CFR Working Paper No Choosing two business degrees versus choosing one: What does it tell about mutual fund managers' investment behavior?

Transcription

1 CFR Working Paper No Choosing two business degrees versus choosing one: What does it tell about mutual fund managers' investment behavior? L. Andreu A. Pütz

2 Choosing two business degrees versus choosing one: What does it tell about mutual fund managers investment behavior? Laura Andreu and Alexander Puetz October 2015 ABSTRACT We analyze what a second business degree reveals about the investment behavior of mutual fund managers. Specifically, we compare performance, risk, and style of managers with both a CFA designation and an MBA degree to managers with only one of these qualifications. We document that the average performance level between these groups does not significantly differ. However, managers with both degrees take less risk, follow less extreme investment styles, and achieve less extreme performance outcomes. We conjecture that these differences in double degree managers investment behavior can be explained by their attitude towards their career. We rule out several alternative explanations: our results are not driven by the manager s skill, the fund family s investment policy, or by the respective education contents of the MBA and the CFA program. JEL classification: G23 Keywords: Mutual funds, Investment behavior, Manager education, MBA, CFA Laura Andreu Faculty of Economics and Business Studies, Accounting and Finance Department University of Zaragoza, Spain landreu@unizar.es Alexander Puetz (corresponding author) Department of Finance and Centre for Financial Research (CFR) University of Cologne, Albertus-Magnus-Platz, Cologne, Germany Phone: Fax: puetz@wiso.uni-koeln.de

3 1 Introduction An age-old question among those headed into the finance world is whether they need to obtain a CFA, an MBA, or both. Do you think there is a benefit to doing both? This question Bloomberg Businessweek s journalist Alison Damast asked the CFA Institute Managing Director Thomas Robinson. 1 Her question suggests: when it comes to education in the finance industry, everything revolves around having an MBA or a CFA. Among mutual fund managers, these are the most common degrees with about 74 percent of the managers having at least one of them. 2 Several academic studies so far have analyzed the distinct impact of each single degree on mutual fund performance and risk (see, e.g., Shukla and Singh 1994; Golec 1996; Chevalier and Ellison 1999a; Gottesman and Morey 2006). In this paper, we examine whether the investment behavior of mutual fund managers who decide to earn both degrees, i.e., a CFA designation additionally to an MBA degree and vice versa, differs from those who earn only one degree. Gottesman and Morey (2006) document that among managers with at least an MBA or a CFA, 55 percent have only one of these degrees, i.e., either an MBA or a CFA. This suggests that having one of those degrees is typically sufficient to work in the mutual fund industry. Nevertheless, the other 45 percent of those managers gathered both degrees. 3 Given the additional effort in terms of time and money to attain a second degree, this raises the question what it reveals about the managers if they decide to go the extra mile and earn the second degree. There are basically two motivations to gather additional education: First, people invest more in education to improve their career prospects, e.g., faster promotion or higher wages 1 See Damast (2011), URL: 2 See Gottesman and Morey (2006). In our sample there are even 83 percent of the managers with at least an MBA or a CFA. 3 The numbers from Gottesman and Morey (2006) are calculated from conditional probabilities based on the descriptive statistics given in their article. These numbers are consistent with our sample where, from those managers with at least one degree, 57 percent have one degree and 43 percent have two degrees. 2

4 (see, e.g., OECD 2014a). 4 Second, they invest in education to lower their risk of unemployment (see, e.g., Mincer 1991; Cohen et al. 1997, OECD 2014b). Both motivations suggest that managers who gather two qualifications are generally more concerned about their career than their single-qualification peers. Eventually, their investment behavior might differ. For example, Chevalier and Ellison (1999b) document that career concerns express in managers risk-taking and the conventionality of their investment style. Managers with a stronger desire to avoid termination will take lower risk levels and follow more conventional investment styles because failing with high risk and unconventional styles is more detrimental to them than failing with low risk and conventional styles. 5 On the other hand, Chevalier and Ellison (1999b) also show that unconventional investment styles and higher risk levels increase the probability of being promoted when these managers beat the market. Thus, we have two opposing hypotheses for the risk and investment style of double degree managers: If they are more concerned about promotion, they will take more risk and follow unconventional investment styles. In contrast, if they are more concerned about termination, they will take less risk and follow conventional investment styles. Furthermore, there is a third possibility: If double degree managers see their education as an insurance against termination or, alternatively, as a guarantee for promotion, their behavior should not differ from single degree managers. It is an empirical question whether double degree managers take more, less, or the same amount of risk compared to their single degree peers and whether their investment style differs. In this paper, we analyze the performance, risk, and investment style of managers with two degrees and compare it to those managers with only one degree. We find that the average performance level does not differ between both groups. However, managers with two degrees 4 However, anecdotal evidence suggests that the gain in compensation from the second degree is rather small, compared to the gain in compensation from the first degree (a gain of percent for the first degree in contrast to only about 10 percent for the second degree). See 5 This is also consistent with Scharfstein and Stein (1990). They motivate their analysis with the words of Keynes (1936): Worldly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally. 3

5 take lower risk and follow more conventional, i.e., less extreme investment styles than managers with only one degree. Consistently, they also achieve less extreme performance outcomes. To verify our conjecture that the managers career concerns are responsible for our results, we test some alternative explanations. First, we show that our results are not driven by the contents managers learn in a CFA or an MBA program. Double degree managers invest more cautiously compared to both managers with only an MBA and managers with only a CFA. Furthermore, they do not change their behavior after they earned their second degree. Second, we rule out that differences in skill between double and single degree managers drive our results. It is possible that especially low skill managers choose to earn a second degree, e.g., to compensate for a poor undergraduate degree or an MBA degree from a less prestigious school. We can show that skill, measured by managers SAT and GMAT scores, does not significantly differ between both manager groups. Furthermore, our initial results still hold after controlling for the manager s skill. Third, we examine the possibility that our results are the consequence of some unobservable fund family policy. In particular, it is possible that some families only employ managers with two degrees or they advise their managers to get the second degree. If these families also restrict their funds risk level and investment style, we would spuriously attribute this family policy to the managers investment behavior. Therefore, we control for an unobservable family policy using family fixed effects. Alternatively, we add the family s size as control variable to our regressions. Our results regarding risk and style remain the same. Finally, we perform the following robustness checks. We analyze whether our results differ between bull and bear years; we repeat our analyses at the manager-year level; and we employ a propensity score matching approach. Our results are robust to these variations. 4

6 Our paper is related to a growing literature on the influence of manager characteristics on investment behavior. As stated above, several studies have analyzed the distinct impact of single educational degrees on fund performance and risk: Shukla and Singh (1994) find that funds with a manager holding a CFA outperform those without a CFA. Golec (1996) shows that fund managers with an MBA show higher risk-adjusted performance than managers without an MBA. In contrast, Chevalier and Ellison (1999a) do not find a difference in riskadjusted performance between MBA managers and non-mba managers. Gottesman and Morey (2006) even document lower performance for managers with an MBA from a low prestige institution, but higher performance of managers with an MBA from a top school. Additionally, they find that a CFA designation or a non-mba master s degree is unrelated to mutual fund performance. Instead of proxying for ability by the quality of education, Grinblatt et al. (2012) use the IQ as a direct measure for skill and show that high-iq investors exhibit superior investment performance. Besides education and skill, also age and experience are related to fund managers behavior. Chevalier and Ellison (1999b) find that younger managers hold less unsystematic risk and have more conventional portfolios. Similarly, the results of Avery and Chevalier (1999) suggest that managers may herd early in their careers and diverge in their actions later. Ding and Wermers (2009) document that more experienced managers outperform if they manage large funds. Also the managers gender plays a role in explaining differences in investment behavior. Niessen-Ruenzi and Ruenzi (2015) show that female fund managers follow more persistent investment styles than male managers, and performances are virtually identical. We contribute to this literature as our study is the first to analyze how investment behavior of managers with both an MBA and a CFA differs from that of manager with only one of these qualifications. To our knowledge, so far, only Dincer et al. (2010) explicitly 5

7 controlled for having both an MBA and a CFA at the same time while analyzing the distinct impact of each single degree on performance. However, they do not analyze the incremental impact from having two compared to having one business degree and do not investigate differences in extremity of those managers investment behavior. Furthermore, they only study a short period of three years from The remainder of this paper is organized as follows. In Section 2, we describe the data and give an overview on the differences of fund and manager characteristics between managers with one and two business degrees. In Section 3, we analyze performance, risk, and style differences between both groups. Section 4 presents alternative explanations. Section 5 contains robustness analyses and Section 6 concludes. 2 Data 2.1 Data sources For our empirical analysis, we mainly rely on two data sources: First, we gather information on fund returns, total net assets, investment objectives, and other fund characteristics from the CRSP Survivor Bias Free Mutual Fund database which covers virtually all US open-end mutual funds. 6 Second, to collect information on fund managers characteristics, we use a set of Morningstar Principia CDs which provide information on the managers name, the date on which a manager assumed responsibility for the fund, their educational degrees, the schools a manager attained, and the job history of the manager. As the Morningstar information on manager characteristics is available from 1996 on, our sample starts in 1996 and ends in We use the Strategic Insight (SI) objective codes provided in the CRSP database to define the market segment in which a fund operates. We focus on actively managed, domestic 6 Source: CRSP, Center for Research in Security Prices. Graduate School of Business, The University of Chicago. Used with permission. All rights reserved. 6

8 equity funds and exclude bond, money market, and index funds. We exclude bond and money market funds because they are not directly comparable to equity funds. We analyze funds from the following six domestic equity fund segments: Aggressive Growth (AG), Balanced (BL), Growth and Income (GI), Income (IN), Long Term Growth (LG), and Sector Funds (SE). 7 Many funds offer multiple share classes which are listed as separate entries in the CRSP database. They usually only differ with respect to their fee structure or minimum purchase requirements. However, the different share classes of a fund are always backed by exactly the same portfolio of assets and have the same portfolio manager. Thus, to avoid multiple counting, we aggregate all share classes of the same fund. 8 To gather information on the managers characteristics, we match all funds from the CRSP database to the funds in the Morningstar database using fund ticker, fund name, and manager name. Through this, we get Morningstar s information on the managers educational degrees, e.g. whether the manager holds an MBA, a CFA, a non-business master s degree, or a PhD, the school from which a manager attained a specific degree, as well as the year in which they earned their MBA and their undergraduate s degree. 9 Furthermore, for all managers with an undergraduate degree we obtain information on the average matriculates SAT score of the institution where they earned their undergraduate degree. Similarly, we collect the GMAT score for all managers with an MBA degree. Information on SAT scores is from the website collegeapps.about.com, while information on GMAT is from the websites 7 Unfortunately, the SI classification is only available till Thus, we use an alternative classification (the Lipper objective codes) to classify funds after To get consistent segment classifications over our entire sample period, we match each Lipper objective code to a SI objective code based on the frequency with which funds of a specific Lipper objective code belong to one of the SI objective codes in those consecutive years in which the availability of the SI codes ends and the availability of Lipper begins. The resulting translation table is presented in the Appendix. 8 Through 2002 we identify the share classes of a fund by matching fund names and characteristics such as fund management structures, turnover, and fund holdings in asset classes. From 2003 on, the CRSP database reports a unique portfolio number for each fund, which is used to aggregate share classes from 2003 through Unfortunately, Morningstar does not report the year in which the managers earned their CFA designation. 7

9 mba.com, businessweek.com, and entrepreneur.com. We calculate the managers industry tenure from the year that Morningstar reports for a manager to be the first year managing a fund in the Morningstar database. As the managers age is not explicitly given in Morningstar, we compute their age from the year in which they got their college degree. To do this, we follow Chevalier and Ellison (1999b) and assume that a manager was 21 upon college graduation. Finally, to collect information on the managers gender, we follow Niessen- Ruenzi and Ruenzi (2015) and compare the managers first name to a list published by the United States Social Security Administration (SSA) that contains the most popular first names by gender for the last 10 decades. Additionally, we identify the gender of managers with ambiguous first names from several internet sources like the fund prospectus, press releases, or photographs that reveal their gender. We focus on single managed funds because Bär et al. (2011) show that team managed funds and single managed funds behave differently and it is not clear how the skills and education of single team members translate into the skills and education of a team. This allows us to cleanly analyze the impact of choosing two business degrees on managerial behavior without being influenced by the fund s management structure. Thus, we exclude fund year observations for which Morningstar reports a management team or gives multiple manager names. We also exclude fund-years in which the fund manager changes because we cannot clearly attribute the investment behavior in that year to any of the changing managers. Furthermore, for some managers Morningstar does not report any educational degree. As it is likely that these managers have some educational degree which is simply not reported in the database, we only keep those observations where the database reports at least an undergraduate degree. Next, we check our variables for outliers. For the turnover ratio, we 8

10 identify some extreme outliers and thus we set 1% of the highest turnover ratios to missing. 10 Finally, we only keep fund year observations for which 12 months of return data is available. 2.2 Sample characteristics Our final sample consists of 7,241 fund year observations which come from a total of 1,520 distinct funds. Table 1 reports summary statistics for funds total net assets (TNA), turnover ratio, expense ratio, and funds age. Please insert TABLE 1 approximately here The number of funds in our sample increases from 529 in 1996 to 637 in From 2003 the number of funds in the sample monotonically decreases which is likely because team management became more popular and several funds might have changed their management structure. 11 These funds drop out of our sample, because we only look at single managed funds. The average fund size in our sample has nearly monotonically increased from about one billion USD in 1996 to about 1.7 billion USD in 2007 with break downs around the dotcom bubble and the subprime crisis with their minima in 2002 and The turnover ratio moves around its overall mean at 90%. The funds expense ratio slightly increased from 1.30 percent in 1996 to 1.47 percent in 2003, but then declined again to 1.27 percent in The funds age slightly increased throughout the sample period from 13 to 19 years. 10 In doing so, we exclude all funds with a turnover ratio greater than 752%. 11 See Bär et al. (2011) for an overview on the development of team-managed mutual funds. 9

11 3 Results 3.1 Fund and manager characteristics The purpose of this study is to analyze what it reveals about managers investment behavior if they decide to gather a second business degree compared to those managers that gather only one. To conduct our first analyses, we group managers by the number of business degrees, i.e., exactly one degree (MBA Xor CFA) or both degrees (MBA and CFA). We assign managers to one of these groups based on all degrees that are reported by Morningstar in any year of the sample period. Thus, also managers who start with one degree, but earn their second degree during the sample period are assigned to the group of managers with two business degrees. This is based on the intuition that we also want to capture the managers' personal attitudes, i.e., their career concerns, which they reveal with the educational degrees they attain over time. Since we assume the manager s personal attitudes to be time-invariant, we can infer them from the manager s decision for a second degree, even if the degree is earned in the future. To get a first sense about differences in the manager and fund characteristics, Table 2 reports summary statistics for both groups. Please insert TABLE 2 approximately here From all fund-year observations in our sample, about 36 percent are managed by managers with both degrees and 47 percent by managers with exactly one degree. Thus, 83 percent have at least one degree. Double degree managers seem to manage larger funds, but the difference is not statistically significant. The funds of these managers have lower expense ratios, lower turnover ratios, and are on average two years older than those funds of managers with only one business degree. As we would expect, managers with two degrees are on average older and have a longer tenure in the fund industry. The fraction of female managers 10

12 is significantly lower among managers with two business degrees. Furthermore, managers with two business degrees are significantly less likely to also have any other master s degree, but are significantly more likely to additionally have a PhD degree. 3.2 Performance We start our empirical investigation by analyzing the impact of a second business degree on fund performance. First, we examine whether having two business degrees is related to better average performance. Second, we study whether performance extremity, i.e. the deviation from the market segment s average, differs between double and single degree managers. We use four different measures for a fund s average yearly performance: return, Jensen (1968) s one-factor alpha, Fama and French (1993) s three-factor alpha, and Carhart (1997) s four-factor alpha. 12 To quantify performance extremity we follow the approach of Bär, Kempf, and Ruenzi (2011) and calculate an extremity measure EM ( P ) in each year: EM ( P) it, 1 N k t P P it, kt, N k t j 1 P P j, t k, t (1) where P stands for the respective performance measure. Performance extremity EM ( P ) it, represents the absolute deviation of a fund i s performance from the average performance of all funds in the same market segment k, divided by the average absolute deviation of all funds in the segment. k N t represents the number of funds in a specific market 12 The latter three alpha measures are determined based on a yearly estimation of the respective factor models. The factor-mimicking portfolio returns for the respective factors and the risk-free rate were taken from Kenneth French s website, 11

13 segment k in a given year t. We calculate the extremity measure for each of the four performance measures. To examine how double degree managers differ from single degree managers with respect to performance, we regress the above measures on the dummy variable MBA and CFA that equals one if both business degrees are reported for a specific manager and zero otherwise. To directly compare double degree managers to single degree managers, we also add the dummy No MBA or CFA to our regressions. No MBA or CFA equals one if a manager has neither an MBA nor a CFA. Thus, our reference group consists of managers who have exactly one business degree and we can interpret the coefficient on the MBA and CFA dummy as the difference in performance between double and single degree managers. We add further control variables to our regressions: As shown in Table 2, managers with two business degrees typically manage funds with different characteristics compared to managers with only one business degree. Thus, we add the logarithm of a fund s lagged size, the fund s lagged expense ratio, its lagged turnover ratio, and the fund s age as additional explanatory variables. Second, we take into account that managers with two business degrees typically exhibit different personal characteristics. In particular, we also control for a manager s other educational degrees by adding dummy variables for any other master s degree and a PhD. Next, we add a dummy that equals one for a female manager and zero otherwise. Furthermore, we add the managers industry tenure and the managers age to the regressions. The managers age is calculated from the year in which they earned their undergraduate degree. If this year is missing, we follow Gottesman and Morey (2006) and set the manager s age to zero while adding a dummy to the regression which equals one for observations with missing manager age. Additionally, to control for any unobservable time or segment effects that could equally affect all funds in a given year or a particular market segment, respectively, we 12

14 include time- and segment-fixed effects in the performance level regressions. Since performance extremity measures are already adjusted for time and segment, we do not include time- and segment-fixed effects in the extremity regressions. All analyses are done at the fund-year level using pooled OLS. To take into account the panel structure of our data, we cluster the standard errors in our regressions by fund following Rogers (1993). Results are presented in Table 3. Please insert TABLE 3 approximately here The results suggest that there is no significant difference in average performance between double degree managers and single degree managers. In the first four columns, none of the coefficients of MBA and CFA is significantly different from zero. However, double degree managers show significantly less extreme performance outcomes. In the last four columns, the coefficients of MBA and CFA are significantly negative at the 1%- and 5%- level. Regarding the control variables, the coefficients typically show a lower average performance for larger funds and funds with higher turnover ratio which both is consistent with previous literature. 13 Furthermore, larger funds realize less extreme performance outcomes, whereas funds with a higher expense ratio and funds with a higher turnover ratio exhibit more performance extremity. Female managers and younger managers show significantly less extreme performance outcomes. To get a better intuition of the results, in Figure 1 we show the fraction of double degree managers in five performance quintiles. Please insert FIGURE 1 approximately here 13 See, e.g., Chen et al. (2004), Berk and Green (2004), or Carhart (1997). 13

15 In each year and for each market segment, funds of which managers have at least one business degree are sorted into performance quintiles. Quintile 1 represents the lowest performance and Quintile 5 represents the highest performance. For all four performance measures, the fraction of double degree managers in the extreme quintiles is particularly low compared to the fraction in the middle quintiles. Thus, single degree managers more often achieve performance outcomes in the extreme quintiles while double degree managers more often achieve performance outcomes in the middle quintiles. 3.3 Risk and style In the next set of analyses, we examine the differences in risk-taking behavior and investment style. Having established that managers with two degrees achieve less extreme performance outcomes, we conjecture that these managers also take lower risk levels and follow more conventional investment styles, i.e., they take less extreme style bets. We measure a fund s risk in four ways: total risk, systematic risk, unsystematic risk, and tracking error. To determine a fund s total risk, we calculate its volatility in each year as the annualized standard deviation of its fund s monthly returns. To measure the fund s systematic and unsystematic risk, we first estimate the Jensen (1968) one-factor model for each fund in each year. We then use the fund s beta as measure for systematic risk and compute its unsystematic risk by the standard deviation of the regression s residuals. The tracking error is defined as the standard deviation of the difference between the fund return and its benchmark index return as in Petajisto (2013). 14 To measure the extremity of a fund manager s investment style, we first estimate Carhart (1997) s four-factor model for each fund in each year. The beta-exposures to the factors of this model (the size factor, SMB, the value factor, HML, and the momentum factor, 14 The data for tracking error were taken from Antti Petajisto s website, 14

16 MOM) represent the fund s investment style. Then, following Bär, Kempf, and Ruenzi (2011), we construct three style-extremity measures, one for each style, as: EM ( S) it, 1 N k t S S it, kt, k Nt S S 1 j, t j k, t (2) where S represents the investment style analyzed (SMB, HML, and MOM, respectively) and k N t gives the number of funds in a specific market segment k in a given year t. EM( S ) it, shows high values for funds which strongly deviate in their exposure to a S specific style ( it, ) from the average exposure of their market segment ( S ) in absolute kt, terms. Thus, a significantly higher or a significantly lower factor loading as compared to the market segment s average will result in a large extremity measure. To normalize the extremity measure, we divide it by the average style deviation in the corresponding market segment and respective year. This normalization makes our style extremity measure comparable across styles, segments, and time. Additionally, we compute an average style extremity measure for each fund across the three investment styles as: 1 EM( style) EM ( SMB) EM( HML ) EM ( MOM ) 3 it, it, it, it, We then regress our risk and style extremity measures on the MBA and CFA dummy and the control variables as in the previous section. Results are reported in Table 4. Please insert TABLE 4 approximately here Managers with two business degrees choose significantly lower risk levels and follow less extreme investment styles than managers with one business degree. The coefficient of MBA and CFA is significantly negative for all types of risk and all styles, typically at the 1%- significance level. 15

17 Regarding the control variables, larger funds show higher total risk and higher systematic risk, but lower unsystematic risk, lower tracking error, and less extreme style exposures. Funds with higher expense ratio and higher turnover ratio show higher risk levels and more style extremity. Regarding the managers personal characteristics, female managers consistently show lower risk and less extreme styles. Furthermore, younger managers take less unsystematic risk and follow more conventional styles, which is consistent with Chevalier and Ellison (1999b). To visualize the risk- and style-results, in Figures 2 and 3 we again show the fraction of double degree managers in five quintiles, this time sorted by the risk and style variables, respectively. Please insert FIGURE 2 approximately here Please insert FIGURE 3 approximately here In case of risk in Figure 2, Quintile 1 represents the lowest risk level while Quintile 5 represents the highest risk level. Typically, the fraction of double degree managers decreases with increasing risk quintiles. Regarding style exposure in Figure 3, Quintiles 1 and 5 represent opposing styles. In particular, for the SMB factor, Quintile 1 stands for a low SMBbeta, i.e., a high exposure to large companies while Quintile 5 stands for a high SMB-beta, i.e., a high exposure to small companies. For the SMB factor, the lowest fraction of double degree managers is in the most extreme quintiles. For the HML factor, the lowest fraction of double degree managers is in the lowest HML beta quintile, i.e., double degree managers avoid growth stocks, but not value stocks. Furthermore, double degree managers mostly avoid extreme momentum stocks, but also contrarian stocks, as can be seen from the quintiles of the momentum beta. 16

18 Overall, our results suggest that managers with both an MBA and a CFA are more concerned about termination than managers with exactly one of these qualifications. They take significantly lower risk levels, follow more conventional investment styles and, consequently, achieve less extreme performance. This is consistent with our conjecture that there is a self-selection among double degree managers, i.e., only managers with stronger career concerns decide to get both degrees. To support our conjecture, in the following section we rule out several alternative explanations. 4 Alternative explanations 4.1 Contents of education An alternative explanation is that the managers learn something during their education which lets them take lower risk and follow more conventional investment styles. Thus, in this section we differentiate between self-selection of managers that are generally more focused on their career and the contents of the MBA and CFA education using two different tests. First, we compare double degree managers separately to managers with only an MBA and to managers with only a CFA, respectively. Second, we compare investment behavior of the same manager before and after the manager attained the second qualification Specific degree If our results were driven by the contents that managers learn during their education, only one specific degree should drive the results. For example, if the CFA program teaches managers to take lower risk, but not the MBA program, we could find lower risk of double degree managers compared to single degree managers which was solely driven by the difference between double degree managers and managers with only an MBA. To the 17

19 contrary, if both programs teach to take lower risk, we would not expect to see a difference between double degree managers and single degree managers because all of them learned to take lower risk. 15 It follows that if managers learn to take lower risk during their education, there should be exactly one degree which is responsible for our results. Consequently, if there is no single degree driving the result, the managers personal attitudes, i.e., their career concerns should be responsible for the investment behavior of double degree managers. To test this alternative explanation, in the following analysis we compare double degree managers separately to managers with an MBA and a CFA, respectively. In particular, we use two new regression specifications. In the first we add a dummy CFA only and in the second we add a dummy MBA only as control variable. Hence, in the first regression specification, our new reference group consists of managers with only an MBA and in the second specification our reference group consists of managers with only a CFA. The MBA and CFA dummy represents the difference to the respective reference group. Results are reported in Table 5. Panel A shows results on performance extremity, Panel B on risk level, and Panel C on style extremity. 16 Please insert TABLE 5 approximately here The results show that our findings are not driven by a specific degree. The MBA and CFA dummy has a significantly negative impact on performance extremity, risk, and style extremity in 21 out of 24 cases. The results do not depend on whether single degree managers have an MBA or a CFA. Moreover, we can see that there is no difference between managers with only an MBA degree and managers with only a CFA degree. Nearly none of the 15 This is based on the assumption that both degrees have an identical impact on risk that does not sum up. For example, if both MBA managers and CFA managers learn to take less risk and reduce it by 1%, a manager with both degrees will still reduce risk by 1%. Theoretically, the impact of both degrees could sum up so that the manager with both degrees reduces risk by 2%. However, we do not think that this is plausible and thus discard this interpretation. 16 Since for the average performance level we did not find any differences, we do not report results for average performance anymore. Furthermore, for the sake of brevity, we resign to report the coefficients for the control variables. Results are available from the authors upon request. 18

20 coefficients of MBA only and CFA only is significantly different from zero, with the only exception for momentum extremity Investment behavior before and after second degree In the second approach, we analyze whether managers change their behavior after they got their second degree (while managing the same fund). In total there are 21 manager-fund combinations where the managers got their second degree during our sample period. For each of these 21 observations, we calculate the mean of the performance, risk and style variables across the years before and after the manager got the second degree, respectively. 17 Then we compare the mean of the respective variable before the second degree to the mean of the variable after the second degree using a paired t-test. Results are reported in Table 6. Please insert TABLE 6 approximately here None of the variables is significantly smaller in the years after the manager attained the second degree compared to the years before the second degree. Thus, the above results suggest that the observed differences in performance extremity, risk-taking, and style extremity are not due to the education contents of the MBA and CFA programs but are most likely a result of managers with stronger career concerns choosing to gather both degrees. 4.2 Manager skill Another possible alternative explanation could be that managers with both degrees differ from managers with one degree with respect to their skill. 18 For example, managers 17 Since Morningstar does not report the year in which the managers earned their CFA designation, we proxy for this year with the first year the CFA designation is mentioned for this manager in the database. 18 We did not control for skill in our regressions in Section 3, because we use two different specifications to measure skill and we have several missing observations for our skill measures. Thus, we decided to present the skill control separately from the other manager characteristics. 19

21 might decide to get a second degree to compensate for either a weak undergraduate degree or a weak MBA degree. As these managers know about their low skill, they might try to avoid deviations from the crowd and invest close to the benchmark as well as follow conventional investment styles. To test this alternative explanation, we first check whether double degree managers differ from single degree managers with respect to their skill. We use two different measures for skill. First, we use the average matriculates SAT score of the school that the managers got their undergraduate degree from. Second, for all managers with at least an MBA, we use the average matriculates GMAT score of their business school. Thus, the first specification covers all managers for which we have information on the undergraduate school, the second specification covers all managers with at least an MBA and information on the respective business school. As in Table 2, we simply compare the means of GMAT and SAT between double degree managers and single degree managers. Results are reported in Table 7. Please insert TABLE 7 approximately here The results show that GMAT is significantly higher for double degree managers while there is no significant difference for SAT scores. 19 Thus, it is not likely that double degree managers take lower risk and follow more conventional investment styles due to lower skill. Nevertheless, we add the managers skill as additional control variable to our regressions. In the first specification, we control for the SAT score, in the second specification, we control for the GMAT score (divided by 100, respectively). Results are reported in Table 8. Panel A shows results on performance extremity, Panel B on risk level, and Panel C on style extremity. Please insert TABLE 8 approximately here 19 The comparison in Table 7 is based on fund-year observations. If we calculate the difference directly from the manager level, we do not find a significant difference, neither in GMAT nor in SAT scores. 20

22 Managers with both an MBA and a CFA still show less extreme performance, lower risk, and less extreme investment styles than their single degree peers. For performance extremity, the coefficients of MBA and CFA are significantly negative in six out of eight cases. For risk and style extremity the coefficient is significantly negative in all cases. The coefficients of the skill variables suggest that skill is typically unrelated to risk, style extremity, and performance extremity. 4.3 Family effects Next, we examine the possibility that the decision to get both degrees is driven by the fund family. In particular, it is possible that particular families only employ managers with both degrees or they advise their managers to get the second degree. If these families also restrict the risk and investment style of their managers, we might spuriously attribute this to the second business degree. In our sample, there are 440 families that employ managers with at least and MBA or a CFA. From these families, 95 employ only double degree managers, 167 employ only single degree managers, and 178 employ both. Thus, the relatively large number of families with only one kind of manager suggests that some family characteristics might play a role. Our first conjecture is that double and single degree managers could be employed by families of different size. To examine this possibility, we compare the average number of funds per family as well as the family s assets under management between both manager groups. Results are reported in Table 9. Please insert TABLE 9 approximately here The results show that the funds of double degree managers belong to smaller families on average. Both the number of funds in the family as well as the assets under management is 21

23 significantly smaller. 20 To preclude that our results are driven by the size of the fund family, we add the number of funds in the family as an additional control to our regressions. In a second specification, we replace the number of funds in the family by the logarithm of the family s assets under management. In a third specification, to capture any unobservable but time-invariant heterogeneity in fund families as, e.g., different employment policies, we add family fixed effects to our regressions. Results are reported in Table 10. Panel A shows results on performance extremity, Panel B on risk level, and Panel C on style extremity. Please insert TABLE 10 approximately here Overall, our results do not change. Regarding risk and style in Panel B and C the coefficient on MBA and CFA is significantly negative in 23 out of 24 cases. Regarding performance extremity in Panel A, results get weaker only if we control for family fixed effects. This might be an indication that part of the lower performance extremity is related to some time-invariant family fixed effects. Our control variables suggest that larger families show less performance extremity, lower risk and less style extremity. 5 Robustness In this section we test for the robustness of our results. 21 In Section 5.1, we analyze whether our results differ between bull and bear years. In Section 5.2, we use a different level for the observations in our regressions, i.e., manager-year observations instead of fund-year observations. Finally, in Section 5.3, we employ a propensity score matching approach. 20 In unreported tests, we find that this result is mainly driven by a lower fraction of double degree managers in extremely large families. Comparing the median shows that funds of double degree managers more often belong to large families. 21 Again, for the sake of brevity, we do not report coefficients for control variables and do not report results for the average performance level. In unreported analyses, we ran all of our robustness tests also for the average performance level. The performance difference remains insignificant in all of these tests. 22

24 5.1 Bull versus bear years Our sample covers several bear years which might have an impact on our results. Therefore, we add a bear year dummy to our regressions and also interact this dummy with the MBA and CFA dummy. The bear year dummy equals one for years with a negative market return (according to the CRSP market factor calculated by Kenneth French) and zero otherwise. In particular, we declare the years 2000, 2001, 2002, and 2008 as bear years. The other years of our sample are declared as bull years. Results are reported in Table 11. Panel A shows results on performance extremity, Panel B on risk level, and Panel C on style extremity. Please insert TABLE 11 approximately here Our results support the idea that double degree managers are generally more riskaverse. The coefficient of the MBA and CFA dummy is significantly negative in 9 out of 12 specifications. The significantly negative coefficient of the interaction term further supports the conjecture that double degree managers are more concerned about not losing their job: they take even lower risk in bear years in which managers face a higher risk to be laid off (see, e.g., Kempf et al. 2009). 5.2 Manager-year level Next, we repeat our analyses at the manager-year level. In our analyses so far, we used fund-year observations, assuming that managers make their decisions at the fund level and, consequently, could make different decisions for different funds. However, since several managers manage more than one fund at the same time, in a specific year multiple observations belong to the same fund manager. Thus, in the following analysis, for each variable we calculate the mean across all funds of the same manager in a specific year and re- 23

25 run the regressions. 22 Because we have a manager-year panel instead of a fund-year panel now, we cluster the standard errors by manager. Results are reported in Table 12. Please insert TABLE 12 approximately here Our results remain constant. The coefficient for MBA and CFA is significantly negative in all but one specification. Only in the tracking error analysis, the coefficient gets insignificant. 5.3 Propensity score matching Finally, we employ a propensity score matching approach to verify our prior results. To do this, we only keep managers with at least one business degree in our sample. Thus, we are left with two groups: one group consists of single degree managers, the other of double degree managers. To each observation of one group we match exactly one similar observation from the respective other group. Similarity between observations is based on the propensity score, i.e., each observation is matched to the one observation from the other group whose propensity score is closest. 23 The propensity score is the predicted probability to belong to the double degree group. To predict the propensity score, we use a logit regression with the control variables from Section 3 as independent variables. 24 Finally, for each variable of interest, we calculate the average difference between the matched pairs. Results are reported in Table 13. Please insert TABLE 13 approximately here 22 Since the risk variables are not yet adjusted for segment specific effects, we subtract the segment average from these variables before calculating the mean per manager. 23 We ensure that the propensity score between two observations differs by less than In unreported tests, we also add the variables from the alternative explanations in Section 4 to estimate the propensity score. Our results remain qualitatively the same. 24

26 The results confirm our previous findings. Those managers with both degrees achieve significantly less extreme performance results, take lower risk, and follow less extreme investment styles compared to similar managers with similar funds who do not choose to gather both degrees. 6 Conclusion The impact of education on fund managers investment behavior has been examined in several academic studies. These studies usually focused on the distinct impact of an MBA degree and a CFA designation as these are the most common qualifications among fund managers. One aspect that has yet been neglected in these studies is the question whether those managers who decide to get both degrees differ from those who only get one degree. In this paper, we compare managers who get both an MBA degree and a CFA designation at the same time to managers which only choose one of these qualifications. We document several new findings: First, the average performance does not significantly differ between the two groups. Second, managers who get both degrees show less extreme performance outcomes than managers with only one degree. Third, managers with both degrees show significantly lower risk levels and less extreme investment styles compared to managers with exactly one business degree. We conjecture that the decision to get both an MBA and a CFA reveals the personal attitudes of a mutual fund manager, i.e.., stronger career concerns. We rule out several alternative explanations. Our results are not driven by fund characteristics such as size, expense ratio, turnover ratio, or fund age. Similarly, manager characteristics like the managers age, their tenure, their gender, or the managers skill do not explain our results. 25

27 Furthermore, we rule out that the contents learned from a specific degree or family specific effects are responsible for the observed behavior of double degree managers. Overall, our results contribute to a growing strand of research on the impact of manager characteristics on investment behavior. We show that managers who get both an MBA as well as a CFA show less extreme investment outcomes which are desirable from an investor s point of view. 26

28 Appendix: Matching of market segments This table contains the rules according to which Strategic Insights segment classifications are matched with the Lipper segment classifications. Market Segment SI Code Name SI Code Lipper Code Lipper Code Name AG Equity USA Aggressive Growth AGG CA Equity USA Small Companies SCG SG Capital Appreciation Funds Small-Cap Funds Micro-Cap Funds MR BL Asset Allocation USA Balanced BAL B Balanced Funds GI Equity USA Growth & Income GRI GI Growth and Income Funds IN Equity USA Income & Growth ING EI Equity Income Funds LG Equity USA Growth GRO G Growth Funds SE Equity USA Environmental Equity USA Financial Sector Equity USA Health Equity Natural Resources & Energy Equity USA Real Estate Equity USA Misc Sectors Equity USA Technology Equity USA Utilities ENV FIN HLT NTR RLE SEC TEC UTI FS H NR RE S TK TL UT Financial Services Funds Health/Biotechnology Funds Natural Resources Funds Real Estate Funds Specialty/Miscellaneous Funds Science & Technology Funds Telecommunication Funds Utility Funds 27