Are Fund Managers Better in Financial Centers?

Susan Christoffersen Sergei Sarkissian*

McGill University and CIRANO McGill University

First draft: June 2000

This version: August 2004

__________________________________

* Both authors are from the Faculty of Management, McGill University, Montreal QC, H3A 1G5, Canada.
Christoffersen maybe reached at (514) 398-4012 or susan.christoffersen@mcgill.ca. Sarkissian maybe reached at
(514) 398-4876 or sergei.sarkissian@mcgill.ca. We are grateful to Roger Edelen, Wayne Ferson, Simon Gervais,
David Hirshleifer, Tobias Moskowitz, Stefan Nagel, Michael Schill, René Stulz, and Ivo Welch for helpful
comments and to Brad Barber who’s discussion on the subject has motivated one of the authors to start this
project. This paper has also benefited from the feedback of participants of the 2001 European Finance Association
Meeting in Barcelona, the 2002 Conference on Mutual Funds in Zurich and workshops at the Copenhagen
Business School, Laval University, McGill University, and University of Washington. We thank John Bromley,
Lance Dexter, Eric Turner, and Maximo Aybar for help organizing the data and Jason Heinhorst of Lipper
Analytical for providing some of the data. The authors acknowledge financial support from the BSI GAMMA
Foundation. This paper was previously circulated under the title “Location Overconfidence.”
 Are Fund Managers Better in Financial Centers?

ABSTRACT

The economic literature predicts that the average skill level in any industry is higher in larger
cities. The existing labor sorting mechanism may also increase overconfidence of managers in
financial centers if they attribute too much of their successes to their ability rather than luck.
This article examines the relation between managerial ability and behavior across cities using
U.S. equity mutual fund data and the Gervais and Odean (2001) model of overconfidence.
Funds in financial centers on average perform better in terms of both gross and risk-adjusted
returns. These funds however exhibit under-diversification and excessive trading, which is
particularly high among young funds and managers and becomes detrimental for performance
for high-performing funds and following bull markets. Our results suggest that financial
centers attract more sophisticated, yet overconfident money managers.

JEL Classification: G14; G23

Keywords: Ability level; Labor market; Mutual funds; Overconfident trading; Performance
evaluation
1. Introduction

... Yet it was a wealthy city, being situated at the junction of

several important roads and trade routes. The connection
between Sardis and money – easy money – was well known in the
ancient world. ... The people of Sardis had always had a tendency
to become soft and complacent. They lived in luxury and
splendor, and were a proud, arrogant, and overconfident
people... [The Holy Bible, Revelation 3:1-6, Commentaries.]

The economic literature provides extensive evidence that high human capital individuals are
attracted to metropolitan areas. For instance, Glaeser (1999) finds that workers in New York,
Chicago, and Los Angeles are 10% more likely to be college graduates than in other U.S. cities.
Wheeler (2001) shows that increase in the size of a city increases the wage return to education
and proportion of college graduates. There are two basic explanations for these findings: both
assert that cities enhance workers’ productivity by offering them a range of positive
externalities. Some studies, including Mills (1967) and Dixit (1973) emphasize the role of
scale economies in agglomeration processes. Others, such as Jacobs (1969), Glaeser, Kallal,
Scheinkman, and Shleifer (1992), Audretsch and Feldman (1996), and Gehrig (1998), point out
that population density enables better transfer of information and knowledge spillover that
enhance growth and attract those who are most likely benefit from extensive information flows.
The literature therefore predicts that for any occupation group or industry the average skill level
and information flows are higher in large cities. 1
The existing sorting mechanism in the labor market where more skilled people end up
working in large cities may however negatively impact managerial behavior. Gervais and
Odean (2001) model an attribution bias in learning where traders attach too much of their
success to their ability rather than to luck or the overall market movement. In their model,
recent high performance signals ability of a trader, but because of a bias in learning, the trader
attributes too much of this success to his/her own ability and becomes overconfident. With a

1
This may explain why Coval and Moskowitz (2001) find that mutual funds from U.S. metro areas, unlike funds
from smaller locations, do not exhibit local biases in portfolio holdings. Fund managers in large cities may be
more knowledgeable about distant assets and thus have no particular preference in investing locally. For example,
an article in the Wall Street Journal (WSJ, 09/11/2002) discusses the difficulty of hiring fund managers away from
financial centers stating that “...some managers insist on working in New York for professional reasons, such as
wanting to tap into the broad array of company executives who pass through the city.”

1
similar learning bias, individuals may overweight the signal received from the labor market.
For instance, managers who work in New York as opposed to Wichita, Kansas, may interpret
this as a proof of their ability given the evidence that higher ability workers concentrate in
larger metropolitan areas. This suggests that large cities have more able but possibly more
overconfident individuals.
The goal of this paper is to determine whether there are measurable differences in
ability and trading behavior for fund managers working in large metropolitan areas, such as
financial centers, versus other places. In economic studies, wage differentials and output per
worker are often used as evidence of varying productivity and ability across workers in
different regions (e.g., see Glaeser and Maré, 2001). We compare abilities across geographic
regions by looking at the output of money managers as measured by their performance and how
trading influences performance rather than focusing on wage differentials. Overconfidence is
measured using the predictions of Gervais and Odean (2001) where one would expect
managers in financial centers to trade more and this excessive turnover should decrease over
time, particularly after recent high performance. We focus on all equity mutual funds in the
United States (i.e., funds that hold information-sensitive securities) that existed between 1992
and 2002.2
In the first set of tests, we find that the turnover is higher for funds in financial centers
and that it significantly decreases with fund age and especially with fund manager age. Then
we condition turnover on fund performance and find that the relation between turnover and
fund age is significantly negative for high-performing funds in financial centers. We also
examine turnover-age relations for different market conditions and observe that the turnover of
funds in financial centers is significantly higher than in other locations among young funds
following strong market performance. All these findings fully support Gervais and Odean’s
(2001) model.
In the second set of tests, we examine performance differences between funds located in
and outside of financial centers. Consistent with the arguments that larger cities attract more

2
Unlike our study, many papers have focused on either the time-series characteristics of fund returns, or cross-
sectional differences linking them to fund managers’ characteristics. See Grinblatt and Titman (1992), Goetzmann
and Ibbotson (1994), and Carhart (1997) as examples of mutual fund performance studies based on unconditional
asset pricing models or Ferson and Schadt (1996) and Christopherson, Ferson, and Glassman (1998) as those bases
on conditional models. Chevalier and Ellison (1999a,b) examine differences in the performance of funds across a
range of personal characteristics of managers.

2
skilled workers (e.g., Jacobs, 1969), we find that funds located in financial centers outperform
funds located in other places in terms of gross and risk-adjusted returns, both unconditional and
conditional. The difference in risk-adjusted returns is about seven basis points per month or
0.85% per year. There are no differences between these two groups of funds however in their
market timing ability or trading strategies based on publicly available information. We also
find some evidence that funds in financial centers are more exposed to market risk – an
observation consistent with the limited portfolio holdings data showing that funds in financial
centers hold less diversified portfolios.
In the last set of tests, we tie our performance findings with turnover. We find that the
abnormal returns of funds in financial centers are strongly and positively related to their
turnover following down markets and among low-performing funds. This implies that at least
a portion of their trading is based on superior ability or some informational advantage, as found
in Grinblatt and Titman (1994). However, this advantage of funds in financial centers
diminishes when their managers become more overconfident. The return-turnover relation
becomes negative and statistically very significant following good markets or for high-
performing funds. These results are particularly strong for New York funds.
We interpret our results as providing evidence that individuals working in financial
centers are different from those working elsewhere. We support the prevailing view that larger
cities have on average more productive workers, including money managers. However, the
superior ability of fund managers in financial centers materializes in higher returns or more
positive return to trading relation only in times when overconfidence is less likely to appear,
i.e., after a period of low returns. When overconfidence level is at its highest, i.e., after a
period of high returns, the performance of money managers in financial centers does not speak
in favor of their superior ability.3
We should point out that our results are completely consistent with a hypothesis of
various informational spillovers that may enhance manager’s ability to perform and learn. In

3
There are some studies that attempt to find differences in performance across investors depending on the size of
their locations. For example, Hau (2001) finds no performance differences between traders in Frankfurt, the
German financial center, and those outside that city, including locations outside Germany. This result is likely to
be affected by at least two facts: first, distances in Germany are smaller than in the U.S., making it more difficult
to clearly identify the financial center relative to the other areas, and, second, investors in that study are separated
by political and cultural boundaries. Hau does not consider a possibility of overconfident trading in financial
centers. Kumar (2004) finds evidence of overconfidence (underconfidence) among individual investors in
metropolitan areas (remote locations).

3
particular, Daniel, Hirshleifer and Subramanyam (1998) argue that investors are overconfident
about their private information, rather than general public information. If managers have
access to better private information in financial centers, this would be consistent with our
findings. Alternatively, managers may learn from each other in an enhanced learning
environment as modeled in Glaeser (1999) or simply exhibit behavioral herding.4 Hong, Kubik
and Stein (2002) present an example of behavioral herding, observing that stock market
participation is influenced by social interaction. Hong, Kubik and Stein (2003) directly show
how information and learning may be transferred within a city as they provide evidence that
fund managers in the same city hold similar portfolios and imitate each other. Hence, for the
remainder of the paper we maintain a very generic definition of ability which one could
interpret as the innate ability of the manager or as a manager’s enhanced ability resulting from
access to better (e.g., private) information or superior learning environment.
The rest of the article is organized as follows. Section 2 describes the data. Section 3
formulates two hypotheses about the performance differential between mutual funds located in
and outside of financial centers. Section 4 presents our main findings. It starts with the
analysis of the relation between turnover and fund and fund manager age. Then it presents the
results of performance evaluation tests and relates these findings to turnover. Section 5
concludes.

2. Data

The data on equity mutual funds comes from CRSP. It contains information not only on fund
returns and total net assets but also on fund’s year of organization, the name of its managers, as
well as its annual turnover. Our sample covers the period from January 1992 to September
2002 because turnover data is not available prior to 1992. We select all U.S. domestic equity
funds that have the following investment objectives: aggressive growth, growth & income,
income, and large growth.

4
For the models of behavioral herding or informational cascades see Banerjee, (1992), Bikhchandani, Hirshleifer,
and Welch (1992), and Welch (1992). These models discuss situations where people in a group make the same
choices after observing each others behavior.

4
 The CRSP data does not provide the physical addresses of funds. To determine each
fund’s location we used in part the data from the Lipper Analytical that provides the
information on where the headquarters of each fund are located. The following six cities are
defined to be financial centers: Boston, Chicago, Los Angeles, New York, Philadelphia, and
San Francisco. This classification corresponds to certain established criteria for identifying a
financial center (see Gehrig, 1998, and references therein).5 We classify a fund to be in one of
these financial centers if the distance of its headquarters from the city proper is no more than 50
miles.
Table 1 shows the summary statistics of our mutual fund data. Panel A shows the fund
distribution across locations and investment objectives. The total number of funds is 5282
resulting in 27024 fund-year observations. There are more funds in financial centers than in
other places, 3016 versus 2266. New York has the largest number of funds, 1467, followed by
Boston with 615 funds. Among investment objectives, large growth funds constitute the largest
proportion of all funds followed by aggressive growth funds. The panel also shows the number
of fund management companies as well as the number of distinct managers for each location.
The sample contains 321 management companies out of which 156 are located in financial
centers and 165 are located in other places. The number of identified distinct fund managers is
2445 in financial centers and 1966 in other places. New York has more than a thousand
different fund managers over our sample period. In computing these numbers we excluded all
team-managed funds.
Panel B of Table 1 shows the average fund characteristics across calendar years, such as
the number of funds, fund size, fund and manager age, turnover, gross returns and expenses.
The size of the fund is measured in terms of its total net assets. The fund age is the difference
in years between the current year and the year of organization of the fund. The manager age
(the tenure of the manager with the fund) is the difference in years between the current year and
the year when the manager was first assigned to the fund. The turnover of the fund is defined
as the assets traded as a percent of the net asset size of the fund over the year. Returns are
shown in basis points per month. Expenses are defined as the annual total expense ratio of the
fund in percentage points. The panel shows a steady increase in the number of funds across

5
The same cities are identified as the largest mutual fund centers in other studies (e.g., see Hong, Kubik, and
Stein, 2003).

5
the sample period with a slight drop in 2002. The average fund size however has not grown,
except during the late 1990’s. We can also observe that both fund age and fund manager age
remain relatively steady over most of our sample period. The average fund age is markedly
higher than the overall level only in 1992-1993. The fund turnover has substantially increased
from the early 1990’s. Interestingly, it remained essentially at the same level of 83-85% during
the entire bull market of 1995-1999. The last three years in the sample are characterized by the
average turnover close to or exceeding 100%. There are no unexpected patterns in average
returns across all funds. As for the fund expenses, they too show slight increase over our
sample period.
Panel C of Table 1 shows the fund characteristics across locations in and outside of
financial centers. It shows the number of observations, the mean, standard deviation, median
for each location, as well as the difference between locations. We can see that funds in
financial centers are on average almost twice as large as those in other places and this
difference is statistically very significant. There are no big differences in fund age or manager
age between the two locations since the corresponding median values are exactly the same even
though managers in financial centers tend to be a bit younger on average (difference is less than
two months). Interestingly, the turnover in financial centers is significantly larger than in other
places (92% versus 86.3%) and the same relation holds for fund returns: funds in financial
centers outperform other funds by 7 basis points per month (about 85 basis points per year).
Notice that the outperformance of funds in financial centers relative to those located elsewhere
is observed also in the median. At last, the average expenses of funds in financial centers are
six basis points lower than those in other places, but the medians in both location groups are
exactly the same. This panel thus shows that there are significant differences in turnover and
returns between funds located in and outside of financial centers but these differences are
unlikely to be related to fund age or manager age. In addition, knowing from previous studies
that larger funds usually trade less often than small ones, the excessive turnover of funds in
financial centers cannot be explained by their size.

3. The two hypotheses

6
In this section, we formulate the two competing but not mutually exclusive hypotheses about
why there may be differences in managerial performance and behavior in large financial centers
versus other places. If the theories of urban agglomeration are correct, we should observe more
able individuals in large financial centers than elsewhere. According to two models of
agglomeration (scale economies and information spillovers), higher ability in cities could
reflect (i) an individual’s innate skills which are better rewarded in a thick job market (e.g., see
Wheeler, 2001), and/or (ii) an enhanced learning experience due to frequent knowledge
transfers in highly populated areas (e.g., see Glaeser, 1999). We cannot distinguish between
these two possible explanations for higher ability people concentrating in large cities; however,
both provide the same testable hypothesis which we are interested in. In particular, one would
expect that funds in financial centers provide higher returns to investors, at least, before fees
and expenses. Moreover, if there is more information in financial centers then trading based on
this information should enhance fund performance.6 This line of reasoning leads to the
following hypothesis,

Fund manager ability hypothesis: Mutual funds located in financial centers have more skilled
managers.

This hypothesis implies that:

• Mutual funds in financial centers perform better than funds in other places,
• The turnover of funds in financial centers is positively related to returns.

The second hypothesis is whether the existing selection mechanism in the labor market
where more skilled people choose larger and therefore wealthier cities to work and live results
in overconfidence.7 Because having a managerial job in a financial center provides a signal of
ability, individuals may overweight this success and attribute too much to their skills rather
than to general market conditions or luck. We test overconfidence in financial centers using the
predictions of Gervais and Odean (2001). In their model successes in trading enhance

6
For example, Loughran and Schultz (2004) find that stocks of firms located in large cities are traded more
heavily than those in rural areas supporting diffusion of information from metro areas to rural locations.

7
overconfidence and cause investors to trade more frequently, but, over time, investors realize
their true ability and their trading decreases. Odean (1999) and Barber and Odean (2000, 2001)
show that overconfidence trading lowers returns.8 Therefore, in the short run, higher turnover
erodes performance but this disappears as managers learn their true ability. Moreover, there
should be more signs of overconfidence when individuals are most likely to suffer from self-
attribution and learning bias, for instance, in times of overall good performance. This
motivaties our second hypothesis,

Fund manager overconfidence hypothesis: Mutual funds located in financial centers have
more skilled but overconfident managers.

This hypothesis implies that:

• Turnover is higher in financial centers,
• Turnover of funds in financial centers is negatively related to fund and manager age,
• The relation between turnover and performance is more negative in financial centers
conditional on high performance of the fund and strong markets,
• Portfolio holdings are less diversified in financial centers than elsewhere.

Ability and overconfidence have opposite implications for managerial performance in

financial centers. In particular, trading will erode performance for an overconfident manager,
potentially negating the affects of his/her ability.9 Therefore, the implications of the first
hypothesis hold only in the absence of overconfidence or, at least, when its level is sufficiently
low. This would include periods when performance of the fund is low or when the market
performance is poor.

7
The link between the status (prestige) of the city and the general attitude of its inhabitants goes as far back as to
The Holy Bible. In the Book of Revelation (e.g., see Commentaries of Scott, 1906), we find that self-confident
people of the once wealthy ancient city of Sardis were twice the victims of their own arrogance.
8
In general, one cannot say that overconfident investors trade more than non-overconfident. Grinblatt, Titman and
Wermers (1995), Ferson and Khang (2000) and others find that fund style matters in trading patterns: value funds
tend to be “contrarian,” while growth funds are primarily “momentum” chasers. Thus, an overconfident value
fund manager might trade less than a non-overconfident growth fund manager. However, since we compare
performance of funds within the same broad category, this issue is not of concern.
9
Note that if overconfident investors have access to private signals, their profits may potentially exceed those of
fully rational investors. For example, Kyle and Wang (1997), Daniel, Hirshleifer, and Subramanyam (2001), and

8
 Thus, in this paper we test whether, consistent with economic theories of urban
agglomeration, financial cities attract more skilled fund managers or whether geographic
sorting in the labor market creates in large cities a pool of more able yet overconfident people.
Our tests therefore, investigate the link between the labor market geography and investor
behavior and performance.

4. Empirical results

Our study of the behavior and performance differentials between funds located in and outside
of financial centers can be divided into three general categories. The first is the analysis of
turnover differentials and the relation between turnover and fund and fund manager age. The
second is the evaluation of performance differentials for which we use unconditional and
conditional performance evaluation models. The third is the examination of the relation
between performance and turnover differentials for different fund and market performance
levels.

4.1. Turnover differentials

In this sub-section, we focus on the relation between portfolio turnover and fund and manager
age for funds located in and outside of financial centers. Portfolio turnover is a relatively good
proxy for the frequency of trading by fund managers because it measures the minimum of total
sales or total purchases made over the asset size of the fund. For instance, a fund that had
100% growth in new purchases but did not sell anything would have reported a zero turnover.
To assess the basic statistical properties of the turnover and fund age relation, we can
use simple panel regression. In this regression model, the dependent variable is the log of the
fund’s excess turnover. The excess turnover is the ratio of the turnover of each fund in a given
year over the median turnover, which is the median turnover across all funds for each year in a
given fund category. The independent variables are the log age of the fund, the dummy for
funds in financial centers, F, and the interaction variable composed of the log age of the fund
and the financial center dummy. The results of the regression with robust standard errors are:

Hirshleifer and Luo (2001) present models where by trading more aggressively on accurate (private) information,
overconfident traders may outperform those who trade less frequently.

9
Turniex,t = −0.172 − 0.005 Age _ Fund i,t + 0.131Fi − 0.034( Age _ Fund i,t Fi ) + ei ,t . (1)
(-8.68) (-0.48) (5.01) (-2.48)

First, the positive and significant slope coefficient on the dummy implies that funds in financial
centers have significantly higher turnover relative to other funds. Second, the negative and
significant slope on the interaction term implies that funds in financial centers trade
significantly more when they are young than when they are old.
It is important to find out whether the relation between turnover and fund age is unique
to funds or whether it reflects a similar relation among fund managers. To assess the statistical
properties of the turnover and manager age relation, we use the same panel regression
approach. In the regression model, the dependent variable is the log of the fund’s excess
turnover. The independent variables now are the log age of the fund manager (the tenure of the
manager with the fund), the dummy for funds located in financial centers, F, and the interaction
variable composed of the log age of the fund manager and the financial center dummy. The
results of the regression with robust standard errors are as follows:

Turniex,t = −0.012 − 0.160 Age _ Mgri,t + 0.133Fi − 0.064( Age _ Mgri,t Fi ) + ei ,t . (2)
(-0.60) (-10.51) (5.01) (-3.23)

As in model (1), model (2) yields a positive and significant slope coefficient on the financial
center dummy implying that fund managers in financial centers have significantly higher
turnover relative to other funds. However, the turnover-age relation is now much stronger in
both economic and statistical terms than in the previous regression. First, the slope on manager
age is negative and statistically highly significant, indicating that old managers trade less then
young ones in locations outside financial centers. Second, the negative and significant slope on
the interactive age variable implies that that old managers in financial centers trade much less
frequently than the young ones relative to their peers in smaller townships.
While the statistical results in regressions (1) and (2) are useful, due to a highly non-
normal nature of the variables in those models, especially the age variable which remains very
different from normal distribution even in its log form, it is important to address the above
issues in a regression-free framework. Moreover, it is well known that turnover decreases with

10
the size of the fund and so the turnover-age relation observed above may simply reflect the
turnover-size relation. Figures 1 and 2 address these concerns. They show the excess turnover
and size differentials (in log scale) between funds located in financial centers and those located
outside depending on the fund or fund manager age. The excess size is the average difference
between the size of each fund in a given year and the median fund size, which is computed for
each year and fund category. We consider five fund age cohorts: 1 to 2, 3 to 5, 6 to 9, 10 to 14,
and 15 plus years of age. Figure 1 gives the patterns for two relations depending on fund age.
Plot A shows the turnover-age and size-age relations for the entire sample of funds. We can
see that funds in financial centers maintain a positive excess turnover differential over funds
located in other places for all the age cohorts but the last one, which contains about only 10%
of all observations. Since the former funds are also larger than the latter ones in every fund age
cohort, it is impossible to explain the excessive trading of funds in financial centers, especially
younger funds, by their smaller size. It may appear however that the overall decrease in the
turnover differential with age is simply due to the observable trend for funds in financial
centers to become even bigger with age relative to those located in other places.
Plot B of Figure 1 illustrates the relation between turnover and fund age for two sub-
samples of funds: those with high performance and those with low performance. The high
(low) performing fund is a fund whose average return in a given year is greater than (less than)
the median return for its fund category in that year. This analysis is motivated by Gervais and
Odean’s (2001) model which shows that overconfidence increases with successful investments
but then decreases more rapidly with investor age, as traders start realizing their genuine ability
in making investment decisions. The plot shows that the turnover-fund age relation for high-
performing funds follows the model’s prediction quite well. In addition, as expected, there is
no convincing pattern for low-performing funds. The corresponding size-age relations show
that the faster decrease in excess turnover differential among high-performing funds cannot be
explained by the faster increase in their size relative to low-performing funds. In fact, we can
see that the excess size differential for high-performing funds barely grows with fund age but
low-performing funds in financial centers become much bigger than funds located in other
places over time.
Now we turn our attention to the relation between turnover and fund manager age.
Figure 2 presents similar plots as those in Figure 1. We can see from Plot A that funds in

11
financial centers maintain a positive excess turnover differential over the fund located in other
places across all age cohorts. It appears once more that the overall decrease in turnover
differential with manager age may be because older managers in financial centers run even
bigger funds than in other places relative to their young peers. However, Plot B again shows
that the faster decrease in excess turnover differential among high-performing funds cannot be
explained by the faster increase in their size relative to low-performing funds.
The variables that we used in the turnover-age relations above are adjusted for fund
investment objective and the year of operation. However, there may be some residual impact
on our findings from a relative change over time in the proportion of funds in different
categories between locations in and outside of financial centers. If the proportion of funds with
generally high turnover (e.g., aggressive growth funds) decreases with age more rapidly in
financial centers than in other places then, one might attribute the observed decrease in excess
turnover differential to such a change. Therefore, in Figure 3 we show the proportion of funds
by fund investment objective for the same five fund age and manager age cohorts. Plot A
gives this relation for fund age, Plot B – for manager age. In both plots we observe no clear
trends that the ratio of the number of funds in financial centers over the number of funds in
other places for each specific fund objective changes with the age cohort.
We conclude this sub-section with another group of tests that differentiates between
fund trading behavior after poor and strong market conditions. This analysis is motivated by
the insights of Gervais and Odean (2001) who argue for more overconfidence among investors
after market gains and lower overconfidence after market losses.10 Therefore, if younger funds
and funds in financial centers are more overconfident, then we should be able to observe larger
turnover differentials between funds located in and outside of financial centers subsequent to
strong markets and, particularly among younger funds and managers.
In Table 2 we rank all the years in our sample plus 1991 based on the yearly market-
wide performance and classify them into poor and strong performance years. The worst year in
the sample is 2002 with the negative annual excess return of almost –23%; the best year is 1995
with the annual excess return of more than 31%. We added the year of 1991 to the table since
we will conduct our analysis on the years subsequent to the given market performance level
and the first year in our sample is 1992.

10
Statman, Thorley, and Vorkink (2003) confirm this on the aggregate market data.

12
 The test results are reported in Table 3. It shows the differences in fund excess turnover
depending on lagged market conditions: poor or strong. The turnover-age sample following
poor market conditions covers the years of 1993-1995 and 2001-2002, while that following
strong market conditions covers the years of 1992 and 1996-2000. Panel A shows the excess
turnover by two fund age groups: young and old. The young (old) funds are those funds whose
age is less than or equal to (greater than) four – the median age across all funds. After poor
market performance, the turnover difference is about 4.6%-4.7% for both young and old funds
but it is statistically significant only for the latter group of funds. The situation changes quite
drastically following strong market performance years. Now the excess turnover differential is
8.5% for young funds and statistically highly significant, while the same differential for old
funds is economically small and insignificant.
Panel B of Table 3 shows the excess turnover by two fund manager age groups: young
and old. The young (old) managers are those fund managers whose tenure with the fund is less
than or equal to (greater than) two – the median tenure across all fund managers. After poor
market performance, the turnover difference is about 6.7% for young funds and 3.0% for old
ones but both of them are statistically insignificant. As with the fund age scenario, the situation
changes dramatically following strong market performance years. The excess turnover
differential increases in magnitude and becomes statistically significant for both manager age
groups. For funds with young managers the difference is 7.8% and significant at the 1% level;
for funds with old managers the difference is 3.52% and significant at the 5% level. Larger
turnover differentials between funds located in and outside of financial centers therefore are
observed subsequent to strong markets and, especially among younger funds and funds with
young managers.
Thus, the patterns of relations between fund turnover and fund age and especially
manager age are fully consistent with theoretical predictions of the model of overconfident
behavior in Gervais and Odean (2001). First, funds and managers in financial centers trade
more often than funds and managers located elsewhere. Second, the turnover of funds
diminishes with fund and manager age and this relation is most profound for high-performing
funds in financial centers. Third, the turnover difference between funds located in and outside
of financial centers among younger funds and managers increases more relative to that among
older funds and managers following strong market performance.

13
4.2. Performance differentials
Table 1 has shown that funds in financial centers outperform other funds in terms of average
gross returns. Therefore, our primary attention in this sub-section is on the differences in risk-
adjusted returns between funds in financial centers and in other locations. We conduct two sets
of tests. The first set deals with unconditional performance evaluation, the second with
conditional.

4.2.1. Unconditional performance evaluation

We consider four unconditional performance evaluation models. The first one uses the market
as the benchmark and so it can be represented as:

ri ,t = α i + β i rM ,t + ei ,t , (3)

where ri and rM are the returns on fund i and the U.S. market portfolio less the one-month U.S.

T-bill rate, respectively. The risk-adjusted return, α, based on this model is the standard
Jensen’s alpha.
Model (3) does not take into account that fund managers may have different exposures
not only to market risk but also to some firm-specific characteristics, such as the book-to-
market ratio and firm size, that have been shown in recent literature to be related to risk (see
Fama and French, 1993, 1996). Carhart (1997) considers one more risk measure in evaluating
mutual fund performance – the momentum factor. Therefore, our second model is based on the
Carhart’s (1997) four-factor model, namely:

ri ,t = α i + β i rM ,t + s i SMBt + hi HMLt + miUMDt + ei ,t , (4)

where SMB and HML are the Fama-French book-to-market and size factors, and UMD is the
momentum factor. The risk-adjusted return, α, based on (4) is the Carhart’s alpha.
To decompose the potential effects of stock selection ability from market timing ability
we also consider two market timing models. The first model follows Treynor and Mazuy
(1966), i.e.,

ri ,t = α i + β i rm ,t + γ i rm2,t + ei ,t , (5)

14
where γi is market timing measure of fund i. A positive and significant γ indicates market
timing ability.
The second model is based on Merton and Henriksson (1981) method, that is, in this
case:

ri ,t = α i + β i rM ,t + γ i rM+ ,t + ei ,t , (6)

where rM+ ,t = rM ,t if rM ,t > 0 and is zero otherwise. The idea here is that good market timers

should become more exposed to the market only when the expected market return is positive.
The test results on the performance differences in unconditional setting between funds
located in financial centers and other places are shown in Table 4. For each fund the
performance results are obtained using the entire fund return history between 1992 and 2002.
The results are then averaged based on the fund’s location. For each performance evaluation
mode, the table shows the point estimates on all the slope coefficients, the number of
observations, and the average adjusted regression R-squared. For the estimates of alphas only
the table also shows the proportion of funds with a performance level within the specific
critical value of the t-test. Panel A presents the results for models (3) and (4). Notice that the
average alphas from both models and for both locations are negative, consistent with most of
the previous studies on mutual fund performance. Both models yield also a similar difference
in risk-adjusted returns: funds located in financial centers outperform funds located elsewhere
and this outperformance is statistically highly significant with the spread of about 6 basis points
per month or about 0.7% annually. When we look at the proportion of alphas within a given
range of the t-test, we observe that there are more funds in financial centers with positive and
statistically significant alphas than in other places. For example, based on the four-factor
model results, the proportion of funds with positive and significant risk-adjusted returns is
3.5% in financial centers but only 1.9% in other places. When we look at the worst performers,
then we can see, for instance, that the proportion of funds with negative and significant alphas
is 11.0% in financial centers but 12.8% in other places. Therefore, the outperformance of funds
in financial centers relative to those located elsewhere in terms of risk-adjusted returns
generally holds for the entire distribution of alphas.
We observe no differences in their exposure to market risk across funds in different
locations based on the single-factor model, but the four-factor model reveals that funds in

15
financial centers take significantly more market risk: the average four-factor beta for funds in
financial centers is 1.00, while the same beta for funds in other places is 0.98. Along with this,
the other significant difference is found in the loadings on the book-to-market portfolio. It
appears that funds in financial centers invest significantly more in value stocks than other
funds. Finally, notice that the four-factor model provides a substantial improvement in the
explanatory power over the one-factor model: the increase in the adjusted R-squared is about
13.5% for funds in both location groups.
Panel B presents the results for models (5) and (6). The estimates of the market timing
measure from both models indicate market mis-timing across all funds, consistent with
previous studies. There are no significant differences in market (mis-) timing ability across
funds in different locations. The selection ability of the funds in financial centers is again
higher than that of funds located in other places, although only the difference based on the
Treynor and Mazuy (1966) model is statistically significant. The proportion of funds with
significantly positive (negative) alphas is again higher (lower) for funds in financial centers
than funds located elsewhere, similar to the estimation results from models (3) and (4). The
average adjusted R-squared for both timing models are only marginally higher than those from
the single-factor model, indicating that these models have very little incremental explanatory
power beyond the simplest single-factor performance benchmark model.

4.2.2. Conditional performance evaluation

There is a very important reason to examine risk-adjusted returns in not only unconditional but
also conditional setting. While any fund performance evaluation based on an alternative
framework can be considered simply as a natural robustness exercise, conditional approach
adds a key dimension to the interpretation of fund managers’ abilities, as it pertains to his/her
use of public information signals versus private signals and stock picking skills. As, Ferson
and Schadt (1996) and Christopherson, Ferson, and Glassman (1998) point out, the conditional
performance evaluation framework can control for possible biases in unconditional
performance measures when managers trade on publicly available information. We follow
these studies and present four conditional performance evaluation models.

16
 The first one is the conditional version of the single-factor model (3), in which market
betas are represented as a linear function of the lagged information variables. The resulting
model, which is called the conditional beta model, has the following form:

ri ,t = α i + β i rM ,t + Bi (Z t −1 rM ,t ) + ei ,t , (7)

where, Z t −1 is the vector of information variables available to investors at time t-1. If the out-

performance of funds in financial centers is related to their time-varying risk exposure in

response to changes in some economy-wide information variables, then model (7) should
reflect that in terms of markedly different average B for funds in financial centers relative to
those located in other places.
Our second conditional model is also based of the single-factor model (3).11 However,
in this case, not only market betas but also the intercept are represented as a linear function of
the lagged information variables. The resulting model, which is called the conditional alpha
and beta model, has the following form:

ri ,t = α i + Ai Z t −1 + β i rM ,t + Bi (Z t −1 rM ,t ) + ei ,t . (8)

Model (8) allows us to directly test if the out-performance of funds in financial centers results
from better trading strategies on public information. If the average A in financial centers
significantly differs from that for funds located elsewhere, one could infer different stock-
picking skills conditional of public signals across fund managers in different locations.
The third model is the conditional version of the Treynor and Mazuy (1966) model. In
this model, similar to model (5), the market beta is assumed to be linear in lagged information
variables. The resulting model therefore can be represented as:

ri ,t = α i + β i rM ,t + Bi (Z t −1 rM ,t ) + γ i rM2 ,t + ei ,t . (9)

The coefficient B captures the fund manager’s shifts in market risk exposure due to public
information. As a result, γ now is the sensitivity of the fund’s beta to the private timing signal.
Finally, the fourth and the last model is the conditional version of the Merton and
Henriksson (1981) model:

ri ,t = α i + β i rM ,t + Bi (Z t −1 rM ,t ) + γ i rM* ,t + ∆ i (Z t −1 rM* ,t ) + ei ,t , (10)

11
The conditional four-factor model becomes quite unmanageable, especially in yearly estimations of risk-
adjusted fund performance. Therefore, we do not consider this specification.

17
where rM* ,t = rM ,t if {rM ,t − E [ rM ,t | Z t −1 ]} > 0 and is zero otherwise. The coefficient ∆

captures the market timing ability of fund in response to lagged information variables. 12 If the
timing ability of funds across different locations is similar we should find no significant
differences not only in average γ but also in average ∆ .
Our set of information variables contains the lagged values of the one-month U.S. T-bill
rate and the term spread. This choice is motivated by the recent evidence on stock returns
predictability that shows that most of the commonly used instruments may have spurious
relation to stock returns. It appears that the short-term interest rate and the term-structure
spread have emerged as the most genuine predictors. (See Avramov, 2002; Ferson, Sarkissian,
and Simin, 2003; and the references therein).
The test results on the performance differences in conditional setting between funds
located in financial centers and other places are shown in Table 5. As before, the test results
are based on the entire return history of each fund. The format of all the reported point
estimates is similar to those in Table 4. Panel A presents the results for models (7) and (8).
Notice first of all that the average alphas from both models and for both locations are less
negative than the corresponding estimates obtain within unconditional evaluation framework.
This outcome corroborates, for instance, with the evidence in Ferson and Schadt (1996) who
also report that risk-adjusted performance of funds in conditional setting leads to more
evidence of managerial skill than in unconditional. The second important observation is that
the difference in the alphas between funds located in and outside of financial centers is now
seven basis points per month (or 0.85% per year) exceeding that in unconditional setting by
more than one basis point per month. This difference is statistically significant at the 1% level
for both conditional beta and conditional alpha and beta models. The proportion of alphas
above or below certain critical values of the t-statistic resembles that in unconditional setting.
The main message here is that the proportion of funds in financial centers with significantly
positive (negative) alphas is larger (smaller) than that in other places.
When looking at the average estimates of the coefficients on the interactive market
terms or the lagged values of instruments, we find no statistically significant differences
between funds located in and outside of financial centers. The only marginally significant

12
See Ferson and Schadt (1996) for more detailed description of the conditional timing models.

18
difference is observed for the slope on the interactive term composed of the lagged value of the
T-bill rate and the market (in the conditional alpha and beta model). Therefore, we can state
that there are no differences in response to public information between fund managers in
different locations.
Thus, the out-performance of mutual funds in financial centers is a result of better stock
selection ability and possibly higher risk taking (based on unconditional four-factor model).
These factors may speak in favor of genuine stock-picking skills among managers of funds in
financial centers. For example, funds in financial centers may outperform funds located
elsewhere because they attract fund managers with superior education and skill level (see
Chevalier and Ellison, 1999a). Alternatively, the outperformance of funds in financial centers
may be due to the existence of private information there (e.g., Gehrig, 1998). In both cases
however, our findings provide evidence that fund managers in financial centers are potentially
more able investors than their peers in smaller locations, supporting the predictions of
agglomeration models in economics literature. Our results are also consistent with Wermers
(2003) who finds that fund managers who trade more frequently (in our setting those in
financial centers) have more persistent stock-picking talents. In the next section, we address all
these issues by directly linking fund performance results to the frequency of trading.

4.2.3. Portfolio concentration differentials

In this sub-section we illustrate that differences between funds located in and outside of
financial centers exist with respect to not only portfolio turnover rates but also concentration of
portfolio holdings. Kacperczyk, Sialm, and Zheng (2004) find that mutual funds that hold
more concentrated portfolios perform on average better than funds with more diversified
holdings and they attribute this superior performance to better information. In addition, Odean
(1998) argues that more overconfident traders hold less diversified portfolios. If fund managers
in large cities are better able, yet more overconfident, then it follows that financial center
managers should also hold more concentrated portfolios.
From Lipper Analytical, we have portfolio holdings for September 1996 and we report
their concentration across funds for different locations in Table 6. The table shows that mutual
funds located in financial centers hold significantly more concentrated investments than those
located in other places, and this is consistent with some evidence of excessive market risk

19
taking by managers of funds in financial centers (see Panel A of Table 4). The difference in
portfolio concentration is the largest for the first industry, about 1.8%, and, even though
somewhat diminished in economic terms, is still present for the other top five industry
investments. In a recent study on individual investors data from Sweden, Goetzmann, Massa,
and Simonov (2004) also find that urban portfolios are less diversified than rural portfolios and
they attribute this to broader “knowledge spillover” processes in larger cities. We find the
same result holds amongst U.S. mutual fund managers.

4.3. Performance-turnover differentials

In this sub-section, we directly examine the possible informational advantage of financial
centers by linking funds’ turnover rates to their returns and managers’ stock selection ability.
To make the full use of our annual turnover data, in this section we employ the following fund
performance evaluation procedure. The risk-adjusted returns for each fund and year are
computed by regressing its twelve monthly excess returns of the year on the corresponding
benchmark portfolios observed during the same twelve-month period. Funds that have less
than a twelve-month history for a given year are disregarded. Since fund return data in our
sample is not available for the last three months of 2002, the last year for which we compute
risk-adjusted returns is 2001. Furthermore, for the sake of conciseness, we use one
unconditional performance evaluation model and one conditional. The unconditional risk-
adjusted returns are based on the Carhart’s (1997) four-factor model, the conditional – on the
CAPM with time-varying alphas and betas. These are the two best models in terms of their
adjusted R-squared from among our two sets of four conditional and unconditional models.
Recall that in Table 3 we observed substantial differences in turnover differentials
between funds located in financial centers and other places following specific market
conditions. A relevant question then is whether there exist differences in returns depending on
turnover level. Table 7 shows gross and risk-adjusted fund returns in financial centers and
other places across high and low turnover funds as well the results of the difference test for the
performance level in different locations conditional on market conditions. After poor market
conditions when the overconfidence level is low, an increase in turnover, at least in part, maybe
associated with some informed trading and therefore should be more beneficial for funds in
financial centers where information flows are larger. More importantly, if fund managers in

20
financial centers become more overconfident than their peers elsewhere after strong market
performance, then the turnover increase in these conditions should be less beneficial for these
(financial center) funds.
Panel A gives the results following poor market performance. For gross returns, we
find that an increase in turnover among funds in financial centers benefits them less than funds
located in smaller places. However, in terms of both unconditional and conditional risk-
adjusted returns, funds in financial centers gain more from larger turnover than funds in other
places. For example, when moving from low to high turnover, funds in financial centers reduce
their unconditional performance by 0.02 basis points a month compared to 0.17 for funds in
other locations. This corresponds to a relative gain of more 0.15 basis points a month or 1.80%
annually. In the conditional setting the relative gain is smaller but still economically significant
(more than 0.3% a year).
Panel B of Table 7 shows the results following strong market performance. Now the
relative differences in the benefits from increased turnover across funds in different locations
are consistent for both gross and risk-adjusted returns. For gross returns, we find that an
increase in turnover among funds in financial centers benefits them less than funds located in
smaller places, 0.23% per month and 0.31% per month, respectively. For risk adjusted returns,
the performance difference becomes even bigger. High turnover adds only 0.07% in
unconditional performance to funds in financial centers, yet almost three times as much, 0.18%,
to fund performance in smaller locations. With conditional alphas the overall picture is similar:
the corresponding performance gains are 0.10% for funds in financial centers and 0.23% for
funds in other places.
The results in Table 7 may speak in favor of more able fund managers in financial
centers but, even more strongly, they point to the existence of more overconfident trading
among those managers. The impact of superior skills on risk-adjusted performance is the
greatest when there is less room for overconfidence, i.e., after poor market conditions. On the
other hand, the potential benefits of higher ability level diminishes when the there is more
overconfidence, i.e., after good market performance years.
We elaborate further on our findings in Table 7 by testing for the incremental impact of
turnover of funds in financial centers on risk-adjusted returns under different market
performance and fund performance levels in Tables 8 and 9. We use only risk-adjusted returns

21
since one should take into account the fact that funds in financial centers appear to take more
risk and are less diversified (see Tables 4 and 6). Holding more concentrated investments in
bear markets usually lead to more underperformance relative to the market as a whole and that
may negate the positive effect on gross returns of more informed trading. In good market
conditions, the effect is the opposite. Having more risky investment in bull markets may
increase gross returns partially offsetting the negative effect of overconfident trading.
In these tests we use the following panel regression model for risk-adjusted returns,
namely:

αˆ i ,t = d o + d1Turniex,t + d 2 Sizei ,t + d 3 Flowsi ,t +

( )
, (11)
+ d 4 Fi + d 5 Turniex,t Fi + d 6 (Sizei ,t Fi ) + d 6 (Flowsi ,t Fi ) + ei ,t

where α̂ i,t is the estimated risk-adjusted return for fund i in year t. Given some evidence on the

importance of changes in fund flows on fund performance (e.g., Edelen, 1999), regression (11)
controls for fund flows, Flowsi ,t , defined as the change in fund’s i total net assets from year t-1

to year t. All the independent variables except for the dummy for financial centers are in the
log form. Also, for each level of market or fund performance, we conduct our estimation not
only on the entire available set of data but also on the sub-sample of funds from the largest
financial center – New York. This allow us to examine whether any relation between turnover
and returns that we might observe are common to all financial centers or are unique to the most
sophisticated financial center. In the next two tables we report our results using Carhart’s
alphas because their properties across locations are in general qualitatively similar to
conditional alphas and, from Tables 4 and 5 we observe that the four-factor unconditional
performance evaluation model produces the highest adjusted R-square across all models.
Table 8 shows the relation between risk-adjusted returns and excess turnover for a given
market performance level. Panel A gives the results for the years following poor market
performance. Regression (1) represents a reduced form of model (11) and provides the
properties of the relation across all locations. In this case, similar to gross returns after bad
markets, on average the excess turnover is negatively impacting performance: the slope
coefficient on turnover is negative and significant. Regression (2) shows however that this
relation for funds in financial centers is now significantly different from that for funds located
in other places. Given the magnitude of slope coefficients on the two turnover terms, one can

22
observe that all the negative impact from extra turnover is present only among funds located
outside of financial centers. This effect strengthens in economic terms in Regression (3) when
funds in New York City are compared against funds in smaller locations. Now the slope on the
interactive turnover variable is positive 0.13 instead of 0.11, and it is larger in magnitude than
the negative slope of 0.11 on turnover for funds located in smaller places. Regression (4)
shows that while trading in New York has the most positive impact on performance, it is not
limited to the largest financial center but is also present in the sub-sample of other large hubs of
mutual fund industry.
If there is any informed trading in financial centers then its presence should be more
observable among high-turnover funds. Therefore, the last two columns of the panel report the
results of the regressions similar to (2) and (3) but using a sub-sample restricted only to funds
which above median turnover for a given year and fund objective category. Consistent with the
prediction of more able trading in financial centers, we indeed observe a larger positive slope
on the interactive turnover term, 0.20 and 0.16 for Regression (5) and (6), respectively. These
two slopes however are much smaller in magnitude than negative slopes in the corresponding
regression models on the turnover variable for funds located in smaller places. Therefore,
increased turnover is not beneficial in absolute terms.
Panel B of Table 8 presents the results for the years following strong market
performance. Overall, the results are diametrically opposite to those in Panel A. Regression
(1) is a reduced form model giving the properties of the relation across all locations. Now, the
turnover coefficient shows that more trading has on average an enhancing effect on risk-
adjusted returns. However, consistent with the overconfidence hypothesis, the excess turnover
is significantly less useful for financial center funds in Regression (2). Moreover, as
Regressions (3) and (4) shows, the detrimental impact of excessive turnover is larger in both
economic and statistical terms in New Your than in other financial centers. When we consider
only the funds with the above-median turnover (columns 5 and 6), the negative impact of
excessive trading becomes even more profound, especially for New York. The negative slope
of the interactive turnover variable in Regression (6) is larger than the corresponding one in
Regression (5) and is substantially larger in magnitude than the positive slope on the turnover
variable for funds outside of financial centers (-0.48 versus 0.35). This implies that funds in
New York engage in more useless trading than all other funds.

23
 At last, in Table 9, we partition all funds on high and low-performing funds, as we did
earlier and again reexamine the turnover-risk-adjusted returns relation. Panel A reports the
results for low-performing funds. Note that in this case, the sign predictions are less obvious.
Low-performing funds are unlikely to be overconfident (also recall Figures 1 and 2); they are
also not very likely to be managed by managers with superior portfolio selection skills.
Regression (1) shows that on average more trading leads to worse performance across the
whole population of funds. Regression (2-4) indicate that even low-performing funds in
financial centers, especially in New York, may still be better in enhancing their returns from
more trading. However, the situation is opposite with the sub-sample of funds with above-
median trading. Regressions (5-6) show that these funds in financial centers have particularly
poor turnover-returns relation, although it is not significant.
Panel B of Table 9 reports the regression results for high-performing funds, which is a
more interesting case, since similar to the case of different market conditions, one can
formulate clear predictions based on our ability and overconfidence hypotheses. High
performance enhances overconfidence and therefore it should be detectable among funds from
financial centers, in particular those from New York. As before, regression (1) reports average
results. Among high-performing funds, more trading produces better risk-adjusted returns.
Regressions (2-4) shows that funds in financial centers underperform in this dimension funds
located in other places, and the largest underperformance occurs in funds from New York.
Furthermore, Regressions (5-6) confirm our predictions that high turnover funds in financial
centers, most notably in New York, are even more inferior to high turnover funds incorporated
in smaller places.
Thus, it appears that market conditions and fund performance levels affect the turnover-
alpha relation of funds in financial centers in a way consistent with the hypothesis that
managers of those funds are more able but also more overconfident individuals than their peers
in less prominent cities. In poor market conditions and to some extent among low-performing
funds, when the overconfidence level is low, the turnover of funds in financial centers is more
positively related to their returns indicating that at least part of this turnover is based on ability
or useful information, possibly private. In strong markets or for high-performing funds, when
the overconfidence level is high, the excessive turnover of funds in financial centers negatively
impacts their returns, highlighting the damaging effect of overconfidence on performance.

24
These results are the particularly striking for New York funds, consistent with the expectation
that funds in portfolio managers in New York maybe among some of the most sophisticated
ones, yet undoubtedly more overconfident as well.

4.4. Alternative explanations

Are their alternative rational explanations for our findings? First, several papers emphasize the
importance of competition among fund managers for clients and information. Dow and Gorton
(1997) develop a model in which portfolio managers can trade simply to show to their clients
that they are working. This would be consistent with our observation that funds located in
financial centers rebalance their portfolios more frequently since they may face more rivalry
than other funds. However, competition for clients does not explain other results, e.g.,
turnover-age and return-turnover relations, portfolio concentration differentials, etc. Coval and
Moskovitz (2001) suggest that there is greater competition for local information in metropolitan
areas that can adversely affect performance of funds in these places. This again leaves many of
our results hard to explain.
Second, Reuter (2004) finds that mutual funds that make large brokerage payments to
underwriters hold a larger share of Initial Public Offerings (IPOs). Given high positive first-
day returns on many IPOs in the 1990’s, funds in financial centers could potentially boost their
returns due to their proximity to leading underwriters. While this is consistent with their
overall better performance, it still cannot explain many relations we observe, especially the
results provided in Table 9 where past performance of the fund influences the turnover-return
relation. Thus, rational explanations do not support for the full set of observed discrepancies in
the trading behavior and performance between funds located in and outside of financial centers.

5. Conclusions

The urban and labor economics literature has long been arguing that large cities enhance
productivity level of workers due various positive externalities, including better knowledge
transfers and other information spillovers. As a result of this process, cities are believed to host
people with higher average ability than small townships and this phenomenon exists in any
industry with comparable presence in both metropolitan areas and smaller locations. Although

25
previous literature has relied on testing this hypothesis using wage differentials, we provide a
new test of this hypothesis using performance differentials of fund managers located in
financial centers versus other places. We further postulate that the existing labor selection
mechanism may induce overconfidence among managers of funds located in financial centers.
Our results support the hypothesis that fund managers in financial hubs are more able,
yet overconfident individuals. On the one hand, we find that on average both gross and risk-
adjusted returns are higher for funds in financial centers. Also, in times when overconfidence
is likely to be at the lowest, such as after down markets, the stock selection ability of funds
located in financial centers is more positively related to their trading than that of funds located
elsewhere, implying that some portion of their turnover is based on useful information, possibly
including private.
On the other hand however, we observe that funds in financial centers trade more often
and hold more concentrated portfolios than other funds. Moreover, we confirm in the data the
theoretical predictions of the model of investor overconfidence developed by Gervais and
Odean (2001). We observe that the turnover of funds is negatively related to fund and fund
manager age, especially for funds in financial centers. Also, conditional on the fund’s
performance, the relation between turnover and age is significantly more negative for high-
performing funds in financial centers. We further show that, consistent with the prediction of
the model, the impact of overconfidence on fund performance is most profound among high-
performing funds, and in the years following strong market. Funds in New York exhibit more
overconfidence than all other funds, including those in the secondary financial centers.
Thus, we document that geographic location of a fund has an important impact on fund
manager’s behavior and that it can have both positive and negative effects on fund
performance. The positive effect is related to the fact that financial centers may enhance the
stock picking ability of fund managers and their trading skills. The negative effect is associated
with overconfidence that is present more in financial centers due to the very perception that on
average these locations attract more able people.

References:

Audretsch, D.B., and M.P. Feldman, 1996, Knowledge spillovers and the geography of
innovation and production, American Economic Review 86, 630-640.

26
Avramov, D., 2002, Stock return predictability and model uncertainty, Journal of Financial
Economics 64, 423-458.

Banerjee, A.V., 1992, A simple model of herd behavior, Quarterly Journal of Economics 107,
797-818.

Barber, B., and T. Odean, 2000, Trading is hazardous to your wealth: The common stock
investment performance of individual investors, Journal of Finance 55, 773-806.

Barber, B., and T. Odean, 2001, Boys will be boys: gender, overconfidence, and common stock
investment, Quarterly Journal of Economics 116, 261-292.

Bikhchandani, S., D. Hirshleifer, and I. Welch, 1992, A theory of fads, fashion, custom, and
cultural change as informational cascades, Journal of Political Economy 100, 992-1026.

Carhart, M.M., 1997, On persistence in mutual fund performance, Journal of Finance 52, 57-
82.

Coval, J., and T. Moskowitz, 2001, The geography of investment: Informed trading and asset
prices, Journal of Political Economy 109, 811-841.

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, Quarterly
Journal of Economics 114, 389-432.

Chrisopherson, J.A., Ferson W., and D. Glassman, Conditioning manager alphas on economic
information: Another look at the persistence of performance, Review of Financial Studies
11, 111-142.

Dixit, A., 1973, The optimum factory town, Bell Journal of Economics and Management
Science 4, 647-654.

Dow, J., and G. Gorton, 1997, Noise trading, delegated portfolio management, and economic
welfare, Journal of Political Economy 105, 1024-1050.

Edelen, R., 1999, Investor flows and the assessed performance of open-end mutual funds,
Journal of Financial Economics 53, 439-466.

Fama, E., and K. French, 1993, Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics 33, 3-56.

Ferson, W.E., and R.W. Schadt, 1996, Measuring fund strategy and performance in changing
economic conditions, Journal of Finance 51, 425-461.

Ferson, W.E., and K. Khang, 2002, Conditional performance measurement using portfolio
weights: Evidence for pension funds, Journal of Financial Economics, forthcoming.

27
Ferson, W.E., S. Sarkissian, and T. Simin, 2003, Spurious regressions in financial economics?,
Journal of Finance 68, 1393-1413.

Daniel, K., D. Hirshleifer, and A. Subramanyam, 1998, Investor psychology and security
market under- and overreactions, Journal of Finance 53, 1839-1885.

Daniel, K., D. Hirshleifer, and A. Subramanyam, 2001, Overconfidence, arbitrage, and

equilibrium asset pricing, Journal of Finance 56, 921-965.

Gehrig, T., 1998, Cities and the geography of financial centers, CERP DP 1894.

Gervais S., and T. Odean, 2001, Learning to be overconfident, Review of Financial Studies 14,
1-27.

Glaeser, E.L., H.D. Kallal, J.A. Scheinkman, and A. Shleifer, 1992, Growth in cities, Journal of
Political Economy 100, 1126-1152.

Glaeser, E.L., 1999, Learning in cities, Journal of Urban Economics 46, 254-277.

Glaeser, E.L., and D.C. Maré, 2001, Cities and skills, Journal of Labor Economics 19, 316-
342.

Goetzmann, W.N., and R.G. Ibbotson, 1994, Do winners repeat? Patterns in mutual fund
performance, Journal of Portfolio Management 20, 9-18.

Goetzmann, W.N., M. Massa, and A. Simonov, 2004, Portfolio diversification and city
agglomeration, NBER Working paper #10343.

Grinblatt, M., and S. Titman, 1992, The persistence of mutual fund performance, Journal of
Finance 47, 1977-1984.

Grinblatt, M., and S. Titman, 1994, A study of monthly mutual fund returns and performance
evaluation techniques, Journal of Financial and Quantitative Analysis 29, 419-444.

Grinblatt, M., S. Titman, and R. Wermers, 1995, Momentum investment strategies, portfolio
performance, and herding: A study of mutual fund behavior, American Economic Review
85, 1088-1105.

Hau, H., 2001, Location matters: An examination of trading profits, Journal of Finance 56,
1959-1983.

Hayashi, Y., 2002, For funds in the hinterland, hiring pros can be a chore, The Wall Street
Journal, Wednesday, September 11, D9.

Henrikson, R., and R. Merton, 1981, On market timing and investment performance II:
Statistical procedures for evaluating forecasting skills, Journal of Business 54, 513-534.

Hirshleifer, D., and Y. G. Luo, 2001, On the survival of overconfident traders in a competitive
securities market, Journal of Financial Markets 4, 73-84.

28
Hong, H., J.D. Kubik, and J. Stein, 2002, Social interaction and stock-market participation,
Journal of Finance, forthcoming.

Hong, H., J.D. Kubik, and J. Stein, 2003, Thy neighbor's portfolio: Word-of-mouth effects in
the holdings and trades of money managers, Working paper, Harvard University.

Jacobs, J., The Economy of Cities, New York: Vintage, 1969.

Kacperczyk, M., C. Sialm, and L. Zheng, 2004, On the industry concentration of actively
managed equity mutual funds, Journal of Finance, forthcoming.

Kumar, A., 2004, Is the local bias of individual investors induced by familiarity or information-
asymmetry?, Working paper, University of Notre Dame.

Kyle, A., and F.A. Wang, 1997, Speculation duopoly with agreement to disagree: Can
overconfidence survive the market test?, Journal of Finance 52, 2073-2090.

Loughran, T., and P. Schultz, 2004, Dissemination of information: Urban versus rural stock
return patterns, Working paper, University of Notre Dame.

Mills, E., 1967, An aggregative model of resource allocation in metropolitan areas, American
Economic Review 57, 197-210.

Odean, T., 1998, Volume, volatility, price, and profit when all traders are above average,
Journal of Finance 53, 1887-1934.

Odean, T., 1999, Do investors trade too much?, American Economic Review 89, 1279-1298.

Reuter, J., 2004, Are IPO allocations for sale? Evidence from mutual funds, Working paper,
University of Oregon.

Scott, C.A, 1905, The book of the revelation, Hodder and Stoughton, London.

Statman, M., S. Thorley, and K. Vorkink, 2003, Investor overconfidence and trading volume,
Working paper, Santa Clara University.

Treynor, J., and K. Mazuy, 1966, Can mutual funds outguess the market? Harvard Business
Review 44, 131-136.

Wermers, R., 2003, Is money really “smart”? New evidence on the relation between mutual
fund flows, manager behavior, and performance persistence, Working paper, University of
Maryland.

Welch, I., 1992, Sequential sales, learning, and cascades, Journal of Finance 47, 695-732.

Wheeler, C.H., 2001, Search, sorting, and urban agglomeration, Journal of Labor Economics
19, 879-899.

29
 Table 1
Summary statistics
This table shows the summary statistics of domestic equity mutual funds in the U.S. The data are from CRSP and
cover the period between January 1992 and September 2001. Panel A shows the fund distribution across fund
types and locations in and outside of financial centers. The fund types are aggressive growth, AG, growth and
income, GI, income, IN, and large growth, LG. The panel also shows the number of management companies and
the number of different fund managers. Panel B gives the summary of average fund characteristics across calendar
time. Panel C provides more detailed information on fund characteristics across two locations – financial centers,
F, and other places, O. Panel D gives the details on turnover and gross returns by fund investment objective. The
fund is considered to be in a financial center if its headquarters are located within 50 miles of one of the six cities
defined as financial centers. The size of the fund is measured in terms of its total net assets. Flow is the log
annual change in the fund’s size. The fund age is the difference in years between the current year and the year of
organization of the fund. The manager age (the tenure of the manager with the fund) is the difference in years
between the current year and the year when the manager was first assigned to the fund. Manager age excludes all
team-managed funds. The turnover is the annual turnover of the fund. The returns are average returns for each
year and fund and are shown in basis points per month. The expenses are the annual total expenses of the fund in
percentage points.

Panel A: Fund distribution

Number of Funds
Obs All AG GI IN LG Comp. Mgr.
Financial centers 15572 3016 856 680 146 1334 156 2445
New York 7330 1467 440 329 51 647 67 1108
Boston 3427 615 142 127 35 311 16 509
Chicago 1502 288 85 54 11 138 22 284
Philadelphia 1223 240 74 58 14 94 11 191
Los Angeles 1067 209 48 63 19 79 17 127
San Francisco 1023 197 67 49 16 65 23 231
Other places 11452 2266 656 543 134 933 165 1966
Total 27024 5282 1512 1223 280 2267 321 4411

Panel B: Average fund characteristics across calendar years

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Num. of Funds 939 1367 2251 2744 3232 3839 4198 4400 4806 5202 5088
Size (bln $) 0.44 0.47 0.35 0.45 0.52 0.62 0.71 0.84 0.72 0.56 0.41
Fund Age 11.82 9.62 7.89 7.15 6.90 6.42 6.54 6.81 7.00 6.81 7.50
Manager Age 4.89 3.95 3.15 2.96 3.10 2.94 3.23 3.58 3.51 3.59 4.06
Turnover (%) 72.2 75.9 78.5 84.2 83.9 84.3 85.5 85.4 92.6 102.5 97.4
Returns (%/m) 0.91 1.04 -0.11 2.29 1.53 1.84 1.31 1.86 0.03 -0.73 -3.35
Expenses (%) 1.31 1.32 1.42 1.42 1.44 1.43 1.43 1.44 1.43 1.47 1.48

30
 Table 1 (continued)

Panel C: Fund characteristics across locations

Obs Mean S.D. Median Diff (F-O)
Size (bln $) F 15027 0.73 3.54 0.07 0.34
O 11069 0.39 1.61 0.04 (9.98)
Fund Age F 15451 7.26 10.80 4 0.15
O 11346 7.11 10.35 4 (1.14)
Manager Age F 13192 3.40 3.91 2 -0.14
O 9295 3.54 4.31 2 (-2.66)
Turnover (%) F 13632 92.0 96.5 72.0 5.70
O 10099 86.3 90.3 64.9 (4.62)
Returns (%/m) F 15416 0.35 2.25 0.77 0.07
O 11278 0.28 2.16 0.65 (2.66)
Expenses (%) F 15374 1.40 0.67 1.33 -0.06
O 11286 1.47 0.78 1.33 (-6.62)

31
 Table 2
Ranking of market-wide performance
The table shows the performance of the U.S. aggregate equity market in excess of the risk free rate over the period
of 1991-2002. The performance is ranked from the worst performance year, 2002, to the best performance year,
1995.

1 2 3 4 5 6 7 8 9 10 11 12

Poor market performance Strong market performance

2002 2000 2001 1994 1992 1993 1996 1998 1999 1997 1991 1995

-22.91 -16.71 -14.78 -4.11 6.41 8.36 16.25 19.42 20.2 26.07 29.1 31.04

32
 Table 3
Turnover and fund and manager age for a given market performance level
The table shows the differences in fund excess turnover depending on lagged market conditions: poor or strong.
The excess turnover is the average difference between the turnover of each fund in a given year and the median
turnover. The median turnover for each fund is the median turnover of funds for each year and fund category.
The turnover and age samples cover the period from 1992 to 2002. The poor market performance is observed in
the following years: 1992-1994 and 2000-2002. The strong market performance is observed in 1991 and 1995-
1999. The turnover-age sample following poor market conditions covers the years of 1993-1995 and 2001-2002,
while that following strong market conditions covers the years of 1992 and 1996-2000. Panel A shows the
turnover by two fund age groups: young and old. The young (old) funds are those funds whose age is less than or
equal to (greater than) four – the median age across all funds. Panel B shows the turnover by two fund manager
age groups: young and old. The young (old) managers are those fund managers whose tenure with the fund is less
than or equal to (greater than) two – the median tenure across all fund managers. The table also shows the number
of observations, the results of the difference test for the turnover in different locations, as well as the difference
between excess turnover differentials under different market conditions.

Panel A: Excess turnover in percent by fund age group

Young funds Old funds

Financial centers Other places Financial centers Other places

Obs Turn. Obs Turn. Diff Obs Turn. Obs Turn. Diff

Poor market 2888 29.07 2255 24.47 4.60 3336 17.97 2397 13.26 4.72
(1.36) (2.32)
Strong market 3997 23.99 2858 15.48 8.51 3160 13.54 2589 11.65 1.89
(3.64) (0.99)

Panel B: Excess turnover in percent by manager age group

Young managers Old managers

Financial centers Other places Financial centers Other places

Obs Turn. Obs Turn. Diff Obs Turn. Obs Turn. Diff

Poor market 2318 38.32 1729 31.57 6.76 3906 14.10 2923 11.08 3.03
(1.71) (1.54)
Strong market 3127 29.89 2102 22.08 7.81 4281 11.89 3345 8.37 3.52
(2.73) (2.13)

33
 Table 4
Risk-adjusted returns and timing ability in unconditional framework
The table shows the unconditional risk-adjusted performance of U.S. domestic equity mutual funds located in and
outside of financial centers as well as the test of the difference in that performance. It also shows the number of
observations for the each estimation. Panel A shows the results of unconditional performance evaluation, such as
the risk-adjusted returns (alphas) and the loadings on the benchmarks, based on the two performance models. The
first is based on the CAPM, the second – on the four-factor model of Carhart (1997), namely:
1-factor: ri ,t = α i + β i rM ,t + ei ,t ,
4-factor: ri ,t = α i + β i rM ,t + s i SMBt + hi HMLt + miUMDt + ei ,t ,

where ri and rM are the returns on fund i and the market portfolio less the one-month U.S. T-bill rate, SMB and
HML are the Fama-French portfolios, and UMD is the momentum portfolio. Panel B shows the results of the
estimation of the funds’ unconditional timing ability based on the two models: the Treynor and Mazuy (1966)
model and the Merton and Henriksson (1981) model, namely:

TM: ri ,t = α i + β i rM ,t + γ i rM2 ,t + ei ,t ,
MH: ri ,t = α i + β i rM ,t + γ i rM+ ,t + ei ,t ,

where rM+ ,t = rM ,t if rM ,t > 0 and is zero otherwise. For each fund the performance results are obtained using the
entire fund return history between 1992 and 2002. The results are then averaged based on the fund’s location. For
the risk-adjusted returns the table also shows the proportion of funds with a performance level within the specific
critical value of the t-test for the alphas.

Panel A: Unconditional performance

Single-factor model Four-factor model

F O Diff (F-O) F O Diff (F-O)

Obs 2742 2016 2742 2016

α -0.0086 -0.0685 0.0599 -0.1110 -0.1682 0.0572

(2.90) (3.61)
tα>1.96 0.0511 0.0382 0.0357 0.0188
1.65< tα<1.96 0.0233 0.0198 0.0248 0.0198
tα<-1.96 0.0941 0.1066 0.1098 0.1275
-1.65>tα>-1.96 0.0361 0.0402 0.0470 0.0645

β 0.9897 0.9852 0.0046 0.9991 0.9810 0.0181

(0.49) (3.03)
s 0.1698 0.1795 -0.0097
(-0.81)
h 0.0516 0.0291 0.0225
(1.96)
m 0.0434 0.0418 0.0015
(0.25)
R2 0.7335 0.7411 0.8725 0.8722

34
 Table 4 (continued)

Panel B: Unconditional timing ability

TM model MH model

F O Diff (F-O) F O Diff (F-O)

Obs 2742 2016 2742 2016

α 0.0269 -0.0281 0.0550 -0.0012 -0.0382 0.0370

(2.09) (1.11)
tα>1.96 0.0562 0.0387 0.0383 0.0263
1.65< tα<1.96 0.0306 0.0298 0.0215 0.0174
tα<-1.96 0.0861 0.0873 0.0708 0.0660
-1.65>tα>-1.96 0.0321 0.0496 0.0274 0.0407

β 0.9847 0.9808 0.0039 0.9893 0.9926 -0.0033

(0.40) (-0.33)
γ -0.0010 -0.0015 0.0005 -0.0013 -0.0147 0.0134
(0.60) (1.22)
R2 0.7399 0.7469 0.7393 0.7463

35
 Table 5
Risk-adjusted returns and timing ability in conditional framework
The table shows the conditional risk-adjusted performance of U.S. domestic equity mutual funds located in and
outside of financial centers as well as the test of the difference in that performance. It also shows the number of
observations for the each estimation. Panel A shows the results of conditional performance evaluation, such as the
risk-adjusted returns (alphas) and the loadings on the benchmarks, based on the two performance models. The
first is based on the conditional beta CAPM, the second – on the conditional alpha and beta CAPM, namely:

Cond. β: ri ,t = α i + β i rM ,t + BiTBILL (Z TBILL,t −1 rM ,t ) + BiTERM (Z TERM ,t −1 rM ,t ) + ei ,t ,

Cond. α & β: ri ,t = α i + AiTBILL Z TBILL,t −1 + AiTERM Z TERM ,t −1 + β i rM ,t + BiTBILL (Z TBILL,t −1 rM ,t ) + BiTERM (Z TERM ,t −1 rM ,t ) + ei ,t ,

where ri and rM are the returns on fund i and the market portfolio less the one-month U.S. T-bill rate, Z TBILL and
Z TERM are the two information variables – the lagged values of the U.S. T-bill rate and the term spread, which is
the difference in yields on long-term U.S. government bonds and short-term bills. Panel B shows the results of the
estimation of the funds’ conditional timing ability based on the two models: the conditional version of the Treynor
and Mazuy (1966) model and the conditional version of the Merton and Henriksson (1981) model, namely:

Cond. TM: ( ) ( )
ri ,t = α i + β i rM ,t + BiTBILL Z TBILL,t −1 rM ,t + BiTERM Z TBILL,t −1 rM ,t + γ i rM2 ,t + ei ,t ,
ri ,t = α i + β i rM ,t + B TBILL
i (Z r )+ B
TBILL,t −1 M ,t (Z TERM
i r
TERM ,t −1 M ,t )+
Cond. MH: ,
+γ r *
i M ,t +∆ TBILL
i (Z *
r
TBILL,t −1 M ,t ) + ∆ (Z TERM
i r*
TBILL,t −1 M ,t )+ e i ,t

where rM* ,t = rM ,t if {rM ,t − E [ rM ,t | Z t −1 ]} > 0 and is zero otherwise. The estimates of E [ rM ,t | Z t −1 ] are obtained
from the fitted values from the regression of rM ,t on lagged instruments. For each fund the performance results
are obtained using the entire fund return history between 1992 and 2002. The results are then averaged based on
the fund’s location. For the risk-adjusted returns the table also shows the proportion of funds with a performance
level within the specific critical value of the t-test for the alphas.

36
 Table 5 (continued)

Panel A: Conditional performance

Conditional beta model Conditional alpha and beta model

F O Diff (F-O) F O Diff (F-O)

Obs 2742 2016 2742 2016

α 0.0077 -0.0635 0.0712 -0.0020 -0.0676 0.0656

(3.33) (2.82)
tα>1.96 0.0565 0.0461 0.0573 0.0496
1.65< tα<1.96 0.0274 0.0248 0.0281 0.0268
tα<-1.96 0.0930 0.1042 0.0966 0.1101
-1.65>tα>-1.96 0.0383 0.0451 0.0416 0.0501

ΑTBILL 0.2987 -0.5481 0.8468

(1.44)
ΑTERM -0.4056 -0.4878 0.0822
(0.10)
β 0.9989 0.9913 0.0076 0.9968 0.9900 0.0068
(0.84) (0.75)
ΒTBILL 0.4355 0.4563 -0.0208 0.2513 0.4410 -0.1897
(-0.26) (1.74)
ΒTERM 0.6970 0.7181 -0.0211 0.5530 0.6346 -0.0816
(-0.20) (-0.56)
R2 0.7559 0.7619 0.7706 0.7765

Panel B: Conditional timing ability

TM model MH model

F O Diff (F-O) F O Diff (F-O)

Obs 2742 2016 2742 2016

α 0.0708 0.0027 0.0681 0.0031 -0.0396 0.0427

(2.34) (1.12)
tα>1.96 0.0853 0.0645 0.0445 0.0342
1.65< tα<1.96 0.0339 0.0342 0.0405 0.0188
tα<-1.96 0.0879 0.0982 0.0755 0.0804
-1.65>tα>-1.96 0.0346 0.0511 0.0408 0.0357

β 0.9930 0.9868 0.0062 0.9804 0.9976 -0.0171

(0.66) (-0.83)
BTBILL 0.4474 0.4919 -0.0445 0.4549 0.2684 0.1865
(-0.52) (0.51)
BTERM 0.7547 0.8260 -0.0713 -0.1588 0.4871 -0.6459
(-0.61) (-0.68)
γ -0.0024 -0.0029 0.0005 0.0253 0.0015 0.0239
(0.67) (0.98)
∆TBILL 0.1558 0.4061 -0.2503
(-0.58)
∆TERM 1.1769 0.5833 0.5937
(0.60)
R2 0.7629 0.7682 0.7742 0.7798

37
 Table 6
Portfolio holdings concentration
The table shows the average concentration (in percent) of top five portfolio holdings by industry for all funds
located in and outside of financial centers, the difference between them (F−O) with the corresponding t-statistics.
The financial centers are the cities of Boston, Chicago, Los Angeles, New York, Philadelphia, and San Francisco.
Chicago and Philadelphia are shown as one group and Los Angeles and San Francesco as another. The data are
the quarterly portfolio holdings from Lipper Analytical as of September 1996 and cover all the domestic equity
funds in the Lipper’s growth and growth and income fund category.

Industry rank

Obs 1 2 3 4 5
Financial Centers 381 18.36 13.18 10.79 8.81 7.37
New York 154 17.22 13.03 10.87 8.82 7.31
Boston 81 18.42 13.48 11.12 9.27 7.75
Chicago-Philadelphia 72 19.39 13.56 10.83 8.77 7.29
Los Angeles-San Francisco 74 19.67 12.76 10.21 8.33 7.15
Other places 503 16.57 11.90 9.57 7.97 6.74
Difference (F-O) 1.79 1.27 1.22 0.84 0.63
t-stat 2.93 3.91 5.02 4.53 4.15

38
 Table 7
Fund returns for a given market performance
The table shows gross and risk-adjusted fund returns for funds with low and high turnover located in financial
centers (F) and other places (O) for specific market conditions, poor (Panel A) and strong (Panel B). The poor
market performance is observed in the following years: 1992-1994 and 2000-2002. The strong market
performance is observed in 1991 and 1995-1999. The return sample following poor market conditions covers the
years of 1993-1995 and 2001-2002 (2002 for gross returns only), while that following strong market conditions
covers the years of 1992 and 1996-2000. The table also shows the results of the performance difference test
between low and high turnover funds. Low (high) turnover funds are funds with negative (positive) excess
turnover. The excess turnover is the average difference between the turnover of each fund in a given year and the
median turnover. The median turnover for each fund is the median turnover of funds for each year and fund
category.

Panel A: Poor market performance

Gross returns Unconditional alphas Conditional alphas

Low High Diff Low High Diff Low High Diff

Financial centers -0.7769 -0.8403 -0.0634 -0.1639 -0.1856 -0.0217 -0.0373 -0.2719 -0.2346
(-1.10) (-0.87) (-7.13)
Other places -1.0035 -0.7486 0.2549 -0.1189 -0.2952 -0.1763 -0.0801 -0.342 -0.2619
(-3.98) (-6.02) (-6.83)
Difference (F-O) -0.3183 0.1546 0.0273

Panel B: Strong market performance

Gross returns Unconditional alphas Conditional alphas

Low High Diff Low High Diff Low High Diff

Financial centers 1.1298 1.3624 0.2326 0.0381 0.1057 0.0676 -0.1211 -0.0233 0.0978
(6.48) (2.59) (2.72)
Other places 1.0435 1.3537 0.3102 -0.0015 0.1786 0.1801 -0.2358 -0.0012 0.2346
(8.07) (5.66) (5.50)
Difference (F-O) -0.0776 -0.1125 -0.1368

39
 Table 8
Relation between risk-adjusted returns and excess turnover for a given market performance level
This table shows the estimation results from the panel regression of risk-adjusted fund returns on excess turnover
and location for different market conditions while controlling for fund size. The excess turnover is the average
difference between the turnover of each fund in a given year and the median turnover. The median turnover for
each fund is the median turnover of funds for each year and fund category. The unconditional risk-adjusted
returns are based on the Carhart’s (1997) four-factor model, the conditional – on the CAPM with time-varying
alphas and betas. The risk-adjusted returns for each fund and year are computed by regressing its twelve monthly
excess returns of the year on the corresponding benchmark portfolios observed during the same twelve-month
period. Funds that have less than a twelve-month history for a given year are disregarded. The regression model
has the following form:

αˆ i ,t = d o + d1Turniex, t + d 2 Sizei , t + d s Flowsi , t + d 4 Fi + d 5 (Turniex, t Fi ) + d 6 (Sizei , t Fi ) + d 7 (Flowsi , t Fi ) + ei , t ,

where α̂ i,t is the estimated risk-adjusted return for fund i in year t, while Turn iex,t , Sizei ,t , and Flows i ,t are the
fund’s i log excess turnover, log size, and flows, respectively. Panel A gives the results for the years in which the
turnover followed poor market performance, Panel B – for strong market performance. F denotes the finance
center dummy, and NY – for the city of New York. The sample of risk-adjusted returns covers the period from
1992 to 2001.

Panel A: Following poor market performance (1993-1995, 2001)

(1) (2) (3) (4) (5) (6)
All All New York ex New York All High Turn NY High Turn
Observations 5375 5375 3831 3876 2703 1898

Intercept -0.1826 -0.2257 -0.2257 -0.2257 -0.1162 -0.1162

(-10.76) (-8.21) (-8.21) (-8.21) (-2.08) (-2.08)
Turnex -0.0551 -0.1127 -0.1127 -0.1127 -0.2791 -0.2791
(-3.98) (-7.59) (-7.59) (-7.59) (-3.94) (-3.93)
Size 0.0277 0.0184 0.0184 0.0184 0.0249 0.0249
(5.54) (2.35) (2.35) (2.35) (2.07) (2.07)
Flows -0.0157 0.0136 0.0136 0.0136 -0.0032 -0.0032
(-1.21) (0.62) (0.62) (0.62) (-0.11) (-0.11)
F 0.0690 0.0111 0.1401 -0.0118 -0.0402
(1.98) (0.28) (3.37) (-0.16) (-0.45)
Turnex *F 0.1064 0.1332 0.0818 0.1997 0.1614
(3.97) (3.03) (3.38) (2.09) (1.33)
Size*F 0.0182 0.0302 0.0050 0.0013 0.0129
(1.74) (2.11) (0.46) (0.09) (0.74)
Flows*F -0.0512 -0.0375 -0.0730 -0.0210 -0.0047
(-1.91) (-1.26) (-2.21) (-0.56) (-0.11)

40
 Table 8 (continued)

Panel B: Following strong market performance (1992, 1996-2000)

(1) (2) (3) (4) (5) (6)
All All New York ex New York All High Turn NY High Turn
Observations 10099 10099 6866 7515 5030 3278

Intercept 0.0642 0.1020 0.1020 0.1020 -0.0090 -0.0090

(3.56) (3.45) (3.45) (3.45) (-0.16) (-0.16)
Turnex 0.0693 0.1193 0.1193 0.1193 0.3471 0.3471
(5.57) (6.42) (6.42) (6.42) (3.97) (3.96)
Size 0.0339 0.0412 0.0412 0.0412 0.0577 0.0577
(6.84) (5.00) (4.99) (4.99) (4.32) (4.32)
Flows 0.0032 0.0027 0.0027 0.0027 0.0035 0.0035
(0.24) (0.13) (0.13) (0.13) (0.10) (0.10)
F -0.0596 -0.0891 -0.0389 0.0434 0.0571
(-1.60) (-2.10) (-0.92) (0.59) (0.66)
Turnex *F -0.0854 -0.1258 -0.0538 -0.3088 -0.4823
(-3.43) (-4.23) (-1.80) (-2.80) (-3.36)
Size*F -0.0125 -0.0137 -0.0126 -0.0258 -0.0267
(-1.22) (-1.04) (-1.12) (-1.58) (-1.39)
Flows*F 0.0007 0.0044 -0.0013 0.0220 0.0308
(0.03) (0.15) (-0.04) (0.49) (0.66)

41
 Table 9
Relation between risk-adjusted returns and excess turnover for a given fund performance level
This table shows the estimation results from the panel regression of risk-adjusted fund returns on lagged turnover
and location for high and low-performing funds while controlling for fund size. The high (low) performing fund is
a fund whose average return in a given year is greater than (less than) the median return for its fund category in
that year. Only the unconditional risk-adjusted returns based on the Carhart’s (1997) four-factor model are
considered. The risk-adjusted returns for each fund and year are computed by regressing its twelve monthly
excess returns of the year on the corresponding benchmark portfolios observed during the same twelve-month
period. Funds that have less than a twelve-month history for a given year are disregarded. The regression model
has the following form:

αˆ i ,t = d o + d1Turniex, t + d 2 Sizei , t + d s Flowsi , t + d 4 Fi + d 5 (Turniex, t Fi ) + d 6 (Sizei , t Fi ) + d 7 (Flowsi , t Fi ) + ei , t ,

where α̂ i,t is the estimated risk-adjusted return for fund i in year t, while Turn iex,t , Sizei ,t , and Flows i ,t are the
fund’s i log excess turnover, log size, and flows, respectively. Panel A shows the results for low-performing
funds, Panel B – for high-performing funds. F denotes the finance center dummy, and NY – the city of New York.
The sample of risk-adjusted returns covers the period from 1992 to 2001. The estimates on the year control
dummies are not shown in the table. The regression also includes the year control dummies but they are not
shown in the model or in the table.

Panel A: Low-performing funds

(1) (2) (3) (4) (5) (6)
All All New York ex New York All High Turn NY High Turn
Observations 7683 7683 5393 5695 3897 2682

Intercept -0.2196 -0.2256 -0.2252 -0.2349 -0.2022 -0.2133

(-8.34) (-7.22) (-6.34) (-6.93) (-3.51) (-3.43)
Turnex -0.0237 -0.0385 -0.0389 -0.0386 -0.1187 -0.1181
(-1.99) (-2.62) (-2.64) (-2.62) (-1.95) (-1.93)
Size 0.0234 0.0166 0.0163 0.0164 0.0319 0.0316
(5.03) (2.15) (2.11) (2.12) (2.84) (2.80)
Flows -0.0021 0.0079 0.0068 0.0086 0.0257 0.0250
(-0.19) (0.47) (0.41) (0.51) (1.16) (1.12)
F 0.0058 -0.0482 0.0537 0.0779 0.0845
(0.20) (-1.32) (1.60) (1.27) (1.06)
Turnex *F 0.0288 0.0194 0.0396 -0.1078 -0.2045
(1.26) (0.57) (1.76) (-1.26) (-1.65)
Size*F 0.0128 0.0293 -0.0003 -0.0115 -0.0051
(1.32) (2.25) (-0.03) (-0.85) (-0.32)
Flows*F -0.0174 -0.0039 -0.0285 -0.0216 -0.0234
(-0.80) (-0.15) (-1.13) (-0.72) (-0.69)

42
 Table 9 (continued)

Panel B: High-performing funds

(1) (2) (3) (4) (5) (6)
All All New York ex New York All High Turn NY High Turn
Observations 7791 7791 5304 5696 3836 2494

Intercept 0.1872 0.2424 0.2121 0.2340 0.0295 0.0773

(3.35) (4.03) (4.89) (3.05) (0.45) (1.10)
Turnex 0.0685 0.0995 0.0989 0.0991 0.3857 0.3851
(5.63) (5.63) (5.64) (5.63) (4.56) (4.59)
Size 0.0222 0.0286 0.0276 0.0292 0.0500 0.0477
(4.99) (4.02) (3.88) (4.09) (4.36) (4.16)
Flows -0.0064 -0.0162 -0.0161 -0.0169 -0.0315 -0.0313
(-0.47) (-0.73) (-0.72) (-0.75) (-0.87) (-0.87)
F -0.0831 -0.1103 -0.0586 0.0103 0.0068
(-2.34) (-2.89) (-1.42) (0.16) (0.09)
Turnex *F -0.0516 -0.0695 -0.0419 -0.2604 -0.3207
(-2.12) (-2.55) (-1.39) (-2.60) (-2.82)
Size*F -0.0099 -0.0113 -0.0117 -0.0347 -0.0343
(-1.09) (-1.02) (-1.12) (-2.36) (-1.93)
Flows*F 0.0174 0.0020 0.0295 0.0434 0.0257
(0.63) (0.07) (0.90) (0.99) (0.55)

43
 0.3 1.5

0.2
Excess Turnover Differential

Excess Size Differential

1.0
0.1

0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+ 0.5
-0.1

-0.2
0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+
-0.3

Fund Age (Years) Fund Age (Years)

0.3 1.5
High High
Low Low
0.2
Excess Turnover Differential

Excess Size Differential

1.0
0.1

0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+ 0.5
-0.1

-0.2
0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+
-0.3
Fund Age (Years) Fund Age (Years)

Figure 1. Relation between excess turnover and size differentials and fund age. The figure shows the excess
turnover and size differentials (in log scale) between funds located in financial centers and those located outside
depending on the fund age. The excess turnover is the average difference between the turnover of each fund in a
given year and the median turnover, which is computed for each year and fund category. The excess size is the
average difference between the size of each fund in a given year and the median fund size, which is computed for
each year and fund category. The size of the fund is measured in terms of its total net assets. The fund age is the
difference in years between the current year and the year of organization of the fund. Plot A depicts the two
relations across all funds; Plot B – separately for high and low-performing funds. The high (low) performing fund
is a fund whose average return in a given year is greater than (less than) the median return for its fund category in
that year.

44
 0.2 1.5
Excess Turnover Differential

Excess Size Differential

1.0
0.1

0.5
0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+

0.0
-0.1 1 to 2 3 to 5 6 to 9 10 to 14 15+

Manager Age (Years) Manager Age (Years)

0.2 1.5
High High
Low Low
Excess Size Differential
Excess Turnover Differential

1.0
0.1

0.5
0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+

0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+
-0.1
Manager Age (Years) Manager Age (Years)

Figure 2. Relation between excess turnover and size differentials and manager age. The figure shows the
excess turnover and size differentials (in log scale) between funds located in financial centers and those located
outside depending on the manager age. The excess turnover is the average difference between the turnover of each
fund in a given year and the median turnover, which is computed for each year and fund category. The excess size
is the average difference between the size of each fund in a given year and the median fund size, which is
computed for each year and fund category. The size of the fund is measured in terms of its total net assets. The
manager age is the difference in years between the current year and the year when the manager was first assigned
to the fund. Plot A depicts the two relations across all funds; Plot B – separately for high and low-performing
funds. The high (low) performing fund is a fund whose average return in a given year is greater than (less than)
the median return for its fund category in that year.

45
 2.5
Ratio of Funds by Fund Objective
2.0

1.5

1.0
AG
GI
0.5
IN
LG
0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+
Fund Age (years)

2.5
Ratio of Funds by Fund Objective

2.0

1.5

1.0
AG
0.5 GI
IN
LG
0.0
1 to 2 3 to 5 6 to 9 10 to 14 15+
Manager Age (years)

Figure 3. Proportion of funds by fund investment objective for different age cohorts. The figure shows the
ratio of the number of funds in financial centers versus those in other places by each of the four fund investment
objectives and for different age cohorts. Plot A shows this ratio across fund age groups, Plot B – across fund
manager age groups. The fund investment objectives are aggressive growth, AG, growth and income, GI, income,
IN, and large growth, LG.

46