvist et al. (2006) in several important dimensions. First,unlike Ljungqvist et al. (2006), we examine the role of investment bankers in M&A deals, as well as, equity transactions. This isimportant because relationships between investment bankers and client firms are critical in the M&A market because acquirerstypically engage in a large number of deals. For instance, Netter et al. (2010) examine all CRSP-listed firms from 1992 to 2009 andfind that 91% are involved in at least one type of acquisitionwith the average firm engaged in 16 acquisitions over this time period.Thus, it is not surprising that most of our results are found for M&A deals. Second, we incorporate banker specific variables likeexperience into our analysis. Third, we examine the industry market share impact to investment banks that lose or gain bankers.Fourth, we track individual deal flow from the gaining to losing banks and vice versa. Finally, we examine the stock price reactionto the departure and arrival of key bankers for our sample of publicly traded investment banks, which is an exogenous way toexamine the market's perception of the switch.
The rest of this paper proceeds as follows. Section 2 provides the motivation of this paper and a discussion of the relatedliterature. Section 3 explains the data and empirical methods, while Section 4 reports our empirical results. Finally, Section 5concludes.
2. Motivation
There is substantial evidence indicating that investment banking is characterized by on-going relationships rather than arms-length transactions. For example, Corwin and Schultz (2005) note that approximately 70% of firms stay with their initialunderwriters from their IPO to their subsequent equity offers. Additionally, Burch et al. (2005) examine underwriting fees forrepeat issuers of new securities to determine the relation between loyalty to an underwriting bank and the fees charged. For equityofferings, they find that loyalty is associated with lower fees for common stock offers, which is consistent with valuablerelationship capital being built.
While relationships are assumed to be important for investment banking, there is little direct evidence examining the extent towhich individual bankers are the gate keepers to clients. To the best of our knowledge, Ljungqvist et al. (2006) is the only paper toaddress this important issue. They note that the high frequency with which investment bankers switch employers has thepotential to create relationship shocks. While the focus of their paper is on the relation between aggressive analystrecommendations and the likelihood of winning investment banking business, they do control for investment banker jobchanges in their models. Their empirical evidence is mixed. In the absence of analyst coverage, they find gaining bankers increasedand losing bankers decreased the chances of winning an equity mandate. However, in the presence of analyst coverage,movements of key bankers had little effect on the likelihood of winning an equity mandate. For debt mandates, movements of keybankers had a substantial impact. The chances of a bank winning a mandate were lower when it had lost key members of its debtteam and higher when it had recently hired debt professionals from other banks.
586 D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
Like Ljungqvist et al. (2006), we hypothesize that the human capital of bankers plays a critical role in generating deal flow.Thus, when bankers defect to a competing firm, the banks gaining (losing) such prominent bankers should see an increase(decline) in industry market share as their ability to attract and retain clients is strengthened (diminished). Any such changesshould be related to the quality of the banker. If job switches are meaningful, the market should respond favorably upon theannouncement of a hire and negatively to the departure for the gaining and losing banks, respectively. Finally, if relationships areimportant, we should observe clients following their banker to their new firm.
3. Data and methods
3.1. Sample description
We hand-collect our sample of investment bankers switching firms from Investment Dealer's Digest (IDD). IDD is published on aweekly basis and has been used as a data source in a number of academic studies including Carhart (1997). The sample is compiledfrom the “Out Takes” section of the magazine, which discusses movement of key personnel within the investment community.From this section, we are able to compile a list of investment bankers who moved from one investment bank to another over theperiod August 1998 to December 2006. Our sample begins in August of 1998, because this is when the data first became availableonline. In addition to the name of the banker and the investment banks switched to and from, the magazine also identifies thebanker's title (i.e., Global Head, Managing Director, etc.). We begin with 617 observations. We restrict our analysis to only thosecases where we can identify an industry of specialization for the banker and clearly identify that the banker is engaged in M&A orequity banking. Our search yields 339 cases of bankers switching firms. For each banker, we also searched theWho'sWho directory,performed Google searches, and collected biographical information provided on IDD to obtain information on the banker'sexperience. We find information on experience for approximately 62% of the sample.3
Panel A of Table 1 shows the frequency of banker turnover by year. The highest incidence of banker turnover occurs in 2000,when there were 56 cases, or about 16.5% of the sample. Panel B reports the titles of the bankers that switched. Themajority of thereported cases occurred at the Managing Director level or higher. Only 74 of our 339 observations were at the director level orlower.4,5
While Clarke et al. (2007) classify analysts switching employers into one of 59 Global Industry Classification Standard (GICS)industries it appears that investment bankers specialize in broader industries. Therefore, we classify bankers into one of 10 GICSsectors.6 Panel C shows that about 75 percent of the observations are classified into one of the following industries: InformationTechnology, Financials, Consumer Discretionary, or Health Care sectors.
In unreported results, we find that approximately 37% of the switching bankers moved to investment banks with a similarindustry reputation, while 46% of the bankers moved to investment banks with a lesser industry presence. The remaining movedto a higher rated bank.
3.2. Market share calculations
The key variable of interest throughout this paper is industry-level market share.7 For each case of banker turnover in oursample, we separately calculate market share at the industry level for equity deals, M&A deals, and a combined equity and M&Amarket share. Ideally, we would like to calculate market share using only those deals associated with an individual banker.However, this data isn't available and we are forced to examine industry level market share. We describe the procedure forcalculating this value below.
For equity deals, we calculate industry market share using gross proceeds as in Dunbar (2000).8 Specifically, we compile acomprehensive database of equity deals between 1996 and 2008 from Thompson Financial Securities New Issues databases.9 Equityissues include 6460 SEOs and 5035 IPOs. From the new issues database, we obtain the following information for every initial publicoffering and seasoned equity offering: the issuer name and CUSIP; the filing, and the issue or withdrawn date; the identity of the
3 We also collected individual information on each banker including undergraduate and graduate institutions attended, whether or not they hold the CFAdesignation, and the number of prior employers they had. This information is available for a much smaller portion of the sample. In unreported regression results,these variables are not significantly related to market share changes.
4 Our results are qualitatively similar if we consider only managing directors or higher.5 We acknowledge that IDD does not contain the full population of banker switches and therefore sample selection bias may be a concern. For completeness,
we also did an exhaustive search on Lexis Nexus for banker defections using various search parameters. This did yield several more banker changes that were notreported in IDD. However, the information contained for each announcement was not uniform and in most cases significantly lacking, particularly for theindustry classification. Our analysis suggests that IDD reports on key personnel changes, which also explains the lack of observations at the director level orlower. Thus, our sample likely contains bankers that have the most important impact on the bank's deal flow.
6 We use GICS codes in preference to other classification schemes used in the literature (such as Standardized Industry Classification System (SIC) codes, NorthAmerican Industry Classification System (NAICS) codes or the Fama and French (1997) industry groupings (FF)) since Bhojraj et al. (2003) show that GICSclassifications are significantly better at explaining stock return co-movements, as well as cross-sectional variations in valuation-multiples, forecasted growthrates, and key financial ratios.
7 Eccles and Crane (1988) argue that market share is highly correlated with bank profit.8 Calculating market share based on number of deals yields qualitatively similar conclusions.9 We focus on equity and M&A advising, because these deals generate the highest fees. Nonetheless, we examine the impact of banker job changes on bond
underwriting market share. We find no evidence that banker switches influence the bond market share of either the gaining or losing bank. To conserve space,we do not report these results.
587D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
investment bank retained by the issuer; the size of the deal; and the fee structure. We use the above databases to calculate industrymarket share for the bank the banker is switching from (the losing bank) and the bank the banker is switching to (the gaining bank).Industrymarket share is calculated as the gross proceeds raised by an investmentbank in the banker's GICS sector dividedby total grossproceeds of all deals completed in that particular sector.We calculatemarket share using twoyears of data prior to and after the switch.
For those bankers that are active in multiple sectors, we add up gross proceeds across each sector to compute market shareacross the multiple industries. The average banker in our sample operates in 1.1 sectors. In the case of multiple advisors, we crediteach advisor with a 1/N portion of the deal. Our sample is characterized by a number of large mergers between investment banks.We follow Ljungqvist et al. (2006) to adjust for these mergers.
We follow a similar process to calculate M&Amarket share. We focus on acquisitions where the deal value and advisor(s) weredisclosed in Thomson Financial's M&A database. In all, we consider 24,992 transactions counting acquirers and targets separately.For these observations, we obtain information on the identity of the target and acquirer, the announcement and effective orwithdrawn dates of the transaction, the size of the deal, and the advisors to both sides of the transaction. We then calculateindustry market share using deal value.
3.3. Measuring a bank's overall reputation
As a control variable in various reported results, we consider the investment bank's overall reputation. In order to proxy for theinvestment bank's overall reputation, we use the bank's Carter–Manaster ranking as updated by Loughran and Ritter (2004). Theranking ranges from 1 to 9, with 9 representing the highest possible prestige ranking.
In this section, we examine the impact of a banker job change on the gaining and losing bank's market share. We investigatemarket share changes for equity underwriting deals (initial public offerings and seasoned equity offerings) and merger andacquisition transactions in the banker's primary industry of specialization.
Table 1Sample descriptive statistics. This table presents various descriptive statistics for our sample of 339 investment bankers who switched investment banks betweenAugust 1998 and December 2006. The sample is compiled from the weekly “Out Takes” section published in Investment Dealers Digest. Industry classifications arebased on the GICS Sector Classifications.
588 D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
Although our sample has 339 cases of banker job changes, several of these cases involve teams of bankers switching. Forpurposes of examining market share changes, we condense these cases into a single observation. For example, three bankers (aManaging Director and two senior VPs) covering the Financial Sector departed Goldman Sachs for Lehman Brothers on March 1st,1999. Rather than treating this as three independent observations, we compress it into a single observation.10 In some cases, abank hires a number of bankers from different banks on the same date. For example, Goldman Sachs hired two VPs covering theHealth Care Sector from Lehman Brothers and Morgan Stanley on August 4th, 2003. We condense observations like these into asingle observation for the gaining bank, but treat it as two distinct observations for the losing bank. After making this adjustmentwe are left with 245 gaining bank observations and 261 losing bank observations.
We report our initial findings on industry level market share changes surrounding the job switch for both the gaining andlosing banks in Table 2. We separately consider stock offerings and M&A deals in the first several columns, and combine them inthe last two columns under the heading investment banking. Panel A of Table 2 reports the results for all investment bankers in oursample. For stock offerings, the gaining bank increases its industry market share by an insignificant 29 basis points, while thelosing bank loses an insignificant 15 basis points. For M&A, the gaining bank increases its industry market share by an insignificant20 basis points while the losing bank loses a statistically significant 63 basis points. The combined investment banking results(equity and M&A) are not surprisingly similar to the M&A results since M&A deals are more frequent and typically much largerthan equity deals.11
In Panel B of Table 2, we report market share for only those cases in which the bank losing the banker had a positive industrymarket share in the two years prior to the switch. Our intent with this filter is to focus on the subset of banks that have a trackrecord of generating investment banking deal flow over the past two years. This restriction eliminates banks that don't generateunderwriting business on a consistent basis. We term this subset of observations the “restricted” sample. The findings arequalitatively similar to those presented in the full sample.
The results in Table 2 indicate that the losing bank experiences a significant decline in M&A market share within the banker'sindustry of specialization following the switch, while the gaining bank does not experience a statistically significant increase. Let'sconsider the economic impact of such changes. In a typical year, the average bank does about $10 billion of gross proceeds inindustry-level equity business. A 0.29 percent increase equates to approximately $29 million in incremental proceeds. If weassume that the average spread is 5%, this suggests that the gaining bank increases its industry-level fee revenue by $1.45 million.The losing bank would experience a reduction of about half that in fee revenue, but again, neither is statistically significant. ForM&A, the typical bank does $135 billion in deal value. This implies the gaining bank increases its market share by about$270 million, while the losing bank sees a loss of $0.8 billion in deal value. Assuming a fee of 0.9% of deal value (Walter et al.(2005)), the gaining bank increases its banking revenue by about $2.4 million whereas the losing bank loses about $7.7 million infee revenue.
Are the numbers above realistic? Consider Morgan Stanley who employed approximately 300 investment bankers at theManaging Director level in 2006. According to Morgan Stanley's annual report, they did $4.76 billion in investment bankingrevenue in 2006. If we assume that Managing Directors were solely responsible for this revenue, this implies average revenue ofapproximately $16 million per Managing Director. Of course, Heads and higher (which are significantly less numerous than
10 One could argue that team switches should not be counted as individual events. If we treat teammovements as individual moves, our results are qualitativelyunchanged. As another test we simply control for team movements using an indicator variable in our regressions if the observation involved a team switch. In allmodels, this variable is not statistically significant.11 We also examine one-year market share changes and find similar results.
Table 2Industry level market share surrounding the switch. This table shows the change in average bank industry market share from the two years prior to the switch ofthe investment banker (Pre) to two years after the switch (Post). Panel A presents results for the full sample and Panel B presents “Restricted” results where thebank losing the banker had a positive market share in the two years preceding the switch. Market share is calculated using deal proceeds for IPO, SEO and M&Atransactions. In the case of multiple advisors on a particular deal, we give each advisor a 1/n share of the deal, where n is the number of advisors.
Marketshare
Equity offering M&A Investment banking
Gaining bank Losing bank Gaining bank Losing bank Gaining bank Losing bank
589D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
Managing Directors) and lower-level investment bankers are not included in this estimation, but it provides an upper bound as totheir impact on banking revenue. Given that bonuses forManaging Directors at large banks are typically in themulti-million dollarrange, this indeed seems reasonable.
To augment the above back-of-the-envelope calculations, we obtain detailed data on underwriter spreads from ThomsonFinancial's SDC database. Unfortunately, detailed data on merger and acquisition fees is sparsely reported in the SDC database.Nonetheless, we use the available date and find that the firm hiring the prominent banker experiences an increase in equityunderwriting fees of $460,000, while the bank losing the banker experiences a decline in equity underwriting fees of $1.39 million.These changes are not statistically significant.
In an effort to back out an estimate of the fees associatedwithM&A transactions, we use the “Double Lehman” scale. Variants ofthis scale have been used since the 1970s. Specifically, we assume that fees for M&A transactions are related to transaction value inthe following manner: 10% up to $1 million; 8% of the second million; 6% of the third million; 4% of the fourth million; and 2% ofanything higher than $4 million. We find that the loss of a prominent banker results in approximately $24 million less in advisoryfees over the two years following the job change. This result is consistent with the market share data and suggests that prominentbankers can have a significant impact on the bottom line.
The results in this subsection suggest that the loss of a key investment banker has a particularly large negative impact on theM&A market share of the losing bank, but not much of an impact for equity deals. It's important to note that these results do nottake into account cross-sectional variation in banker characteristics nor prior deal characteristics. In the next section, we examinethis issue in greater detail.
4.2. Determinants of market share changes
In this section, we examine banker and bank-related characteristics that may influence changes in market share at theindustry-level. We first define four banker-related variables. We capture whether the banker moves from a more reputable bank(Higher Bank Reputation) based on the Carter and Manaster (1990) rankings updated by Loughran and Ritter (2004). This is adichotomous variable that takes on a value of one if the banker moves to a more reputable bank, and zero otherwise. Though weinclude it as a control, the effect is ambiguous. Less prestigious banks may have more difficulty in generating deal flow. Forinstance, Fernando et al. (2005) argue that high quality issuers naturally pair with high quality underwriters. Thus, holding all elseconstant, a banker may simply be more successful at a more reputable firm. On the other hand, hiring a banker from a morereputable firm might allow a smaller firm to significantly increase its market share by tapping into the banker's establishednetwork.
We include a variable to measure whether the banker comes from a bank with a more established industry market share,Higher Industry Reputation. This variable takes the value of one if the bank losing the banker has a higher industry market sharethan the gaining bank, and zero otherwise. As mentioned earlier, some banks have niche investment banking departments. Inthese cases, the bank's reputation within a particular industry may differ significantly from that of the bank as a whole. We defineManaging Director to take the value of one if the switching banker is a Managing Director or higher. The assumption is that moresenior bankers should have a larger impact on market share changes. We also control for differences in the number of analystswithin the banker's industry at the gaining and losing banks.Difference in # All-stars is the number of all-star analysts at the gainingbank in the banker's industry minus the corresponding number for the losing bank. Finally, we include the banker's experience,which is measured in years. Like bankers with the Managing Director or higher title, more experience should be positively relatedto market share changes.
We next consider firm and deal-level characteristics. Trend takes the value of one if the percentage market share at the newbank increases relative to that at the original bank between year−2 and year−1, and zero otherwise. This is to control for banksthat are either trying to beef up or shave off their banking business.12 Dunbar (2000) examines market share changes in the U.S.IPOmarket and finds that abnormal underpricing, stock price performance, fees, and percentage of withdrawn deals are related tomarket share changes for reputable banks. Therefore, we include measures for both IPOs and SEOs in our regressionmodels. Thesevariables are estimated at the industry-level following the approach in Dunbar (2000).13 Abnormal IPO fees and Abnormal SEO feesgauge the relative pricing structure of banks. Banks that offer lower fees may see an increase in their business. Abnormal IPO 1-yearreturn and Abnormal SEO 1-year return measure the 1-year performance of each type of security offering by the underwriter.Investment banks that have positive abnormal performancemight see an increase inmarket share if their deals performwell post-issuance. Abnormal IPO First day return and Abnormal SEO First day return are relative measures of how much money is left on thetable. Leaving toomuchmoney on the table may cause banks to losemarket share. Finally,Withdrawn IPO andWithdrawn SEO takethe value of one if the bank is to withdraw either an IPO or an SEO over the two years prior to the job change and zero otherwise.Dunbar (2000) finds that investment banks experiencing a high percentage of withdrawn deals lose market share.
12 In unreported results, we also include two separate indicator variables for whether the gaining bank or losing bank experienced a merger surrounding thebanker switch. Neither of the coefficients is significant.13 To estimate abnormal spreads for IPOs, for example, we pool all IPOs that occurred in the banker's industry in the two years prior to the switch. For each IPO,we regress the spread on gross proceeds and the logarithm of gross proceeds, using IPOs from three years before the offer through the year of the offer. Theresidual from this regression is taken to be a proxy for the abnormal spread. We average the abnormal spread for each IPO where the banker's firm acted as theunderwriter. We follow a similar approach to estimate the abnormal spread for SEOs.
590 D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
Similar to equity transactions, a number of deal characteristics for mergers and acquisitions are also considered. Tomeasure theannouncement effects of M&As, abnormal announcement day returns are calculated. Abnormal Target Announcement Day Return isthe actual minus predicted announcement day return. Predicted values are estimated using a regression of Percentage TargetAnnouncement-day return on Gross deal values and the Logarithm of gross deal values, using target deals advised by the bank fromfour years before the deal through the year of the deal. Abnormal Acquirer Announcement Day Return is the actual minus predictedannouncement day return for the acquirer.We follow a similar process to estimate the predicted announcement day return for theacquirer.Withdrawn Target takes the value of one if a bank acted as the advisor to a target in any unsuccessful transaction prior tothe job change. Withdrawn Acquirer is similarly defined.
Panel A of Table 3 examines changes in market share for equity deals. For both the gaining and losing banks, we only reportregression results for the restricted case. Results are qualitatively similar for the unrestricted case. For each regression specificationpresented in Table 3, we include year fixed effects. The first specification includes each of the deal and banker characteristics,except for banker experience. Since inclusion of our experience variable results in a non-trivial reduction in sample size due tomissing observations, we estimate a separate regression with this variable.
In the first model, which examines market share changes for the gaining bank, we find that Higher Industry Reputation is highlysignificant, whereas the rest of the banker-related variables are not. Trend is insignificant, suggesting that bankers are not simplyfollowing momentum. That is, bankers are not switching to better performing banks. None of the other control variables issignificant.
Table 3Regression analysis of industrymarket share changes. This table presents results from OLS regressions where the dependent variable is the change in industrymarketshare from the two years prior to the switch to two years after the switch. Panel A presents equity results and Panel B presents M&A results.Wemeasure the change inmarket share at the industry-level for both the bank gaining the banker and the bank losing the banker. Industry classifications are based on 9 GICS sectors.Higher BankReputation takes the value of one if the bank losing the banker had a higher Carter–Manaster ranking than the bank gaining the banker.Higher Industry Reputation takesthe value of one if the bank losing the banker had a higher industrymarket share prior to the switch than the bank gaining the banker.Managing Director is an indicatorvariable that takes the value of one if the bankerwas aManaging Director or higher. Difference in the # All-stars is equal to the number of all-star analysts at the gainingbank in the banker's industryminus the corresponding number for the losing bank. Experiencemeasures the banker's experience in years. Trend takes the value of one ifthemarket share at the new bank increases relative to that at the original bank between year−2 and year−1, and zero otherwise.Measures of abnormal performanceand fees are estimated as in Dunbar (2000). See the text for a more detailed description.Withdrawn IPO (SEO) takes the value of one if the bankwas associatedwith awithdrawn IPO (SEO) and zero otherwise. We include year indicator variables in each regression. P-values are reported in parentheses.
591D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
In the second specification, the experience indicator variable is insignificant indicating that hiring a more experienced bankerdoes not increase market share for the gaining bank. The remaining variables are essentially unchanged. In particular, HigherIndustry Reputation remains highly significant. Holding all else constant, the coefficient suggests that hiring a banker from aninvestment bank with a stronger industry presence leads to a 1.3 (model 1) to 1.8 (model 2) percent increase in market share.
For the losing bank in the last two columns, each of the banker-related variables is insignificant. This suggests that losing aprominent banker does not have a significant impact on the equity market share of the losing bank. Abnormal SEO 1-yr return isnegative and significant suggesting that the losing bank losesmarket share when SEO deals from the previous year perform poorly.
In Panel B, we examine M&A deals. We introduce M&A-related controls previously discussed into the regressions. HigherIndustry Reputation is highly significant with a coefficient of 1.1 (model 1) to 2.0 (model 2) implying an increase in market share ofbetween 1.1 and 2% to the bank that hires a banker from a rival investment bankwith amore prominent industry IB department. Inthe secondmodel, whileHigh Experience has the correct sign, the p-value of .11 is not statistically significant at conventional levels.
For the bank losing the banker, Higher Industry Reputation is negative and significant in both models. The coefficient rangesfrom −2.6% to −2.8% suggesting that, ceteris paribus, losing a banker to a weaker industry IB department ultimately negativelyimpacts the losing bank's market share by approximately 3% based on our estimates.
Overall, the results from Table 3 suggest that the most important determinant of market share gains and losses is the ability tohire an investment banker from a rival in a more industry-dominant position. In the next subsection, we determine if businessfollows the banker from the old firm to the new firm.
4.3. Does business switch with the banker?
Thus far, we have demonstrated significant market share losses at the industry-level for the losing bank, but hiring a moreinfluential industry-level banker is strongly related to market share changes for both gaining and losing banks. In this section, weexamine deal flow to isolate the source of themarket share changes.We assume that business is generated in one of threeways. First,business can come from existing clients. Such repeat business is most likely due to preexisting relationships the firm had before thebanker made the switch. Second, business can come from entirely new clients. A third category, which is a subset of new clients, isbusiness that is stolen away from the banker's old firm.
For this analysis,wedefine anexisting client to be one that did a dealwith thebankwithin the banker's industry of specialization inthe two years prior to the switch. If a firmwithin the banker's industry of specialization did not have a transaction two years before aninvestment banker switched, but did participate in a deal two years afterwards, we assume that the bank receiving the mandate is anew client of the bank.
In order to test if somebusiness follows the investment banker fromonefirm to the other, we need an appropriate benchmark. Forexample, suppose a banker left Credit Suisse forGoldmanSachs. Cliff andDenis (2004) andothers suggest that about 35%of IPO issuersswitch underwriters from their IPO to SEO. Assuming that this percentage is in the correct ballpark for our sample, then a substantialnumberof clientswill leaveCredit Suisse. If a switchoccurs at random,GoldmanSachswill likely receive someof Credit Suisse's formerclients due to chance. Thus, it would not be appropriate to use 0 percent market share as a benchmark. Instead, in this example, wecompare the percentage of business Goldman Sachs receives from Credit Suisse's former clients and vice versa (the percentage ofbusiness Credit Suisse receives from Goldman Sachs).
Although we present all three categories of client types, the one that is of particular interest is the last row, Rival. This is thepercentage of business that switched from the gaining to losing bank and vice versa. For equity deals, 4.7% of the gaining bank's equitybusiness switched from the losing bank. In comparison, 0.5% of the losing bank's equity business switched from the gaining bank. Thisdifference of 4.2% is economically and statistically significant.
Likewise, for M&A transactions, 3.0% of the gaining bank's M&A business switched from the losing bank. This compares to 1.4% ofthe losing bank's M&A business that switched from the gaining bank. This difference of 1.6% is also economically and statisticallymeaningful. The total investment banking results are similar.14
4.4. Model of switching
The results in Table 4 indicate that bankers bring deal flow from their old firms. In this section, we expand on the literatureinvestigating the determinants of the decision by client firms to switch investment banks by introducing banker-related attributes. Inorder to be considered in this analysis, thefirmmust have completed either an equity orM&Adeal in the two years before and the twoyears after the banker switch. Similar toKrigmanet al. (2001) andCliff andDenis (2004), thedependent variable takes on a value of 1 ifthe firm switches underwriters. We run separate logistic regressions for equity and M&A deals. We use the same set of explanatoryvariables as in Table 4, with two exceptions. First, we drop the high experience variable, since inclusion of this variable diminishes thesample size. Additionally, we add a variable that captures the number of deals the client firm completedwith the investment bank inthe two years prior to the switch. This variable is designed to capture the strength of the relation between a banker and the client firm.We would expect that the greater the number of deals completed, the stronger the client relationship and the more likely the clientfirm is to switch. We again include year fixed effects.
14 Apointworthmentioning is thatmany investmentbankers are forced to signnon-compete andnon-solicit agreements,which typically are inplace up to18 monthsonce they leave theirfirm. However, there is debate about the enforceability of these contracts. In all of our analyses, the post-transaction period lasts for two years andthus the banker may not legally be allowed to solicit former clients during most of our post-switch period. This has the effect of biasing against our findings.
592 D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
The results are presented in Table 5. For equity deals, we find strong evidence that the number of prior banking deals is asignificant determinant of the decision to switch. The coefficient is positive and significant indicating that the greater the strengthof the relation between the banker and the client firm, the greater the likelihood the client firmwill switch banks. No other banker-related variables are statistically significant. For example, hiring a banker from a bank with a greater industry reputation is notsignificantly associated with the decision to switch. Among deal characteristics, we find a negative and significant coefficient onthe Difference in Fraction Withdrawn SEOs. No other deal characteristics are significant. These results suggest that the strength ofthe relationship between a banker and a client firm is a key determinant in the decision to switch underwriting business.
ForM&Adeals,wefindsimilar results. The coefficient onNumber of PastDeals is positive and significant, suggesting that the greaterthe strength of the relation between the banker and the client firm, the greater the likelihood the clientfirmwill switch.We alsofind apositive relation betweenManagingDirector and the decision to switchM&Aadvisors. This suggests that if a banker ismore senior, thefirm is more likely to switch business to the banker's new employer. None of the deal characteristics is statistically significant.
4.5. Event study results
If market participants view investment banker job changes as important events, the returns surrounding the announcement ofthe change should be positive for banks hiring a banker and negative for the bank losing the banker. Our market analysis is limitedto investment banks in our sample that are publicly traded, reducing our sample size to 183 observations for the bank losing thebanker and 139 observations for the bank gaining the banker.
Table 5Likelihood of switching business. This table examines the likelihood of switching underwriters. The sample includes only those firms that did a deal at the old bankin the two years prior to the banker switching and returned to themarket for a deal in the two-year post turnover period. The dependent variable takes the value ofone if the firm switched to the new banker's firm and zero otherwise. Managing Director is an indicator variable that takes the value of one if the banker was aManaging Director or higher. Number of Past Deals is the number of deals completed by the firm at the bank losing the key banker during the past two years.Difference in the # All-stars is equal to the number of all-star analysts at the gaining bank in the banker's industry minus the corresponding number for the losingbank. Higher Bank Reputation takes the value of one if the bank losing the banker had a higher Carter–Manaster ranking than the bank gaining the banker. HigherIndustry Reputation takes the value of one if the bank losing the banker had a higher industry market share prior to the switch than the bank gaining the banker.Measures of abnormal performance, fees, and withdrawn deals are estimated as in Dunbar (2000). See the text for a more detailed description. For each of theseabnormal measures, we take the difference between the gaining and losing banks. P-values are reported in parentheses.
Higher Bank Reputation −0.29 (0.48) −1.30 (0.01)Higher Industry Reputation 0.90 (0.18) −0.04 (0.93)Managing Director 1.04 (0.14) 0.98 (0.09)Number of Past Deals 0.77 (0.00) 0.93 (0.00)Difference in the # of All-Stars 0.02 (0.35) −0.02 (0.18)Trend 0.06 (0.89) −0.66 (0.09)
Equity characteristicsDifference in Abnormal IPO Fees −1.27 (0.17)Difference in Abnormal SEO Fees −0.47 (0.22)Difference in Abnormal IPO 1-yr Return 0.13 (0.80)Difference in Abnormal SEO 1-yr Return 0.52 (0.54)Difference in Abnormal IPO First day return 0.01 (0.19)Difference in Abnormal SEO First day return −0.00 (0.87)Difference in Fraction Withdrawn IPOs −0.54 (0.48)Difference in Fraction Withdrawn SEOs −3.83 (0.05)
M&A characteristicsDifference in Abnormal Target Announcement Day Returns 0.01 (0.42)Difference in Abnormal Acquirer Announcement day returns 0.04 (0.47)Difference in Fraction Withdrawn Target 1.37 (0.12)Difference in Fraction Withdrawn Acquirer 1.40 (0.15)
# of observations 1399 2102PrNChiSq 0.00 0.00
Table 4Where does business come from after the switch? This table examines the gross proceeds for over 35,000 equity and M&A transactions in the two years followingthe banker job change. For each bank, we calculate the percentage of business coming from new clients (New), existing clients (Existing), and a subset of newclients that are clients who switched to the gaining and losing banks (Rival).
593D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
We calculate market-adjusted returns using the date the job change was reported in Investment Dealer's Digest as the eventdate.15 Since IDD is published on a weekly basis, the announcement date could be off by several days. Therefore, we perform aLexis-Nexis search for each of the bankers in our sample in an effort to find a more accurate departure date. We are able to matchapproximately 1/3 of our sample using this additional source. In the reported results, we use the earlier of the Lexis-Nexis date orthe IDD publication date as the event date. We calculate market adjusted returns using the value-weighted index.
In Panel A of Table 6, we examine all banks in our sample that have data available. We require that both the gaining and thelosing banks have stock price data available. Thus, our analysis is based on a paired t-test. Both the gaining bank and losing bankexperience significantly positive returns several months before the switch. The (−180, −31) and (−30, −6) returns for thegaining bank are 8.73% and 2.14%, respectively, and the corresponding returns for the losing bank are 7.85% and 1.70%. This findingis consistent with Fee and Hadlock (2003) who find that CEOs switch firms after experiencing positive performance. However, therelative performance of the two banks is not significantly different in the months preceding the switch.
The announcement period return is a significantly positive 0.98% for the gaining bank and insignificantly negative 0.37% for thelosing bank. Thepaireddifference between the twobanks is 1.35%,which is significant at the 6 percent level. Thus, themarket perceivesthe hiring of an investment banker as wealth increasing. This is contradictory to the negative wealth effect found by Groysberg et al.(2008) for banks gaining all-star analysts. In contrast, our results suggest that hiring a prominent banker is not a value destroyingactivity. Finally, there are no significant abnormal returns in the period following the switch for either the gaining or losing banks.
In Panel B of Table 6, we report a regression analysis of the announcement period return differential. We find no evidence thatthe differential is related to whether a team of analyst switches or if it is an individual banker. Nor do we find evidence that theeffect is related to bank reputation. While this result may seem at odds with earlier findings, the focus on publicly traded banks inthis analysis skews our analysis to larger, more reputable banks. Thus, there is less variation in this variable. The departure of thefirst banker in a given year generates a much larger price reaction than subsequent departures. Thus, the market reaction isnonlinear and appears to suggest that market participants anticipate the loss of additional bankers.
4.6. Robustness checks
Our interpretation of the results to this point has been that investment banker job changes cause changes in market share. Thesechanges may not only be the result of the human capital of the banker, but related to externalities associated with the switch such asboosting (lowering) morale and productivity of fellow colleagues or signaling an increase (decline) in perceived quality to prospectiveclients for thegaining (losing)banks.Of course, these externalities arenotmeasurable, but are at least indirectly tied to thebanker switch.
An alternative explanation for our results is that prominent bankers are adept at timing the market and switch from banks thatare on a downward trajectory to those that are on an upward trajectory. Oyer (2008) finds that market conditions stronglyinfluence the career paths of newly minted MBAs on Wall Street. MBA students are likely to time the market and go directly intoinvestment banking during a bull market when there is high demand for bankers. If seasoned bankers time the market also, thiscould possibly give the appearance of a causal relation between banker turnover and market share when none is present. On theother hand, Fee and Hadlock (2003) examine executives who move to the CEO position at another employer. They find that theseexecutives experience above average stock price performance at their previous employer prior to the switch. In other words, theyare not merely jumping to a better performing company.
First, we indeed find that the prior stock price performance was significantly positive at the mover's investment bank severalmonths before the switch, on average. However, both the gaining and losing banks had positive performance before the switch andthe difference between the two is not significant. Thus, our results partially agree with Fee and Hadlock (2003) that bankers movewhen market conditions are favorable. However, our results do not indicate that they are moving to a better performing bank.
If switches are voluntary during bull markets, but involuntary during bear markets, we might expect differential effects onmarket share during these periods. Thus, we split our sample into two periods: 1998–2000 and 2001–2006, which coincided witha bull market followed by a bear market. Our results are similar regardless of the period examined.
Finally, it's worth noting that we included the control variable Trend in our regression models, which measures the relativemarket share changes in the year preceding the switch. If bankers were moving to better performing firms, this variable should besignificant. We find it is not. Thus, we believe these findings should mitigate concerns about reverse causality.
It is likely that our events overlap and thus we violate the independence assumption in our statistical analysis. Therefore, weexamine a “clean” sample of non-overlapping events in the two-year prior switch period. Although our sample size is reduced ourresults are qualitatively unchanged.
Another concern is that teams of investment bankers and analysts switch banks simultaneously. Thus, it could be that staranalysts are driving our results and not investment bankers. We examine cases of analyst and banker switching and find only 5instances where an analyst follows a banker to the same bank during a year period beginning six months before the bankerswitches. We rerun our analyses and find that these 5 cases do not influence our results. Another possibility is that the gaining(losing) bank also adds (loses) a star analyst from (to) another firm. In unreported results, we create a dummy variable equal toone for each of these instances where a bank hires a star analyst from another firm and rerun our regressions in Table 4. We findthat this has no impact on our results.16
15 We obtain similar results using a market-adjusted return specification with either the value-weighted or equally-weighted index.16 We also note that the findings in Clarke et al. (2007) suggest that all-star job changes have only a small impact on equity market share. Our evidence isstrongest for M&A transactions suggesting all-star switching and banker switching are independent events.
594 D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
Finally, job changes by investment bankers could signal that a bank is shifting focus and no longer providing advisory servicesin certain industries. In this case, it would not be surprising that the gaining bankwould increase their market share at the expenseof the losing bank. While it is difficult to assess the motivation behind job changes for investment bankers, the effect describedabove would more than likely apply to switches by teams of bankers. An investment bank switching focus would most likelydismiss all bankers within the affected industry. As a robustness check, we eliminate teams of bankers from our sample and findsimilar results.
5. Conclusions
Our paper adds to the growing literature on the determinants of investment banking deal flow by showing that human capitalis an important factor. In particular, we examine the impact on investment banks in cases where a prominent banker switchesfrom one firm to another. We find that the bank losing the banker experiences a decrease in their industry-level market share. Thiseffect is statistically and economically important and driven primarily by a decline in M&A activity. We find that the ability to hirean investment banker from a bank with a higher industry reputation is the most important determinant of market share changesfor both equity and M&A transactions. We find a significant positive wealth transfer to the hiring bank suggesting the marketfavorably views hiring a banker. Finally, we find a significant number of clients follow the banker to the new bank from the oldbank. Collectively, our evidence indicates that human capital is a critical component of investment banking deal flow.
Acknowledgments
We thank Samuel Bulmash, Jack Cooney, Tyler Henry, Delroy Hunter, Bill Johnson, Beverly Marshall, Rajesh Narayanan, JeffNetter, Jay Ritter, Ninon Sutton, Larry Wall, Lei Wedge, Donghang Zhang and seminar participants at the 2008 FinancialManagement Association and 2008 Eastern Finance Association conferences, the University of Georgia, Virginia Tech, and theUniversity of South Florida for helpful comments. This work was supported by the Hongik University new faculty research supportfund. We are responsible for all remaining errors.
References
Anand, Bharat, Galetovic, Alexander, 2000. Information, nonexcludability, and financial market structure. J. Bus. 73, 357–401.Bhojraj, Sanjeev, Lee, Charles, Oler, Derek, 2003. What's my line? A comparison of industry classification schemes for capital market research. J. Acc. Res. 41,
745–774.Burch, Timothy, Nanda, Vikram, Wartha, Vincent, 2005. Does it pay to be loyal? An empirical analysis of underwriting relationships and fees. J. Financ. Econ. 77,
673–699.Carhart, Mark M., 1997. On persistence in mutual fund performance. J. Finance 52, 57–82.Carter, Richard, Manaster, Steven, 1990. Initial public offerings and underwriter reputation. J. Finance 45, 1045–1067.Clarke, Jonathan, Khorana, Ajay, Patel, Ajay, Rau, Raghavendra, 2007. The impact of all-star analyst job changes on their coverage choices and investment banking
deal flow. J. Financ. Econ. 84, 713–737.Cliff, Michael, Denis, David, 2004. Do initial public offering firms purchase analyst coverage with underpricing? J. Finance 59, 2871–2901.Corwin, Shane, Schultz, Paul, 2005. The role of IPO underwriting syndicates: pricing, information production, and underwriter competition. J. Finance 60, 443–486.
Table 6Wealth effects of gaining or losing a banker. In this table, we examine the wealth effects to the bank gaining or losing a banker around the time of the job changes.Our sample consists of only those cases where the gaining bank or losing bank is publicly traded and has available data on CRSP. Panel A provides results for theentire sample of gainers and losers that meet this criteria and provides a pooled t-test. The announcement date is the earlier date that the switch was reported inInvestment Dealers Digest or Lexis-Nexus. Market adjusted abnormal returns are reported over various windows. Panel B presents a regression analysis of theannouncement period return differential between the gaining bank and the losing bank. We include year dummies in the regression specification. P-values arereported in parentheses.
Panel A: Paired T-tests
Event window Bank gaining the banker Bank losing the banker Difference
Panel B: Regression analysis of announcement period return differential
Variable Coefficient
Intercept 0.34 (0.91)Higher Bank Reputation −0.45 (0.77)Higher Industry Reputation 1.59 (0.40)Managing Director −0.65 (0.72)First departure in a given year 2.96 (0.08)Team of departures −0.10 (0.96)Adj R2 0.05
595D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
Dunbar, Craig, 2000. Factors affecting investment bank initial public offering market share. J. Financ. Econ. 55, 3–41.Eccles, Robert, Crane, Dwight, 1988. Doing Deals: Investment Banks at Work. Harvard Business School Press, Boston, MA.Fama, Eugene, French, Kenneth, 1997. Industry costs of equity. J. Financ. Econ. 43, 153–193.Fee, Edward, Hadlock, Charles, 2003. Raids, rewards, and reputations in the market for managerial talent. Rev. Financ. Stud. 14, 1315–1357.Fernando, Citro, Gatchev, Vladimir, Spindt, Paul, 2005. Wanna dance? How firms and underwriters choose each other. J. Finance 60, 2437–2470.Fernando, Citro, Anthony, May, William, Megginson, 2010. The value of investment banking relationships: evidence from the collapse of Lehman Brothers.
Forthcoming J. Finance.Groysberg, Boris, Lee, Linda-Elling, Nanda, Ashish, 2008. Can they take it with them? The portability of star knowledge workers' performance. Manag. Sci. 54,
1213–1230.Krigman, Laurie, Shaw, Wayne, Womack, Kent, 2001. Why do firms switch underwriters? J. Financ. Econ. 60, 245–284.Ljungqvist, Alexander, Marston, Felicia, Wilhelm, William J., 2006. Competing for securities underwriting mandates: banking relationships and analyst
recommendations. J. Finance 61, 301–340.Loughran, Tim, Ritter, Jay R., 2004. Why has IPO underpricing changed over time? Financ. Manage. 33, 5–37.Netter, Jeffry, Mike, Stegemoller, Babajide, Wintoki, 2010. Implications of data screens on merger and acquisition analysis: a large sample study of mergers and
acquisitions from 1992–2009. Forthcoming Rev. Financ. Stud.Morrison, Alan, Wilhelm, William, 2007. Investment banking: past, present and future. J. Appl. Corp. Finance 19, 8–20.Oyer, Paul, 2008. The making of an investment banker: stock market shocks, career choice, and lifetime income. J. Finance 63, 2601–2628.Walter, Terry, Alfred, Yawson, Charles, Yeung, 2005. Does a quality premium exist in M&A advisory fees? Working paper. University of New South Wales.Yasuda, Ayako, 2005. Do bank relationships affect the firm's underwriter choice in the corporate-bond underwriting market? J. Finance 60, 1259–1292.
596 D. Bradley et al. / Journal of Empirical Finance 18 (2011) 585–596
