CEO Successions and Firm Performance in the US Financial Industry

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. CEO Successions and Firm Performance in the US Financial Industry Antonio Falato and Dalida Kadyrzhanova 2012-79 NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

CEO Successions and Firm Performance in the US Financial Industry Antonio Falato1 Dalida Kadyrzhanova Federal Reserve Board University of Maryland November 2012 1Corresponding Author: Antonio Falato, Federal Reserve Board - Division of Research and Statistics, Washington DC. Phone: (202) 452-2861. Email: antonio.falato@frb.gov. Comments from Nellie Liang, Teodora Paligorova, Gordon Phillips, Steve Sharpe, and seminar participants at the Federal Reserve Board are gratefully acknowledged. Nicholas Ryan, Mercedes Bent, Kristina DeLeon, and Darrell Ashton provided excellent research assistance. All remaining errors are ours. The analysis, conclusions, and discussion in this paper are those of the authors and do not indicate concurrence by other members of the research sta⁄ or the Board of Governors of the Federal Reserve System. References to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the authors to protect the tentative character of these papers.

Abstract ThispaperexaminesthelabormarketforCEOsinthe(cid:133)nancialsectorfrom1988to2007,usinga new hand-collected sample of 1,655 CEO successions. We document that there is a signi(cid:133)cant role of outside successions, as about one out of two successions involves an outside hire. In addition, using di⁄erence-in-di⁄erences estimates, we study the link between the labor market for (cid:133)nance CEOs and (cid:133)rm performance. We document that (1) there is a large performance gap between inside and outside successions, as outside successions are followed by signi(cid:133)cantly larger improvements in (cid:133)rm performance; (2) the performance gap between outside and inside successions is larger for (cid:133)rms with an insider dominated board of directors; (3) the performance gap widened after an important deregulation event (the 1999 Gramm-Leach-Bliley Act). These results are robust to using a battery of (cid:133)rm performance measures (short-run and long-run stock market returns, and several long-run operating performance measures) and a matched sample approach to address selection issues. Overall, our (cid:133)ndings suggest that managerial human capital is very valuable in the (cid:133)nancial industry, and weak internal governance hurts (cid:133)rm performance by limiting the scope of labor market competition.

1 Introduction Internal governance mechanisms andmanagerial labormarket inthe US(cid:133)nancial industryhave recently received considerable attention in the wake of highly publicized CEO successions at the top (cid:133)nancial institutions in the US such as Bank of America, Citigroup, and Wells Fargo. In addition, the debate on the causes and cures of the 2007 (cid:133)nancial crisis has highlighted the key economic role played by top corporate managers at (cid:133)nancial institutions and generated a great deal of interest in the consequences of CEO successions for (cid:133)nancial (cid:133)rm performance. However, while recent work has started to study the role of skill di⁄erentials for non-executive employees (see Philippon and Reshef (2008)), there is surprisingly little evidence on the role of an arguably important input, human capital at the top of the executive ladder in the US (cid:133)nancial industry.1 In order to (cid:133)ll this gap in the literature, we examine the labor market for CEOs in a new hand-collected sample of 1,655 CEO successions in the US (cid:133)nancial sector from 1988 to 2007. In particular, we examine whether CEO succession decisions matter for (cid:133)rm performance. This question has implications for three broad issues. First, evidence for non-(cid:133)nancial (cid:133)rms points to higher incidence of outside successions, which suggests that the importance of the external labormarketforCEOshasincreasedoverthelasttwodecades. Ascon(cid:133)rmedbytheevidencein PhilipponandReshef(2008), (cid:133)nancial(cid:133)rmsarerelativelymorecomplexandrequirespecialized skills, which makes the question of the value of human capital at the top likely to be important. Second,akeychallengefacingcorporateboardsof(cid:133)nancialinstitutionsistoidentifyandattract 1ALthough there is an extensive literature on CEO turnover, the literature is focused on relatively large industrial (cid:133)rms (Forbes, S&P 1500), with very limited coverage of (cid:133)nancials. 1

superior replacement managers. Evidence on whether or not boards actually do so is necessary to address the e⁄ectiveness of internal monitoring. Finally, recent policy interventions have imposed heavy constraints on the ability of large (cid:133)nancial institutions, such as, for example, BankofAmerica, totaptheexternallabormarketforCEOs. Thus, itisimportanttodocument evidence on whether or not these constraints are likely to hurt (cid:133)nancial (cid:133)rm performance. Oursample consists of 1,655 nontakeover-relatedCEOsuccessions inthe US(cid:133)nancial sector over the period 1988 to 2007. An important advantage is that we hand-collected our CEO succession data for the universe of US public (cid:133)nancial (cid:133)rms reporting annual (cid:133)lings (Proxy or 10-K) with the Securities and Exchange Commission (SEC). Thus, we are able to o⁄er a comprehensive picture of the CEO labor market in the US (cid:133)nancial industry, which is in contrast to previous studies that focus on relative large (and mostly industrial) public (cid:133)rms in the S&P 500 or S&P1500 and, thus, include only a handful of the very largest US (cid:133)nancial (cid:133)rms. Our (cid:133)rst set of (cid:133)ndings is about the short-term and long-term stock market reaction to the announcement of CEO succession decisions. We (cid:133)nd that investors expect outside CEOs in the (cid:133)nancialindustrytosigni(cid:133)cantlyoutperforminsiders,asbothshort-termannouncementreturns and average abnormal returns for the three-year post-succession period are signi(cid:133)cantly higher for outside CEO successions. In addition, we (cid:133)nd evidence consistent with both a governance and a human-capital explanation of this (cid:133)nding. In particular, the return di⁄erential is higher for (cid:133)rms with insider-dominated boards of directors (de(cid:133)ned as (cid:133)rms whose boards have 40 percent or more inside directors). This result is consistent with the hypothesis that insiderdominated boards of directors are expected to hurt (cid:133)rm performance by limiting the scope of 2

labor market competition and hiring (cid:146)bad(cid:146)CEOs from inside the (cid:133)rm. In addition, we (cid:133)nd that the return di⁄erential widens after deregulation in 1999 (Gramm-Leach-Bliley Act) that removed barriers separating traditional banking, insurance, and securities underwriting and thereby increased complexity of bank operations. This result suggests that there is a "skillbiased" e⁄ect of deregulation, in the sense that the increased complexity of (cid:133)nancial (cid:133)rms led to a bigger return di⁄erential between inside and outside successions. A well-known issue with event-study results is that their interpretation is di¢ cult, since a management change may signal that (cid:133)rm performance is worse than expected, that (cid:133)rm performance will improve as a result of the management change, or that the (cid:133)rm is "in play" as a takeover target. Moreover, top management changes are likely to be partially anticipated due to the poor pre-turnover (cid:133)rm performance. Thus, in order to provide a more convincing assessment of the e⁄ect of CEO succession decisions on (cid:133)rm performance, we examine changes in several measures of (cid:133)rm performance (operating return on assets, operating returns on sales, and Tobin(cid:146)s Q) around management changes. We (cid:133)nd that, on average, there is a signi(cid:133)cant performance gap between inside and outside successions. This overall result masks considerable di⁄erences between (cid:133)rms with di⁄erent internal governance structures, as, again, the gap is larger for (cid:133)rms with insider-dominated boards. Moreover, the gap in operating performance widens after 1999. Finally, the gap remains signi(cid:133)cant even after 2003, suggesting that it is unlikely to be driven by outsiders taking in the expansion stage of the credit cycle aggressive risks that later materializd during the (cid:133)nancial crisis.2 Althoughourresultsonchangesinoperatingperformanceareakintodi⁄erence-in-di⁄erences, 2We thanks Steve Sharpe for suggesting this additional test. 3

in that we can estimate CEO impact in a setting that explicitly controls for time-invariant differences in (cid:133)rm characteristics that may a⁄ect performance (see Perez-Gonzalez (2006) for a similar approach in the context of family successions), there is an important selection issue we havetoaddress. Infact, insidesuccessionsinvolvebothobservabledi⁄erencesin(cid:133)rmcharacteristics, which might by time-varying, and unobservable di⁄erences, making a direct comparison between inside and outside successions problematic. Ideally, we would like to compare the change in performance of an inside appointment (cid:133)rm to the same (cid:133)rm(cid:146)s performance change had the (cid:133)rm appointed an outside CEO. Since the counterfactual is not observed, we must (cid:133)nd an empirical proxy for the hypothetical performance without succession type change. As our main identi(cid:133)cation strategy, we construct a nearest-neighbor matching estimator, following AbadieandImbens (2007). Weestimatealogit regressiontoidentifyobservable(cid:133)rmcharacteristics that predict inside successions. We then match each inside CEO succession to the outside succession that, at the time of the succession, had the closest predicted probability of being an inside succession, or propensity score (Rosenbaum and Rubin 1983). CEO successions are a natural application for matching since the succession decisions are made by corporate boards who, like the econometrician, have to rely mostly on public information to assess outside CEO quality. Our results for performance changes around CEO successions relative to the matched control sample largely con(cid:133)rm and are somewhat stronger than our baseline results. To the best of our knowledge, our paper o⁄ers the (cid:133)rst direct large sample evidence on the labor market for CEOs in the US (cid:133)nancial industry. Our evidence broadly suggests that (cid:133)nancialCEOselectiondecisionsmatter. Ourresultsstandinsharpcontrasttopreviousstudies that focus on non-(cid:133)nancials and tend to (cid:133)nd mixed and at best weak evidence of performance 4

di⁄erentials between insiders and outsiders. The strong performance gap between inside and outside successions suggests the external labor market for CEOs has a special role in the US (cid:133)nancial industry. This (cid:133)nding has two main implications for the literature. First, our (cid:133)nding that di⁄erences among CEO successions are important for (cid:133)nancial (cid:133)rms, especially after the 1999 deregulation, is consistent with the evidence for non-executive employees in Philippon and Reshef (2008), and supports the notion that the (cid:133)nancial industry is relatively complex and skill-intensive. In addition, our evidence furthers the understanding of the role of the external labor market for CEOs. Existing work is limited to mostly non-(cid:133)nancial (cid:133)rms (e.g. Warner, Watts, and Wruck (1988), Parrino (1997), and Huson, Parrino, and Starks (2001)). Our evidence shows that the external market for CEOs is an important source of value for (cid:133)nancial (cid:133)rms. Given both the broad set of new variables we examine and the large cross-section of (cid:133)rms we include in our hand-collected dataset, our investigation represents to best of our knowledge the (cid:133)rst large-sample study of the impact of the external labor market for CEOs on (cid:133)nancial (cid:133)rm performance. Our study is also complementary to the small but growing literature that attempts to identify the e⁄ect of CEOs on (cid:133)rm performance. Bertrand and Schoar (2003), Bennedsen, Perez-Gonzalez, and Wolfenzon (2006), and Bennedsen, Nielsen, Perez-Gonzalez, and Wolfenzon(2007)presentevidencethatCEOsmatterfor(cid:133)rmperformance. However, thelinkbetween CEOs and (cid:133)rm performance in the (cid:133)nancial industry has been surprisingly overlooked. Thus, our paper is the (cid:133)rst to show that (cid:133)nancial CEOs matter. Second, our(cid:133)ndingthattheperformancegapbetweeninsideandoutsidesuccessionislarger for (cid:133)rms with insider-dominated boards has important implications for the recent governance 5

debate and the standard criticism of board of directors for not doing a good job at monitoring CEOs (see, for example, Bebchuk and Fried (2003)). Our evidence is complementary to the basic premise of this argument, and suggests that identifying and attracting superior CEO replacementsisindeedanimportant, althoughoftenoverlooked, functionof boardsof directors. The remainder of the paper is organized as follows. In Section 2 , we discuss our sample selection procedure and describe our samples of CEO successions in the US (cid:133)nancial industry. In Section 3, we present event-study results documenting announcement-period and long-term abnormal returns associated with our sample management changes. In Section 4, we document changesinaccountingperformancemeasuresaroundmanagementchanges. Section5concludes. 2 Data ToexplorethelinkbetweenCEOsuccessionsand(cid:133)rmperformanceintheUS(cid:133)nancialindustry, we construct a database of the (cid:133)nance CEO labor market that contains detailed information on CEO turnovers, as well as multiple empirical proxies for (cid:133)rm performance. This section details how we constructed the dataset and the collection process for each of our variables. 2.1 Sample selection We hand-collected our CEO succession data for the universe of US public (cid:133)nancial (cid:133)rms reporting annual (cid:133)lings (Proxy or 10-K) with the Securities and Exchange Commission (SEC) for the 1988 to 2007 period. To construct our sample, we start with the universe of all Compustat (cid:133)nancial (cid:133)rms (SIC codes between 6000 and 6999) in existence during the years 1986 to 2007. 6

This is a sample of 31,583 (cid:133)rm-year observations. Next, for every (cid:133)rm-year observation in the Compustat (cid:133)nancials universe, we match the observation to its respective Proxy or 10-K SEC (cid:133)ling that generates the Compustat data. Using these matches, we employ a text-search algorithm to search the (cid:133)lings for the full name and basic biographical characteristics of the CEO in o¢ ce. We cross-check the results of our algorithm for accuracy with information from the same (cid:133)lings in Compact Disclosures. We recognize a CEO turnover for each year in which the CEO name changes (earlier studies for industrial (cid:133)rms, such as Parrino (1997), Huson, Parrino, and Stark (2001), Huson, Malatesta, and Parrino (2004), use Forbes surveys; Jenter and Kanaan (2006) use ExecuComp which only includes S&P 1500 (cid:133)rms). This gives us a (cid:133)rst sample of 1,995 candidate CEO succession events. We then search the Factiva news database in order to collect information about the circumstances around each succession. We exclude 340 successions that are directly related to a takeover. The (cid:133)nal sample contains 1,655 CEO succession events. Our sample is broadly representative of the US (cid:133)nancial universe in Compustat: about 40 percent of our successions involve bank holding companies (BHC), about 25 percent are commercial banks, about 15 percent are insurance companies, and about 5 percent are broker dealers. We classifyeachCEOturnoveraccording towhetherit was forcedorvoluntaryandwhether the incoming CEO is an insider or an outsider to the (cid:133)rm, following standard criteria in the literature (Parrino (1997), Huson, Malatesta, and Parrino (2004)). We classify successor CEOs whohadbeenwiththeir(cid:133)rmsforoneyearorlessatthetimeoftheirappointmentsasoutsiders. All other new CEOs are classi(cid:133)ed as insiders. Finally, for each succession we determine exact announcement dates - which are the earliest dates of the news about incumbent CEOdeparture 7

andsuccessorCEOappointment. DeparturesforwhichthepressreportsstatethattheCEOhas been (cid:133)red, forced out, or retired or resigned due to policy di⁄erences or pressure, are classi(cid:133)ed as forced. All otherdeparturesforCEOsaboveandincludingage60areclassi(cid:133)edas not forced. All departures for CEOs below age 60 are reviewed further and classi(cid:133)ed as forced if either the article does not report the reason as death, poor health, or the acceptance of another position (including the chairmanship of the board), or the article reports that the CEO is retiring, but does not announce the retirement at least six months before the succession.3 This careful classi(cid:133)cation scheme is necessary since CEOs are rarely openly (cid:133)red from their positions. Table 1 presents an overview of our CEO succession data set for the US (cid:133)nancial industry with descriptive statistics on total CEO successions, and successor type (inside vs. outside) for each year (Panel A) and for two sub-periods ((cid:133)rst and second half of the sample) covered by our sample (Panel B). We are able to give a signi(cid:133)cantly more comprehensive picture of the CEO labor market in the US (cid:133)nancial industry than previous studies since our sample includes a more detailed collection and considerably larger cross-section of (cid:133)rms (Compustat universe) than S&P500, S&P 1500, or Forbes sub-samples, which have been the standard focus of the literature on industrial (cid:133)rms.4 Our statistics con(cid:133)rm results for non-(cid:133)nancials suggesting that the nature of the CEO labor market has changed signi(cid:133)cantly in the last two decades with respect to the 1970s and 1980s. The likelihood that the new CEO comes from outside the (cid:133)rm 3The cases classi(cid:133)ed as forced can be reclassi(cid:133)ed as voluntary if the press reports convincingly explain the departureasduetopreviouslyundisclosedpersonalorbusinessreasonsthatareunrelatedtothe(cid:133)rm(cid:146)sactivities. 4Studies covering earlier periods use Forbes Compensation Surveys, which roughly include S&P 500 and S&P MidCap 400 (cid:133)rms. Denis and Denis (1995) covers a sample of 908 CEO successions between 1985 and 1988. Huson, Parrino, and Starks (2001) and Huson, Malatesta, and Parrino (2004) have 1,316 and 1,344 CEO successions, respectively, between 1971 and 1994. Murphy and Zabojnik (2007) have 2,783 appointments between 1970 and 2005, which is a larger, but less detailed dataset than ours. 8

are much higher than what it had been documented in previous decades. Both Panels in Table 1 show that, as it has been documented for non-(cid:133)nancial (cid:133)rms, the (cid:133)nancial industry was also subject to an important recent trend in the CEO labor market: there is an increased prevalence of (cid:133)lling CEO openings through external hires rather than through internal promotions, suggesting that there has been a material change in the CEO selection process in the 1990s. About (cid:133)fty percent of the departing CEOs in the last two decades are replaced by executives who have been employed at the (cid:133)rm for one year or less. This frequency of outside appointments is about in line with recent studies of industrial (cid:133)rms, although somewhat higher. In fact, studies that focus on non-(cid:133)nancials have (cid:133)gure that range between 35 and 40 percent. This di⁄erence is due not only to possible di⁄erences between (cid:133)nancials and non-(cid:133)nancials, but also to the fact that the samples typically used in previous studies only includes relatively larger (cid:133)rms, which tend to rely more on inside hires. Finally, these (cid:133)gures are striking if contrasted against earlier decades. For example, Murphy and Zabojnik (2007) and Huson, Parrino, and Starks (2001) report that during the 1970s and 1980s outside hires accounted for only 15% to 17% of all CEO replacements, only half as large as our (cid:133)gures since 1998. 2.2 Firm Performance and Firm-Level Controls Wesupplementourdatawithseveralmeasuresof(cid:133)rmstockmarketandoperatingperformance, as well as a variety of (cid:133)rm-level controls whose importance in the CEO labor market has been documented in the literature. All measures are at calendar year-end. Our stock market-based measure of performance is based on stock returns from CRSP 9

(Parrino (1997), Warner, Watts, and Wruck (1988), and Huson, Parrino, and Starks (2001)). We use three measures of (cid:133)rm operating performance from Compustat: (1) operating return on assets(OROA),de(cid:133)nedastheratioofoperatingincometothebookvalueofassets(2)returnon assets(ROA),de(cid:133)nedastheratioofnetincometothebookvalueofassets; (3)operatingreturn on sales (OROS), de(cid:133)ned as the ratio of operating income to sales. For each of these measures, we de(cid:133)ne its industry-adjusted counterpart by subtracting the median of the relevant industry (2-digit SIC) and year, and its industry and performance-adjusted counterpart by subtracting the median of the relevant variable of a control group of (cid:133)rms with similar industry-adjusted performance. The control groups are created by dividing COMPUSTAT (cid:133)rms into deciles sorted by the relevant variable (e.g. industry-adjusted OROA) the year prior to transition. The yearly median of the relevant group of (cid:133)rms (ex-event) is then used as the control for each (cid:133)rm-year observation (see Barber and Lyon (1996) for more details on the construction of the performance-adjusted variables). Our main set of controls includes (cid:133)rm size (logarithm of total assets), and CEO age. The role of (cid:133)rm size in the CEO labor market is an important implication of competitive models such as ours (see Gabaix and Landier (2008) and Tervio (2007)). Previous research suggests that CEO pay and turnover rates are a function of CEO age (see, for example, Milbourn (2003) and Chevalier and Ellison (1999)(cid:146)s study of the sensitivity of mutual funds manager turnover to performance). Finally, we include in our data set several measures of (cid:133)rm internal governance. In particular, we include the size and independence of the board of directors (see Weisbach (1988) and KaplanandMinton(2006)forevidenceonboardsandCEOsuccessions). Ourmainvariablefor 10

board independence is a dummy that takes the value of one if the board is insider-dominated (top quartile of the distribution of board independence in our sample, which correspons to a proportion of insiders of 40 percent or more). 3 Event-Study Results OurresearchsettingallowsustoimplementdirecttestsoftherelationbetweenCEOsuccessions and (cid:133)rm performance. In this section, we o⁄er event-study evidence of the impact of CEO succession decisions on performance for (cid:133)rms in the US (cid:133)nancial industry. 3.1 Short-Term and Long-Term Event Studies of CEO Succession Decisions Before moving on to our main analysis, we examine announcement and long-term abnormal stock returns around CEO successions. Investor perception is an informative and intuitive indicator of anticipated future performance conditional on all relevant information (Warner, Watts, and Wruck (1988), Denis and Denis (1995), Huson, Malatesta, and Parrino (2004), Perez-Gonzalez (2006)). Thus, for example, we expect to see positive abnormal returns for outside hires at the time of the hiring announcement, if the market expects them to outperform inside hires. Table 2 shows evidence that indeed investors expect a positive impact of CEOs hired from outside the (cid:133)rm on performance. In particular, we present mean abnormal returns for a two-day event window around CEO succession announcements for all successions, and for successions when management changes are broken down by internal and external successions 11

(top panel), and by insider-dominated board and post-1999 deregulation period (lower panel).5 Column 1 of Table 2 shows that on average CEO successions are associated with a statistically signi(cid:133)cant (albeit small at 0.8%) abnormal return. The positive average return is in contrast to previous studies that use earlier samples of larger and non-(cid:133)nancial (cid:133)rms and tend to (cid:133)nd insigni(cid:133)cant returns on average (see, for example, Huson, Malatesta, and Parrino (2004)). However, as shown in Column 2 of Table 2, this di⁄erence is likely explained by the fact that internal appointments, which constitute a much larger fraction of the total sample in earlier studies, are associated with abnormal returns that are not di⁄erent from zero. By contrast, investors react positively to appointments of outside CEOs, which constitute a larger fraction of our sample and on average are associated with a signi(cid:133)cant 1.7% return. Overall, outside successions carry a statistically signi(cid:133)cant 1.8% excess return with respect to inside successions, consistent with the market(cid:146)s anticipation that outside hires will outperform inside hires. This performance di⁄erential is much larger than documented in studies of CEO successions for non-(cid:133)nancial (cid:133)rms, which is consistent with the notion that managerial human capital is relatively more valuable in the (cid:133)nancial industry (see Philippon and Reshef (2008)), but also with the idea that governance issue might be more severe. The lower panel of Table 2 explores the merit of two main explanations for the performance gap between insiders and outsides, di⁄erences in the value of human capital and internal governance issues. To explore the role of internal governance issues, we ask whether the performance gap is larger among (cid:133)rms with insider-dominated boards (de(cid:133)ned as (cid:133)rms where more than 40 5Abnormal returns are calculated using the capital asset pricing model (CAPM) and standard event study methodology (see MacKinlay (1997) for a detailed review). We use the market model and CRSP equallyweighted return as the market return to estimate the market model parameters from event day -210 to event day -11. 12

percent of the members of the board of directors are insiders), which would be consistent with thehypothesisthatinsider-dominatedboardsofdirectorsareexpectedtohurt(cid:133)rmperformance by hiring underperforming insiders. Consistent with this hypothesis, we see a somewhat larger 2.1% excess return for outsiders. Finally, inordertoevaluatethehypothesisthatdi⁄erencesinthevalueofhumancapitalare drivingthegapbetweeninsidersandoutsiders,weaskwhetherthegapwidensafterderegulation in 1999 (Gramm-Leach-Bliley Act) that increased complexity of bank operations by removing barriersseparatingtraditionalbanking, insurance, andsecuritiesunderwriting. Consistentwith di⁄erences in the value of human capital between insiders and outsiders, the bottom panel of Table 2 shows that outside CEOs(cid:146)excess returns is higher (2.6%) after 1999. In summary, our short-term event study shows that investors expect outside CEOs in the (cid:133)nancialindustrytosigni(cid:133)cantlyoutperforminsiders, andmoresoforCEOsuccessiondecisions madebyinsider-dominatedboardsandafter1999., whichisconsistentwithbothhumancapital and governance factors potentially driving the expected performance di⁄erential. 3.1.1 Long-Term Event Study A potential concern with shot-term announcement returns is that, as emphasized by Khurana (2002), an anticipated positive impact of CEOs does not necessarily imply a realized positive impact since investors (and board of directors themselves) might simply irrationally over-react to the appointment of a popular and charismatic CEO and thus lead to a positive stock market reaction that is unrelated to actual CEO performance. In order to partially addresses this overreaction concern, we consider long-term abnormal returns, which are more likely to capture 13

subsequent information on the value of CEO human capital that is revealed slowly over time. As it is standard in the literature (see, for example, Huson, Malatesta, and Parrino (2004) and Perez-Gonzalez (2006)), we calculate monthly calendar-time portfolio returns for portfolios that buy shares in (cid:133)rms subject to a CEO transition within the following 36 months, as well as for portfolios invested in (cid:133)rms that underwent a succession in the preceding 36 months. We estimate abnormal returns using the four-factor market-model (see Fama and French (1993); and Jegadeesh and Titman (1993)). Table 3 reports the resulting average abnormal returns for the three-year pre-succession period(Panel A)andthethree-yearpost-successionperiod(Panel B).6 BeforeCEOtransitions, (cid:133)rms tend to earn signi(cid:133)cant negative abnormal returns, which is in line with the standard (cid:133)nding in the literature that underperforming (cid:133)rms are more likely to replace their CEO. Presuccession underperformance tends to be more pronounced for outside successions, consistent with another standard result in the literature that (cid:133)rms are more likely to appoint an outsider when they are relatively more underperforming. Turning to Panel B of Table 3, the portfolio of post-CEO transition (cid:133)rms earned on average statistically signi(cid:133)cant abnormal returns of about 5%, signi(cid:133)cant at the (cid:133)ve-percent level. 6Abnormal returns are estimated using calendar-time portfolio regressions. In each month t, all (cid:133)rms subject to a CEO succession within the next (prior) 36 months are included in that month(cid:146)s pre (post)transitionportfolio. Meanportfolioreturns,rp areusedtoestimateabnormalreturnsusingthe t following regression: (rp rf )=(cid:11)+(cid:12) (rm rf )+(cid:12) SMB +(cid:12) HML +(cid:12) UMD +" ; where rf is the t (cid:0) t 1 t (cid:0) t 2 t 3 t 4 t t t risk-freeratecalculatedusingone-monthTreasury-billrates,(rm rf )isthemarketriskpremium,calt t (cid:0) culatedasthedi⁄erencebetweenthevalue-weightedreturnonallNYSE,AMEX,andNASDAQstocks from CRSP less the risk-free rate, SMB is the return di⁄erence between portfolios of small stocks and t big stocks, HML is the return di⁄erence between portfolios of high book-to-market stocks and low t book-to-marketstocks,andUMD isthereturndi⁄erencebetweenportfoliosofhighprior-returnstocks t andlowprior-returnstocks. Thereportedabnormalreturnsaretheintercept((cid:11))estimatedfromtheregressionabove. Theimpliedone-yearabnormalreturniscalculatedas[(1+(cid:11))12 1]. Dataonthefactors (cid:0) were obtained from Ken French(cid:146)s website: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french. 14

Consistent with our previous (cid:133)nding of a more positive short-term announcement return for outside appointments, portfolios of (cid:133)rms that appoint outside CEOs earn higher abnormal returns after transitions relative to (cid:133)rms that appoint inside CEOs. The performance gap between outsiders and insiders in terms of one-year excess returns is of about 4%. Thus, there is about 4% return premium earned by investors of (cid:133)rms that appoint outside CEOs. Finally, consistent with both governance and human capital factors driving the performance gap, the returnpremiumislargerinthesub-sampleofappointmentsbyinsider-dominatedboards(about 7%) and for appointments made after 1999 (about 7%). 4 Main Results Overall, both short-term and long-term abnormal returns support the view that outside CEOs are more likely to have a positive impact on (cid:133)rm performance. However, these results are only suggestive and do not establish that there is indeed a positive impact of outside CEOs on (cid:133)rm performance since the results might also be driven by anticipation e⁄ects, such as the fact that appointment decisions reveal information related to (cid:133)rms(cid:146)prospects, irrespective of the value of CEO human capital. Moreover, lower signi(cid:133)cance might also be driven by the fact that inside transitions were expected and already incorporated into prices. In order to address these concerns, in our main analysis we pursue an alternative strategy based on using changes in operating performance. Thus, our strategy is testing whether there are signi(cid:133)cant di⁄erences in (cid:133)rm performance before and after CEO successions for (cid:133)rms that appoint inside vs. outside CEOs. 15

The advantage of this approach, which is akin to di⁄erence-in-di⁄erences, is that we can estimate CEO impact in a setting that explicitly controls for time-invariant di⁄erences in (cid:133)rm characteristics that may a⁄ect performance (see Perez-Gonzalez (2006) for a similar approach in the context of family successions). We use three di⁄erent measures of operating performance which are standard in the CEO turnover literature: (1) operating return on assets (OROA), (2) operating return on sales (OROS), and (3) valuation ratios (Tobin(cid:146)s Q), which addresses the concern that, while e⁄ective at addressing anticipation issues, one potential limitation of operating performance measures is that they only capture current pro(cid:133)tability. We report results for the di⁄erence between these measures three years after and one year prior to CEO appointment. We consider industry-adjusted and industry- and prior performance-adjusted versions of the three measures to address potential concerns with the results being driven by industry-wide trends or simply mean-reversion with respect to prior performance.7 The results are reported in Table 4, which reports mean di⁄erences in (cid:133)rm performance before and after CEO transitions. Consistently across our three di⁄erent measure of performance, the average di⁄erence in performance three-yearafterCEOsuccessions minus performance oneyear before transitions for the entire sample is not statistically di⁄erent from zero. This result is in line with studies of CEO successions for non-(cid:133)nancial (cid:133)rms (see, for example, Huson, Malatesta, and Parrino (2004)). However, again consistently across our three di⁄erent measure of performance, when we classify (cid:133)rms by succession type (inside vs. outside succession), we 7To construct control-group adjusted performance,we follow Barber and Lyon (1996) matching method. In particular,eachsample(cid:133)rmismatchedtocomparison(cid:133)rmswiththesametwo-digitCompustatSICcodewhose performancemeasuresovertheyearbeforetheturnoverarewithin10%ofthesample(cid:133)rm(cid:146)sperformance. Each sample(cid:133)rm(cid:146)sperformanceisadjustedbysubtractingthemedianperformanceofitscontrolgroup.Changesover time in adjusted performance are then calculated. 16

(cid:133)nd large di⁄erences-in-di⁄erences between inside and outside successions. In economic terms, our estimates indicates a gap in performance between insiders and outsiders ranging from 25 to 50 percent of the pre-transition unadjusted level of performance. Finally, consistent with both governance and human capital factors driving the performance gap, the gap is signi(cid:133)cantly larger in the sub-sample of appointments by insider-dominated boards and for appointments made after 1999. Firm Decisions and Risk Our (cid:133)nding of a signi(cid:133)cant positive impact of CEO succession decisions on (cid:133)rm performance opens the intriguing question of what it is exactly that outside CEOs manage to do better than insiders. This question is related to the evidence in Bertrand andSchoar(2003)thattherearesigni(cid:133)cantdi⁄erencesin(cid:133)rmpoliciesacrossCEOs. Inaddition, thebuildupofriskintheUS(cid:133)nancialsectorpriortothe(cid:133)nancialcrisisraisesanotherimportant question: does the positive impact of outside successions on (cid:133)rm performance simply re(cid:135)ect outside CEOs(cid:146)higher propensity to take risks in the expansion stage of the credit cycle which later materialized into sub-par performance once the (cid:133)nancial crisis hit? Thus, we next ask whethersuccessiondecisionshaveexplanatorypowerforchangesin(cid:133)rmpoliciesandriskpro(cid:133)le around CEO successions. Table 5 reports results that are aimed at answering this question. In particular, we now use our di⁄erence-in-di⁄erences strategy to consider a variety of (cid:133)nancial, operating, and risk taking (cid:133)rm policies (these (cid:133)rm decisions are analogous to the ones studied in Bertrand and Schoar (2003)). Our results on the impact of CEO talent on (cid:133)rm decisions paint a picture that (cid:133)ts remarkably well anecdotal accounts of outside CEOs as aggressive professional turnaround 17

specialists. In particular, outside CEOs are more likely to cut leverage, to increase internal (cid:133)nancing (cash - not reported), and to generate higher cash (cid:135)ows. In addition, and perhaps surprisingly, outside CEOs do not appear to improve performance by increasing (cid:133)rm risk. In fact, appointments of outside CEOs lead to larger reductions in (cid:133)rm total risk (as measured by total return volatility) compared to appointments of inside CEOs. Finally, con(cid:133)rming our resultsontheperformancegap, di⁄erencesin(cid:133)rmpoliciesbetweeninsideandoutsideCEOsare more pronounced for appointments by insider-dominated boards and for appointments made after 1999, which lends further support to the idea that both governance and human capital factors are important. A potential concern with this evidence is that although outside CEOs took more risks, these risks need not necessarily have materialized within three years from their appointments. In order to address this concern, we consider the sub-set of post-1999 appointments starting from 2003. Since we track performance up to three years subsequent to these appoitments, this sub-sample includes CEO successions for which we observe at least one one of subsequent performance overlaps with the (cid:133)nancial crisis since 2006. In unreported results available upon request, we have repeated our analysis in Panel B of Table 4 for this post-2003 subsample. In summary, the results con(cid:133)rm our (cid:133)ndings for the post-1999 period of strong outperformance by outside CEOs. In particular, we (cid:133)nd a statistically signi(cid:133)cant performance di⁄erential of 0.208 when performance is measured based on Tobin(cid:146)s Q, 0.015 based on OROA, and 0.054 based on OROS. Overall, these results are inconsistent with the hypothesis that outside CEOs outperformed insiders by taking more aggressive risks. 18

4.1 Identi(cid:133)cation An important concern with our main results on inside vs. outside successions is that, even though we control for pre-succession performance, there are other variables that can e⁄ect di⁄erential (cid:133)rm performance around CEO successions, including, for example, (cid:133)rm size. Thus, partof ourestimatedimpactof CEOsuccessiondecisionsmightbeattributedtothesevariables rather than type of succession itself. For example, since large (cid:133)rms are more likely to hire insiders, it might be that part of the subsequent performance improvement is simply due to outside CEOs being chosen to run smaller (cid:133)rms that are easier to turn around. In the ideal empirical experiment, we would compare the performance of an inside appointment (cid:133)rm to the same (cid:133)rm(cid:146)s performance had the (cid:133)rm appointed an outside CEO. Since the counterfactual is not observed, we must (cid:133)nd an empirical proxy for the hypothetical performance without succession type change. Our approach is a natural starting point since we compare average ex-post changes in performance of (cid:133)rms that appoint inside CEOs to the expost change in performance of (cid:133)rms that appoint outside CEOs. This di⁄erence-in-di⁄erences approach would provide a valid estimate of the treatment e⁄ect of the treated if assignment to the treatment group were random. However, basic theoretical considerations and previous evidenceonCEOsuccessionssuggestthatthisassumptionisnotlikelytoholdinthedata. Infact, when we test di⁄erences in pre-succession (cid:133)rm characteristics across the two groups (inside vs outside appointments), we (cid:133)nd signi(cid:133)cant di⁄erences in (cid:133)rm size and performance, with inside appointments associated with larger and relatively less underperforming (cid:133)rms. Economically, these di⁄erences re(cid:135)ect the endogeneity of CEO succession decisions. In order to isolate the real e⁄ects of CEO succession type on corporate performance from 19

selection e⁄ects, our main strategy is to construct a nearest-neighbor matching estimator, following Rosenbaum and Rubin (1983) and Abadie and Imbens (2007). While we do not observethecriteriausedtoselectinsidevs. outsideCEOs, thematchingprocedurereconstructs this information using observable characteristics. We construct the control sample in two steps. First,werunalogitregressiontopredictsuccessiontype(insidevs. outsideCEO)basedon(cid:133)rm characteristics. We set the binary dependent variable to 1 if the (cid:133)rm appoints an inside CEO. We then regress the inside CEO indicator on controls for (cid:133)rm characteristics. Based on the results in the existing literature for non-(cid:133)nancials, we include (cid:133)rm size (the natural logarithm of marketcapitalizationatthebeginningof theyearbeforetheappointment), (cid:133)rmperformance (asmeasuredbyourthreeproxiesatthebeginningoftheyearbeforetheappointment),and(cid:133)rm board characteristics (size and insider-dominated dummy). We also include dummies for years. In unreported results available upon request, we (cid:133)nd a weak relation between pre-transition (cid:133)rm performance and the likelihood of observing an inside appointment (with underperforming (cid:133)rms being less likely to appoint an insider), which is consistent with previous studies. Also consistent with previous studies, we (cid:133)nd that larger (cid:133)rms and (cid:133)rms with insider-dominated boards are signi(cid:133)cantly more likely to appoint inside CEOs (see Weisbach (1988) for similar evidence) Next, we use the predicted values from the logit regression (propensity scores) to construct a nearest-neighbor matched sample for inside CEO appointments. In each year, we choose, with replacement, the outside CEO appointments with propensity scores closest to those of each inside appointment. We use the propensity score as the match variable to reduce the 20

dimensionality of the matching problem.8 Table 6 contains the results. For each performance measure, the (cid:133)rst column shows the di⁄erence-in-di⁄erences estimates from Table 4 (Panel A), while the second column reports the di⁄erence with respect to the matched sample. Clearly, robustly across the three performance measures, we continue to (cid:133)nd a signi(cid:133)cant impact of CEO succession decisions on (cid:133)rm performance and a large performance gap between inside and outside CEO successions. Finally, con(cid:133)rming our previous results, the performance gap of insiders with respect to the matched sample is more pronounced for appointments by insider-dominated boards. 5 Conclusion CEO successions are important instances when managerial human capital is in play. We argue that focusing on the labor market for CEOs in the US (cid:133)nancial industry can augment our understanding of the role of managerial human capital and (cid:133)rm internal governance as determinants of (cid:133)nancial (cid:133)rm performance. In a large hand-collected sample of CEO turnovers over the last two decades, we (cid:133)nd robust evidence that outside CEOs performed signi(cid:133)cantly better than inside hires. Moreover, this result is stronger for CEOs hired by insider-dominated boards and after an important (cid:133)nancial deregulation event in 1999. Our results cannot be explained by temporary over-reaction or anticipation e⁄ects as they are derived using long-term 8We also use the procedure of Abadie and Imbens (2007) to correct for remaining bias due to (ex ante) di⁄erences between the treatment and control samples. The procedure estimates an auxiliary OLS regression of the e⁄ect of the match variables on the outcome variable (in the control sample) and uses the estimates to adjust for di⁄erences in the match variables between the treatment and control samples. This correction ensures, for example, that an outlier insider with a propensity score too high to closely match does not drive our results. In unreported results available upon request, we (cid:133)nd that this adjustment has a very small e⁄ect on our result, which are essentially unchanged. 21

measures of operating performance and are robust to addressing selection on observable size or pre-transition performance. Finally, we o⁄er suggestive evidence that the impact of outside CEO successions is related to classical turnaround skills. 22

References [1] Abadie, A. and G. W. Imbens, 2007, "Bias Corrected Matching Estimators for Average Treatment E⁄ects,(cid:148)Mimeo, 2007. [2] Aggarwal, R. and A. Samwick, 1999, "Executive Compensation, Strategic Competition, and Relative Performance Evaluation: Theory and Evidence," The Journal of Finance, 54, pp. 1999-2043. [3] Baker, G. P., and B. J. Hall, 2000, "CEO Incentives and Firm Size," NBER WP6868. [4] Bebchuk, L. A. and J. M. Fried, 2003, (cid:148)Executive Compensation as an Agency Problem,(cid:148) Journal of Economic Perspectives 17(3), pp.71-91. [5] Bebchuk, L. A., J. M. Fried and D. I. Walker, 2002, (cid:148)Managerial Power and Rent Extraction in the Design of Executive Compensation,(cid:148)University of Chicago Law Review, 69, pp.751-846. [6] Bennedsen, M., F. Perez-Gonzalez, and D. Wolfenzon, 2006, "Do CEOs Matter?" mimeo, NYU Stern. [7] Bennedsen, M., K.M. Nielsen, F. Perez-Gonzalez, and D. Wolfenzon, 2007, "Inside the Family Firm: The Role of Families in Succession Decisions and Performance," Quarterly Journal of Economics, forthcoming. [8] Bertrand, M. and S. Mullainathan, 2001, "Are CEOs Rewarded for Luck? The Ones without Principles are," Quarterly Journal of Economics 116, 901-932. [9] Bertrand, M. and A. Schoar, 2003, (cid:147)The Managing with Style: The E⁄ect of Managers on CorporatePolicy,(cid:148)TheQuarterlyJournalofEconomics, November2003, 118(4), pp.1169- 1208 [10] Chevalier, J., and G. Ellison, 1999, "Career Concerns of Mutual Fund Managers," Quarterly Journal of Economics, 114, pp. 389-432. [11] Denis, D. J., D. K. Denis, and A. Sarin, 1997, "Ownership Structure and Top Executive Turnover," Journal of Financial Economics, 45, pp. 193-222. [12] Fisman R., R. Khurana, and M. Rhodes-Kropf, 2008, "Governance and CEO Turnover: Do Something or Do the Right Thing?" mimeo, Columbia University. [13] Francis, J., A. H. Huang, S. Rajgopal and A. Y. Zang, 2004, "CEO Reputation and Reporting Quality," mimeo, Duke University. [14] Frydman,C.,2005,"RisingThroughtheRanks.TheEvolutionoftheMarketforCorporate Executives, 1936-2003," mimeo, MIT. [15] Gabaix, X., and A. Landier, 2006, (cid:147)Why has CEO Pay Increased so Much?(cid:148)NBER WP #12365. 23

[16] Garvey, G. and T. Milbourn, 2006, "Asymmetric Benchmarking in Compensation: Executives are Rewarded for Good Luck but not Penalized for Bad," forthcoming, Journal of Financial Economics. [17] Gibbons, R. and L. F. Katz, 1991, "Layo⁄s and Lemons," Journal of Labor Economics, 9, pp. 351-380. [18] Gibbons, R. and K. J. Murhy, 1992, "Optimal Incentive Contracts in the Presence of Career Concerns: Theory and Evidence," Journal of Political Economy, 100, pp. 468-505. [19] Gompers, P. A., J. L. Ishii and A. Metrick, 2003, "Corporate Governance and Stock Prices," Quarterly Journal of Economics (forthcoming). [20] Greenwald, B., 1986, "Adverse Selection in the Labour Market," Review of Economic Studies, 53, pp. 325-347. [21] Hermalin, B. E., 2005, "Trends in Corporate Governance," Journal of Finance, forthcoming. [22] Himmelberg, C. P., and R. G. Hubbard, 2000, "Incentive Pay and the Market for CEOs: An Analysis of Pay-for-Performance Sensitivity," mimeo, Columbia University. [23] Holmstrom, B., 1979, (cid:147)Moral Hazard and Observability,(cid:148)Bell Journal of Economics, 10, pp. 74(cid:150)91. [24] Holmstrom, B. and Milgrom, P., 1992, (cid:147)Multitask Principal Agent Analysis(cid:151)Incentive Contracts, Asset Ownership, and Job Design,(cid:148)Journal of Law Economics and Organization, 7, pp. 24(cid:150)52. [25] Holmstrom, B. and Tirole, J., 1993, (cid:147)Market Liquidity and Performance Monitoring(cid:148), Journal of Political Economy, 101, pp. 678(cid:150)709. [26] Huson, M., Parrino, R., and Starks, L., 2001, (cid:147)Internal Monitoring Mechanisms and CEO Turnover: A Long Term Perspective,(cid:148)Journal of Finance, 56, pp. 2265-2297. [27] Jensen, M.C., and W.H. Meckling, 1976, (cid:148)Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure,(cid:148)Journal of Financial Economics 3, pp.305-360. [28] Jensen, M. C., and K. J. Murphy, 1990, "Performance Pay and Top-management Incentives," Journal of Political Economy, 98, pp. 225-264. [29] Jensen, M., K. J. Murphy, and E.G. Wruck, 2004, "Remuneration: Where We(cid:146)ve Been, How We Got Here, What Are the Problems, and How to Fix Them," WP HBS 04-28. [30] Jenter, D. and F. Kanaan, 2006, "CEO Turnover and Relative Performance Evaluation," MIT WP 4594-06 [31] Kaplan, S., M. M. Klebanov, and M. Sorensen, 2007, "Which CEO Characteristics and Abilities Matter?" mimeo, University of Chicago. 24

[32] Kaplan, S. and A. Minton, 2006, "How Has CEO Turnover Changed?" mimeo, University of Chicago [33] Kaplan S. and J. Rauh, 2006, "Wall Street and Main Street: What Contributes to the Rise in the Highest Incomes?" mimeo, University of Chicago GSB. [34] Khurana, R. 2002, Searching for a Corporate Savior: The Irrational Quest for Charismatic CEOs, Princeton University Press. [35] Milbourn, T., 2003, "CEO Reputation and Stock-Based Compensation," Journal of Financial Economics, 68, pp. 233-262. [36] Murphy, K. J., 1999, (cid:148)Executive Compensation,(cid:148)in ed. Ashenfelter O. and D. Card, Handbook of Labor Economics 3, North Holland. [37] Murphy, K. J. and J. ZÆbojn(cid:237)k, 2003, (cid:147)Managerial Capital and the Market for CEOs,(cid:148) Working Paper, USC. [38] Oyer, P., 2004, "WhydoFirmsuseIncentivesthathavenoIncentiveE⁄ects?"TheJournal of Finance, 59, pp. 1619-1649. [39] Parrino, R., 1997, "CEO Turnover and Outside Succession: A Cross-sectional Analysis," Journal of Financial Economics, 46, pp. 165-197. [40] Philippon, T and A. Reshef, 2008, "Wages and Human Capital in the US Financial Industry," mimeo, NYU Stern. [41] Rajgopal, S., T. Shevlin, and V. Zamora, 2006, "CEOs(cid:146)Outside Employment Opportunities and the Lack of Relative Performance Evaluation in Compensation Contracts," Journal of Finance, 61, pp. 1813-1844. [42] Rosen, S., 1981, (cid:147)The Economics of Superstars.(cid:148)American Economic Review, 71(3), pp. 845-858. [43] Rosen, S., 1982, "Authority, Control and the Distribution of Earnings," Bell Journal of Economics, 13, pp. 311-323. [44] Rosen, S., 1992, (cid:147)Contracts and the Market for Executives,(cid:148)in Lars Werin and Hans Wijkander, eds., Contract Economics, Cambridge, MA and Oxford: Blackwell, pp.181(cid:151) 211. [45] Rosenbaum, P. and D. Rubin, 1983, (cid:147)The Central Role of the Propensity Score in Observational Studies for Causal E⁄ects,(cid:148)Biometrika 70, pp. 41-55. [46] Tervio, M., 2007, "The Di⁄erence that CEOs Make: An Assignment Model Approach," American Economic Review, forthcoming. [47] Waldman, M., 1984, "Job Assignment, Signaling, and E¢ ciency," Rand Journal of Economics, 15, pp. 255-267. 25

[48] Warner, J. B., R. L. Watts, and K. H. Wruck, 1988, "Stock Prices and Top Management Changes," Journal of Financial Economics, 20, pp. 461-492. [49] Weisbach, M. S., 1988, "Outside Directors and CEO Turnover," Journal of Financial Economics, 20, pp. 431-460. [50] Zhao, M., andK.Lehn, 2005, (cid:148)CEOTurnoverafterAcquisitions: AreBadBiddersFired?(cid:148) Journal of Finance, forthcoming. 26

Table 1: Sample Distribution by Year The sample consists of 1,655 CEO successions between 1988 and 2007 for (cid:133)rms in the (cid:133)nancial industry (SIC 6000-6999). This table presents an overview of the data set by showing the number and the frequency of internal successions in the sample. Successions are classi(cid:133)ed as internal when incoming CEOs were hired by the (cid:133)rm earlier than a year before succession, and external otherwise. Successions due to mergers and spin-o⁄s are excluded. Panel A: Sample Distribution by Year Number of Percent Firms Number of Year insiders with successions appointed successions 1988 39 23 (57.9%) 3.6% 1989 61 30 (48.8%) 5.8% 1990 91 52 (56.7%) 8.6% 1991 81 36 (44.4%) 7.5% 1992 86 50 (57.7%) 7.7% 1993 83 45 (54.2%) 4.6% 1994 96 50 (51.8%) 5.2% 1995 94 54 (57.1%) 5.1% 1996 70 38 (53.6%) 3.9% 1997 88 55 (62.5%) 5.1% 1998 105 57 (54.2%) 6.0% 1999 116 56 (48.5%) 6.4% 2000 142 65 (45.9%) 8.2% 2001 91 43 (47.5%) 5.4% 2002 87 39 (45.0%) 5.3% 2003 73 34 (46.7%) 4.4% 2004 74 32 (42.9%) 4.7% 2005 53 23 (43.2%) 3.5% 2006 57 23 (41.1%) 4.0% 2007 68 29 (42.6%) 5.2% Total 1655 832 (52.0%) 5.2% Panel B: Annual Averages by Sub-Period Number of Percent Firms Number of Period insiders with successions appointed successions 1987-97 789 431 (54.6%) 5.5% 1998-07 866 402 (46.4%) 5.4% 27

Table 2: Short-Run Cumulative Abnornal Returns around Succession Announcements This table reports short-run cumulative abnormal returns around CEO successions for (cid:133)rms in the (cid:133)nancial industry (SIC 6000-6999) during the period from 1988 to 2007. Abnormal returns are calculated using the capital asset pricing model (CAPM). The (0,+1) window of analysis is relative to actualannouncementdatesofCEOappointments(indays),wheret=0isthedayoftheannouncement. Stock returns data are from CRSP. Row [1] reports results for all sample, Row [2] restricts the sample to (cid:133)rms with 40% or more insiders on the board (upper quartile of the distribution), and Row [3] restricts the sample to years after the 1999 Gramm-Leach-Bliley Act. Robust standard errors are in parentheses. Levels of signi(cid:133)cance are denoted by , , and for statistical signi(cid:133)cance at the 1%, (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3) 5%, and 10% level, respectively. CAR [0,+1] All By Type of Succession Internal External Di⁄erence (1) (2) (3) (4) [1] All Appointments 0.008 -0.001 0.017 0.018 (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (0.004) (0.005) (0.006) (0.008) {t-stat} {2.238} Board Independence: [2] Insider Dominated Boards 0.005 -0.005 0.016 0.021 (cid:3)(cid:3) (cid:3)(cid:3) (0.005) (0.006) (0.008) (0.010) {t-stat} {1.995} Deregulation: [3] (Gramm-Leach-Bliley Act) Post 1999 0.017 0.002 0.029 0.026 (cid:3)(cid:3) (cid:3)(cid:3) (cid:3) (0.008) (0.011) (0.012) (0.014) {t-stat} {1.857} 28

Table 3: Long-Run Abnormal Stock Returns Around CEO Transitions This table reports long-run abnormal returns around CEO transitions for (cid:133)rms in the (cid:133)nancial industry (SIC 6000-6999) during the period from 1988 to 2007. Abnormal returns are estimated using calendar-time portfolio regressions. In each month t, all (cid:133)rms subject to a CEO succession within the next (prior) 36 months are included in that month(cid:146)s pre (Panel A) and post (Panel B) transition portfolio. Mean portfolio returns, rp are used to estimate abnormal returns using the t following regression: (rp rf ) = (cid:11) + (cid:12) (rm rf ) + (cid:12) SMB + (cid:12) HML + (cid:12) UMD + " ; t t 1 t t 2 t 3 t 4 t t (cid:0) (cid:0) where rf is the risk-free rate calculated using one-month Treasury-bill rates, (rm rf ) is the t t t (cid:0) market risk premium, calculated as the di⁄erence between the value-weighted return on all NYSE, AMEX,andNASDAQstocksfromCRSPlesstherisk-freerate,SMB isthereturndi⁄erencebetween t portfolios of small stocks and big stocks, HML is the return di⁄erence between portfolios of high t book-to-market stocks and low book-to-market stocks, and UMD is the return di⁄erence between t portfolios of high prior-return stocks and low prior-return stocks. The reported abnormal returns are the intercept ((cid:11)) estimated from the regression above. In each panel, Row [1] reports results for all sample, Row [2] restricts the sample to (cid:133)rms with 40% or more insiders on the board (upper quartile of the distribution), and Row [3] restricts the sample to years after the 1999 Gramm-Leach-Bliley Act. Robust standard errors are in parentheses. Levels of signi(cid:133)cance are denoted by , , and (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3) for statistical signi(cid:133)cance at the 1%, 5%, and 10% level, respectively. Panel A: Pre-transition portfolio All By Type of Succession Internal External (1) (2) (3) [1] All Appointments -0.0063 -0.0018 -0.0090 (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (0.0019) (0.0019) (0.0026) Implied 1-year abnormal return (%) -7.83 -2.18 -11.35 Board Independence: [2] Insider Dominated Boards -0.0108 -0.0013 -0.0097 (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (0.0027) (0.0026) (0.0035) Implied 1-year abnormal return (%) -13.76 -1.57 -12.28 Deregulation: [3] (Gramm-Leach-Bliley Act) Post 1999 -0.0030 -0.0014 -0.0045 (0.0032) (0.0037) (0.0042) Implied 1-year abnormal return (%) -3.66 -1.69 -5.54 29

Panel B: Post-transition portfolio All By Type of Succession Internal External (1) (2) (3) [1] All Appointments 0.0050 0.0036 0.0064 (cid:3)(cid:3) (cid:3)(cid:3) (0.0024) (0.0026) (0.0030) Implied 1-year abnormal return (%) 6.17 4.41 7.96 Board Independence: [2] Insider Dominated Boards 0.0029 0.0022 0.0074 (cid:3)(cid:3)(cid:3) (0.0025) (0.0024) (0.0021) Implied 1-year abnormal return (%) 3.54 2.67 9.25 Deregulation: [3] (Gramm-Leach-Bliley Act) Post 1999 0.0071 0.0030 0.0086 (cid:3) (cid:3)(cid:3) (0.0038) (0.0047) (0.0030) Implied 1-year abnormal return (%) 8.86 3.66 10.82 30

snoisseccuS OEC dnuora ecnamrofreP mriF :4 elbaT ot 8891 morf doirep eht gnirud )9996-0006 CIS( yrtsudni laicnan(cid:133) eht ni smr(cid:133) rof ecnamrofrep mr(cid:133) ni segnahc stroper elbat sihT eW .noisseccus OEC retfa sraey eerht ot erofeb raey eno morf detaluclac si ecnamrofrep ni egnahc ehT .snoisseccus OEC dnuora 7002 )3( dna ;)SORO( selas no nruter gnitarepo )2( ;)AORO( stessa no nruter gnitarepo )1( :serusaem ecnamrofrep eerht rof stluser troper erusaem ecnamrofrep naidem eht gnisu detsujda-yrtsudni era serusaem eseht A lenaP nI .)Q s(cid:146)niboT( stessa fo eulav koob ot tekram era slortnoc ecnamrofreP .detsujda-ecnamrofrep dna yrtsudni era serusaem eseht B lenaP nI .)CIS tigid-owt( yrtsudni tnaveler eht fo ehT .noitisnartroirpraey ehtnielbairav detsujda-yrtsudnitnavelerehtybdetrosselicedotnismr(cid:133)TATSUPMOCgnidividyb detaerc stluser stroper ]1[ woR ,lenap hcae nI .lortnoc sa desu neht si )tneve-xe( smr(cid:133) fo puorg ecnamrofrep tnaveler eht fo naidem launna dna ,)noitubirtsid eht fo elitrauq reppu( draob eht no sredisni erom ro %04 htiw smr(cid:133) ot elpmas eht stcirtser ]2[ woR ,elpmas lla rof fo sleveL .sesehtnerap ni era srorre dradnats tsuboR .tcA yelilB-hcaeL-mmarG 9991 eht retfa sraey ot elpmas eht stcirtser ]3[ woR .ylevitcepser ,level %01 dna ,%5 ,%1 eht ta ecnac(cid:133)ingis lacitsitats rof dna , , yb detoned era ecnac(cid:133)ingis (cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) serusaeM ecnamrofreP detsujdA-yrtsudnI ,snoisseccuS OEC dnuora ecnamrofreP mriF laitnere⁄iD :A lenaP )Q s(cid:146)niboT( eulaV )SORO( selas no nruter gnitarepO )AORO( stessa no nruter gnitarepO ecnamrofreP ])1-=t(-)3+=t([ noisseccuS fo epyT yB llA noisseccuS fo epyT yB llA noisseccuS fo epyT yB llA ecnere⁄iD lanretxE lanretnI ecnere⁄iD lanretxE lanretnI ecnere⁄iD lanretxE lanretnI )21( )11( )01( )9( )8( )7( )6( )5( )4( )3( )2( )1( 790.0 670.0 020.0- 810.0 750.0 940.0 800.0- 810.0 120.0 510.0 600.0- 400.0 stnemtnioppA llA ]1[ (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3) )240.0( )330.0( )620.0( )120.0( )220.0( )220.0( )900.0( )110.0( )010.0( )900.0( )500.0( )500.0( }313.2{ }175.2{ }161.2{ }tats-t{ :ecnednepednI draoB 081.0 261.0 810.0- 150.0 080.0 360.0 710.0- 610.0 530.0 320.0 210.0- 300.0 sdraoB detanimoD redisnI ]2[ (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3) (cid:3) )460.0( )350.0( )930.0( )330.0( )130.0( )130.0( )410.0( )510.0( )710.0( )310.0( )600.0( )900.0( }097.2{ }795.2{ }640.2{ }tats-t{ :noitalugereD )tcA yelilB-hcaeL-mmarG( ]3[ 731.0 060.0 770.0- 120.0- 760.0 260.0 500.0- 920.0 230.0 910.0 410.0- 300.0 9991 tsoP (cid:3)(cid:3) (cid:3)(cid:3) (cid:3) (cid:3) (cid:3)(cid:3) (cid:3) )360.0( )450.0( )730.0( )120.0( )630.0( )230.0( )510.0( )810.0( )510.0( )110.0( )900.0( )800.0( }161.2{ }248.1{ }751.2{ }tats-t{ 31

serusaeM ecnamrofreP detsujdA ecnamrofreP dna yrtsudnI ,snoisseccuS OEC dnuora ecnamrofreP mriF laitnere⁄iD :B lenaP )Q s(cid:146)niboT( eulaV )SORO( selas no nruter gnitarepO )AORO( stessa no nruter gnitarepO ecnamrofreP ])1-=t(-)3+=t([ noisseccuS fo epyT yB llA noisseccuS fo epyT yB llA noisseccuS fo epyT yB llA ecnere⁄iD lanretxE lanretnI ecnere⁄iD lanretxE lanretnI ecnere⁄iD lanretxE lanretnI )21( )11( )01( )9( )8( )7( )6( )5( )4( )3( )2( )1( 201.0 260.0 040.0- 200.0 710.0 020.0 300.0 110.0 110.0 600.0 500.0- 4000.0 stnemtnioppA llA ]1[ (cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) )260.0( )150.0( )730.0( )030.0( )800.0( )700.0( )500.0( )400.0( )600.0( )500.0( )500.0( )300.0( }656.1{ }950.2{ }137.1{ }tats-t{ :ecnednepednI draoB 991.0 871.0 220.0- 150.0 420.0 720.0 300.0 410.0 510.0 200.0 310.0- 600.0sdraoB detanimoD redisnI ]2[ (cid:3)(cid:3) (cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) )090.0( )690.0( )040.0( )440.0( )210.0( )110.0( )800.0( )600.0( )700.0( )600.0( )600.0( )500.0( }502.2{ }589.1{ }550.2{ }tats-t{ :noitalugereD )tcA yelilB-hcaeL-mmarG( ]3[ 403.0 911.0 481.0- 750.0- 720.0 620.0 200.0- 210.0 610.0 400.0 210.0- 400.0- 9991 tsoP (cid:3) (cid:3) (cid:3) (cid:3)(cid:3) (cid:3) )161.0( )421.0( )401.0( )180.0( )610.0( )110.0( )110.0( )800.0( )900.0( )500.0( )800.0( )500.0( }388.1{ }247.1{ }847.1{ }tats-t{ 32

snoisseccuS OEC dnuora snoisiceD mriF :5 elbaT 7002 ot 8891 morf doirep eht gnirud )9996-0006 CIS( yrtsudni laicnan(cid:133) eht ni smr(cid:133) rof seicilop mr(cid:133) ni segnahc stroper elbat sihT troper eW .noisseccus OEC retfa sraey eerht ot erofeb raey eno morf detaluclac si seicilop ni egnahc ehT .snoisseccus OEC dnuora stroper ]1[ woR .)ytilitalov nruter kcots latot( ksir mr(cid:133) )3( dna ;wo(cid:135) hsac )2( ;egarevel tekram )1( :seicilop mr(cid:133) eerht rof stluser ,)noitubirtsid ehtfo elitrauqreppu( draobeht no sredisni eromro %04htiw smr(cid:133)ot elpmas eht stcirtser]2[woR ,elpmasllarof stluser sleveL .sesehtnerap ni era srorre dradnats tsuboR .tcA yelilB-hcaeL-mmarG 9991 eht retfa sraey ot elpmas eht stcirtser ]3[ woR dna .ylevitcepser ,level %01 dna ,%5 ,%1 eht ta ecnac(cid:133)ingis lacitsitats rof dna , , yb detoned era ecnac(cid:133)ingis fo (cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) ksiR latoT wolF hsaC egareveL yciloP ])1-=t(-)3+=t([ noisseccuS fo epyT yB llA noisseccuS fo epyT yB llA noisseccuS fo epyT yB llA ecnere⁄iD lanretxE lanretnI ecnere⁄iD lanretxE lanretnI ecnere⁄iD lanretxE lanretnI )21( )11( )01( )9( )8( )7( )6( )5( )4( )3( )2( )1( 751.0- 432.0- 670.0- 051.0- 610.0 800.0 800.0- 100.0- 320.0- 010.0- 210.0 300.0 stnemtnioppA llA ]1[ (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3) (cid:3) )260.0( )150.0( )730.0( )130.0( )800.0( )600.0( )500.0( )400.0( )310.0( )510.0( )010.0( )900.0( }045.2{ }131.2{ }167.1{ }tats-t{ :ecnednepednI draoB 102.0- 322.0- 220.0- 011.0- 720.0 120.0 600.0- 400.0 540.0- 120.0- 420.0 600.0 sdraoB detanimoD redisnI ]2[ (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3) )290.0( )780.0( )660.0( )350.0( )210.0( )810.0( )500.0( )700.0( )420.0( )720.0( )610.0( )510.0( }881.2{ }671.2{ }678.1{ }tats-t{ :noitalugereD )tcA yelilB-hcaeL-mmarG( ]3[ 581.0- 534.0- 942.0- 243.0- 020.0 410.0 700.0- 400.0 430.0- 530.0- 200.0- 810.0- 9991 tsoP (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3) (cid:3)(cid:3) (cid:3) (cid:3) )880.0( )270.0( )250.0( )540.0( )210.0( )700.0( )010.0( )600.0( )020.0( )120.0( )610.0( )310.0( }590.2{ }537.1{ }307.1{ }tats-t{ 33

noitac(cid:133)itnedI snoisseccuS OEC dnuora ecnamrofreP mriF laitnere⁄iD :6 elbaT ot 8891 morf doirep eht gnirud )9996-0006 CIS( yrtsudni laicnan(cid:133) eht ni smr(cid:133) rof ecnamrofrep mr(cid:133) ni segnahc stroper elbat sihT .noisseccus OEC retfa sraey eerht ot erofeb raey eno morf detaluclac si ecnamrofrep ni egnahc ehT .snoisseccus OEC dnuora 7002 ;)SORO( selas no nruter gnitarepo )2( ;)AORO( stessa no nruter gnitarepo )1( :serusaem ecnamrofrep eerht rof stluser troper eW erusaem ecnamrofrep naidem eht gnisu detsujda-yrtsudni era serusaem esehT .)Q s(cid:146)niboT( stessa fo eulav koob ot tekram )3( dna OEC lanretxe dna lanretni neewteb ecnere⁄id troper ew ,erusaem ecnamrofrep hcae roF .)CIS tigid-owt( yrtsudni tnaveler eht fo ytisneporp robhgien-tseraen a dna stnemtnioppa OEC lanretni neewteb ecnere⁄id dna )"waR"( nmuloc tsr(cid:133) eht ni stnemtnioppa fo raey hcae ni enod si gnihctaM .)"elpmaS dehctaM"( nmuloc dnoces eht ni stnemtnioppa OEC lanretxe fo elpmas dehctam erocs sredisnieromro%04htiwsmr(cid:133)otelpmasehtstcirtser]2[woRdnaelpmasllarofstluserstroper]1[woR .tnemecalperhtiw ,noisseccus yb detoned era ecnac(cid:133)ingis fo sleveL .sesehtnerap ni era srorre dradnats tsuboR .)noitubirtsid eht fo elitrauq reppu( draob eht no .ylevitcepser ,level %01 dna ,%5 ,%1 eht ta ecnac(cid:133)ingis lacitsitats rof dna , , (cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) Q s(cid:146)niboT SORO AORO ecnamrofreP ])1-=t(-)3+=t([ dehctaM waR dehctaM waR dehctaM waR elpmaS elpmaS elpmaS )6( )5( )4( )3( )2( )1( 880.0 790.0 650.0 750.0 810.0 120.0 stnemtnioppA llA ]1[ (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) )340.0( )240.0( )220.0( )220.0( )800.0( )010.0( }730.2{ }313.2{ }735.2{ }175.2{ }790.2{ }161.2{ }tats-t{ :ecnednepednI draoB 191.0 081.0 370.0 080.0 140.0 530.0 sdraoB detanimoD redisnI ]2[ (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3)(cid:3) (cid:3)(cid:3) )860.0( )460.0( )330.0( )130.0( )910.0( )710.0( }297.2{ }097.2{ }022.2{ }795.2{ }151.2{ }640.2{ }tats-t{ 34