a tale of two types: generalists vs. specialists in mutual funds

A Tale of Two Types: Generalists vs.Specialists in Mutual Funds Asset

Management

Rafael Zambrana∗ and Fernando Zapatero†

October 31, 2015

Abstract

We study the criteria that lead mutual funds to appoint specialists -managers whorun funds with a single investment style– or generalists -managers who run severalfunds with different investment styles. We identify managers with either stock-picking or market-timing ability. Managers who display stock-picking ability aremore likely to be specialists and managers with market-timing ability are morelikely to be generalists. In addition, we find that such assignments are optimal sincestock-pickers earn higher returns than other managers as specialists, and similarlymarket-timers as generalists. Finally, according to this optimal human capitalallocation, specialists with timing ability are more likely to switch to generalists.

Keywords : Mutual Fund, Asset Management, Human Capital, Portfolio Manager, Spe-cialist, Generalist.

JEL classification: G20, G23, J24, M51.

∗Nova School of Business and Economics, Lisboa, Portugal. E-mail: [email protected]†FBE, Marshall School of Business, USC. E-mail: [email protected]

1. Introduction

We show that it is optimal to assign portfolio managers with market-timing ability

to generalist responsibilities, and managers with stock-picking ability to specialist re-

sponsibilities. Generalists are managers who run one or several funds comprising several

investment objectives, while specialists focus on just one investment objective.

One of the most controversial subjects in financial economics is the ability of asset

managers (or lack thereof) to achieve returns higher than the market on a risk-adjusted

basis. Authors going back to Jensen (1968) and before argue that actively managed

funds do not achieve higher returns after fees than passively managed funds. However,

a good part of the literature, especially more recently, argues that there is such a thing

as portfolio management ability, and it helps explains flows of funds –for example, the

influential work of Berk and Green (2004). In a step further, Kacperczyk, Nieuwerburgh

and Veldkamp (2014) study two different types of managerial ability, stock-picking and

market timing, and show that skilled portfolio managers display stock-picking ability in

economic expansions and market-timing ability in recessions.

Independently of this debate, we observe that within the asset management industry

there are different organizational structures. In particular, mutual fund families histor-

ically have assigned asset managers in what appears to be somehow different functions.

In particular, some asset managers run funds with a single investment objective1 while

other managers combine different objectives under their supervision. We will call spe-

cialists the fund managers who run either just one fund or several funds with the same

investment objective, and generalists the managers in charge of several funds with more

than one investment objective among them. There is already a literature that consid-

ers this distinction between specialists and generalists, but among CEOs. Murphy and

Zabojnık (2004) document that generalist skills, i.e., transferable across industries, have

become more important for CEOs. More recently, Custodio, Ferreira and Matos (2013),

show that generalist CEOs are paid a premium over specialist CEOs.

In this paper, using a version of one the standard methodologies, we identify portfolio

managers who have either of the two abilities, market timing or stock selection –many

seem to have neither and very few both– and study whether it is optimal for the manage-

ment company to deploy them as specialists or as generalists, depending on their type

of ability. We conjecture that managers with market-timing ability are more suited to

work as generalists, while those with stock-piking ability are a better fit for specialist

assignments. In addition to verifying if our conjecture is correct, we are interested in

studying whether funds assign managers according to this criterion and, if that is the

case, we want to quantify the effect.

1There are different possible classifications of investment objectives. To avoid possible selection problemsour study is based on the classification established by the SEC. All mutual funds must declare theirobjective according to this classification in the NSAR form they are legally required to file semi-anually.

1

To assess whether managers have stock picking or market timing ability we use the

Treynor-Mazuy (1966) market-timing model, augmented with multi-risk factors.2 We

find more managers with market timing ability among the set of generalists and more

managers with stock picking ability among specialists. Of course, market-timing ability

and stock-pricing ability are both valuable and we find that managers with either of

these two skills outperform managers that lack them, regardless of whether they are

specialists or generalists. However, the difference in performance is significantly higher

-both in economic and statistical terms– when market timers are generalists and when

stock pickers are specialists. We also find that companies are more likely to re-assign a

market-timer working as specialist to a generalist position than a non-market-timer. In

addition, management companies that assign market-timers to generalist roles outperform

other companies.

We conjecture that pickers, who are meant to perform fundamental analysis, narrow

down their focus into segments in which they have expertise, whereas timers possess a

more general view of the market and benefit from a wider access to information that

allows a better allocation among different security classes.

We explore whether there are individual characteristics associated with each type of

ability. We find that generalist timers are more likely to have a PhD and/or quantitative

background, while specialist pickers are more likely to be MBAs with business related

studies. Gottesman and Morey (2006) study the effect of GMAT and school ranking on

performance. Our work also contributes to the literature on the organization and person-

nel decisions in mutual funds, especially the work on human capital assignment strategies

in mutual funds. Evans (2009, 2010) argues that companies use measures of risk-adjusted

performance to promote or demote their managers. Massa, Reuter and Zitzewitz (2010)

study the trade-offs between publicizing the names of their fund managers and keeping

them anonymous. Fang, Kemp and Trapp (2014) document that the most skilled man-

agers are assigned to market segments that are less efficient where ability has a higher

expected payoff.

Overall, our findings are consistent with the idea that there is a certain degree of

efficiency in the mutual funds industry. In particular, we find that: (i) mutual funds

that assign managers depending on their type of skill achieve higher performance; (ii)

therefore, there seems to be an optimal strategy that consists in assigning managers

with timing ability to generalists positions and pickers to specialists positions; (iii) many

management companies follow this strategy.

The paper is organized as follows. First we describe the data. Then we introduce the

2Using novel measures of ability, a number of papers argue that some fund managers are better thanothers (Daniel, Grinblatt, Titman, and Wermers (1997), Cohen, Coval, and Pastor (2005), Bollen andBusse (2005), Kacperczyk, Sialm and Zheng (2005), Kacperczyk and Seru (2007), Cremers and Petajisto(2009), Baker, Litov, Wachter, and Wurgler (2010), Berk and van Binsbergen (2012), Koijen (2012),Kacperczyk, Nieuwerburgh and Veldkamp (2014), Ferson and Mo (2015)).

2

notions of timers and pickers, as well as the functions of generalist and specialist, and

identify fund managers in our database accordingly. In the following section we present

our main results. Section 5 provides a number of robustness tests. We close the paper

with some conclusions.

2. Data Description

We use three sources. First, the CRSP Survivorship-bias free Mutual Funds Database.

It provides names of the money managers, funds returns, total net assets, funds inceptions,

turnover, expenses, and other fund and family characteristics. Since it is not clear how

the skill of the team members translates into the skill of a team and our focus is on the

ability and specific role of each individual manager, we restrict our dataset to funds run

by a single manager, as opposed to a team.3 We filter manager names manually, since

in some cases they appear under their middle name, a shortened first name, or simply

by their family name. We manually correct manager names with different spellings and

code them with a unique identifier.

Next, we merge this information with Morningstar Direct. This database provides

comprehensive information about both professional and academic backgrounds of the

portfolio managers. To merge them we use text matching and we check manually those

unmatched. We also examine managers’ websites and web-search for managers’ resumes

when necessary.

We exclude index funds, funds with less than $5 million in assets under management,

and funds in which the observation date is prior to the inception date.4 The CRSP

database has information about multiple share classes issued by a particular fund. These

classes have the same underlying portfolio and the main difference among them is the

fee structure. Thus, for mutual funds with different share classes, we aggregate all the

observations from different classes, grouping them at the fund level.5

Third, we use the NSAR forms required by the SEC to be filled by all U.S. mutual

funds and other regulated investment management companies. Mutual funds file this

form every six months. NSAR filings provide a substantial amount of information about

the Management Company, advisory arrangements, fund investment objectives, and fund

compensation characteristics.6 Although certain funds file reports starting in 1993, the

3Some new research focuses on individual managed funds (i.e Fang, Kempf and Trapp (2014), Kempf,Manconi and Spalt (2014).

4Some papers discuss the possible existence of an incubation bias (Evans, 2010).5We group data by observation at the fund level, following the literature (i.e., Nanda, Narayan andWarther (2000) or Gaspar, Massa and Matos (2006)). We aggregate returns, turnover and expensesweighting each class by their total net assets (TNA) where the fund TNA is the sum of TNA over allclasses. For the qualitative attributes of the funds such as age, names or styles, we choose that of theoldest among all classes.

6A key variable we need for our analysis is investment objective of the fund. CRSP provides different

3

data appear to be more reliable for all funds after mandatory disclosure begins in 1996.

We merge NSAR filings with CRSP by text matching and check manually. To mitigate

any possible selection bias, our time series starts in 1996. Our final dataset contains

monthly-fund observations, from 3,005 U.S. open-ended domestic equity, 2,832 fixed in-

come, 349 balanced and 897 international mutual funds. This corresponds to a total of

521 management companies and 4,625 portfolio managers from 1996 to 2011.

3. Functions and Types of Portfolio Managers

Our primary objective is to study if mutual funds allocate portfolio managers to

different functions depending on their skills. In particular, we focus on two different

functions, generalists and specialists, and two different abilities, stock-picking and market-

timing. Generalists are managers that during a given period manage funds with more

than one investment objective –which we will proxy by the style reported by the fund; we

discuss this later. Specialists either manage just one fund or manage funds with the same

investment style. With respect to the abilities, we call managers who have extraordinary

ability at picking stocks “pickers” and managers with extraordinary ability at timing

the market “timers.” Of course, some managers are neither. Pickers and timers are the

types, as opposed to the functions. We will show evidence that pickers perform better

as specialists and timers perform better as generalists, and mutual funds improve their

performance when they allocate managers accordingly.

3.1. Specialists and Generalists

Table 1 describes our sample. Portfolios are classified attending to 9 different in-

vestment objectives as defined in the NSAR filings and included in the fund prospectus

(capital appreciation, growth, income, total return, government short-term debt, govern-

ment long-term debt, corporate debt, balance and international stocks).7 While equity

funds seem to be more concentrated on capital appreciation and growth objectives, the

most frequent fixed-income funds invest in government long-term debt, followed by funds

that invest in government short-term. For each investment objective, the number of funds

in our sample seems to follow a similar pattern of growth, increasing until 2003-2004 and

decreasing afterwards. Since we focus on funds managed by an individual portfolio man-

ager, the recent decrease in the number of funds in our sample is the result of the new

trend of mutual funds managed by a team rather than a single portfolio manager.8

classifications of fund style. However, we believe that NSAR filings are more reliable since they providethe actual investment objective described in the prospectus. In the appendix, we describe in detail theobjectives included in NSAR filings.

7A full description of these investment objectives is in the Appendix.8Bliss, Potter and Schwarz (2008) and Bar, Kempf and Ruenzi (2011) study the growth in team-managedfunds.

4

[Insert Table 1 here]

We also provide more information about our sample in Table 2. In particular, we

report the number of funds run by a generalist, the number of families that have funds

run by generalists, and the number of generalists, compared to the totals (generalists plus

specialists) in each category. The total number of funds grew to over 2000 funds by 2004,

and subsequently dropped to under 1500 by 2010. We observe a similar pattern on the

number of funds managed by generalist managers; it reaches 561 in 2002, is above 400

until 2004, and decreases to below 300 by 2010. The total number of managers in our

sample starts at 1176 and ends at 657, with a maximum of 1390 on 2000; meanwhile, the

number of generalists starts at 133 and finishes at 60, with a maximum of 175. Finally,

the number of management companies starts at 261 in 1996 and ends at 197 in 2011.

Out of them, 89 in 1996 and 44 in 2011, were offering funds run by generalists.


In Table 3, we present characteristics and differences between generalists and spe-

cialists, as well as between the funds they run and the fund families to which they are

affiliated. In Panel A, we show the average characteristics of funds managed by special-

ists and generalists, as well as the magnitude and significance of their differences. On

average, generalists run funds that are smaller, younger and cheaper, with higher flows

and turnover, and with similar cumulative past returns. Panel B shows that smaller

management companies (less assets, fewer funds and fewer managers) are more likely

to employ generalist managers. Specialist managers are more likely to be working for

companies that offer their management services to other firms (sub-advisors). This is

consistent with the literature on outsourcing portfolio management decisions that find

that sub-advising contracts allow fund families to gain market share by partnering with

specialized external management firms (Cashman and Deli, 2009; Moreno, Rodriguez and

Zambrana, 2015). Panel C summarizes the mean and differences between specialist and

generalist characteristics. Specialists are more likely to hold a MBA degree and have held

more jobs in the past, while managers with PhD studies are more likely to be generalists.

On average, generalists manage a larger number of funds and have a higher volume of

assets under management. They manage their funds longer, have been affiliated with the

management company longer and have shown a better past performance track record.9


9Past Manager Skill is the TNA-weighted cumulative return of the objective-adjusted before fees returnsof all the funds run by the manager during the past 24 months. We can see that specialists do not doa very good job, in general. However, as we will show later, their performance is substantially betterwhen they have some type of skill, especially if they are pickers. Table A1 in the appendix sectionprovides further tests.

5

3.2. Pickers and Timers

Next, we consider two possible skills of portfolio managers: stock-picking and market

timing. Assuming some portfolio managers have either of these skills, we want to analyze

if this affects their assignment as generalist or specialist in in the fund. First we need to

decide whether a fund manager has picking or timing ability. For that purpose, we run the

Treynor-Mazuy (1966) market-timing model (hereafter referred to as TM), augmented

with multi-risk factors, and sorted by asset class. Prior research have also considered a

multi-factor version of the Treynor-Mazuy (1966) and the Henriksson and Merton (1981)

approach (i.e. Bollen and Busse, (2001, 2005)).10 In particular, for the equity funds we

use the following model:

rit = αi + βrm,irmt + γrm2,irm2t + βsmb,ismbt + βhlm,ihlmt + βmom,imomt + εit (1)

where rit is equity fund i’s before-expense return in month t in excess of the 30-day risk-

free interest rate; rmt is the market portfolio return in excess of the risk-free rate; smbt,

hlmt and momt are the size, book-to-market and momentum factors commonly used in

the literature.11 For the fixed income funds we use:

rit = αi + βrm,iABt + γrm2,iAB2t +Bj,iBFt + εi,t for j = 5 (2)

where rit is bonds fund i’s before-expense return in month t in excess of the 30-day

risk-free interest rate; ABt is the U.S. Aggregate Bond Index return in excess of the risk

free rate and its squared value; BF , for “bond factors”. We follow the methodology in

Blake, Elton, and Gruber (1993) and add six bond index returns, all in excess of the

1-month treasury rate. Those bond indices include three for government bond (Barclays

U.S. Treasury Long, Barclays U.S. Treasury Intermediate, and Barclay U.S. Treasury Bill

36m), two for corporate bonds (Barclays U.S. Corp Investment Grade, and Barclays U.S.

High-Yield Composite), and one for agency bonds (Barclays GNMA 30-Year). Finally,

for the international stock funds we use this model:

rit = αi + βrm,iGMt + γrm2,iGM2t +Bj,iGFt + εi,t for j = 3 (3)

which is similar to (2), with the only difference that fund returns GM , for “global mar-

kets,” are from international stocks funds, and risk factors, GF , are the Fama-French

10Many other studies use portfolio holdings to determine timing and selection ability. However, given theshortcomings of the existing databases, we choose to follow the basic TM model of portfolio returns.In particular, the database with mutual fund portfolio holdings most frequently used by academics isThomson Reuters, however it does not serve our purposes. Thompson provides holdings only for equityfunds, but we need to observe the ability of managers across all the different asset classes. Besides, itonly reports portfolio holdings quarterly.

11From Kenneth French’s website.

6

global factors.12

Next, we classify as pickers the managers of funds for which αi is greater than 0

and statistically significant, and as timers the managers of funds for which γi is positive

and statistically significant.13 For portfolio managers who manage more than one fund

simultaneously, we use the TNA-weighted average of these coefficients.14

In Table 4, we present the proportion of funds managed by generalists sorted by the

investment objective. We list the proportion for each style when the manager has picking

ability, timing ability or no ability at all, as well as the total. We find that a higher

proportion of generalists run total return and balanced funds, regardless of their ability.

Also, all balance funds are managed by generalists; this is not surprising, as this category

allows funds to invest in both equity and fixed-income assets. We also observe that

generalists are more frequent among timers across all categories, except for government

long-term and foreign funds. Thus, it seems that generalists are more likely to be timers

than pickers, and very unlikely that they are unskilled.15


A possible explanation for this finding is that pickers are better suited to work as

specialists, while management companies can profit more from assigning timers to work

as generalists. That could explain why a substantial number of mutual funds assign them

accordingly. The argument seems straightforward in the case of specialists: by definition,

specialists have to invest within a narrow class of securities, and they benefit from an

ability to choose the best performers within that class. Generalists, on the other hand,

manage several funds and have a wider range of securities to cover. That role might suit

timers better. Since they manage a large and diverse number of securities, their ability

to predict market trends might allow them to shift money across groups of assets with

different cyclical characteristics, instead of individual securities. Their strategy would

rely on predicting market trends and decide across the different funds on what sets of

securities to bet. Besides, timers might benefit from access to larger set of information

within the family and hence use it to invest across different assets.

12The global factors include all 23 countries in the four regions: Australia, Austria, Belgium, Canada,Denmark, Finland, France, Germany, Greece, Hong Kong, Ireland, Italy, Japan, Netherlands, NewZealand, Norway, Portugal, Singapore, Spain, Switzerland, Sweden, United Kingdom, United States.

13For each period, we estimate all the coefficients using data covering the previous 24 months (with aminimum of 20 observations). As a robustness check, we also estimate them using 36 months, with aminimum of 30 observations, and results remain unchanged.

14Breen, Jagannathan and Ofer (1986) show that the Henriksson and Merton (1981) regression mayexhibit heteroscedasticity and therefore might be less accurate than the TM approach, both in termsof size and power. Nevertheless, we replicate of our test using the Henriksson and Merton (1981)approach and the main results remain unchanged.

15Consistent with Kacperczyk, Nieuwerburgh and Veldkamp (2014), portfolio managers display eithersecurity selection or market timing abilities, but not both at the same time. We find that very fewmanagers are both timers and pickers simultaneously.

7

We explore our conjecture in Table 5. In Panel A we show that stock-picking skilled

managers working as specialists produce better results than those working as generalists.

While, on average, funds managed by specialist pickers have 11.1 (11.3) bps of gross (net)

return per month higher than the average fund in that style, funds managed by generalist

pickers only show a 3.9 (5) bps of gross (net) excess return. This difference is even

larger when we compute the average performance over all the funds run by the manager

and it also exists when we consider the performance of the whole family performance.

In Panel B we study portfolios with managers with market-timing ability. When the

managers are specialists, their average style-excess return is a mere 0.1 (gross) and 0.3

(net) bps. These numbers, though, go up to 4.8 and 4.5 bps, respectively, when the

managers are generalists. Furthermore, the average manager performance of generalist

timers is about 11.7 (12.5) bps of monthly gross (net) investment objective-adjusted

return. Finally, Panel C displays average performances of funds managed by unskilled

managers. Predictably, these figures are negative or close to zero. Also, it seems that

unskilled managers are less harmful as specialists than as generalists.16


4. Empirical Results

4.1. Managerial Type and Performance

Our main hypothesis is that portfolio managers with a certain skill (timing or picking)

are better suited to perform a specific function -generalist or specialist. To test this, we

estimate the following model:

OARi,t = a0 + a1Gj,t + a2MSj,t + a3Gj,t ×MSj,t + a4Xi,t−1 + δt + ei,t (4)

where OARi,t, the fund performance, measures the investment objective-adjusted return

of fund i at time t, using the excess return of the portfolio over the average return of

all funds in their style. Gj,t –for “Generalist”– measures the level of diversification of

fund i run by manager j in month t (i might represent several funds, if the manager

runs more than one). In particular, G is a dummy variable equal to 1 if the Herfindahl

index:∑9

s=1

(TNAs,j,t

TNAj,t

)2

is below 1, and 0 otherwise. The subindex ‘s corresponds to

the “fund style” as defined in the NSAR-B filings (capital appreciation, growth, income,

total return, government short-term debt, government long-term debt, corporate debt,

balance and international stocks)17 and TNAs,j,t is the total net assets managed by

16See Table A2 in the appendix for further tests about differences in fund performance between managerswith timing or stock-picking ability versus those without any skill.

17A full description of these investment objectives is in the Appendix.

8

manager j according to investment style s at time t. Therefore, for funds managed by

specialist managers, who manage funds in a single investment style, G = 0. MSj,t–for

“Manager Skill”– denotes whether the manager has timing or picking ability; we use a

dummy variable for each of these abilities, with values 1 or 0.18 X is a vector of fund,

manager and family-specific control variables, including size, age, turnover, expenses,

flows and past return of the portfolio, size and number of funds within the family, number

of managers in the family, whether the family offers or demands sub-advising services,

manager background information (PhD, MBA, the number of prior positions, college

type), number of funds and assets the manager is currently managing as well as the time

the manager has been affiliated with the portfolio.19 We also control for style of the

fund (dummy variables for each investment objectives) and the period (year) in which

the manager is evaluated (δt) to rule out the possibility that the results are driven by a

correlation between a given fund style or time period and the fund performance –that is,

style and year fixed effects. We estimate equation (4) using Pooled-OLS regressions. We

also adjust for serial correlation by clustering standard errors at the fund level.20

Table 6 shows the results depending on whether we measure the performance at the

portfolio i or manager j level. Columns 1 to 3 result from OAR computed at the portfolio

level. Columns 4 to 6 consider as dependent variable the OAR of the manager: TNA-

weighted average gross (before deducting fees and expenses) style-adjusted return of all

the funds managed by that manager. Whereas the relationship between generalist (as

opposed to specialist) and fund performance is practically nonexistent, there is a strong

negative relation for funds managed by generalist pickers, and significantly positive for

funds managed by generalist timers. In economic terms, generalist managers with market-

timing ability have an abnormal fund (manager) return of 324 (215) bps per year (.27

and .179 monthly, respectively) greater than those managed by specialists. On the other

hand, generalist managers with stock-picking ability yield a fund (manager) performance

of 247 (377) yearly bps (.206 and .314 monthly, respectively) lower than those with a

specialist role. This means that management companies can achieve better performance

by allowing managers with picking ability to manage similar funds and allocating market-

timers to manage funds with different investment styles.


18As we argued before, we will get 1 and 0, 0 and 1 and, very often, 0 and 0; 1 and 1 is exceptional.19We winsorize all the control variables at the 1% level.20These results are robust to different additional tests such as including continuous rather than dummy

variables, risk-adjusted performance measures, fund, family and manager fixed effects, clustering bytime, as well as Fama-MacBeth (1973) regressions. See Appendix for more details.

9

4.2. Managerial Type and Performance: Fixed Effects

To reinforce our previous conclusions, we repeat the analysis with additional fixed

effects. In Panel A of Table 7 we control for fund fixed effects, which allows us to compare

differences in performance across different portfolio managers with different skills (timers

or pickers) and different assignments (specialists or generalists). The coefficients of the

interactions Generalist×Timer and Generalist×Picker are even larger. Funds managed

by Generalist-Timers or Specialist-Pickers return on average about 353 and 312 bps more

per year than Specialist-Timers and Generalist-Pickers, respectively. We also control for

manager and family fixed effects to rule out the possibly that the results are driven by

specific portfolio managers or management company characteristics. Panels B and C of

Table 7 confirm the results.


4.3. Managerial Type and Performance: Subsamples

To achieve further insight in our results, we split our sample into funds managed by

timers, funds managed by pickers and funds managed by unskilled portfolio managers,

and we estimate the following model for each subsample:

OARi,t = a0 + a1Generalistj,t + a2Xi,t−1 + δt + ei,t (5)

The dependent variable OARi,t, is performance measured at different levels: fund and

manager. For fund performance we use the investment objective adjusted return described

in the previous section. Manager performance is the TNA-weighted average OAR across

all the funds run by the manager.21 Generalistj,t is a dummy variable that captures

whether the manager performs a generalist or a specialist role within the management

company. X is a vector of control variables at the fund, family and manager levels.22 We

include time and investment objective dummies (δt). We cluster standard errors at the

fund level and estimate equation (5) using Pooled OLS regression.23

Table 8 shows the results of estimating (5) in two parts –columns (1)-(3) and (4)-

(6), respectively. On the left side we estimate the objective-adjusted performance at the

fund level, and on the right side at the portfolio manager level. Each part sorts the

sample into funds managed by timers, pickers and funds managed by unskilled portfolio

21There is a wide range in the number of funds managed by the same person. In our sample they varyfrom 1 to 26, with a mean of 3.05 and a standard deviation of 3.5.

22For a detailed description of these variables, see the appendix.23We apply the Petersen (2009) approach to estimate, in an efficient way, the standard errors of our

regression. The SE clustered by funds are dramatically larger than the white SE, while the SE clusteredby years are only slightly larger than the white SE. Besides, clusters by funds and years are similar toclusters by funds. Then, the importance of time effect (after including time dummies) is small, and inthe presence of a fund effect, white and Fama-MacBeth SE are significantly biased.

10

managers. Whereas funds managed by generalists with timing skills perform about 19.9

bps per month better than specialists with timing skill, funds managed by specialist

with stock-picking ability yield 14.3 bps per month more than other generalists with

equivalent stock-picking ability. Unskilled portfolio managers seem to be performing

similarly, regardless of their managerial function. We find similar results when we consider

manager performance. Thus, we conclude that pickers are better suited to manage funds

with a single investment objective and timers to portfolios from different styles, because

they both contribute to improve the performance of the funds they run, as well as the

overall performance of all their portfolios.


4.3.1. Portfolio Management Misallocation

According to our evidence, it seems optimal to assign timers to generalist positions,

and pickers to specialist roles. Yet, in some cases managers are not allocated according

to this rule. In this section we study what might drive suboptimal decisions, and we

estimate the following logistic model:

Prob(Msai,t = 1) =exp(a0 + a1Xi,t−1 + δt + ei,t)

1 + exp(a0 + a1Xi,t + δt + ei,t)(6)

The dependent variable Msaj,t –for misallocation – represents funds that are run

by either a generalist with stock-picking ability (left panel of table 9) or a specialist

with market-timing skill (right panel). X is a set of fund, manager and family-specific

explanatory variables, including size, age, turnover, expenses, flows and past return of the

portfolio, size and number of funds within the family, number of managers in the family,

and whether the family offers or demands sub-advising services, manager background

information (PhD, MBA, the number of prior positions, college reputation), number of

funds and amount of assets under management, as well as the time the manager has been

affiliated with the fund.

Table 9 presents the results of estimating specification (6) for all the U.S. open-

end funds in our dataset. The unconditional probability of misallocating a picker as a

generalist is 5.4% and 8.9% for a timer to a specialist role. The probability of assigning

a stock-picker as a generalist is larger for small funds, with high turnover and good past

performance. These managers usually have a graduate degree (MBA or PhD), manage

a good number of funds, amounting to a high value of assets, and work in families with

few managers. It is possible that small firms lack employees with timing ability and use

their more qualified pickers to manage several small funds. On the other hand, managers

without a PhD degree, running a relatively low total of assets spread out across several

expensive funds with poor past performance, are more likely to be specialists with timing

11

ability. This type of misallocation seems more frequent among families with a large value

of assets under management in which managers are in charge of several funds and the firm

offers its services as sub-advisor. Providing specialized managing services as sub-advisors

might require the firm to assign timers to specialist functions. Lack of enough pickers

might lead to employ the less quantitatively qualified timers to manage expensive funds

with a poor past record.

Overall, human capital misallocation seems to be associated with a lack of qualified

portfolio managers to run the volume of assets the firms control.


4.3.2. Influence of skills on Promotions: Specialist to Generalist

We now study whether market-timing or stock-picking skills affect a portfolio man-

ager’s switch from specialist to generalist. We estimate the following logistic model:

Prob(yi,t = 1) =exp(βfzi)

1 + exp(βfzi)(7)

where βfzi = (a0+a1Skillj,t+a2MPSj,t+a3Xi,t−1+δt+ei.t). The dependent variable (yi,t)

is a dummy variable, equal to 1 when portfolio manager j in charge of fund i switches from

specialist at t to generalist at t+1, and equal to 0 when the manager remains a specialist.

Skillj,t stands for the two different types of portfolio manager ability: Timerj,t if manager

j successfully timed the market from t-25 to t-1 and Pickerj,t if manager j showed stock-

picking skill during the prior 24 months. We also include MPSj,t (“Manager Past Skill”)

as the past 24 months cumulative OAR of the manager (TNA-weighted average of the

objective-adjusted return of all the funds run by the manager j). X is a vector of manager-

related control variables lagged one period. We include time and investment objective

dummies (δt) –year and style fixed effects– and cluster standard errors at the fund level.

Table 10 shows the results of estimating (7). We observe in Models 3 and 4 that

both timers and pickers are more likely to go from specialist to generalist than unskilled

managers. The marginal effects of the Timer and Picker coefficients are about 0.3% and

0.2%, whereas the unconditional probability of the change from specialist to generalist

is 1.3%. Therefore, these managers are about (0.003/0.013) 23% and (0.002/0.013) 15%

more likely to switch to generalist than unskilled specialists. In Model 4 we interact

the type of managerial skill with the overall abnormal return of the manager during the

past 24 months. We find that the probability of changing from specialist to generalist

for top performers increases in a significant way when they are timers and decreases for

top pickers. In economic terms, an increase of one standard deviation on MPS (4.611)

makes a timer (0.003 + 0.001 ∗ 4.611)/0.013 = 58.5% more likely to switch than other

specialists. Arguably, a switch from specialist to generalists is, in general, a promotion

12

since, as we showed previously, generalists manage a significantly larger amount of assets

than specialists.24

These results suggest that management companies base a change from specialist to

generalist on the ability of the manager to predict market trends. Overall, these results

provide additional evidence on the importance of managerial skills in determining the

function of the manager; the results are also consistent with an optimal assignment of

managers to functions. Top specialist-pickers, although less likely to switch from specialist

to generalist, are likely to be highly compensated by the management company.


4.3.3. Performance around Managerial Reassignment: Event Study

Next we study the effects on performance of reassignment of specialists to generalists.

We conduct an event study and average the fund performance during the six months

before and twelve months after the switch. We measure performance using the 6-months

cumulative objective-adjusted return (OAR) before deducting expenses and fees. We

divide our sample into funds managed by timers, funds managed by pickers, and funds

run by unskilled managers.

Table 11 Panel A displays the results of reassignment on funds performance. In

particular, we present the effect on return of a change of management from specialist to

generalist –the manager of the fund becomes a generalist or is replaced with a generalist

with the same type of skill. On average, the performance of the funds run by a timer

improve significantly in the quarter after the manager went from specialist to generalist

–it could be the same timer or a different timer. This improvement seems to be persistent

even twelve months after the switch. On the other hand, when the manager has picking

skill, the performance in the quarter before the switch is positive and might even improve

in the short-run after the fund is managed by a picker-generalist rather than specialist,

but it drops overtime and remains negative twelve months after the event. We find similar

results for funds run by unskilled managers. They get a significant improvement in the

first months after being run by a generalist, but this positive performance disappears in

the second quarter. Furthermore, we compute the difference in cumulative performance

before and after the event and its statistical significance.25 We find that funds managed by

specialists experience a significant improvement after the manager becomes a generalist

only for funds managed by timers. The return of funds managed by pickers decreases.

The change in return of funds run by unskilled managers is not significant.

24Prior research has classified promotions and demotions based on total assets under management (i.e.Chevalier and Ellison (1999a), Hu, Hall, and Harvey (2000), and Baks (2003)).

25We compare the average performance of funds two months prior the switch and the performance twelvemonths after the event. We consider t-2 because the month right before the event might not be veryrepresentative of the managerial behavior due to the proximity of the switch. Similarly, we considert+12 as managers might need some time to be adapted to their new functions.

13

In Panel B of Table 11 we examine the effect on the performance of the manager.

So we compare the performance of the funds managed by the same managers before and

after their functions change. We classified each funds based on the type of skills (timer,

picker or unskilled) the manager had before being upgraded to generalist.26 We find

similar results, specialists who become generalists improve their overall performance only

they were timers. Pickers and Unskilled managers had an improvement in the short-run

that get reduced in the second quarter post the event.


5. Robustness and Alternative Interpretations

5.1. Ability or Selection: Propensity Score Matching

We want to rule out that timers might perform better as generalists for a reason other

than their timing skill, and similar for pickers as specialists. To eliminate this concern we

carry out a propensity score matching exercise. We use two different propensity matching

techniques: the Nearest Neighbor procedure of Rosenbaum and Rubin (1983), and the

Kernel Matching of Heckman, Ichimura and Todd (1997, 1998). We first identify a con-

trol sample of funds managed by specialist-timers that exhibit no observable differences

in characteristics relative to the funds managed by generalist-timers. Thus, each pair

of matched funds is almost identically to one another, except for the main variable of

interest: the function of the manager. Similarly, we also identify pairs of funds managed

by generalist-pickers that are identical to specialist-pickers, except for the type of ability

of the managers.

More explicitly, we calculate the probability (i.e., the propensity score) that a fund

with certain characteristics is managed by a generalist. To calculate the propensity score

we use characteristics of the fund, management company and portfolio manager. In

particular, we estimate this probability as a function of the following factors: size, age,

turnover, expenses, flows and past returns of funds; volume of assets and number of

funds and managers of the family; number of prior positions, length of time the manager

has been run the fund, number of funds and total amount of assets the manager has

currently under management. We require that the maximum difference between the

propensity scores of the funds does not exceed 0.1% in absolute value.

Next, we compare fund performance between the two groups of matched funds. As

the control funds are a set of peers almost identical in terms of observable characteristics,

unless timing ability matters, the funds managed by generalist-timers should perform at

a level similar to the funds managed by specialist-timers. Similarly for the group of funds

26Unlike in Panel A, this panel focus on funds managed by the same manager and allow for changes intheir skills after the managerial switching

14

run by portfolio managers with stock picking skills. We calculate fund performance as

the excess return over the mean of the style. We use returns before and after fees.

In Table 12 we compare performance of the two groups and report the value of the

difference (Generalist-Specialist) and the statistical significance using bootstrapped stan-

dard errors associated to that difference. We also group the portfolios into quintiles

based on the timing and picking skills of the managers running the funds during the

period 1996-2011.

Panel A contains all funds sorted into quintiles according to timing ability, from

lowest to highest. The matching resulting from the propensity score methodology uses

a control fund from the same quintile. In every quintile, funds managed by generalist–

timers outperform their specialist peers, specially in the highest quintile in which we

observe that such outperformance averages 42.2 and 68.5 bps per month -depending on

the propensity score method. Panel B reports the differences between generalist and

specialist managers depending on different levels of stock-picking ability. As expected,

stock pickers are more effective at managing funds within the same investment objectives.

The greater the picking skill the greater the differences between specialist and generalist

managers. For the top timing quintile the difference averages 89.4 bps and 73.8 bps per

month, depending on the score propensity measure.27

We have conjectured that the optimal portfolio manager allocation strategy is to

assign stock-picking skilled managers to specific investment objective funds and timer

skilled managers to different investment style funds. We would like to study who collects

the rents of the superior management: maybe the better managed funds charge higher

fees and the difference in performance is irrelevant for investors after we take fees into

account. In Table 12 we include the differences on net performance, that is, we use

returns after fees. Overall, the results in Panel A and B do not change substantially.

We conclude that investors are better off purchasing funds managed by generalists with

timing skills and specialists with stock-picking skills.


5.2. Selection Bias: Heckman’s (1979) two-step procedure

By definition, a generalist has to manage more than one fund at a time. However,

this is not always the case and we can find families that only allocate one manager per

fund. These firms can have some policy of a manager per fund or simply because they

have too many managers for too few funds. Nevertheless, they will not be promoted

from specialist to generalist, not because of a lack of timing skills but because of some

other family characteristics. Thus, we are under a selection bias problem in which, only

27These results are also robust when we use the radius and stratification matching methods.

15

a subsample of managers will be specialist candidates to be promoted. To address this

issue, we conduct the Heckman selection model28, in which first we obtain the probability

that a manager runs several funds simultaneously, and in the second stage, we estimate

the probability of becoming a generalist.

Table 13 reports the regression results from the first-stage of the Heckman’s selection

estimation. In the this regression, we model the probability that a manager has more

than one fund using the same set of manager control variables previously used, and a

variable that measure the ratio of funds per manager that a family has on average29:

Prob(MFMi,t = 1) = φ(β0 + β1wf,t−1 + β2xi,t−1 + δt + εi,t) (8)

where φ(·) is the cdf (cumulative density function) of the standard normal distribution.

The dependent variable MFMi,t, Multi-Funds Manager, is a dummy that takes the value

1 if the fund i is managed by manager j who runs more than one fund, and 0 otherwise.

β0 is a constant and wf,t−1 is the variable Funds per Manager, defined as the number of

funds in the family f divided by the number of managers on that family. Xi,t−1 is the set

of fund, family and manager control variables for each portfolio we have used previously.

We also include year and investment objective dummies (δt), and the standard errors

are clustered at the fund level. In the selection model of Table 13, we show that the

probability of managing more than one fund clearly depends positively on the ratio of

funds per manager of the family.


In Table 14, we show the estimates from the second stage Heckman’s two step proce-

dure presented in equation (8). We find in Model 5 that conditioned on the probability

of being a multi-fund manager, the specialist-timers are about 60% (0.012/0.020) more

likely to become generalists than other specialists. Similarly, specialist pickers are 45%

more likely than unskilled portfolio managers. A possible explanation is that becoming

a generalist is a promotion which rewards ability.

Additionally, we also interact the type of skill with the cumulative past performance

record of the manager. The higher the quality of the timer, the greater the probability

of reassignment to generalist, while top pickers are more likely to stay as specialist.


28The original Heckman Correction (1979) was used for continuous depend variable, in our case, thedependent variable is discrete and thus we are using a newer version of these procedure.

29The variable Fund per Manager clearly affects the probability that a manager will run more than afund, however there does not seem to be any theoretical reason why it will have any effect on thedecision to assign fund managers to generalist functions.

16

5.3. Family Expansion

One possible explanation behind the decision to reassign a manager to a generalist role

might be that it is efficient for firms that are trying to increase the range of investment

objectives they offer. In particular, a recent study on product differentiation and market

share concludes that fund families gain market share by offering a wider variety of fund

investment objectives (Khorana and Servaes, 2012). Thus, it is possible that firms that

are trying to increase their market share decide to offer a wider range of products. In

order to staff the new funds, the firms can assign some of their existing talent to manage

the new funds, and therefore naturally reassign some of their specialists to generalist

functions.

In Table 15 we again analyze the probability of a reassignment from specialist to gen-

eralist, but this time we divide our sample based on the concentration level of the family.

We sort families into four different groups every month and then group them into the low

concentrated ones if their Herfindahl index across investment objectives funds are within

the first quartile and highly concentrated if they are in the forth quartile. We show that

for funds that belong to families with either low or high levels of concentration, the timing

skill of the manager has a strong effect on the probability of switch to generalist, while

picking skills do not seem to matter for this decision. Therefore, family concentration is

an important characteristic to take into consideration. 30


5.4. Downturn Markets

Kacperczyk, Nieuwerburgh and Veldkamp (2014) provide evidence that outperforming

fund managers excel at stock picking in bull markets and at timing in recessions. We

wonder then if there is a cyclical component in the decision to switch a manager from

specialist to generalist, and that is the driving factor, and not the ability as timer. For

example, it could be that the management companies are trying to save costs by having

portfolio managers run different funds, possibly with different objectives.

For that reason, in Table 16 we estimate the probability of being transfered from

specialist to generalist for different market conditions. We classified bull and bear market

based on whether Chicago Fed National Activity Index (CFNAI) is on the forth or first

30In an unreported table we show that the level of concentration affects negatively the probability of areassignment to generalist. However, after controlling for this factor, timing and picking abilities arestill statistically relevant. We observe again that timing ability (23% more likely than a managers withno ability) is a stronger predictor of the switch than picking ability (15% more likely). Our resultsalso show that specialists with timing skills and outstanding performance record are more likely to bereassigned than managers without ability, but we cannot make the same claim for pickers.

17

quartile, respectively. 31 We can observe that timing ability is still the main factor

affecting the decisions of switching to generalist functions, independently on the market

being in recession or expansion. On the other hand, Table 16 shows that picking skills

matter only for bear market. 32


6. Conclusions

Our paper supports the literature that argues that there are actively managed mu-

tual funds that outperform passively managed funds. In particular, we identify portfolio

managers with two types of skills, stock-picking or market-timing –although we agree

that many portfolio managers do not display either type of skill. We argue that pick-

ers are a better fit for positions as specialists: managers who run funds with a single

investment style, while timers perform better as generalists, running several funds with

different investment styles. Consistent with an optimal allocation of human capital,

we find more timers among the generalists and more pickers among the specialists. In

addition, management companies tend to switch timers from specialists to generalists

functions, specially when they have been performing extremely well in the past. This is

considered as a promotion within the firm and we observe that overall manager perfor-

mance improve after such event. On a side note, we find that market-timers are more

likely to have a PhD degree and a quantitative background, while stock-pickers tend to

have MBA degree. Overall, management companies make rational decisions based on

measurable skills.

This result has important implications for the organization decisions of management

companies. Many studies have shown inefficiencies of this industry, in which portfolio

managers that are highly compensated are unable to outperform a given benchmark

or performance persists for poor managers but not for top performers. Our evidence

presents a picture in which funds management companies allocate their employees to

exert the maximum productivity. Further research might focus at the portfolio manager

level rather than a the fund level to understand the costs and benefits associated with

manager assignments given their skills as well as the role that human capital and industrial

organization play in this industry.

31The CFNAI is a coincident indicator of national economic activity comprising 85 existing macroe-conomic time series. It is constructed to have an average value of zero and a standard deviation ofone. As in Kacperczyk, Nieuwerburgh and Veldkamp (2014), we use the headline three-month movingaverage to measure the market conditions.

32In an unreported table, we verify that market conditions affect the probability of reassignment, butas in our previous tests, having some ability and specially the interaction between timing ability andprevious outstanding performance explains in a significant way the probability of reassignment.

18

References

Baker, M., Litov L., Wachter, J. A., and Wurgler J., 2010, Can mutual fund managers

pick stocks? Evidence from their trades prior to earnings announcements, Journal of

Financial and Quantitative Analysis 45, 1111-1131.

Baks, K. P., 2003, On the Performance of Mutual Fund Managers, Working paper.

Bar, M., Kempf, A. and S. Ruenzi, 2011, Is a Team Different from the Sum of its Parts?

Evidence from Mutual Fund Managers, Review of Finance 15, 359–396.

Berk, Jonathan B., and Richard C. Green, 2004, Mutual fund flows and performance in

rational markets, Journal of Political Economy 112, 1269-1295.

Berk, Jonathan B., and Jing Xu, 2005, Persistence and fund flows of the worst perform-

ing mutual funds, Working paper, U.C. Berkeley.

Berk, Jonathan, and Jules van Binsbergen, 2012, Measuring managerial skill in the

mutual fund industry, Working paper Stanford University.

Blake, Christopher R., Edwin J. Elton, and Martin J. Gruber, 1993, The performance

of bond mutual funds, The Journal of Business 66, 371-403.

Blake, D., Rossi, A., Timmermann, A., Tonks, I., and R. Wermers, 2013. Decentralized

investment management: evidence from the pension fund industry, Journal of Finance

68, 1133–1178.

Bliss, R., Potter, M. and C. Schwarz, 2008. Performance Characteristics of Individually-

Managed versus Team-Managed Mutual Funds, The Journal of Portfolio Management

34, 110–119

Bollen, Nicolas P. B., and Jeffrey A. Busse, 2001, On the timing ability of mutual fund

managers, Journal of Finance 56, 10751094

Bollen, Nicolas P. B., and Jeffrey A. Busse, 2005, Short-term persistence in mutual fund

performance, Review of Financial Studies 18, 569-597.

Boyson, Nicole M., 2003. Why do experienced hedge fund managers have lower returns?,

working paper.

Bradley, Michael, Alon Brav, Itay Goldstein, and Wei Jiang, 2005, Costly communica-

tion, shareholder activism, and limits to arbitrage, Working paper, Duke University

Brauer, Gregory A., 1984, Open-ending closed-end funds, Journal of Financial Eco-

nomics 13, 491507.

19

Brauer, Gregory A., 1988, Closed-end fund shares’ abnormal returns and the information

content of discounts and premiums, Journal of Finance 43, 113-128.

Breen, W., R. Jagannathan and A. Ofer (1986), Correcting for Heteroscedasticity in

Tests for Market Timing Ability, Journal of Business, Vol. 59, No. 4, pp. 585-98.

Brickley, James A., and James S. Schallheim, 1985, Lifting the lid on closed-end invest-

ment companies: A case of abnormal returns, Journal of Financial and Quantitative

Analysis 20, 107-117.

Brown, Stephen J., and William N. Goetzmann, 1995, Performance persistence, Journal

of Finance 50, 679-698

Carhart, Mark M., 1997, On persistence in mutual fund performance, Journal of Finance

52, 57-82.

Cashman, George D. and Daniel N. Deli, 2009, Locating decision rights: Evidence from

the mutual fund industry, Journal of Financial Markets 12, 645-671.

Chen, Hsiu-Lang, Narasimhan Jegadeesh, and Russ Wermers, 2000, The value of active

mutual fund management: An examination of the stockholdings and trades of fund

managers, Journal of Financial and Quantitative Analysis 35, 343-368.

Chevalier, J., and G. Ellison, 1999a, Are Some Mutual Fund Managers Better than

Others? Cross-Sectional Patterns in Behavior and Performance, Journal of Finance

54, 875- 899.

Chevalier, J., and G. Ellison, 1999b, Career Concerns of Mutual Fund Managers, Quar-

terly Journal of Economics 105, 1167-1200.

Chordia, T., 1996. The Structure of Mutual Fund Charges, Journal of Financial Eco-

nomics 41, 3-39.

Cohen, Randolph B., Joshua D. Coval, and Lubos Pastor, 2005, Judging fund managers

by the company they keep, Journal of Finance 60, 1057-1096.

Cremers, K. J. Martijn, and Antti Petajisto, 2009, How active is your fund manager? A

new measure that predicts performance, Review of Financial Studies 22, 3329-3365.

Custodio, C., Ferreira, M. and P. Matos, 2013. Generalists Versus Specialists: Lifetime

Work Experience and Chief Executive Officer Pay, Journal of Financial Economics

108, , 471–492.

Daniel, K., Grinblatt, M., Titman, S., and R. Wermers, 1997. Measuring Mutual Fund

Performance with Characteristic-Based Benchmarks, Journal of Finance 52, , 1035–

1058.

20

Ding, B. and R. Wemers, 2004, Mutual Fund Stars: The Performance and Behavior of

U.S. Fund Managers, Working Paper.

Ding, Bill, and Russell R. Wermers. 2012, Mutual fund performance and governance

structure: The role of portfolio managers and boards of directors, working paper.

Drazin, R. and H. Rao, 2002, Harnessing Managerial Knowledge to Implement Product

line Extensions: How Do Mutual Fund Families Allocate Portfolio Managers to Old

and New funds? The Academy of Management Journal 45, 609-619.

Evans, Richard B., 2009, Does Alpha Really Matter? Evidence from Mutual Fund

Incubation, Termination and Manager Change, working paper, University of Virginia.

Evans, R., 2010. Mutual fund incubation, Journal of Finance 65, 1581–1611.

Fang, Jieyan, Kempf, Alexander, and Monika Trapp, 2014, Journal of Financial Eco-

nomics 111, 661–674.

Fee, C. and C. Hadlock, 2003, Raids, Rewards, and Reputations in the Market for

Managerial Talent, Review of Financial Studies 16, 1315-1357.

Ferreira, D. and R. Kumar Sah, 2012. Who gets to the top? Generalists versus specialists

in managerial organizations, RAND Journal of Economics.

Ferson, Wayne, and Haitao Mo, 2015, Performance measurement with selectivity, market

and volatility timing, Journal of Financial Economics, forthcoming.

Fortin, Rich, Stuart Michelson, and James Jordan-Wagner, 1999. Does mutual fund

manager tenure matter?, Journal of Financial Planning, August Issue, 72-79.

Frydman, C., 2009. Rising through the ranks: the evolution of the market for corporate

executives, 19362003. Unpublished working paper.

Gallaher S., R. Kaniel and L. Starks, 2006, Madison Avenue Meets Wall Street: Mutual

Fund Families, Competition, and Advertising, Working Paper.

Gaspar, J., M. Massa, and P. Matos, 2006, Favoritism in Mutual Fund Families? Evi-

dence on Strategic Cross-Fund Subsidization, Journal of Finance 61, 73-104.

Golec, J., 1996, The Effects of Mutual Fund Managers’ Characteristics on their Perfor-

mance, Risk and Fees, Financial Services Review 5, 133-147

Gottesman, A. and M. Morey, 2006, Manager Education and Mutual Fund Performance,

Journal of Empirical Finance 13, 145-182.

21

Jain, P. and J. Wu, 2000, Truth in Mutual-Fund Advertising: Evidence on Future

Performance and Fund Flows, Journal of Finance 55, 937-958.

Jensen, Michael C., 1968, The performance of mutual funds in the period 1945-1964,

Journal of Finance 23, 389-416.

Heckman, J., 1979. Sample selection bias as a specification error, Econometrica 47,

153-61.

Heckman, J., Ichimura, H., and P. Todd, 1997. Matching As An Econometric Evaluation

Estimator: Evidence from Evaluating a Job Training Programme, Review of Economic

Studies 64, 605–654.

Heckman, J., Ichimura, H., and P. Todd, 1998. Matching as an Econometric Evaluation

Estimator, Review of Economic Studies 65, 261–294.

Hu, Fan, Alastair R. Hall, and Campbell R. Harvey, 2000. Promotion or demotion?

An empirical investigation of the determinants of top mutual fund manager change,

working paper.

Kacperczyk, Marcin, Clemens Sialm, and Lu Zheng, 2005, On the industry concentration

of actively managed equity mutual funds, Journal of Finance 60, 1983-2011.

Kacperczyk, Marcin, and Amit Seru, 2007, Fund manager use of public information:

New evidence on managerial skills, Journal of Finance 62, 485-528.

Kacperczyk, M., van Nieuwerburgh, S. and L. Veldkamp, 2014. Time-Varying Fund

Manager Skill, Journal of Finance 69, 1455–1484.

Kempf, E., Manconi, A., Spalt, O.G., 2014. Learning by Doing: The Value of Experience

and the Origins of Skill for Mutual Fund Managers, working paper, Tilburg University.

Khorana, A., 1996. Top Management Turnover: An Empirical Investigation of Mutual

Fund Managers, Journal of Financial Economics 40, 403-427.

Khorana, A., Sunil Wahal, and Marc Zenner, 2002. Agency conflicts in closed-end funds:

The case of rights offerings, Journal of Financial and Quantitative Analysis 37, 177-

200

Khorana, A. and H. Servaes, 2012. What Drives Market Share in the Mutual Fund

Industry? Review of Finance 16, 81–113.

Koijen, Ralph, 2012, The cross-section of managerial ability, incentives, and risk prefer-

ences, Journal of Finance

22

Lazear, E., 2009. Firm-specific human capital: a skill-weights approach. Journal of Po-

litical Economy 117, 914940.

Litov, Lubomir, Malcolm Baker, Jessica Wachter, and Jeffrey Wurgler. 2005. Can mutual

fund managers pick stocks? Evidence from their trades prior to earnings announce-

ments, Working paper, New York University

Malmendier, U., Tate, G., 2009. Superstar CEOs. Quarterly Journal of Economics 124,

1593-1638.

Mamaysky, H., and M. Spiegel, 2001. A Theory of Mutual Funds: Optimal Fund Ob-

jectives and Industry Organization, Working Paper.

Marris, Robin (1963): A model of the managerial enterprise.” Quarterly Journal of

Economics-77, 185-209.

Massa, M., 2003, How do Family Strategies Affect Fund Performance? When

Performance-Maximization is not the Only Game in Town, Journal of Financial Eco-

nomics 67, 249-304.

Massa, M., J. Reuter and E. Zitzewitz, 2010, When Should Firms Share Credit with Em-

ployees? Evidence from Anonymously Managed Mutual Funds, Journal of Financial

Economics 95, 400-424.

Merton, Robert C., 1980. On estimating the expected return on the market: An ex-

ploratory investigation, Journal of Financial Economics 8, 323-361.

Moreno, David, Rosa Rodriguez, and Rafael Zambrana, 2013, Management sub-

advising: Mutual fund industry, Working Paper.

Murphy, K. J. and J. Zbojnk, 2004. CEO Pay and Appointments: A Market-Based

Explanation for Recent Trends, American Economic Review, 192–196.

Nanda, V., M. P. Narayanan and V. P. Warther, 2000, Liquidity, Investment Ability,

and Mutual Fund Structure, Journal of Financial Economics 57, 417-443.

Rosenbaum, P. D. and Rubin, 1983. The Central Role of the Propensity Score in Ob-

servational Studies for Causal Effects, Biometrika 70, 41–55.

Schultz, Paul, 2010, Rational cross-sectional differences in market efficiency: Evidence

from mutual fund returns, Journal of Financial and Quantitative Analysis 45, 847-881.

Sirri, E. R. and P. Tufano, 1998. Costly Search and Mutual Fund Flows, Journal of

Finance 53, 1589-1622.

23

Treynor, J. and K. Mazuy, 1966. Can Mutual Funds Outguess the Market?, Harvard

Business Review 44, 131–136.

Warner, Jerold, R. Watts, and K. Wruck (1988): Stock prices and top management

changes.” Journal of Financial Economics 20, 461-492.

Wermers, Russ, Youchang Wu, and Josef Zechner, 2005, Portfolio performance, discount

dynamics, and the turnover of closed-end fund managers, Wp, University of Maryland.

24

Table 1: Funds by Investment Objectives

Table 1 displays the number of U.S open-end funds managed by an individual manager (team-managed funds are excluded)according to their investment objectives during the period 1996-2011. Funds are classified attending to the different categoriesprovided by the SEC for each investment company.

Capital Growth Income Return Gov ST Gov LT Corporate Balance Foreign1996 254 137 141 35 355 552 84 61 1861997 254 144 120 40 366 586 90 61 1931998 337 193 152 58 374 549 108 75 2511999 360 205 136 56 351 556 118 73 2572000 465 230 159 78 376 626 115 83 2382001 494 234 143 74 356 558 110 82 1992002 495 283 135 74 432 497 97 84 2362003 453 251 111 62 458 515 96 65 2002004 400 236 119 70 405 481 100 78 1842005 331 196 105 68 268 432 72 58 1482006 345 163 96 69 275 455 73 70 1442007 348 173 84 76 243 460 73 61 1512008 387 175 88 82 199 393 62 57 1642009 387 189 87 65 208 362 69 42 1972010 320 164 72 53 170 342 74 32 2102011 256 120 49 29 83 255 57 16 172

Table 2: Distribution of Funds and Managers by year

Table 2 displays the total number of funds managed by individual managers, the number of funds managed by generalists, thetotal number of individual managers, number who are generalists, the total number of management companies and how manyhave employed generalist managers. The sample covers all equity, fixed income and international U.S open-end funds managed byindividual managers during 1996-2011.

Funds Generalist Funds Managers Generalist Managers Firms Generalist Firms1996 1807 395 1176 133 261 891997 1860 444 1165 144 269 961998 2105 433 1361 142 311 1151999 2118 478 1359 150 299 1142000 2386 541 1390 175 407 1432001 1955 403 1202 138 358 1052002 2341 561 1307 169 339 1212003 2216 456 1241 144 333 1132004 2079 446 1171 138 321 1092005 1681 373 977 113 283 962006 1693 353 943 104 266 882007 1670 331 944 96 247 842008 1608 366 891 102 252 752009 1611 380 901 107 260 732010 1441 265 846 89 255 632011 1037 159 657 60 197 44

25

Table 3: T-Test Analysis: Specialist vs Generalist

This table presents the mean of fund, family and manager characteristics for the samples of specialists (portfolio managers offunds within a single investment objective) and generalists (portfolio managers of funds from different investment objectives) andthe associated difference among the two samples. * denotes significance at the 10% level, ** denotes significance at the 5% leveland *** denotes significance at the 1% level. Panel A contains all the variables at the fund level, Panel B summarizes the familycharacteristics and Panel C the variables at the portfolio manager level. The description of each variable is defined in the appendixsection. The data covers the period 1996 to 2011.

Panel A: Fund CharacteristicsSpecialist Generalist Difference

Fund Size 4.956 4.805 0.150∗∗∗

Fund Age 10.570 9.834 0.736∗∗∗

Fund Turnover 92.681 99.819 -7.138∗∗∗

Fund Expenses 1.152 1.122 0.030∗∗∗

Fund Flows 0.399 0.469 -0.071∗∗∗

Past Year Return 0.073 0.074 -0.001

Panel B: Family Characteristics

Specialist Generalist DifferenceFamily Size 8.120 7.793 0.327∗∗∗

Family Funds 26.689 21.175 5.514∗∗∗

Family Managers 9.789 8.397 1.392∗∗∗

Demand Advising 0.393 0.395 -0.002Supply Advising 0.654 0.579 0.075∗∗∗

Panel C: Manager Characteristics

Specialist Generalist DifferenceIvy League 0.210 0.211 -0.001MBA 0.464 0.423 0.041∗∗∗

PhD 0.028 0.035 -0.007∗∗∗

Past Positions 2.419 2.390 0.029∗∗∗

Manager Size 5.743 6.492 -0.748∗∗∗

Manager Funds 2.683 4.350 -1.667∗∗∗

Fund Affiliation 5.391 5.518 -0.126∗∗∗

Past Manager Skill -0.158 0.600 -0.758 ∗∗∗

26

Table 4: Proportion of Generalist by Style and Skill

This table presents the proportion of funds managed by generalist managers, according to the investment objective and managerialskill (timing, picking or unskilled) for the period 1996-2011. A full description of the type of manager variables is provided in theappendix.

Picker Timer Unskilled TotalCapital 0.33 0.34 0.18 0.20Growth 0.36 0.41 0.26 0.27Income 0.48 0.54 0.33 0.35Return 0.51 0.60 0.40 0.40Gov ST 0.14 0.28 0.24 0.18Gov LT 0.42 0.23 0.13 0.17Corporate 0.32 0.38 0.20 0.23Balance 1.00 1.00 0.42 0.44Foreign 0.37 0.23 0.12 0.14

Table 5: Objective Adjusted Returns by Types and Managerial Skills

This table presents the average investment objective (gross and net) returns of funds managed by generalist and specialist managers,according to their managerial skill (timing, picking or unskilled) for the period 1996-2011. A full description of the return measuresis provided in the appendix.

Gross Objective-Adj Return Net Objective-Adj ReturnPanel A: Picker Fund Family Manager Fund Family Manager

Specialist 0.111 0.117 0.131 0.113 0.123 0.143Generalist 0.039 0.067 0.035 0.050 0.078 0.044Panel B: Timer Fund Family Manager Fund Family ManagerSpecialist 0.001 0.006 0.098 0.003 0.013 0.106Generalist 0.048 0.013 0.117 0.045 0.017 0.125Panel C: Unskilled Fund Family Manager Fund Family ManagerSpecialist -0.015 -0.017 0.005 -0.013 -0.011 0.015Generalist -0.029 -0.029 0.004 -0.024 -0.019 0.014

27

Table 6: Managerial Type and Performance (I)

This table presents the results of monthly Pooled OLS regressions of fund and manager investment objective-adjusted returns onfund, manager and family characteristics. Fund returns are the actual returns before deducting fees and expenses (gross) andmanager returns are the TNA-weighted average return of all the portfolios managed by the same manager at the same time. Thedependent variable are fund and manager performance, measured by substracting the median return of their investment objectivepeers, from the actual return of the fund and manager, respectively. Generalist is a dummy variable equals 1 if the fund is managedby manager that is in charge on funds from different investment styles. Timer is a dummy equals 1 if the fund is managed byportfolio manager that has been able to time the market during the past 24 months. Picker is a dummy variable equals 1 if thefund is managed by a manager that was able to pick stocks efficiently during the past 24 months. All variables are lagged oneperiod. A full description of the remaining variables is in the appendix. Time and investment objective dummies are included butnot reported; t-statistics are reported in parentheses. We adjust for serial correlation by clustering standard errors at the fundlevel. * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Fund Performance Manager Performance(1) (2) (3) (4) (5) (6)

Generalist -0.030 0.037 -0.009 -0.028 0.046 0.016(-0.88) (1.03) (-0.24) (-0.87) (1.41) (0.45)

Timer -0.090∗ -0.067 -0.082∗ -0.061(-1.80) (-1.35) (-1.68) (-1.24)

Generalist × Timer 0.286∗∗∗ 0.270∗∗∗ 0.199∗∗∗ 0.179∗∗

(3.45) (3.25) (2.70) (2.43)Picker 0.294∗∗∗ 0.290∗∗∗ 0.305∗∗∗ 0.302∗∗∗

(5.43) (5.35) (5.62) (5.54)Generalist × Picker -0.214∗∗∗ -0.206∗∗ -0.320∗∗∗ -0.314∗∗∗

(-2.63) (-2.53) (-4.14) (-4.06)Fund Size 0.027∗∗ 0.026∗∗ 0.027∗∗ 0.020∗ 0.019∗ 0.020∗

(2.28) (2.26) (2.34) (1.73) (1.70) (1.74)Fund Age -0.001 -0.000 -0.000 0.000 0.000 0.000

(-0.35) (-0.09) (-0.14) (0.05) (0.30) (0.27)Fund Turnover 0.016∗ 0.015∗ 0.015 0.004 0.004 0.004

(1.69) (1.65) (1.61) (0.57) (0.56) (0.55)Fund Expenses 0.146∗∗∗ 0.136∗∗∗ 0.138∗∗∗ 0.126∗∗∗ 0.117∗∗∗ 0.119∗∗∗

(4.49) (4.16) (4.23) (4.15) (3.81) (3.86)Fund Flows 0.069∗∗∗ 0.067∗∗∗ 0.067∗∗∗ 0.064∗∗∗ 0.063∗∗∗ 0.063∗∗∗

(4.54) (4.45) (4.46) (4.43) (4.34) (4.34)Past Year Return -0.458∗∗∗ -0.488∗∗∗ -0.492∗∗∗ -0.451∗∗∗ -0.481∗∗∗ -0.485∗∗∗

(-3.52) (-3.74) (-3.77) (-3.45) (-3.67) (-3.70)Family Size 0.016∗ 0.016∗ 0.016∗∗ 0.020∗∗ 0.021∗∗ 0.021∗∗

(1.89) (1.95) (1.97) (2.41) (2.47) (2.48)Family Funds 0.000 0.000 0.000 0.000 0.000 0.000

(0.60) (0.45) (0.43) (0.89) (0.69) (0.68)Family Managers -0.001 -0.001 -0.001 -0.001 -0.001 -0.001

(-1.14) (-1.05) (-1.00) (-1.48) (-1.45) (-1.42)Supply Advising -0.045 -0.053∗ -0.051∗ -0.056∗∗ -0.063∗∗ -0.062∗∗

(-1.51) (-1.77) (-1.70) (-2.01) (-2.27) (-2.22)Demand Advising -0.014 -0.009 -0.009 -0.001 0.004 0.004

(-0.60) (-0.39) (-0.39) (-0.05) (0.18) (0.19)MBA 0.009 0.007 0.006 0.005 0.004 0.004

(0.34) (0.29) (0.26) (0.19) (0.17) (0.15)PhD -0.072 -0.072 -0.068 -0.066 -0.062 -0.060

(-1.40) (-1.40) (-1.32) (-1.34) (-1.26) (-1.22)Past Positions -0.012 -0.012 -0.012 -0.011 -0.011 -0.011

(-1.28) (-1.33) (-1.30) (-1.24) (-1.32) (-1.30)Ivy League 0.019 0.019 0.021 0.028 0.028 0.028

(0.59) (0.59) (0.63) (0.90) (0.90) (0.93)Manager Funds 0.003 0.003 0.003 0.001 0.000 0.000

(1.24) (1.16) (1.01) (0.28) (0.11) (0.10)Manager Size 0.009 0.004 0.003 0.009 0.005 0.004

(0.77) (0.35) (0.29) (0.78) (0.41) (0.39)Fund Affiliation -0.001 -0.002 -0.002 0.000 -0.000 -0.000

(-0.49) (-0.69) (-0.62) (0.10) (-0.07) (-0.02)Constant -0.374∗∗∗ -0.362∗∗∗ -0.363∗∗∗ -0.303∗∗∗ -0.298∗∗∗ -0.298∗∗∗

(-3.63) (-3.51) (-3.52) (-3.03) (-2.97) (-2.98)Time Dummies Yes Yes Yes Yes Yes YesStyle Dummies Yes Yes Yes Yes Yes YesObservations 80059 80059 80059 80059 80059 80059r2 0.022 0.022 0.022 0.022 0.022 0.022

28

Table 7: Managerial Type and Performance (II)

This table presents the results of monthly portfolio fixed effect (Panel A), manager (Panel B) and family fixed effect regressions(Panel C) of Objective-adjusted returns on fund, manager and family characteristics. Fund returns are the actual returns beforededucting fees and expenses (gross) and manager returns are the TNA-weighted average return of all the portfolios managed bythe same manager at the same time. The dependent variable are fund and manager performance, measured by substracting themedian return of their investment objective peers, from the actual return of the fund and manager, respectively. Generalist is adummy variable equals 1 if the fund is managed by manager that is in charge on funds from different investment styles. Timer isa dummy equals 1 if the fund is managed by portfolio manager that has been able to time the market during the past 24 months.Picker is a dummy variable equals 1 if the fund is managed by a manager that was able to pick stocks efficiently during the past24 months. All variables are lagged one period. Control variables and time and investment style dummies are included but notreported; t-statistics are reported in parentheses. We adjust for serial correlation by clustering standard errors at the fund level.* denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Panel A: Fund Fixed EffectFund Performance Manager Performance

Generalist -0.121∗∗ -0.037 -0.089 -0.115∗∗ -0.030 -0.069(-2.15) (-0.70) (-1.58) (-2.10) (-0.57) (-1.26)

Timer -0.082 -0.061 -0.083 -0.062(-1.43) (-1.06) (-1.45) (-1.09)

Generalist × Timer 0.315∗∗∗ 0.294∗∗∗ 0.252∗∗∗ 0.227∗∗∗

(3.23) (3.01) (2.96) (2.64)Picker 0.284∗∗∗ 0.279∗∗∗ 0.284∗∗∗ 0.279∗∗∗


(-2.67) (-2.47) (-3.45) (-3.28)r2 0.043 0.044 0.044 0.043 0.043 0.043

Panel B: Manager Fixed EffectFund Performance Manager Performance

Generalist -0.170∗∗∗ -0.085∗ -0.137∗∗ -0.161∗∗∗ -0.078 -0.118∗∗

(-3.17) (-1.67) (-2.52) (-3.12) (-1.59) (-2.29)Timer -0.105∗ -0.085 -0.104∗ -0.084

(-1.89) (-1.53) (-1.88) (-1.52)Generalist × Timer 0.324∗∗∗ 0.303∗∗∗ 0.261∗∗∗ 0.237∗∗∗

(3.48) (3.25) (3.21) (2.90)Picker 0.255∗∗∗ 0.248∗∗∗ 0.263∗∗∗ 0.255∗∗∗


(-2.72) (-2.53) (-3.31) (-3.15)r2 0.039 0.039 0.039 0.040 0.040 0.040

Panel C: Family Fixed EffectFund Performance Manager Performance

Generalist -0.064∗ 0.003 -0.038 -0.063∗ 0.008 -0.022(-1.73) (0.08) (-0.98) (-1.82) (0.25) (-0.61)

Timer -0.074 -0.053 -0.072 -0.051(-1.42) (-1.02) (-1.42) (-1.01)

Generalist × Timer 0.257∗∗∗ 0.242∗∗∗ 0.201∗∗∗ 0.182∗∗

(2.96) (2.78) (2.61) (2.35)Picker 0.285∗∗∗ 0.281∗∗∗ 0.292∗∗∗ 0.289∗∗∗


(-2.69) (-2.53) (-3.62) (-3.48)r2 0.028 0.028 0.028 0.029 0.029 0.029

Control Variables Yes Yes Yes Yes Yes YesTime Dummies Yes Yes Yes Yes Yes YesStyle Dummies Yes Yes Yes Yes Yes YesObservations 80059 80059 80059 80059 80059 80059

29

Table 8: Managerial Skill and Performance: Subsamples

This table presents the results of monthly panel regressions of portfolio performance on manager type (generalist vs specialist)and other characteristics. Fund returns are actual returns before deducting fees and expenses (gross) and manager returns areTNA-weighted average returns of all the portfolios run by the same manager at the same time. The dependent variables are fund–columns (1)-(3)– and manager –(4)-(6)– performance, measured by subtracting the median return of their investment objectivepeers, from the actual return of the fund and manager, respectively. Observations are sorted into funds managed by Timers incolumns 1 and 4, Pickers in columns 2 and 5, and by Unskilled managers in column 3 and 6. Generalist is a dummy variable equalto 1 if the fund is run by a manager who is in charge on funds with different investment objectives. All variables are lagged oneperiod. A full description of the remaining variables is provided in the appendix. Time and Style dummies are included but notreported; t-statistics are reported in parentheses. Standard errors are clustered at the fund level. * denotes significance at the10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Fund Performance Manager PerformanceTimer Picker Unskilled Timer Picker Unskilled

Generalist 0.199∗∗ -0.143∗ 0.004 0.146∗ -0.195∗∗ 0.008(2.45) (-1.83) (0.11) (1.96) (-2.44) (0.24)

Fund Size 0.025 0.054∗ 0.039∗∗∗ 0.017 0.039 0.028∗∗

(0.92) (1.79) (2.87) (0.59) (1.33) (2.09)Fund Age -0.006 0.002 -0.001 -0.004 0.003 -0.000

(-1.57) (0.39) (-0.38) (-1.15) (0.64) (-0.15)Fund Turnover 0.071∗∗ 0.042∗∗∗ -0.001 0.043 0.012 0.001

(2.20) (4.00) (-0.08) (1.54) (0.90) (0.07)Fund Expenses 0.200∗∗ 0.146 0.157∗∗∗ 0.147∗ 0.131 0.139∗∗∗

(2.53) (1.56) (4.14) (1.94) (1.51) (3.77)Fund Flows 0.060∗ 0.061∗∗∗ 0.064∗∗∗ 0.047 0.054∗∗ 0.065∗∗∗

(1.82) (2.69) (4.22) (1.51) (2.50) (4.43)Past Year Return -0.154 -1.761∗∗∗ -0.364∗∗∗ -0.062 -1.586∗∗∗ -0.368∗∗∗

(-0.46) (-4.97) (-3.18) (-0.20) (-4.44) (-3.27)Family Size 0.010 -0.040∗ 0.026∗∗∗ 0.028 -0.014 0.027∗∗∗

(0.42) (-1.79) (2.73) (1.23) (-0.63) (2.87)Family Funds 0.003∗∗ 0.003∗ -0.001 0.003∗ 0.002∗ -0.001

(2.13) (1.96) (-1.58) (1.72) (1.72) (-1.07)Family Managers -0.007∗∗ 0.000 0.000 -0.007∗∗ 0.000 -0.000

(-2.19) (0.07) (0.10) (-2.36) (0.20) (-0.29)Supply Advising -0.116 0.033 -0.088∗∗∗ -0.094 -0.005 -0.097∗∗∗

(-1.29) (0.36) (-2.62) (-1.18) (-0.05) (-2.98)Demand Advising -0.134∗∗ -0.144∗∗ 0.031 -0.055 -0.150∗∗ 0.039

(-1.98) (-2.03) (1.18) (-0.89) (-2.20) (1.53)MBA -0.115 -0.033 0.020 -0.183∗∗ -0.059 0.026

(-1.48) (-0.47) (0.73) (-2.58) (-0.88) (0.94)PhD -0.277 -0.077 -0.035 -0.239 -0.065 -0.026

(-1.58) (-0.55) (-0.65) (-1.44) (-0.47) (-0.50)Past Positions 0.035 -0.051∗∗ 0.000 0.046∗ -0.047∗∗ -0.002

(1.24) (-2.13) (0.03) (1.83) (-2.07) (-0.16)Ivy League 0.028 -0.048 0.016 0.114 -0.029 0.009

(0.32) (-0.54) (0.42) (1.43) (-0.33) (0.26)Manager Funds 0.004 0.001 0.002 -0.008 -0.013∗∗ 0.003

(0.63) (0.15) (0.69) (-1.55) (-1.97) (0.99)Manager Size 0.059∗∗ 0.002 -0.008 0.052∗ -0.002 -0.004

(2.01) (0.08) (-0.59) (1.79) (-0.07) (-0.29)Fund Affiliation -0.001 -0.010 -0.001 0.008 -0.006 -0.000

(-0.09) (-1.11) (-0.39) (0.85) (-0.68) (-0.13)Constant -0.836∗∗∗ 0.998∗∗∗ -0.596∗∗∗ -0.757∗∗ 0.955∗∗∗ -0.525∗∗∗

(-2.65) (3.37) (-5.20) (-2.52) (3.37) (-4.71)Time Dummies Yes Yes Yes Yes Yes YesStyle Dummies Yes Yes Yes Yes Yes YesObservations 11197 15473 57158 11197 15473 57158r2 0.036 0.032 0.019 0.035 0.031 0.019

30

Table 9: Portfolio Manager Mis-allocation

This table presents the results of monthly logistic regressions of portfolio manager misallocation within their management companieson fund, manager and family characteristics. The dependent variable is a dummy variable that equals 1 if a Picker is allocated asa Generalist (Generalist Misalloaction) or a Timer is allocated as a Specialist (Specialist Misallocation). All variables are laggedone period. A description of fund, manager and family variables is in the appendix. The sample contains all U.S. mutual funds runby an individual portfolio manager from 1996 to 2011. Time and Investment Objective dummies are included but not reported;t-statistics are reported in parentheses. Standard errors are clustered at the fund level. * denotes significance at the 10% level, **denotes significance at the 5% level and *** denotes significance at the 1% level.

Generalist Mis-allocation Specialist Mis-allocationCoef/t Mfx/Std Coef/t Mfx/Std

Fund Size -0.248∗∗∗ -0.006∗∗∗ 0.042 0.003(-4.677) 1.792 (1.192) 1.791

Fund Age -0.002 -0.000 0.004 0.000(-0.185) 9.491 (0.784) 9.144

Fund Turnover 0.086∗∗∗ 0.002∗∗∗ 0.004 0.000(3.593) 1.683 (0.282) 1.700

Fund Expenses -0.116 -0.003 0.261∗∗∗ 0.017∗∗∗

(-0.634) 0.537 (3.076) 0.535Fund Flows 0.017 0.000 -0.008 -0.001

(1.412) 2.275 (-0.584) 2.302Past Year Return 0.285∗ 0.007∗ -1.326∗∗∗ -0.085∗∗∗

(1.782) 0.196 (-7.568) 0.198Family Size -0.131∗∗∗ -0.003∗∗∗ 0.044∗∗∗ 0.003∗∗∗

(-4.133) 2.565 (2.896) 2.554Family Funds -0.011∗∗∗ -0.000∗∗∗ 0.001 0.000

(-3.454) 38.983 (0.947) 39.404Family Managers -0.024∗∗∗ -0.001∗∗∗ -0.001 -0.000

(-5.134) 18.511 (-0.248) 18.590Supply Advising -0.114 -0.003 0.022∗∗ 0.001∗∗

(-0.668) 0.473 (2.064) 0.471Demand Advising 0.134 0.003 0.101 0.007

(0.863) 0.492 (1.442) 0.492MBA 0.330∗∗ 0.008∗∗ 0.116 0.007

(2.160) 0.499 (1.591) 0.499PhD 0.570∗∗ 0.015∗∗ -0.384∗ -0.025∗

(2.003) 0.179 (-1.771) 0.176Past Positions -0.021 -0.001 0.019 0.001

(-0.409) 1.364 (0.677) 1.352Ivy League -0.257 -0.007 -0.046 -0.003

(-1.385) 0.428 (-0.563) 0.430Manager Funds 0.051∗∗∗ 0.001∗∗∗ 0.066∗∗∗ 0.004∗∗∗

(3.098) 4.443 (8.617) 4.500Manager Size 0.772∗∗∗ 0.020∗∗∗ -0.094∗∗∗ -0.006∗∗∗

(14.985) 1.846 (-2.754) 1.835Fund Affiliation 0.002 0.000 0.005 0.000

(0.098) 4.446 (0.559) 4.393Time Dummies Yes YesStyle Dummies Yes YesObservations 80059 80059Pseudo R2 0.180 0.091Baseline Predicted Prob 0.054 0.089

31

Table 10: Specialist Manager Switching to Generalist

This table presents the monthly logistic regressions of manager switches from specialist to generalist on manager and othercharacteristics. The dependent variable is a dummy variable that equals 1 if a specialist becomes generalist in the next month, and0 otherwise. Timer is a dummy variable that takes the value of 1 if the manager has been timing the market significantly for thepast 24 months. Picker is a dummy variable that takes the value of 1 if the manager has been selecting stocks successfully for thepast 24 months, and 0 otherwise. Manager Past Skill is the TNA-weighted cumulative returns of the objective-adjusted returns ofall the funds run by the manager during the past 24 months. All variables are lagged one period. A description of the remainingvariables is in the appendix. The sample contains all U.S. mutual funds managed by a specialist from 1996 to 2011. Time andInvestment Objective dummies are included but not reported; t-statistics are in parentheses. Standard errors are clustered at thefund level. * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1%level.

Model 1 Model 2 Model 3 Model 4Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std

Timer 0.711∗∗∗ 0.003∗∗∗ 0.806∗∗∗ 0.003∗∗∗ 0.876∗∗∗ 0.003∗∗∗

(4.800) 0.310 (5.454) 0.310 (5.067) 0.314Picker 0.679∗∗∗ 0.002∗∗∗ 0.769∗∗∗ 0.003∗∗∗ 0.766∗∗∗ 0.002∗∗∗

(4.681) 0.326 (5.288) 0.326 (4.358) 0.332Manager Past Skill 0.004 0.000

(0.962) 17.683Timer × Manager Past Skill 0.021∗∗∗ 0.001∗∗∗

(3.265) 4.611Picker × Manager Past Skill -0.028∗∗∗ -0.001∗∗∗

(-3.915) 10.278Fund Size -0.273∗∗∗ -0.001∗∗∗ -0.276∗∗∗ -0.001∗∗∗ -0.273∗∗∗ -0.001∗∗∗ -0.262∗∗∗ -0.001∗∗∗

(-5.680) 1.773 (-5.733) 1.773 (-5.680) 1.773 (-4.710) 1.810Fund Age -0.001 -0.000 0.001 0.000 0.001 0.000 0.008 0.000

(-0.112) 9.062 (0.242) 9.062 (0.155) 9.062 (1.044) 9.211Fund Turnover 0.040∗ 0.000∗ 0.039∗ 0.000∗ 0.041∗ 0.000∗ 0.033 0.000

(1.844) 1.408 (1.824) 1.408 (1.921) 1.408 (1.166) 1.331Fund Expenses -0.434∗∗∗ -0.002∗∗∗ -0.400∗∗ -0.001∗∗ -0.452∗∗∗ -0.002∗∗∗ -0.517∗∗∗ -0.002∗∗∗

(-2.590) 0.533 (-2.418) 0.533 (-2.699) 0.533 (-2.649) 0.534Fund Flows 0.004 0.000 0.001 0.000 0.002 0.000 0.009 0.000

(0.216) 2.488 (0.065) 2.488 (0.141) 2.488 (0.554) 2.335Past Year Return 0.305 0.001 -0.019 -0.000 0.113 0.000 0.180 0.001

(1.467) 0.196 (-0.090) 0.196 (0.521) 0.196 (0.564) 0.188Family Size -0.148∗∗∗ -0.001∗∗∗ -0.144∗∗∗ -0.001∗∗∗ -0.146∗∗∗ -0.001∗∗∗ -0.145∗∗∗ -0.000∗∗∗

(-4.992) 2.542 (-4.754) 2.542 (-4.895) 2.542 (-4.219) 2.560Family Funds -0.004∗ -0.000∗ -0.004∗ -0.000∗ -0.004∗ -0.000∗ -0.003 -0.000

(-1.812) 40.412 (-1.820) 40.412 (-1.810) 40.412 (-1.417) 39.002Family Managers -0.002 -0.000 -0.002 -0.000 -0.002 -0.000 -0.004 -0.000

(-0.601) 18.531 (-0.575) 18.531 (-0.580) 18.531 (-1.016) 17.831Supply Advising 0.076 0.000 0.034 0.000 0.043 0.000 0.193 0.001

(0.527) 0.470 (0.237) 0.470 (0.297) 0.470 (1.096) 0.474Demand Advising 0.397∗∗∗ 0.001∗∗∗ 0.446∗∗∗ 0.002∗∗∗ 0.419∗∗∗ 0.001∗∗∗ 0.459∗∗∗ 0.002∗∗∗

(3.153) 0.491 (3.492) 0.491 (3.310) 0.491 (3.051) 0.493MBA 0.139 0.001 0.141 0.001 0.130 0.000 0.285∗ 0.001∗

(1.045) 0.499 (1.067) 0.499 (0.977) 0.499 (1.759) 0.499PhD 0.364 0.001 0.317 0.001 0.335 0.001 -0.136 -0.000

(1.154) 0.170 (0.964) 0.170 (1.029) 0.170 (-0.294) 0.158Past Positions 0.027 0.000 0.022 0.000 0.019 0.000 -0.051 -0.000

(0.620) 1.353 (0.514) 1.353 (0.454) 1.353 (-0.960) 1.328Ivy League -0.050 -0.000 -0.071 -0.000 -0.064 -0.000 -0.179 -0.001

(-0.326) 0.427 (-0.461) 0.427 (-0.414) 0.427 (-0.921) 0.434Manager Funds 0.052∗∗∗ 0.000∗∗∗ 0.054∗∗∗ 0.000∗∗∗ 0.050∗∗∗ 0.000∗∗∗ 0.063∗∗∗ 0.001∗∗∗

(2.901) 4.667 (3.116) 4.667 (2.816) 4.667 (3.130) 5.176Manager Size 0.643∗∗∗ 0.002∗∗∗ 0.633∗∗∗ 0.002∗∗∗ 0.622∗∗∗ 0.002∗∗∗ 0.572∗∗∗ 0.002∗∗∗

(12.471) 1.806 (12.055) 1.806 (12.066) 1.806 (9.451) 1.812Fund Affiliation -0.136∗∗∗ -0.000∗∗∗ -0.136∗∗∗ -0.000∗∗∗ -0.137∗∗∗ -0.000∗∗∗ -0.148∗∗∗ -0.001∗∗∗

(-5.904) 4.309 (-5.931) 4.309 (-5.945) 4.309 (-5.461) 4.286Time Dummies Yes Yes Yes YesStyle Dummies Yes Yes Yes YesObservations 62235 62235 62235 62235Pseudo R2 0.116 0.116 0.122 0.140Baseline Predicted Prob 0.012 0.012 0.013 0.013

32

Table 11: Performance around Managerial Switching: Event Study

We present average fund performance in Panel A and manager performance in Panel B over 6 months before and 12 months after the manager goes from specialist to generalist. The performance is expressedin percentage and corresponds to the semi-annual cumulative objective-adjusted return before fees and expenses (gross). The analysis is for the entire sample from 1996 to 2011. Our sample is divided intofunds run by timers, pickers and unskilled portfolio managers. Panel A accounts for funds that go from being managed by a timer, picker or unskilled specialist to be managed by a generalist (the samemanager or new one) with the same type of skill. Panel B considers funds that keep the same manager that were specialist timer, picker or unskilled and become generalist (independently of the new skillthey might have gained). We also report differences between the cumulative performance before and after each event. * denotes significance at the 10% level, ** denotes significance at the 5% level and ***denotes significance at the 1% level.

Panel A: Fund Performancet-6 t-5 t-4 t-3 t-2 t-1 t t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t+10 t+11 t+12

Timer -1.39 -1.14 -0.92 -2.34 -2.52 -0.07 0.92 -3.66 -2.44 -0.47 1.51 1.34 0.05 2.79 7.14 0.99 1.00 2.90 3.68Picker 0.10 -0.08 -0.16 0.30 0.11 0.57 -0.04 0.02 0.49 0.94 0.31 -0.36 -0.84 -0.36 -0.67 -0.94 -0.99 -1.17 -2.10Unskilled -0.95 -0.76 -0.32 -0.54 -0.89 -0.44 -0.96 -0.56 0.15 2.50 2.09 1.24 -0.69 -0.51 -0.02 -0.52 -1.34 -1.10 -2.01

Cumulative Performance DifferencePrior Event - Post Event

Value t-statTimer -6.194 -2.57Picker 2.211 2.32

Unskilled 1.113 1.64

Panel B: Manager Performancet-6 t-5 t-4 t-3 t-2 t-1 t t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t+10 t+11 t+12

Timer -0.61 0.19 0.17 -1.12 -1.70 -0.08 0.77 -0.21 1.04 2.51 3.29 2.66 1.36 1.18 1.20 0.57 0.16 0.43 1.89Picker 0.09 0.25 0.11 0.16 0.17 0.47 -0.03 0.14 0.14 0.88 0.79 -0.25 -1.25 -0.77 -1.03 -1.21 -1.24 -1.32 -1.76Unskilled -0.62 -0.80 -0.35 -0.45 -0.89 -0.43 -0.47 -0.09 0.63 2.28 2.64 1.70 -0.07 0.07 0.12 -0.47 -1.24 -0.70 -2.00

Cumulative Performance DifferencePrior Event - Post Event

Value t-statTimer -3.587 -2.33Picker 1.933 2.15

Unskilled 1.109 1.79

33

Table 12: Propensity Score Matching: Fund Performance and Manager Skill

In this table, we identify a control sample of funds managed by specialist with two different propensity score matching procedures:Nearest Neighbor of Rosenbaum and Rubin (1983), and the Kernel Matching of Heckman, Ichimura and Todd, (1997, 1998). Toestimate the propensity score we use the size, age, turnover, expenses, flows and past returns of the funds, the size and the numberof funds and managers of the family, and the number of prior positions, the length of time the manager has been managing the fund,the number of funds and the total amount of assets the manager has currently under management. We require that the differencebetween the propensity score of the funds managed by a specialist and the matching peer does not exceed 0.1% in absolute value.We then compare the fund performance between the two groups and report the value of the difference (Generalist-Specialist) andthe statistical significance using bootstrapped standard errors associated to that difference. Fund performance is the excess returnover the mean of the style, using the gross and net returns of the portfolio. We group the funds into quintiles according to thetiming ability (Panel A) and picking managerial skills (Panel B) during the period 1996-2011.* denotes significance at the 10%level, ** denotes significance at the 5% level and *** denotes significance at the 1% level

Generalist vs SpecialistPanel A:Timing Q1 Q2 Q3 Q4 Q5

Nearest Kernel Nearest Kernel Nearest Kernel Nearest Kernel Nearest KernelGross Ret 0.110 0.013 0.179 0.278 0.072 0.034 0.314∗∗ 0.375∗∗∗ 0.422∗∗∗ 0.685∗∗∗

Net Ret 0.111 -0.011 0.158 0.250 0.067 0.024 0.137 0.189 0.455∗∗∗ 0.720∗∗∗

Panel B:Picking Q1 Q2 Q3 Q4 Q5

Nearest Kernel Nearest Kernel Nearest Kernel Nearest Kernel Nearest KernelGross Ret 0.342 0.232 -0.010 -0.023 -0.017 -0.032 - 0.136 -0.209 -0.894∗∗∗ -0.738∗∗∗

Net Ret 0.351 0.240 -0.002 -0.031 -0.037 -0.010 -0.139∗∗∗ -0.218∗∗∗ -0.851∗∗∗ -0.716∗∗∗

Table 13: Selection Bias: Heckman’s two-step procedure (1st Stage)

In this table, we show the estimates from the first stage Heckman’s two-step procedure. The model estimates the probability thata manager works for a firms with multi-funds policy. We present the monthly logistic regressions of managers running more thanone fund at the same time on fund and family characteristics. Funds per Manager is the total funds of the family divided by thenumber of manager of that firm. All variables are lagged one period. A complete description of the variables is provided in theappendix. The sample contains all U.S. mutual funds managed by a specialist portfolio manager from 1996 to 2011. Time andStyle dummies are included but not reported; t-statistics are reported in parentheses. Standard errors are clustered at the fundlevel. * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Multi-Funds Policy=1

Coef/t Mfx/StdFund Size 0.003 -0.001

(0.180) 1.762Fund Age -0.000 0.000

(-0.090) 9.232Fund Turnover -0.031 0.000

(-1.614) 1.498Fund Expenses -0.065 -0.004

(-0.921) 0.542Fund Flows 0.001 0.000

(0.179) 2.286Past Year Return -0.060 0.003

(-0.896) 0.201Funds per Manager 0.007∗∗∗ 0.001∗∗∗

(2.707) 8.684Time Dummies YesStyle Dummies YesObservations 90955

34

Table 14: Selection Bias: Heckman’s two-step procedure (2nd Stage)

In this table, we show the estimates from the second stage Heckman’s two-step procedure to examine how different managercharacteristics affect the probability of a specialist being transfered to generalist, conditioned on being a multi-fund manager. Allvariables are lagged one period. A complete description of the variables is provided in the appendix. The sample contains allU.S. mutual funds managed by a specialist portfolio manager from 1996 to 2011. Time and Style dummies are included but notreported; t-statistics are reported in parentheses. Standard errors are clustered at the fund level. * denotes significance at the10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Model 1 Model 2 Model 3 Model 4Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std

Timer 0.215∗∗∗ 0.017∗∗∗ 0.257∗∗∗ 0.015∗∗∗ 0.309∗∗∗ 0.012∗∗∗

(3.235) 0.298 (4.096) 0.298 (4.359) 0.293Picker 0.233∗∗∗ 0.013∗∗∗ 0.265∗∗∗ 0.014∗∗∗ 0.267∗∗∗ 0.009∗∗∗

(3.863) 0.316 (4.313) 0.316 (3.829) 0.317Manager Past Skill 0.003 0.000

(1.544) 18.341Timer*Manager Past Skill 0.010∗∗∗ 0.002∗∗∗

(2.941) 5.148Picker*Manager Past Skill -0.015∗∗∗ -0.001∗∗∗

(-4.839) 9.557Fund Size -0.027 -0.002 -0.030 -0.002 -0.029 -0.002 -0.025 -0.001

(-1.266) 1.755 (-1.380) 1.755 (-1.344) 1.755 (-1.028) 1.762Fund Age 0.005 0.000 0.005 0.000 0.005 0.000 0.007∗ 0.000∗

(1.356) 9.096 (1.390) 9.096 (1.386) 9.096 (1.783) 9.232Fund Turnover 0.024 0.002 0.021 0.001 0.021 0.001 0.014 0.000

(1.354) 1.478 (1.257) 1.478 (1.283) 1.478 (0.812) 1.498Fund Expenses -0.084 -0.006 -0.088 -0.005 -0.100 -0.006 -0.129∗ -0.004∗

(-1.022) 0.540 (-1.215) 0.540 (-1.361) 0.540 (-1.723) 0.542Fund Flows 0.003 0.000 0.002 0.000 0.002 0.000 0.006 0.000

(0.433) 2.260 (0.301) 2.260 (0.340) 2.260 (1.011) 2.286Past Year Return 0.116 0.009 0.028 0.002 0.060 0.003 0.089 0.003

(1.332) 0.203 (0.313) 0.203 (0.678) 0.203 (0.686) 0.201Family Size -0.051∗∗∗ -0.004∗∗∗ -0.050∗∗∗ -0.003∗∗∗ -0.051∗∗∗ -0.003∗∗∗ -0.056∗∗∗ -0.002∗∗∗

(-3.718) 2.598 (-3.793) 2.598 (-3.862) 2.598 (-3.741) 2.614Family Funds -0.002∗∗∗ -0.000∗∗∗ -0.002∗∗∗ -0.000∗∗∗ -0.002∗∗∗ -0.000∗∗∗ -0.002∗∗ -0.000∗∗

(-2.786) 40.769 (-2.803) 40.769 (-2.710) 40.769 (-2.118) 40.481Family Managers 0.002 0.000 0.002 0.000 0.002 0.000 0.002 0.000

(1.595) 17.998 (1.537) 17.998 (1.503) 17.998 (0.948) 17.766Supply Advising -0.018 -0.001 -0.027 -0.002 -0.026 -0.001 0.016 0.001

(-0.295) 0.471 (-0.445) 0.471 (-0.429) 0.471 (0.226) 0.472Demand Advising 0.168∗∗∗ 0.013∗∗∗ 0.196∗∗∗ 0.011∗∗∗ 0.183∗∗∗ 0.010∗∗∗ 0.220∗∗∗ 0.007∗∗∗

(2.933) 0.490 (3.562) 0.490 (3.358) 0.490 (3.521) 0.491MBA 0.084∗ 0.007∗ 0.081 0.004 0.080 0.004 0.119∗ 0.004∗

(1.658) 0.499 (1.544) 0.499 (1.516) 0.499 (1.883) 0.499PhD -0.034 -0.003 -0.062 -0.003 -0.044 -0.002 -0.316 -0.010

(-0.245) 0.177 (-0.425) 0.177 (-0.298) 0.177 (-1.390) 0.167Past Positions 0.012 0.001 0.011 0.001 0.009 0.000 -0.010 -0.000

(0.684) 1.340 (0.592) 1.340 (0.493) 1.340 (-0.484) 1.341Ivy League 0.037 0.003 0.038 0.002 0.035 0.002 -0.004 -0.000

(0.632) 0.424 (0.618) 0.424 (0.575) 0.424 (-0.056) 0.426Manager Funds 0.001 0.000 0.001 0.000 -0.000 -0.000 0.005 0.000

(0.117) 3.911 (0.226) 3.911 (-0.051) 3.911 (0.633) 3.935Manager Size 0.126∗∗∗ 0.010∗∗∗ 0.124∗∗∗ 0.007∗∗∗ 0.120∗∗∗ 0.007∗∗∗ 0.125∗∗∗ 0.004∗∗∗

(4.248) 1.834 (4.737) 1.834 (4.636) 1.834 (4.264) 1.848Fund Affiliation -0.038∗∗∗ -0.003∗∗∗ -0.040∗∗∗ -0.002∗∗∗ -0.039∗∗∗ -0.002∗∗∗ -0.043∗∗∗ -0.001∗∗∗

(-4.284) 4.205 (-4.896) 4.205 (-4.816) 4.205 (-4.591) 4.206Time Dummies Yes Yes Yes YesStyle Dummies Yes Yes Yes YesObservations 90955 90955 90955 90955Baseline Predicted Prob 0.051 0.034 0.035 0.020

35

Table 15: Family Expansion

This table presents the monthly logistic regressions of manager switches from being specialist to generalist on manager and other characteristics. The dependent variable is a dummy variable that equals 1 ifa specialist portfolio manager is transfered to generalist in the next month and 0 otherwise. Timer is a dummy variable that takes the value of 1 if the fund is managed by manager that has been significantlytiming the market during the past 24 months. Picker is a dummy variable that takes the value of 1 if the fund is managed by manager that has been efficiently selecting stock during the past 24 months and 0otherwise. The sample is divided into quartiles based on the Herfindahl index across investment objectives of the fund within the family for each month. We consider Low and High concentrated firms basedon whether the management company is in the first or fourth quantile, respectively. All variables are lagged one period. A description of the remaining variables is provided in the appendix. The samplecontains all U.S. mutual funds managed by a specialist portfolio manager from 1996 to 2011. Time and Investment Objective dummies are included but not reported; t-statistics are reported in parentheses.Standard errors are clustered at the fund level. * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/StdTimer 1.265∗∗∗ 0.002∗∗∗ 1.205∗∗∗ 0.002∗∗∗ 1.320∗∗∗ 0.003∗∗∗ 1.292∗∗∗ 0.002∗∗∗

(4.615) 0.334 (3.256) 0.285 (4.910) 0.334 (3.502) 0.285Picker 0.196 0.000 0.312 0.001 0.396 0.001 0.525 0.001

(0.600) 0.341 (0.723) 0.333 (1.225) 0.341 (1.196) 0.333Fund Size -0.287∗∗∗ -0.001∗∗∗ -0.296∗ -0.001∗ -0.284∗∗∗ -0.001∗∗∗ -0.313∗ -0.001∗ -0.284∗∗∗ -0.001∗∗∗ -0.307∗ -0.001∗

(-3.728) 1.954 (-1.708) 1.589 (-3.586) 1.954 (-1.777) 1.589 (-3.662) 1.954 (-1.744) 1.589Fund Age 0.002 0.000 0.001 0.000 0.002 0.000 0.002 0.000 0.003 0.000 0.002 0.000

(0.177) 10.959 (0.044) 7.014 (0.146) 10.959 (0.081) 7.014 (0.255) 10.959 (0.091) 7.014Fund Turnover -0.105 -0.000 0.081 0.000 -0.115 -0.000 0.080 0.000 -0.139 -0.000 0.086 0.000

(-0.551) 1.533 (1.465) 1.509 (-0.582) 1.533 (1.401) 1.509 (-0.683) 1.533 (1.438) 1.509Fund Expenses -0.190 -0.000 -0.086 -0.000 -0.165 -0.000 -0.059 -0.000 -0.166 -0.000 -0.125 -0.000

(-0.514) 0.448 (-0.281) 0.599 (-0.421) 0.448 (-0.196) 0.599 (-0.447) 0.448 (-0.403) 0.599Fund Flows -0.256∗ -0.000∗ 0.046∗∗∗ 0.000∗∗∗ -0.300∗ -0.001∗ 0.042∗∗∗ 0.000∗∗∗ -0.265∗ -0.001∗ 0.043∗∗∗ 0.000∗∗∗

(-1.735) 2.189 (3.679) 3.225 (-1.687) 2.189 (3.475) 3.225 (-1.758) 2.189 (3.617) 3.225Past Year Return -0.293 -0.001 0.647∗ 0.001∗ -0.557 -0.001 0.446 0.001 -0.306 -0.001 0.575 0.001

(-0.476) 0.173 (1.802) 0.253 (-0.979) 0.173 (1.109) 0.253 (-0.498) 0.173 (1.554) 0.253Family Size -0.217∗∗∗ -0.000∗∗∗ -0.019 -0.000 -0.201∗∗∗ -0.000∗∗∗ -0.009 -0.000 -0.220∗∗∗ -0.000∗∗∗ -0.019 -0.000

(-3.877) 2.399 (-0.159) 2.358 (-3.514) 2.399 (-0.079) 2.358 (-3.918) 2.399 (-0.154) 2.358Family Funds -0.000 -0.000 -0.002 -0.000 -0.000 -0.000 0.001 0.000 -0.000 -0.000 -0.002 -0.000

(-0.149) 56.310 (-0.085) 12.960 (-0.154) 56.310 (0.081) 12.960 (-0.165) 56.310 (-0.085) 12.960Family Managers 0.008 0.000 -0.015 -0.000 0.010∗ 0.000∗ -0.019 -0.000 0.008 0.000 -0.006 -0.000

(1.564) 27.565 (-0.089) 3.770 (1.764) 27.565 (-0.117) 3.770 (1.553) 27.565 (-0.039) 3.770Supply Advising -0.001 -0.000 -0.767∗∗ -0.001∗∗ -0.161 -0.000 -0.766∗ -0.001∗ -0.007 -0.000 -0.797∗∗ -0.001∗∗

(-0.002) 0.402 (-1.994) 0.500 (-0.442) 0.402 (-1.869) 0.500 (-0.018) 0.402 (-2.070) 0.500Demand Advising 1.041∗∗∗ 0.002∗∗∗ 0.579 0.001 1.210∗∗∗ 0.002∗∗∗ 0.576 0.001 1.022∗∗∗ 0.002∗∗∗ 0.570 0.001

(3.441) 0.469 (1.425) 0.479 (3.725) 0.469 (1.396) 0.479 (3.383) 0.469 (1.403) 0.479MBA -0.139 -0.000 -0.039 -0.000 -0.162 -0.000 0.070 0.000 -0.151 -0.000 -0.053 -0.000

(-0.517) 0.499 (-0.106) 0.500 (-0.596) 0.499 (0.186) 0.500 (-0.563) 0.499 (-0.142) 0.500PHD 0.108 0.000 0.692 0.001 -0.059 -0.000 0.701 0.001 0.063 0.000 0.660 0.001

(0.201) 0.160 (1.076) 0.209 (-0.111) 0.160 (1.067) 0.209 (0.114) 0.160 (1.064) 0.209Past Positions 0.030 0.000 0.050 0.000 0.004 0.000 0.078 0.000 0.024 0.000 0.055 0.000

(0.351) 1.320 (0.407) 1.300 (0.048) 1.320 (0.624) 1.300 (0.274) 1.320 (0.446) 1.300Ivy League 0.057 0.000 -0.416 -0.001 0.051 0.000 -0.519 -0.001 0.086 0.000 -0.480 -0.001

(0.213) 0.491 (-0.761) 0.370 (0.191) 0.491 (-0.894) 0.370 (0.322) 0.491 (-0.831) 0.370Manager Funds -0.026 -0.000 0.336∗∗∗ 0.001∗∗∗ -0.019 -0.000 0.345∗∗∗ 0.001∗∗∗ -0.026 -0.000 0.332∗∗∗ 0.001∗∗∗

(-1.108) 7.265 (2.631) 2.100 (-0.793) 7.265 (2.690) 2.100 (-1.111) 7.265 (2.627) 2.100Manager Size 0.764∗∗∗ 0.001∗∗∗ 0.519∗∗∗ 0.001∗∗∗ 0.789∗∗∗ 0.002∗∗∗ 0.545∗∗∗ 0.001∗∗∗ 0.755∗∗∗ 0.001∗∗∗ 0.520∗∗∗ 0.001∗∗∗

(6.449) 1.756 (2.687) 1.666 (6.276) 1.756 (2.834) 1.666 (6.322) 1.756 (2.684) 1.666Fund Affiliation -0.260∗∗∗ -0.001∗∗∗ -0.097∗ -0.000∗ -0.263∗∗∗ -0.001∗∗∗ -0.100∗ -0.000∗ -0.263∗∗∗ -0.001∗∗∗ -0.097∗ -0.000∗

(-5.304) 4.400 (-1.901) 4.539 (-5.287) 4.400 (-1.911) 4.539 (-5.290) 4.400 (-1.912) 4.539Family Concentration Low High Low High Low HighTime Dummies Yes Yes Yes Yes Yes YesStyle Dummies Yes Yes Yes Yes Yes YesObservations 19234 12082 19234 12082 19234 12082Pseudo R2 0.210 0.190 0.195 0.177 0.212 0.192Baseline Predicted Prob 0.011 0.034 0.011 0.033 0.011 0.035

36

Table 16: Downturn Markets

This table presents the monthly logistic regressions of manager switches from being specialist to generalist on manager and other characteristics. The dependent variable is a dummy variable that equals 1 ifa specialist portfolio manager is transfered to generalist in the next month and 0 otherwise. Timer is a dummy variable that takes the value of 1 if the fund is managed by manager that has been significantlytiming the market during the past 24 months. Picker is a dummy variable that takes the value of 1 if the fund is managed by manager that has been efficiently selecting stock during the past 24 months and0 otherwise. Manager Past Skill is measured as the TNA-weighted cumulative returns of the objective-adjusted returns of all the funds run by the manager during the past 24 months. Market Condition ismeasured with the Chicago Fed National Activity Index. All variables are lagged one period. A description of the remaining variables is provided in the appendix. The sample contains all U.S. mutual fundsmanaged by a specialist portfolio manager from 1996 to 2011. Time and Investment Objective dummies are included but not reported; t-statistics are reported in parentheses. Standard errors are clusteredat the fund level. * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/Std Coef/t Mfx/StdTimer 1.135∗∗∗ 0.006∗∗∗ 0.926∗∗∗ 0.007∗∗∗ 1.269∗∗∗ 0.006∗∗∗ 0.912∗∗∗ 0.007∗∗∗

(4.850) 0.288 (3.243) 0.333 (5.248) 0.288 (3.172) 0.333Picker 0.757∗∗∗ 0.004∗∗∗ -0.248 -0.002 0.940∗∗∗ 0.004∗∗∗ -0.097 -0.001

(2.844) 0.311 (-0.709) 0.343 (3.446) 0.311 (-0.274) 0.343Fund Size -0.241∗∗∗ -0.001∗∗∗ -0.496∗∗∗ -0.004∗∗∗ -0.218∗∗∗ -0.001∗∗∗ -0.507∗∗∗ -0.004∗∗∗ -0.215∗∗∗ -0.001∗∗∗ -0.496∗∗∗ -0.004∗∗∗

(-3.119) 1.755 (-5.622) 1.781 (-2.684) 1.755 (-5.651) 1.781 (-2.716) 1.755 (-5.630) 1.781Fund Age 0.000 0.000 0.011 0.000 0.003 0.000 0.012 0.000 0.002 0.000 0.011 0.000

(0.044) 8.622 (1.093) 9.113 (0.300) 8.622 (1.175) 9.113 (0.210) 8.622 (1.071) 9.113Fund Turnover -0.024 -0.000 0.132∗∗ 0.001∗∗ -0.010 -0.000 0.115∗∗ 0.001∗∗ -0.027 -0.000 0.131∗∗ 0.001∗∗

(-0.314) 1.289 (2.451) 1.273 (-0.129) 1.289 (2.228) 1.273 (-0.344) 1.289 (2.448) 1.273Fund Expenses -0.210 -0.001 -0.284 -0.002 -0.105 -0.001 -0.324 -0.002 -0.220 -0.001 -0.279 -0.002

(-0.683) 0.528 (-1.086) 0.521 (-0.345) 0.528 (-1.239) 0.521 (-0.706) 0.528 (-1.062) 0.521Fund Flows -0.006 -0.000 0.023∗ 0.000∗ -0.020 -0.000 0.021∗ 0.000∗ -0.018 -0.000 0.023∗ 0.000∗

(-0.227) 2.630 (1.923) 2.626 (-0.644) 2.630 (1.785) 2.626 (-0.635) 2.630 (1.958) 2.626Past Year Return 1.388∗∗∗ 0.007∗∗∗ 0.600 0.004 0.966∗∗∗ 0.005∗∗∗ 0.533 0.004 1.195∗∗∗ 0.006∗∗∗ 0.627 0.005

(5.940) 0.233 (1.148) 0.169 (3.764) 0.233 (0.997) 0.169 (4.882) 0.233 (1.129) 0.169Family Size 0.006 0.000 -0.181∗∗∗ -0.001∗∗∗ -0.003 -0.000 -0.178∗∗∗ -0.001∗∗∗ 0.001 0.000 -0.181∗∗∗ -0.001∗∗∗

(0.098) 2.536 (-3.428) 2.540 (-0.048) 2.536 (-3.331) 2.540 (0.022) 2.536 (-3.427) 2.540Family Funds -0.004 -0.000 -0.008∗∗ -0.000∗∗ -0.006 -0.000 -0.008∗∗ -0.000∗∗ -0.006 -0.000 -0.008∗∗ -0.000∗∗

(-1.051) 36.516 (-2.312) 42.494 (-1.335) 36.516 (-2.303) 42.494 (-1.268) 36.516 (-2.306) 42.494Family Managers -0.007 -0.000 0.005 0.000 -0.005 -0.000 0.003 0.000 -0.005 -0.000 0.005 0.000

(-0.962) 17.192 (0.780) 18.647 (-0.595) 17.192 (0.530) 18.647 (-0.654) 17.192 (0.759) 18.647Supply Advising -0.351 -0.002 0.893∗∗∗ 0.007∗∗∗ -0.351 -0.002 0.958∗∗∗ 0.007∗∗∗ -0.364 -0.002 0.901∗∗∗ 0.007∗∗∗

(-1.467) 0.471 (3.446) 0.468 (-1.425) 0.471 (3.647) 0.468 (-1.519) 0.471 (3.481) 0.468Demand Advising 0.748∗∗∗ 0.004∗∗∗ -0.335 -0.002 0.875∗∗∗ 0.004∗∗∗ -0.268 -0.002 0.797∗∗∗ 0.004∗∗∗ -0.336 -0.002

(3.497) 0.497 (-1.398) 0.488 (4.048) 0.497 (-1.110) 0.488 (3.690) 0.497 (-1.403) 0.488MBA 0.104 0.001 -0.107 -0.001 0.069 0.000 -0.079 -0.001 0.078 0.000 -0.104 -0.001

(0.508) 0.500 (-0.496) 0.500 (0.335) 0.500 (-0.362) 0.500 (0.377) 0.500 (-0.485) 0.500PHD 0.552 0.003 -2.261∗∗∗ -0.017∗∗∗ 0.671 0.003 -2.217∗∗ -0.017∗∗ 0.533 0.002 -2.238∗∗ -0.016∗∗

(0.970) 0.171 (-2.583) 0.173 (1.195) 0.171 (-2.553) 0.173 (0.924) 0.171 (-2.530) 0.173Past Positions 0.020 0.000 0.013 0.000 0.010 0.000 0.019 0.000 0.008 0.000 0.013 0.000

(0.290) 1.406 (0.169) 1.308 (0.150) 1.406 (0.238) 1.308 (0.107) 1.406 (0.167) 1.308Ivy League 0.201 0.001 0.132 0.001 0.180 0.001 0.122 0.001 0.225 0.001 0.135 0.001

(0.873) 0.428 (0.531) 0.434 (0.769) 0.428 (0.482) 0.434 (0.960) 0.428 (0.541) 0.434Manager Funds 0.009 0.000 0.059∗∗ 0.000∗∗ 0.012 0.000 0.078∗∗∗ 0.001∗∗∗ 0.015 0.000 0.059∗∗ 0.000∗∗

(0.376) 5.192 (2.256) 4.622 (0.530) 5.192 (2.821) 4.622 (0.650) 5.192 (2.279) 4.622Manager Size 0.620∗∗∗ 0.003∗∗∗ 0.728∗∗∗ 0.005∗∗∗ 0.629∗∗∗ 0.003∗∗∗ 0.731∗∗∗ 0.006∗∗∗ 0.590∗∗∗ 0.003∗∗∗ 0.731∗∗∗ 0.005∗∗∗

(7.246) 1.770 (7.948) 1.818 (6.613) 1.770 (7.565) 1.818 (6.745) 1.770 (7.817) 1.818Fund Affiliation -0.155∗∗∗ -0.001∗∗∗ -0.112∗∗∗ -0.001∗∗∗ -0.159∗∗∗ -0.001∗∗∗ -0.113∗∗∗ -0.001∗∗∗ -0.155∗∗∗ -0.001∗∗∗ -0.111∗∗∗ -0.001∗∗∗

(-4.781) 4.459 (-2.977) 4.173 (-4.799) 4.459 (-2.909) 4.173 (-4.773) 4.459 (-2.951) 4.173Market Bear Bull Bear Bull Bear BullTime Dummies Yes Yes Yes Yes Yes YesStyle Dummies Yes Yes Yes Yes Yes YesObservations 12167 6902 12167 6902 12167 6902Pseudo R2 0.183 0.213 0.175 0.204 0.191 0.213Baseline Predicted Prob 0.023 0.049 0.025 0.049 0.025 0.049

37

Supplementary Appendix

A1. Variable definitions

Variable DefinitionFund Characteristics

Fund Size Natural logarithm of TNA under management in millions of dollars.Fund Age Number of years the fund has been offered.Fund Turnover Minimum of aggregate purchases and sales of securities divided by

average TNA over the calendar year.Fund Expenses Total annual expenses and fees divided by year-end TNA.Fund Flow Percentage of new inflows of the fund over the previous year.Alpha 6F Intercept from estimating Carharts model augmented by MSCI World

Index return and the U.S. Aggregate Bond Index return (in excess ofrisk-free).

Objective-adj Returns Portfolio gross return minus the median value of the return of all thefunds within the same investment objective.

Net Return Objective-adj return using net returns instead of gross (before fees).Family Characteristics

Family Size Logarithm of TNA of all funds in the family, excluding the fund itself.Family Funds Logarithm of the number of funds within the fund family.Family Managers Number of portfolio managers within the fund family.Demand Advising Dummy variable equal 1 if the family has at least one fund outsourced

to an unaffiliated firm.Supply Advising Dummy variable equal 1 if the family is managing at least one fund

from an unaffiliated firm.Funds Per Manager Number of total funds of the family divided by total number of man-

agers within the family.Family Concentration Herfindahl index across investment objectives of the fund of the family.

Portfolio Manager CharacteristicsIvy league Dummy variable equals 1 if the manager graduated from an Ivy

League university.MBA Dummy variable equals 1 if the manager holds a MBA degreePhD Dummy variable equals 1 if the manager holds a PhD degreePast Positions Number of prior job positions of the managerManager Funds Number of funds managed simultaneously by a portfolio manager.Manager Size Natural logarithm of the sum of TNA of all the funds the manager is

managing in that period.Fund affiliation Number of years the manager is managing the fund.Family affiliation Number of years the manager is managing funds from the family.Picker Dummy variable equals 1 if the portfolio manager that has been effi-

ciently selecting stock during the past 24 months.Timer Dummy variable equals 1 if the portfolio manager that has been effi-

ciently predicting the market during the past 24 months.Concentration Herfindahl index of concentration among all different investment ob-

jectives of manager funds.Generalist Manager with Concentration variable less than 1.Specialist Manager with Concentration variable equal than 1.Manager Performance TNA-weighted average return of all the funds managed by the same

manager in one period.Manager Past Skill TNA-weighted cumulative return of the objective-adjusted returns of

all the funds run by the manager during the past 24 months.Picking Alpha coefficient for estimating a modified version of the TM Model.Timing Gamma coefficient for estimating a modified version of the TM Model.Multi-Funds Manager Dummy variable equals 1 if the manager is managing more than 1

fund simultaneously.Other Variables

Specialist To Generalist Equals 1 if the manager is specialist in t and generalist in t+1.Generalist To Specialist Equals 1 if the manager is generalist in t and specialist in t+1.Market Condition The Chicago Fed National Activity Index.Generalist Mis-allocation Equals 1 if a manager with picking ability is allocated as generalist.Specialist Mis-allocation Equals 1 if a manager with timing skill is allocated as specialist.

38

A2. Asset classes and investment styles definitions

Under the Investment Act of 1940, an investment company has to register with the

Securities and Exchange Commission (SEC). All U.S. mutual funds and other regulated

investment management companies are required to file Form NSAR (along with other

documents) on a semi-annual basis. According to this Form, the filer must classify the

funds attending to different Asset Classes and Investment Objective. Here is the definition

of these categories that the SEC gives to the registrant and that we will use to classify

the funds of our database:

ASSET CLASS:

• Equity: invests in equity securities, options and futures on equity securities, indices

of equity securities or securities convertible into equity securities.

• Fixed Income: invests primarily in debt securities, including convertible debt secu-

rities, options and futures on debt securities or indexes of debt securities.

• International: have more than 50% of its net assets at the end of the current period

invested in securities located primarily in countries other than the United States.

INVESTMENT STYLE

1. Equity Funds:

• Capital: primarily and regularly seeks short and intermediate-term return

by investing in moderate to high-risk securities, with little or no concern for

receipt of income.

• Growth: seeks long-term growth, with a moderate degree of risk. Receipt of

income may be considered to some degree in selecting investments.

• Income: primarily and regularly makes low risk investments with the objective

of capital growth and income production.

• Return: portfolio includes a varying mix of equity and debt securities.

2. Fixed Income Funds:

• Government Short-Term: Short-Term Maturities of U.S. Treasury, U.S. Gov-

ernment Agency and State and municipal tax-free.

• Government Long-Term: Intermediate & Long-Term Maturities of U.S. Trea-

sury U.S. Government Agency, State and Municipal tax-free.

• Corporate: Intermediate & Long-Term Maturities of Corporate assets.

– Short-term maturities are defined for purposes of this form as securities

with maturities of 12 months or less. Securities having variable or floating

39

interest rates or subject to a demand feature should be considered short-

term if the interest rate adjustment period or the demand period is 12

months or less. Intermediate and long-term maturities include all other

debt securities.

3. Balance: at least 25% of the value of the assets fund should be invested in either

debt securities, preferred stock, or some combination of both. If convertible senior

securities are included in the required 25%, only that portion of their value at-

tributable to their fixed income characteristics may be used in calculating the 25%

figure.

4. Foreign: Invest more than 50% of its net assets in securities located primarily in

countries other than the United States.

A3. Generalist vs Specialist

In this section, we also examine the frequency of generalists across the different in-

vestment objectives sorted by quintiles of portfolio sizes. We find a larger proportion of

generalists among smaller income funds (42%), larger return funds (47%) and above me-

dian balance funds (53%-54%). On the other hand, the largest proportions of specialists

are within large short-term debt funds and international funds.

[Insert Table A1 here]

We further sorts our sample of managers into specialists and generalists and shows

the difference in fund performance between managers with timing ability or stock-picking

skill against those without any type of skill. Since we have fixed income and international

stocks, in addition to domestic equity funds, we measure fund performance using the al-

pha of Carhart’s model augmented by two risk factors –the MSCI World Index return

and the U.S. Aggregate Bond Index return, both in excess of the risk free rate. By defi-

nition, pickers and timers are better performers than managers lacking these skills. More

interesting is the fact that their performance advantage is greater for pickers among spe-

cialists and timers among generalists, except for the international category. For example,

for equity funds, a specialist with stock picking skill delivers an extra 91.2 bps per month,

while a generalist with similar skill only achieves 42.2 bps per month. On the other hand,

timing ability means an extra 12.5 bps per month for generalists, but only 8.1 bps for

specialists. The exception is international funds, where timing ability seems to be very

profitable: specialists who manage international funds achieve an average of 37 bps extra

per month. Intuitively, international funds invest in a broad range of assets and a timer

is in a good position to run them, using the argument we just stated.


40

A4. Switching Between Generalist and Specialist

In Tables A3 and A4 we study the characteristics of the funds run by managers who

switch roles. Table A3 compares the characteristics of the funds managed by managers

who have just switched roles with those of managers who have not switched and have the

same function (specialist or generalist) as the manager in question after the switch. In

Table 10 we compare the difference in characteristics between the funds managed before

and after the switch by a given manager.

In our sample, we have 1149 intra-firm manager switches; in 561 of these cases, a

specialist becomes generalist, and in 588, a generalist becomes specialist. Table A3 shows

that managers reassigned from specialist to generalist within the firm have higher tenure

at the company and hold a PhD degree and a quantitative background, manage funds

with large volume, and on average show higher turnover and flows and lower fees than

those run by other generalists at the time of the switch. In addition, their firms are also

more likely to outsource funds –prior literature has shown that management companies

outsource funds as an attempt to offer a wider variety of investment choices. Thus, there

is a higher probability of switch to generalist in firms that usually demand sub-advisory

services. A possible explanation is that management firms that demand sub-advisory

services are considering an expansion of the number of objectives they cover in-house;

when they decide to start a fund with a new objective, a former specialist is charged

with the management of the new fund, without dropping the funds managed up to that

point; the management firm uses existing in-house talent, instead of hiring outside. On

the other hand, managers who just switched from generalist to specialist run fewer funds

and during shorter periods, and these funds are bigger and older. These moves are more

likely to take place in firms with a larger number of funds that offer external sub-advisory

services.

Between-firms changes of management function are less frequent: In our sample,

we have 165 moves from specialist to generalist and 162 from generalist to specialist.

Changes between firms -without a change in the type of management– are more common

for specialists (1349 times) while there are only 306 changes for generalists. In general

these transfers are more likely for managers that had a shorter tenure at smaller families

and were managing less assets and funds. The funds they were managing were smaller,

younger and more expensive. Those who change companies to be specialists, whether

generalists or specialists previously, are more likely to hold MBA degrees and have more

experience in past positions, while they were managing funds receiving larger flows. Those

who change to be generalists are more likely to have graduated in an Ivy school and end up

in firms that own a wider variety of products for which they demand sub-advising services–

again, consistent with our argument that they adopt a generalist role in a management

firm that is growing.

41


In Table A4 we show that a manager moves to a different family –for the same

or different function– on average will run younger funds, especially when changing to

specialist. The firms of destination, are smaller and offer fewer funds. Of course, new,

growing firms, are more likely to have to hire outside talent. In the same spirit, managers

who change functions within the family end up managing more funds, and when the

change is to generalist, they manage more assets as well.


A5. Managerial Skill and Concentration

Our main hypothesis is that portfolio managers with a certain skill (timing of picking)

are better suited to exert different functions -generalists or specialists. In order to test

this, we replicate equation (4) using continuous variables instead of dummies for manager

function and skill. Generalist is a dummy variable equals 1 if the fund is managed by

manager that is in charge on funds from different investment styles. Timing and Picking

are the gamma and alpha coefficients from estimating a modified version of the TM

Timing model. Concentration measures the level of diversification of fund i managed by

j in month t (i might represents several funds, if the manager runs more than one). In

particular, this variable is a Herfindahl index:∑9

s=1

(TNAs,j,t

TNAj,t

)2

, with s the “fund style”

as defined in the NSAR-B filings (capital appreciation, growth, income, total return,

government short-term debt, government long-term debt, corporate debt, balance and

international stocks)33 and TNAs,j,t total net assets managed by manager j according

to investment style s at time t. Therefore, the higher the index, the more focused the

portfolio is. Timer is a dummy equals 1 if the fund is managed by portfolio manager that

has been able to time the market during the past 24 months. Picker is a dummy variable

equals 1 if the fund is managed by a manager that was able to pick stocks efficiently

during the past 24 months.

Table A5 shows the results of monthly Pooled OLS regressions of fund and manager

investment objective-adjusted returns on fund, manager and family characteristics. Hav-

ing picking skill is always associated to higher fund and manager performance, while

market-timing ability is positive related to performance only for generalist managers.

Additionally, the relationship between concentration and fund performance is highly

positive for funds managed by pickers, and has no effect for funds managed by timers. In

economic terms, one standard deviation increase in concentration (0.15) leads to an ab-

normal return increase of 108 bps per year in fund performance and 153 bps on manager

33A full description of these investment objectives is in the Appendix.

42

performance for funds managed by stock pickers. This means that management compa-

nies can obtain a greater output by allowing managers with picking ability to manage

similar funds.


A6. Managerial Skill by Portfolio Manager Type

We have established that managers with stock picking ability are better suited to work

as specialists, while managers with timing ability are better as generalists. We want to

analyze further the effect of managerial skills on performance. With that goal, we split

our sample into funds managed by generalists and funds managed by specialists, and we

estimate the following model for each subsample:

OARi,t = a0 + a1Timerj,t + a2Pickerj,t + a3Xi,t + δt + ei,t (9)

Table A5 shows the results of estimating (9) using pooled OLS, fund, manager and

family fixed effects, divided in two different Panels. Panel A sorts the sample into funds

managed by generalist and Panel B funds managed by specialist. Generalist with picking

skills do not affect performance while those with timing skills result in an increase in

fund performance from 20.8 bps per month to 30.4 bps per month. Specialist with

timing ability has no influence on fund performance, similar managers with picking skills

contribute to an increase in fund performance that ranges from 26.2 bps to 32.4 bps per

month. Thus, we conclude that pickers are better suited to manage funds with a single

investment objective because they contribute to improve the performance of the funds

they run, whereas timers do a better job at generalist functions.


A7. Risk-adjusted Returns

We replicate equation (4) using as dependent variable a risk-adjusted return.


A8. Fama-MacBeth (1973) regressions

We estimate prior equation following the Fama-MacBeth (1973) approach.


Overall findings point in the direction that management companies that assign pick-

ers to specialist functions and timers to generalist functions improve their performance

regardless the approach followed.

43

Table A1: Proportion of Generalist by Styles and Size

This table summarizes the proportion of funds managed by generalist managers sorted by quintiles of portfolio sizes and accordingto the fund investment objective for all the U.S mutual funds managed by individual managers during 1996-2011.

Small (1) (2) Medium (3) (4) Large (5)Capital 0.23 0.22 0.19 0.16 0.17Growth 0.26 0.29 0.26 0.29 0.23Income 0.42 0.31 0.33 0.29 0.32Return 0.31 0.36 0.43 0.38 0.47Gov ST 0.26 0.21 0.17 0.15 0.09Gov LT 0.17 0.14 0.14 0.19 0.19Corporate 0.28 0.25 0.19 0.18 0.24Balance 0.34 0.35 0.53 0.54 0.46Foreign 0.18 0.13 0.12 0.12 0.13

Table A2: T-Test Analysis: Managers’ role and Skill

This table reports the performance differences between portfolio managed by timers (Market Prediction Skills) and those unskilledin Panel A and the performance difference between fund managed by Pickers (Security Selection Skills) and the unskilled ones inPanel B. For each of the three asset classes (Domestic Equity, fixed income and international stocks), we sort the managers bySpecialist and Generalists and display the fund performance differences measured using the alpha of Carharts model augmentedby two more risk factors (to be more conservative as our sample also contains fixed income and international stocks).

Panel A Panel BPicking vs Unskilled Timing vs Unskilled

Specialist Generalist Specialist GeneralistEquity 0.912∗∗∗ 0.422∗∗∗ 0.081∗∗∗ 0.125∗∗∗

Debt 0.085∗∗∗ 0.070∗∗∗ 0.014∗∗∗ 0.027∗∗∗

Foreign 0.911∗∗∗ 0.170∗∗∗ 0.370∗∗∗ 0.007

44

Table A3: Transition Matrix (I): Role-switched Manager vs other Managers withinthat Role

This table presents the value difference of fund managed by managers that: 1) have switched from specialist to generalist role(and vice versa) within the same company, 2) between different firms and 3) simple changing the firm but keeping the role. Eachcolumn contains the difference between the characteristics of the switched manager and all the other managers within the samerole over the period 1996-2011. The number of switches within each category is also reported on the last row. Full description ofall these variables is provided in the appendix.* denotes significance at the 10% level, ** denotes significance at the 5% level and*** denotes significance at the 1% level.

Intra-Family Between-FamilySpecialist to Generalist to Specialist to Generalist to Specialist to Generalist toGeneralist Specialist Generalist Specialist Specialist Generalist

Fund Size 0.166∗∗∗ 0.235∗∗∗ -0.428 -0.591∗∗∗ -0.521∗∗∗ -0.512∗∗

Fund Age -0.363 0.871∗∗∗ -3.206∗∗ -2.391∗∗∗ -2.587∗∗∗ -0.644Fund Turnover 0.208∗∗∗ 0.098 0.038 0.173 0.058 -0.047Fund Expenses -0.103∗∗∗ -0.092∗∗∗ 0.125 0.228∗∗∗ 0.204∗∗∗ 0.150∗

Fund Flows 0.251∗∗∗ -0.115 -0.141 1.021∗∗ 0.302∗∗ -0.345Family Size 0.452∗∗∗ 0.619∗∗∗ -0.886∗ -0.929∗∗∗ -1.162∗∗∗ -1.101∗∗∗

Family Funds 0.259∗∗∗ 0.295∗∗∗ -0.617∗∗ -0.615∗∗∗ -0.626∗∗∗ -0.632∗∗∗

Family Managers 0.093∗∗∗ 0.107∗∗∗ -0.207 -0.186∗ -0.310∗∗∗ -0.169Demand Advising 0.021∗∗ 0.013 0.156∗ 0.017 0.002 0.089Supply Advising 0.056∗ 0.088∗∗∗ -0.041 0.009 -0.071∗∗∗ 0.001Ivy League 0.013 0.046∗∗∗ 0.106 -0.061 -0.046∗∗∗ 0.159∗∗

MBA -0.007 -0.004 -0.029 0.111∗∗ 0.044∗∗ 0.029PhD 0.017∗∗∗ -0.008 -0.028 0.033∗ 0.000 -0.035Past Positions -0.029 0.084 -0.893∗∗∗ 0.366∗∗ 0.301∗∗∗ -0.390Manager Size 1.073∗∗∗ -0.252∗∗∗ -0.105 -2.189∗∗∗ -1.245∗∗∗ -1.216∗∗∗

Manager Funds 1.607∗∗∗ -0.768∗∗∗ -0.586 -3.283∗∗∗ -1.598∗∗∗ -2.320∗∗∗

Fund Affiliation -0.387∗∗∗ 0.243∗ -0.816 -2.211∗∗∗ -1.652∗∗∗ -1.929∗∗∗

Number of Events 561 588 165 162 1349 306

Table A4: Transition Matrix (II): Before and after Role-switched Manager

This table presents the value difference of fund managed by managers that: 1) have switched from specialist to generalist role(and vice versa) within the same company, 2) between different firms and 3) simple changing the firm but keeping the role. Eachcolumn contains the difference between the characteristics of the switched manager before and after the event of the switch. Thenumber of events within each category is also reported on the last row. Full description of all these variables is provided in theappendix.* denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1%level.

Intra-Family Between-FamilySpecialist to Generalist to Specialist to Generalist to Specialist to Generalist toGeneralist Specialist Generalist Specialist Specialist Generalist

Fund Size 0.174 -0.630∗∗ -0.354 -0.673 -0.252∗∗∗ -0.324Fund Age -0.128 -1.416 -2.562 -3.219 -2.366∗∗∗ -1.638Fund Turnover 0.183 0.387 -0.168 -1.130∗∗ -0.060 -0.153Fund Expenses -0.198∗∗∗ -0.261∗∗∗ 0.029 -0.064 0.019 -0.131Fund Flows -0.127 0.267 -0.118 1.421 0.246 -0.028Family Size 0.889∗∗∗ 0.443 0.097 -1.323 -0.295∗∗∗ -0.452Family Funds 0.531∗∗∗ 0.069 -0.319 -0.597 -0.281∗∗∗ -0.264Family Managers 0.219∗∗∗ 0.437∗∗ -0.008 -0.413 -0.032 -0.123Demand Advising 0.024 -0.061 0.021 -0.088 -0.117∗∗∗ -0.081Supply Advising 0.072∗∗∗ -0.006 0.022 -0.287 -0.036∗ -0.028Ivy League 0.025 -0.012 0.088 0.149 -0.007 0.223MBA -0.045 -0.191∗∗ -0.059 -0.466 -0.004 0.006PhD -0.005 0.027 -0.013 0.068 0.009 0.000Past Positions -0.165 -0.013 -0.884 -0.744 0.001 -1.190Manager Size 1.816∗∗∗ -0.182 0.759∗∗ -1.075 -0.250∗∗∗ -0.292Manager Funds 3.248∗∗∗ 2.011∗∗∗ 1.097∗∗∗ -0.433∗∗∗ 0.013 0.018Fund Affiliation 0.025 0.212 -1.812∗ -2.101 -2.015∗∗∗ -2.153∗∗

Number of Events 561 588 165 162 1349 306

45

Table A5: Managerial Skill and Concentration

This table presents the results of monthly Pooled OLS regressions of fund and manager investment objective-adjusted returns onfund, manager and family characteristics. Fund returns are the actual returns before deducting fees and expenses (gross) andmanager returns are the TNA-weighted average return of all the portfolios managed by the same manager at the same time. Thedependent variable are fund and manager performance, measured by substracting the median return of their investment objectivepeers, from the actual return of the fund and manager, respectively. Generalist is a dummy variable equals 1 if the fund is managedby manager that is in charge on funds from different investment styles. Timing and Picking are the gamma and alpha coefficientsfrom estimating a modified version of the TM Timing model. Concentration is the Herfindahl index of concentration among alldifferent investment objectives of the funds of the manager. Timer is a dummy equals 1 if the fund is managed by portfolio managerthat has been able to time the market during the past 24 months. Picker is a dummy variable equals 1 if the fund is managed by amanager that was able to pick stocks efficiently during the past 24 months. All variables are lagged one period. A full descriptionof the remaining variables is in the appendix. Time and investment objective dummies are included but not reported; t-statisticsare reported in parentheses. We adjust for serial correlation by clustering standard errors at the fund level. * denotes significanceat the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Fund Performance Manager Performance Fund Performance Manager PerformanceGeneralist 0.020 0.006

(0.65) (0.22)Timing -0.003 -0.003

(-1.62) (-1.53)Picking 0.033∗∗∗ 0.034∗∗∗

(5.40) (5.64)Generalist × Timing 0.021∗∗∗ 0.022∗∗∗

(4.39) (4.78)Generalists × Picking -0.003 -0.013

(-0.23) (-1.11)Concentration -0.154 -0.257

(-1.44) (-0.73)Picker -0.315∗ -0.565∗∗∗

(-1.74) (-3.59)Timer 0.133∗∗∗ 0.181∗∗∗

(3.65) (3.49)Concentration × Picker 0.599∗∗∗ 0.849∗∗∗

(2.97) (4.69)Concentration × Timer -0.130 0.083

(-0.58) (0.44)Fund Size 0.023∗∗ 0.016 0.029∗∗ 0.023∗∗

(1.98) (1.46) (2.47) (2.02)Fund Age 0.000 0.001 -0.000 0.000

(0.11) (0.47) (-0.12) (0.27)Fund Turnover 0.018∗∗ 0.006 0.015 0.003

(2.04) (0.92) (1.62) (0.46)Fund Expenses 0.133∗∗∗ 0.114∗∗∗ 0.137∗∗∗ 0.120∗∗∗

(4.03) (3.69) (4.21) (3.90)Fund Flows 0.063∗∗∗ 0.059∗∗∗ 0.067∗∗∗ 0.063∗∗∗

(4.39) (4.27) (4.46) (4.35)Past Year Return -0.659∗∗∗ -0.655∗∗∗ -0.481∗∗∗ -0.473∗∗∗

(-5.02) (-4.92) (-3.68) (-3.61)Family Size 0.016∗ 0.021∗∗ 0.017∗∗ 0.021∗∗

(1.92) (2.44) (1.99) (2.51)Family Funds 0.000 0.000 0.000 0.000

(0.32) (0.60) (0.42) (0.68)Family Managers -0.001 -0.001 -0.001 -0.001

(-1.02) (-1.37) (-0.98) (-1.35)Supply Advising -0.047 -0.057∗∗ -0.047 -0.056∗∗

(-1.60) (-2.07) (-1.60) (-2.04)Demand Advising -0.004 0.009 -0.008 0.005

(-0.19) (0.41) (-0.34) (0.22)MBA 0.016 0.012 0.008 0.005

(0.66) (0.49) (0.32) (0.21)PhD -0.060 -0.050 -0.071 -0.063

(-1.14) (-0.99) (-1.39) (-1.28)Past Positions -0.013 -0.012 -0.013 -0.012

(-1.46) (-1.41) (-1.41) (-1.41)Ivy League 0.017 0.026 0.020 0.029

(0.53) (0.84) (0.62) (0.95)Manager Funds 0.004 0.001 0.002 0.000

(1.44) (0.38) (0.97) (0.09)Manager Size 0.006 0.005 0.002 0.001

(0.52) (0.48) (0.14) (0.05)Fund Affiliation -0.002 -0.000 -0.002 0.000

(-0.72) (-0.10) (-0.57) (0.07)Constant -0.283∗∗∗ -0.214∗∗ -0.219∗ -0.055

(-2.76) (-2.15) (-1.68) (-0.45)Time Dummies Yes Yes Yes YesStyle Dummies Yes Yes Yes YesObservations 80059 80059 80059 80059r2 0.023 0.023 0.022 0.022

46

Table A6: Managerial Skills by Portfolio Manager type: Generalist and Specialist

This table presents the results of monthly Pooled OLS, Fund, Manager and Family fixed effects regressions of fund investmentobjective-adjusted returns on fund, manager and family characteristics. Fund returns are the actual returns before deductingfees and expenses (gross). The dependent variable are fund performance, measured by substracting the median return of theirinvestment objective peers, from the actual return of the fund. Timer is a dummy equals 1 if the fund is managed by portfoliomanager that has been able to time the market during the past 24 months. Picker is a dummy variable equals 1 if the fund ismanaged by a manager that was able to pick stocks efficiently during the past 24 months. Panel A contains only the subsampleof funds managed by generalist managers while Panel B considers only portfolios managed by Specialist managers. All variablesare lagged one period. A full description of the remaining variables is in the appendix. Control variables, time and investmentobjective dummies are included but not reported; t-statistics are reported in parentheses. We adjust for serial correlation byclustering standard errors at the fund level. * denotes significance at the 10% level, ** denotes significance at the 5% level and*** denotes significance at the 1% level.

Panel A: Generalist SamplePooled OLS Fund Fixed Effect Manager Fixed Effect Family Fixed Effect

Timer 0.208∗∗∗ 0.282∗∗∗ 0.304∗∗∗ 0.206∗∗∗

(3.10) (3.31) (3.64) (2.73)Picker 0.106∗ -0.012 -0.048 0.034

(1.67) (-0.13) (-0.54) (0.46)Control Variables Yes Yes Yes YesTime Dummies Yes Yes Yes YesStyle Dummies Yes Yes Yes YesObservations 19833 19833 19833 19833r2 0.024 0.058 0.046 0.039

Panel B: Specialist SamplePooled OLS Fund FE Manager FE Firm FE

Timer -0.053 -0.035 -0.067 -0.041(-1.04) (-0.60) (-1.21) (-0.79)

Picker 0.324∗∗∗ 0.281∗∗∗ 0.262∗∗∗ 0.293∗∗∗

(6.25) (4.40) (4.26) (5.59)Control Variables Yes Yes Yes YesTime Dummies Yes Yes Yes YesStyle Dummies Yes Yes Yes YesObservations 60226 60226 60226 60226r2 0.020 0.042 0.039 0.027

47

Table A7: Managerial Type and Performance: Risk-adjusted Returns

This table presents the results of monthly Pooled OLS (Panel A) and Family Fixed Effect (Panel B) regressions of fund and managerrisk-adjusted returns on fund, manager and family characteristics. Fund returns are the actual returns before deducting fees andexpenses (gross) and manager returns are the TNA-weighted average return of all the portfolios managed by the same manager atthe same time. The dependent variable is Fund Performance (obtained using the 6-factors model previously defined) and ManagerPerformance (TNA-weighted average alpha of all the funds managed by the same manager at the same time). Generalist is adummy variable equals 1 if the fund is managed by manager that is in charge on funds from different investment styles. Timer isa dummy equals 1 if the fund is managed by portfolio manager that has been able to time the market during the past 24 months.Picker is a dummy variable equals 1 if the fund is managed by a manager that was able to pick stocks efficiently during the past24 months. All variables are lagged one period. Control variables, time and investment objective dummies are included but notreported; t-statistics are reported in parentheses. We adjust for serial correlation by clustering standard errors at the fund level.* denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level.

Panel A: Pooled OLSFund Performance Manager Performance

(1) (2) (3) (4) (5) (6)Generalist -0.021 0.014 0.004 -0.031∗ 0.001 -0.009

(-1.13) (0.75) (0.21) (-1.81) (0.08) (-0.51)Timer -0.015 0.025 -0.011 0.031

(-0.76) (1.28) (-0.53) (1.63)Picker 0.519∗∗∗ 0.521∗∗∗ 0.533∗∗∗ 0.535∗∗∗

(16.85) (16.98) (16.82) (16.98)Generalist × Timer 0.070∗∗ 0.049∗ 0.069∗∗ 0.048∗

(2.18) (1.74) (2.50) (1.78)Generalist × Picker -0.316∗∗∗ -0.313∗∗∗ -0.310∗∗∗ -0.307∗∗∗

(-7.76) (-7.70) (-7.94) (-7.87)Observations 70425 70425 70425 70425 70425 70425r2 0.102 0.151 0.152 0.103 0.159 0.159

Panel B: Family Fixed EffectFund Performance Manager Performance

(1) (2) (3) (4) (5) (6)Generalist 0.007 0.029∗∗ 0.027∗ -0.011 0.010 0.007

(0.50) (2.13) (1.85) (-0.73) (0.71) (0.53)Timer -0.016 0.031 -0.011 0.037

(-0.60) (1.08) (-0.41) (1.31)Picker 0.469∗∗∗ 0.475∗∗∗ 0.490∗∗∗ 0.498∗∗∗

(9.88) (10.04) (9.39) (9.56)Generalist × Timer 0.088∗∗∗ 0.079∗∗∗ 0.084∗∗∗ 0.075∗∗∗


(-5.55) (-5.72) (-4.96) (-5.12)Observations 70425 70425 70425 70425 70425 70425r2 0.202 0.249 0.258 0.202 0.256 0.266

Time Dummies Yes Yes Yes Yes Yes YesStyle Dummies Yes Yes Yes Yes Yes Yes

48

Table A8: Managerial Type and Performance: Fama-MacBeth (1973)

This table presents the results of Fama-MacBeth regressions of fund and manager investment objective-adjusted returns on fund,manager and family characteristics. Fund returns are the actual returns before deducting fees and expenses (gross) and managerreturns are the TNA-weighted average return of all the portfolios managed by the same manager at the same time. Generalist is adummy variable equals 1 if the fund is managed by manager that is in charge on funds from different investment styles. Timer isa dummy equals 1 if the fund is managed by portfolio manager that has been able to time the market during the past 24 months.Picker is a dummy variable equals 1 if the fund is managed by a manager that was able to pick stocks efficiently during the past24 months. All variables are lagged one period. A full description of the remaining variables is in the appendix. Investmentobjective dummies are included but not reported; t-statistics are reported in parentheses. We adjust for serial correlation byapplying Newey-West (1987) estimates of standard errors with lags of order three. * denotes significance at the 10% level, **denotes significance at the 5% level and *** denotes significance at the 1% level.

Fund Performance Manager Performance(1) (2) (3) (4) (5) (6)

Generalist -0.021 0.017 0.008 -0.035∗∗ 0.001 -0.008(-1.15) (0.99) (0.44) (-2.04) (0.05) (-0.48)

Timer -0.019 0.019 -0.016 0.023(-0.95) (0.99) (-0.83) (1.20)

Picker 0.486∗∗∗ 0.488∗∗∗ 0.499∗∗∗ 0.501∗∗∗

(16.30) (16.42) (16.32) (16.48)Generalist × Timer 0.070∗∗ 0.053∗ 0.065∗∗ 0.048∗


(-8.11) (-8.02) (-7.99) (-7.89)Fund Size 0.024∗∗∗ 0.024∗∗∗ 0.024∗∗∗ 0.013∗ 0.012∗ 0.013∗

(2.81) (2.90) (2.93) (1.66) (1.72) (1.75)Fund Age -0.004∗∗∗ -0.003∗∗∗ -0.003∗∗∗ -0.003∗∗∗ -0.002∗∗∗ -0.002∗∗∗

(-4.25) (-3.70) (-3.74) (-3.45) (-2.92) (-2.95)Fund Turnover 0.004 0.003 0.003 0.004 0.002 0.002

(0.47) (0.33) (0.31) (0.45) (0.30) (0.27)Fund Expenses -0.023 -0.031 -0.030 -0.017 -0.025 -0.025

(-0.84) (-1.15) (-1.14) (-0.63) (-0.96) (-0.95)Fund Flows 0.029∗∗∗ 0.026∗∗∗ 0.026∗∗∗ 0.027∗∗∗ 0.024∗∗∗ 0.024∗∗∗

(4.52) (4.32) (4.32) (4.54) (4.33) (4.33)Family Size 0.011∗∗∗ 0.010∗∗∗ 0.010∗∗∗ 0.013∗∗∗ 0.012∗∗∗ 0.012∗∗∗

(2.77) (2.58) (2.61) (3.26) (3.09) (3.12)Family Funds 0.000 0.000 0.000 0.000 0.000 0.000

(0.91) (0.96) (0.96) (0.98) (1.04) (1.03)Family Managers -0.001 -0.000 -0.000 -0.001 -0.001 -0.001

(-0.85) (-0.39) (-0.37) (-1.20) (-0.70) (-0.68)Supply Advising -0.012 -0.018 -0.018 -0.024 -0.030 -0.029

(-0.38) (-0.59) (-0.58) (-0.77) (-1.01) (-1.00)Demand Advising 0.010 0.019 0.018 0.015 0.024 0.023

(0.47) (0.95) (0.91) (0.76) (1.30) (1.25)MBA -0.051∗∗∗ -0.050∗∗∗ -0.050∗∗∗ -0.055∗∗∗ -0.054∗∗∗ -0.054∗∗∗

(-2.67) (-2.73) (-2.73) (-2.91) (-2.99) (-2.99)PhD -0.020 -0.027 -0.023 -0.034 -0.042 -0.038

(-0.42) (-0.57) (-0.48) (-0.75) (-0.91) (-0.82)Past Positions 0.010 0.007 0.008 0.009 0.006 0.006

(1.38) (1.08) (1.10) (1.25) (0.90) (0.92)Ivy League 0.032 0.036 0.035 0.038∗ 0.042∗∗ 0.041∗

(1.38) (1.64) (1.61) (1.70) (1.99) (1.96)Manager Funds -0.015∗∗∗ -0.013∗∗∗ -0.014∗∗∗ -0.014∗∗∗ -0.013∗∗∗ -0.014∗∗∗

(-4.26) (-3.86) (-3.99) (-4.51) (-4.11) (-4.27)Manager Size 0.035∗∗∗ 0.024∗∗∗ 0.024∗∗ 0.044∗∗∗ 0.033∗∗∗ 0.033∗∗∗

(3.63) (2.60) (2.58) (4.84) (3.82) (3.80)Fund Affiliation -0.001 -0.001 -0.001 0.002 0.001 0.002

(-0.26) (-0.35) (-0.30) (0.69) (0.63) (0.68)Family Affiliation -0.005∗ -0.005∗∗ -0.005∗∗ -0.007∗∗∗ -0.007∗∗∗ -0.007∗∗∗

(-1.82) (-2.02) (-2.04) (-2.86) (-3.12) (-3.15)Constant -0.471∗∗∗ -0.436∗∗∗ -0.436∗∗∗ -0.476∗∗∗ -0.438∗∗∗ -0.438∗∗∗

(-5.71) (-5.56) (-5.56) (-5.86) (-5.71) (-5.71)Style Dummies Yes Yes Yes Yes Yes YesObservations 70425 70425 70425 70425 70425 70425r2 0.207 0.246 0.247 0.213 0.257 0.257

49

a tale of two types: generalists vs. specialists in mutual funds

Documents