Will Women Be Women?

Analyzing the Gender Difference among Financial Experts

Daniela Beckmann and Lukas Menkhoff*

Discussion Paper No.391

February 2008
ISSN 0949-9962

Abstract
There are robust gender differences in the domains of risk taking, overconfidence and compe-
tition behavior. However, as expertise tends to level these differences, we ask whether finan-
cial experts still show gender dissimilarities in their domains of decision making? We analyze
survey responses of 649 fund managers in the U.S., Germany, Italy and Thailand, and find
that female fund managers tend to behave as expected from gender studies: they are more risk
averse and shy away from competition in the tournament scenario. The expected lower degree
of overconfidence by women is yet so small that it becomes insignificant in fund manage-
ment.

JEL-Classification: J16 (economics of gender), G23 (funds), D8 (information, uncertainty)

Key words: fund managers, gender differences, risk, overconfidence, tournament
behavior

We very much appreciate the contributions made by the fund managers who were available
for interviews and took time to respond to the survey. We thank the Investment Management
Associations in Germany, Italy, and Thailand for the provision of indispensable recommenda-
tion letters. Furthermore, we are grateful to Torben Lütje, Michael Melvin, Luca Rebeggiani,
Maik Schmeling and Ulrich Schmidt in helping us to conduct the survey and giving com-
ments, and we thank two anonymous referees for their support. Last but not least, financial
support by the Volkswagen Foundation is gratefully acknowledged.

* Daniela Beckmann, Lukas Menkhoff, Leibniz Universität Hannover, Department of Eco-

nomics, Königsworther Platz 1, D-30167 Hannover, Germany, beckmann@gif.uni-
hannover.de, menkhoff@gif.uni-hannover.de
 2

Will Women Be Women?

Analyzing the Gender Difference among Financial Experts

I. INTRODUCTION
We tend to have established stereotypes in mind when we think about the behavior of women
and men. A prominent case is behavior towards risk, in which women are often seen as being
more risk averse. Indeed, there are reams of carefully designed experiments which exactly
detect risk-related gender dissimilarities and thus approve it as being fair to say that women
are different from men in this respect (cf. Byrnes et al., 1999, Eckel and Grossman, 2006).
While virtually all of these studies reflect students’ behavior, it is reassuring that survey evi-
dence from households confirms experimental findings (cf. e.g. Barsky et al., 1997, Donkers
et al., 2001). However, groups of students and households may be representative for the popu-
lation but not necessarily for those financial professionals who are experts in managing risks.
As we know that familiarity with risk changes behavior towards the latter (cf. e.g. Slovic et
al., 2000, Dwyer et al., 2002), the question emerges whether the change of focus from laymen
to financial experts also impacts the gender difference. We therefore conduct a survey among
professional fund managers which is the first for this group to analyze differences between
women and men in their behavior towards risk.
Fund managers are not only worth a detailed analysis because they are experts in man-
aging risks but also because they work in a field of financial decision making. There is some
evidence that decisions in financial affairs may differ from decisions in abstract gambles, pos-
sibly because financial decisions involve clear incentives. This framing effect has been found
in experiments and reduces the gender difference in risk aversion (cf. Schubert et al., 1999).
Thus, there seem to be two separate forces which reduce the gender difference in risk aver-
sion, i.e. familiarity with risk and risk decision under financial framing. Both forces apply to
fund managers’ business. Therefore, we expect that gender-specific behavior towards risk is
overruled by the more uniform behavior of being a financial expert.
As this hypothesis of “expertise dominates gender” is not restricted to risk behavior
alone, we also test it with respect to two further phenomena: (over)confidence and tournament
behavior. Both are important for financial decision making, and women have been found to
differ from their male counterparts: people are tentatively overconfident but women less than
men (cf. e.g. Lenney, 1977, Lundeberg et al., 1992). Barber and Odean (2001) regard the
stronger overconfidence of men as absolutely undisputed and thus use gender as a proxy for
 3

overconfidence in order to explain excessive, detrimental trading behavior. Successive studies

have questioned, however, whether it is really male overconfidence that drives the outcome.
Deaves et al. (2004) and Biais et al. (2005) confirm that overconfidence drives trading (and
men trade relatively more) but do not find male students to be more overconfident in their
miscalibration measures. There is thus some doubt as to whether male overconfidence is ro-
bust in financial affairs; indeed, it motivates one to extend evidence to financial experts—here
fund managers—whose performance would be affected by excessive trading. Finally, there is
emerging evidence that women “shy away from competition” to quote Niederle and Vester-
lund (2007). This pattern has also been found for example by Gneezy et al. (2003) or Dohmen
and Falk (2006). As it already seems to exist among young children (cf. Gneezy and Rus-
tichini, 2004), the less competitive behavior of women may be seen as typical. However, this
behavior stands in contrast to tournaments in the fund management industry (cf. Brown et al.,
1996) where fund managers are publicly benchmarked and ranked. This contrast raises the
question whether a gender difference shows up in the environment of fund management tour-
naments.1 Again, we are not yet aware of any evidence on gender-related differences in this
field.
Despite this contrast between “shying away from competition” versus “welcoming tour-
naments”, there is no general presumption that typical female behavior would be less efficient
than male behavior.2 This also applies to higher risk aversion as fund managers do not aim for
maximum returns but for risk-adjusted performance. It seems intriguing in this respect that the
few studies analyzing gender-related performance among U.S. fund managers find that
women achieve about the same risk-adjusted performance as their male colleagues (cf. Atkin-
son et al., 2003, Niessen and Ruenzi, 2006).3 If there are effects from the gender difference on
performance they may go in both directions: some measures of risk aversion that we use indi-
cate (more) distorted risk perception of women and would thus be a disadvantage to invest-

1
It is perfectly possible that there is an ongoing self-selection mechanism among women on becom-
ing fund managers. If so, this supports the hypothesis of “expertise dominates gender”.
2
Interestingly, recent research indicates that women react on incentives—which are extremely strong
in fund management—as men do (Paarsch and Shearer, 2007).
3
Unfortunately, Chevalier and Ellison (1999), who carefully analyze a large sample of fund manag-
ers, do not include a gender variable because they regard their 7%-share of women as being too
small. Atkinson et al. (2003) compare a sample of 70 women managing bonds to a matching sample
of men. Niessen and Ruenzi (2006) rely on a share of female professionals in their sample of around
10%, which equals approximately between 130 and 170 women over the whole sample period.
Green et al. (2007) conclude that female analysts seem to do their job as well as men although they
have a somewhat worse forecasting performance. Kumar (2007) detects female analysts to provide
bolder forecasts but to be relatively more accurate.
 4

ment performance for female analysts, whereas less female overconfidence would probably
benefit performance.
Our research on gender differences extends most former studies towards the realm of fi-
nancial experts, here professional fund managers. As we search for information which is un-
distorted by (unobservable) legal, customer or company restrictions, we ask fund managers
directly via a questionnaire survey.4 In this way, we obtain their version of risk behavior,
overconfidence and tournament behavior and do not need to infer it from fund transactions,
which also reflect the impact from legal restrictions or customer requirements. Accordingly,
survey data are undistorted in this respect and thus provide useful evidence complementing
transaction data.5 The survey approach has another important advantage by compiling the
necessary data, including the above-mentioned three areas of gender differences—each of
them addressed by specifically designed items—as well as a full set of relevant control vari-
ables. Finally, as we are interested in generally valid gender differences, we provide informa-
tion on fund managers in four countries, i.e. the United States, Germany, Italy and Thailand.6
The resulting response from 649 fund managers, among them 125 women, yields one main
finding.
Testing the “expertise dominates gender” hypothesis surprisingly ends in a victory for
the gender difference. Whether we take descriptive statistics or control for a large set of com-
peting influences, the gender variable always shows the sign as expected from the earlier lit-
erature, i.e. women will indeed be women even in the demanding environment of fund man-
agement. However, the economic importance of the gender difference varies. First, female
fund managers keep their more risk averse behavior but on the other hand the effect is com-
paratively weak for the established risk measures. Second, we reject the view that women are
less overconfident than men for the case of fund managers (although coefficients’ signs hint at
the expected direction). This finding is robust to an extension of three different measures of
overconfidence, i.e. overoptimistic self-assessment, illusion of control and miscalibration.
Third, evidence is consistent with the hypothesis that female fund managers shy away from

4
We only know of one study which analyzes some attitudes towards risk of US financial experts,
mainly investment counselors, finding that these female professionals seem to be more risk averse
than men (Olsen and Cox, 2001).
5 The conduct of a new survey is often the only way to get information. Examples include the survey
of 139 compensation executives (Levine, 1993), the survey of 184 firms on reasons for wage rigid-
ity (Campbell and Kamlani, 1997), the survey of 84 central banks on issues of credibility (Blinder,
2000) and the survey of 200 foreign exchange professionals on their time horizons in position tak-
ing (Menkhoff, 2001).
6 The selection of target countries is determined by achieving variety (Europe, the US and an emerg-
ing market) and by achieving reasonable response due to support from investment associations (see
data section).
 5

competition. Fourth, the relative economic importance of the gender-related difference in ex-
plaining behavior is sometimes small in comparison to competing influences, indicating that
indeed financial expertise decreases the gender difference—but does not erase it.
The remainder of the paper is organized as follows: Section II. describes the data set
used. Section III. addresses risk behavior, while Section IV. focuses on overconfidence and
Section V. on tournament behavior. Section VI. concludes.

II. DATA
The utilized data consists of a unique set of responses from 649 fund managers in four coun-
tries particularly compiled for our research.
We conducted the written survey with professional fund managers in the United States,
Germany, Italy, and Thailand between spring 2003 and winter 2004. With the help of the re-
spective investment association7 in three out of the four countries, we generated 649 valid
questionnaires, where gender is split into 125 women and 524 men. Prior oral interviews in all
four countries with fund managers of different hierarchies and of both sexes, previous test
runs for clarity, and guaranteed anonymity were factors to ensure useful responses.
In order to check for representativeness of the country data sets generated, we compare
assets under management covered by the surveyed firms with the market structure in total.
Test results are shown in Table 1. In all four countries, we find that responses by and large
reflect the market with respect to the representation of firm size.
Furthermore, Table 1 reveals firm participation rate, number of questionnaires and share
of female participation for each country. Indeed, as fund managers sometimes passed ques-
tionnaires to colleagues, we cannot report response rates in terms of the distributed question-
naires. We know, however, that we generated a participation rate of 64.7% altogether at the
company level. While participation among contacted firms was highest in Thailand with
93.5% (29 of 31 contacted companies participated), it was relatively the lowest in the United
States, where 74 out of 250 contacted companies responded (29.6%). In absolute numbers,
generated country samples cover between 112 questionnaires in Italy and 263 in Germany.
Regarding representation of gender, the global mutual fund industry can undoubtedly be
seen as a male-dominated field. Nevertheless, the share of female respondents varies consid-
erably among the four countries surveyed. While the U.S. and German samples comprise

7
We received helpful support from the “Bundesverband Investment und Asset Management e.V.”, in
short: “BVI” for Germany, the “Associazione Italiana del Risparmio Gestito”, in short: “Assoges-
tioni” for Italy and the “Association of Investment Management Companies”, in short “AIMC” for
the Thai market. In the USA, Professor Michael Melvin, at that time at Arizona State University,
supported this research.
 6

slightly over ten percent of female participants, in Italy one fifth of the surveyed managers are
women, and in Thailand the sample is almost equally balanced. Indeed, taking the U.S. mu-
tual fund industry on the one hand, female shares have constantly been around 10% within the
last twelve years (cf. Chevalier and Ellison, 1999, Niessen and Ruenzi, 2006). In Thailand on
the other hand, the fund managers interviewed highlighted that the female work force, in con-
trast to some other East Asian countries, has been comparably strong in banking and financial
service sectors for decades and that several women have reached considerable influence in the
financial industry. This is in line with the substantial female share in our sample.
In order to learn more about surveyed fund managers’ characteristics, Table 2 provides
descriptive statistics concerning their demographic, career and fund characteristics, split by
gender and testing the significance of gender-related differences. Table 2 first presents four
demographic characteristics, i.e. age, experience, marital status and education. The U.S. fund
managers lead the others with respect to age and experience, probably matching the fund
management industry’s development in a cross-country comparison. We also find women to
be younger and less experienced than their male counterparts in the U.S., Germany and Italy,
with some statistically significant differences in the last two countries. In Thailand, however,
we find our data corresponding to what fund managers told us during previous interviews:
women are on average as old as their male colleagues and even significantly more experi-
enced. While the minority of female fund managers are married, differing significantly from
their male counterparts in Germany and Italy, the opposite holds only for the U.S., where it is
the majority of (the comparatively older) fund managers of both sexes who are married. In
terms of education, our sample discloses no significant gender difference in any country.
Turning to the three career characteristics, we find that women hold significantly lower
positions than men in Germany and Italy, a fact which goes hand in hand with significantly
lower personal assets under management and shorter working hours. In the U.S. and Thai-
land—by contrast—there is no gender difference observable in these three variables.
Finally, women do not really manage other fund types than men. Although they seem to
work somewhat more often in mutual than in other funds, in equity more than in bond funds
and in actively more than in passively managed funds, the gender difference is only signifi-
cant in two cases. Moreover, this tendency only applies to the three Western countries and is
rather reversed—although not to a statistically significant degree—in Thailand.
Taken together, for the U.S. and for Thailand, there are no significant gender differ-
ences revealed, while for the two other markets, female and male fund managers often differ
significantly in their individual characteristics. Accordingly, women in our total sample tend
 7

to be younger, less experienced, less often married, work in lower positions, manage fewer
assets, work fewer hours, more often manage mutual funds with a higher degree of active
management.
Due to these marked differences—in individual characteristics and between countries—
it appears crucial to control for these variables when analyzing the question of whether
women will be women in terms of risk taking, overconfidence and tournament behavior.

III. RISK BEHAVIOR

Analyzing fund managers’ behavior towards risk by applying three measures, we reveal
women to behave more risk averse than men. This finding holds for comparisons between
sexes for the single measures in each country as well as for multivariate regression evidence
where countries are pooled and the gender variable is controlled by a set of determinants.
We start the analysis of gender differences among professional financial market players
with a first direct measure of risk behavior, i.e. self-stated risk aversion, which has proved to
be a most useful variable (cf. Kapteyn and Teppa, 2002). This survey measure has been vali-
dated in experiments and has thus proved that women are more risk averse (cf. Dohmen et al.,
2005). As a second measure, we assess the degree of loss aversion as in particular modeled by
Kahneman and Tversky’s (1979) prospect theory; experimental work by Schmidt and Traub
(2002) shows women to be more loss averse than men. Finally, we complement these gener-
ally established risk measures by including a questionnaire item on the empirically detected
tendency of investors to hold losing stocks too long and sell winning stocks too early, the so-
called disposition effect (cf. Shefrin and Statman, 1985). We regard the disposition effect as a
kind of distorted risk aversion; Weber and Welfens (2006) do not find a significant gender
difference for individual investors.
Table 3 shows the first results in the form of mean responses for both sexes and a test to
discover whether there is gender homogeneity. Directly asked about risk taking in profes-
sional investment decisions, only in Italy women respond to act more risk averse than men to
a statistically significant degree. Turning to loss aversion, we discover that women show a
significantly higher loss aversion in Thailand. Furthermore, a notably higher number of
women than men agree on their behavior being consistent with the disposition effect. As an
intermediate summary, we can state that there is a tendency by female fund managers to be-
have more risk averse, as all significant results support this tendency (5 out of 12 cases). In-
terestingly, in terms of cross-country considerations, we reveal two significant gender differ-
 8

ences for the Italian and Thai market, only one different item for the German and none for the
U.S. market.
So far, we have not controlled for fund managers’ characteristics. These characteristics,
however, do not only mediate gender differences, as previously described in studies such as
Atkinson et al. (2003) for wealth and investment knowledge, or Sunden and Surrette (1998) in
the case of marital status, but they have also been shown to exert significant influence on an
individual’s risk behavior itself (cf. e.g. Jianakoplos and Bernasek, 1998). In order to be on
the safe side when analyzing gender impact, we use—in addition to marital status—the other
variables introduced in Section 2 as controls: starting with demographic variables, Dohmen et
al. (2005) detect willingness to take risk to increase corresponding to younger age levels (al-
though Dwyer et al., 2002, do not find this effect). Graham (1999) finds that risk aversion
decreases with experience of analysts (cf. Clement and Tse, 2005), which is tentatively con-
firmed by Menkhoff et al. (2006) for fund managers and by Dhar and Zhu (2007) for individ-
ual investors. Moreover, e.g. Dwyer et al. (2002) find that more educated investors tend to
take more risk than their less educated counterparts. As our second group of variables, we use
career proxies to capture the effect of wealth and knowledge on (more) risk taking which has
been identified for individual investors (cf. Graham et al., 2006). In our analysis we thus con-
trol for a higher position, for more personal assets under management and for more working
hours. The third group of control variables considers possible impacts from the type of fund
managed: it has been hypothesized that mutual fund business may be more competitive than
pension funds (Lakonishok et al., 1992) and it seems self-evident that fund managers may to
some extent self-select into preferred fields. More risk-averse fund managers may thus prefer
bonds over equities and passive over active management. Finally, as we pool the total sample
of 649 cases, we have to include dummy variables to control for country-specific effects.
Results of the multivariate ordered probit regressions are given in Table 4. Overall, we
find the gender variable to reveal significant explanatory power over two out of three risk
behavior variables; gender is thus among the most important variables in our setting to under-
stand risk aversion. Although controlling for the ten variables just discussed above, we reveal
that female professionals behave consistently more risk averse than their male colleagues.8 In
detail, directly questioned less risk averse investment behavior (column 1) is significantly
lower for female fund managers (but more pronounced among higher positioned professionals
managing less index-oriented bond funds). By contrast, a lower loss aversion (column 2) is
not significantly different for female fund managers but determined by a few other variables,
8
These results, as well as those reported later, are robust to the exclusion of different variables from
the regressions.
 9

such as being married and having more assets under management. The disposition effect (col-
umn 3) is significantly smaller for male asset managers as well as for those with the relatively
superior career characteristics of being in a higher position with more assets under their per-
sonal responsibility; the latter scenario indicates an efficient selection mechanism as more
career success is related to less distorted decision making (moreover, managers in mutual
funds and equity business express a lower disposition effect).
It seems noteworthy that risk behavior in this approach is not well understood by demo-
graphic characteristics, which makes the explanatory power of gender even more pronounced.
Rather, a self-selection mechanism may be observed, because in particular equity managers
and more successful managers—proxied by position and assets under management—respond
by being more willing to take risks. Finally, country dummies do not reveal much in the mul-
tivariate approach.
Overall, our primary attention is on the gender variable and this gives a clear result: fe-
male fund managers tend to show more risk averse behavior then their male counterparts (as it
is also found in the general population), and a higher degree of the disposition effect, indicat-
ing distorted behavior towards risk. We now turn to the relation between overconfidence and
gender.

IV. OVERCONFIDENCE
We adopt three measures to address overconfidence and find that female fund managers do
not differ significantly from their male colleagues.
Previous studies have shown that different measures of overconfidence are not necessar-
ily correlated.9 Therefore, we choose to address all three forms of overconfidence independ-
ently, respectively covering a measure of the better-than-average effect, illusion of control,
and miscalibration. In detail, we ask the surveyed professionals to self-assess their achieve-
ments in fund management compared to their peer group; we question them about their ap-
proval of the statement that published business news do not surprise them; and finally, taking
into account that overconfident individuals have been found to miscalibrate by giving too nar-
row confidence intervals, we confront them with the task of 90% probability forecasts of their
respective home market equity index for the following month.10 Again, as demonstrated in
Section 3 on risk behavior, we first show comparisons between sexes for three measures in all
the four countries surveyed and then give regression results on the pooled sample.

9
A thorough experimental assessment of different measures is given by Glaser et al. (2005). Neither
they nor e.g. Deaves et al. (2004) and Oberlechner and Osler (2006) find positive correlations be-
tween e.g. miscalibraton and an overly positive self-assessment.
10
Cf. e.g. Biais et al. (2005), Glaser et al. (2005) or Menkhoff et al. (2006) for using similar measures.
 10

Table 5 shows our findings, split by gender and with indications of statistical signifi-
cance of a gender-related difference. Asked about their achievements relative to their peers,
surveyed fund managers across all four countries and both sexes reveal on average overly
positive self-assessments: while a value of four would indicate fund managers to be equally as
good as their benchmark, surveyed professional groups show means between 2.7 and 3.3 and
thus describe themselves on average to be better than their benchmark. Only in the case of
Italy does the gender difference significantly show less overconfident female response. Ad-
dressing control illusion among professionals, we assume that more confident managers will
experience lower surprise about upcoming news in financial markets. However, on average,
we find that both women and men rather contradict the statement that “Most of the published
business news does not surprise me at all”, and we thus find neither any indication for this
form of overconfidence, nor do we detect any significant gender difference. For the third type
of overconfidence, miscalibration, a comparison between female and male professionals con-
cerning estimated width of equity index ranges, calculated from maximum and minimum
level estimations in relation to the actual index level, reveals overconfidence again: estimates
neither include future realization in 90% of cases, nor does the given range seem wide enough
when compared to a GARCH (1,1) forecast. However, it is only in the German case that
women are less overconfident than their male colleagues. To sum up, the gender difference is
rarely significant, i.e. only in two out of 12 cases, although these two cases—as well as sev-
eral signs—point towards less overconfident female fund managers.
In order to control for the possible influence of personal characteristics, we choose the
same set of control variables as before when moving towards the multivariate approach.
Again, controlling for several factors is well justified by previous research and derived con-
troversies: Bengtsson et al. (2005) or Deaves et al. (2004) confirm e.g. age to be a relevant
factor in observed gender differences in their experimental work. While e.g. Gervais and
Odean (2001) theoretically and Locke and Mann (2001) empirically argue that overconfi-
dence is highest for inexperienced traders, experts have been found to be more likely than
inexperienced subjects to show overconfident behavior (cf. e.g. Heath and Tversky, 1991 or
Glaser et al., 2005). Moreover, Barber and Odean (2001) have highlighted the influence of
marital status on their results, revealing differences to be greatest between single women and
men. Finally, Estes and Hosseini (1988) reveal the significant role of education and familiar-
ity—proxied by our career variables—for confidence in investment decisions.
The multivariate estimation results are shown in Table 6 and confirm our previous find-
ing from Table 5. To cut a long story short: we do not find the gender variable to significantly
 11

matter for any of the three overconfidence measures. Our findings are thus in line with the
student experiments by Deaves et al. (2004) and Biais et al. (2005) and indicate that fund
managers behave differently from the individual investors analyzed by Barber and Odean
(2001). Further significant variables indicate that the self-assessment measure of overconfi-
dence may also involve justified confidence as more experienced and harder working fund
managers feel more successful. Moreover, the plausibility of our survey as well as the impor-
tance of control variables appears to be underlined by the last regression in Table 6: fund
managers with a larger responsibility for assets under management indeed give broader—i.e.
here better—equity index estimates. Country dummies do not give a clear pattern here.
Overall, we find some evidence in explaining overconfidence but women are rarely sig-
nificantly different from men in the 12 comparisons in Table 5 and the gender variable is
never significant in the three regressions shown in Table 6.
The third part of our analysis sheds light on a possible gender difference in fund manag-
ers’ competition behavior in a tournament situation.

V. TOURNAMENT BEHAVIOR
Regarding tournament behavior, we confront fund managers with two different realistic sce-
narios and subsequently find women to behave differently from men. Indeed, women are
found to change their strategy more often when they are ahead of or behind the market—they
try to perform closer to the market development than men.
We translate the literature’s finding that women are more likely than men to shy away
from competition into tournament situations which often happen near the end of the invest-
ment period (cf. Brown et al., 1996). In these situations, fund managers are aware of the op-
portunity, on the one hand, to receive disproportionately high new cash flows into “their”
fund for extraordinary performance (cf. Sirri and Tufano, 1998). Seeking such performance
requires deviating from the market benchmark. On the other hand, bad investment deci-
sions—i.e. performance below the market—are punished with negative career consequences
(cf. Chevalier and Ellison, 1999). In the survey, we give two scenarios: either achieved suc-
cess in outperforming the benchmark or underperformance near the end of the valuation pe-
riod.
In both scenarios—out- or underperformance—fund managers face a two-stage deci-
sion. First, they could stick to their earlier investment strategy, which should actually be the
one derived from their (private) information. The only new piece of information that is going
to be considered here is their fund’s position relative to the benchmark. This information
 12

might motivate them to change their strategy. Then there is a second decision to be made: in
the case of outperformance, professionals could either intend to “lock in” their achieved out-
performance or even aggressively welcome competition by increasing the risk level in order
to become top performer. In the second case, underperforming their benchmark might lead
them to try to guard a chance at “catching up” relative to their benchmark or to decrease the
risk level in order to save them from further hurtful performance losses.
Former mutual fund studies do not provide undisputed expectations on fund managers’
behavior in such tournaments (e.g. Busse, 2001, Goriaev et al., 2005). From a gender perspec-
tive, however, the less competitive behavior of women (Gupta et al., 2005, Dohmen and Falk,
2006) may imply a stronger orientation towards the benchmark and less competition seeking
by sticking to their preferred strategy (which obviously deviates from the benchmark). This
expectation concerning women’s behavior is consistent with the finding that U.S. female fund
managers show very good or very bad performance ranks less often than men (Niessen and
Ruenzi, 2006).
Starting with the scenario of outperformance, Table 7 shows female professionals in It-
aly and Thailand to be significantly more likely than their male colleagues to change instead
of keeping their previous strategy. Indeed, they predominantly (55.2% in Thailand and even
81.8% in Italy) “lock in” achieved outperformance. When we look at the “second” decision—
that is the direction of riskiness conditional on change—only a very limited number11 of pro-
fessionals chooses an increase in the risk level.
Turning to the second scenario—previous underperformance—we disclose once more
that women are more likely to vary the risk level than men. This time, findings tend to be con-
sistent across all four countries, although only being significant for Thailand. Concerning the
second decision, whether to increase or decrease riskiness in the future, we reveal interesting
findings: women are more prone than men to increasing the relative risk level, with statistical
significance for Germany (where 20.8% of surveyed female fund managers increase risk) and
Thailand (with as high as 58.6% of women doing so). Women’s willingness to increase risk
may seem surprising at first sight, but is consistent with a strong ambition to stick close to the
benchmark’s performance. Concerning a decrease in risk level, gender evidence is mixed,
with a few U.S. female professionals (11.8% compared to only 3.1% of their male counter-
parts) significantly preferring this behavior.
As in the previous sections, we test the impact of gender in a multivariate approach—
here a binary probit model—controlling for the usual set of variables. Table 8 shows that our
11
We refrain from integrating this specific scenario in following multivariate regressions as the num-
ber of observations is too small for meaningful estimations.
 13

findings on gender differences also hold in the extended framework. Gender is the only vari-
able to consistently explain the first decision to be made as women more often change their
strategy in both scenarios. Looking at the second decision, i.e. the sign of change, women
consistently decrease the risk level in the outperformance scenario. When confronted with
underperformance near the end of the valuation period, women are significantly more likely
to increase the risk level. Only in terms of a risk level decrease is the gender variable not sig-
nificant. Besides, significant country variables reinforce effects in Germany, Thailand and
Italy in comparison to the United States.
Overall, female fund managers seem to confirm the kind of competitive behavior which
is found among women in general, i.e. they shy away from competition in the assumed tour-
nament scenarios more often than men.

VI. CONCLUSIONS
The question whether women behave differently from men has been extensively researched
and reveals robust gender differences. Women are significantly more risk averse, tend to be
less overconfident and behave less competitively oriented. Thus, women behave differently in
these three domains, which are highly important in the financial industry as risk taking, confi-
dent decision making and tough competition are part of fund managers’ jobs. Moreover, we
know that expertise in a certain field tends to level the gender difference. We thus get a most
interesting setting: do we reject in our sample of financial experts the gender difference as it is
usually found in the three domains of risk taking, overconfidence and competition behavior?
In short, does expertise dominate gender?
We address this question head on by providing fresh evidence. What is most interesting
is that our survey covers fund managers, thus extending beyond students or individual inves-
tors, who are usually considered. Moreover, the survey covers four countries, i.e. the U.S.,
Germany, Italy and Thailand. From a methodological point of view it has two advantages:
first, the answers reflect the fact that the respondents’ opinions are undistorted by restrictions
that co-determine their observable investment behavior. Second, the survey approach allows
controlling for a large set of variables that have been revealed to be potentially relevant in
earlier studies.
We find that fund managing women will be women in their profession: they are more
risk averse and shy away from competition in the tournament. Regarding the lesser overconfi-
dence of women found in most cases, however, the gender difference is so weak in fund man-
agement that it becomes insignificant in the regression approach.
 14

Although the overall outcome of obviously existing gender differences in behavior

could have been expected from former gender studies, it is nonetheless surprising when
evaluated from the studies focusing on experts’ behavior. To summarize, we thus conclude a
rejection of the “expertise dominates gender” hypothesis for our sample.
The persistence of gender differences in investment-related behavior may raise concerns
as to whether this is compatible with market efficiency. What we can say on this issue is, in-
deed, that our analysis does not provide any indication that women may be less successful as
fund managers than men: a different risk aversion is mediated by the convention in fund man-
agement to focus on risk-adjusted returns. Women’s tendency in tournaments to change their
position more than men towards the market may lead to less extreme outcomes but does not
imply lower returns (cf. Nissen and Ruenzi, 2006). Finally, less overconfidence of women—
even if insignificant—will hardly deteriorate performance.
So the discovery of systematic differences in behavior—in combination with equal per-
formance—stimulates thoughts about consequences for the industry: female fund managers
may be better suited to female customers who share their pattern in behavior than do men and
female fund managers may possess a greater ability to exercise certain investment styles than
their male colleagues.
 15

REFERENCES
Atkinson, Stanley M., Samantha Boyce Baird and Melissa B. Frye (2003). Do Female Mutual
Fund Managers Manage Differently, Journal of Financial Research. 26(1): 1-18.
Barber, Brad M. and Terrance Odean (2001). Boys Will Be Boys: Gender, Overconfidence,
and Common Stock Investment, Quarterly Journal of Economics. 116(1): 261-292.
Barsky, Robert B., Thomas F. Juster, Miles S. Kimball and Matthew D. Shapiro (1997). Pref-
erence Parameters and Behavioral Heterogeneity: An Experimental Approach in the
Health and Retirement Study, Quarterly Journal of Economics. 112(2): 537-579.
Bengtsson, Claes, Mats Persson, and Peter Willenhag (2005). Gender and Overconfidence,
Economics Letters. 86(2): 199-203.
Biais, Bruno, Denis Hilton, and Karine Mazurier (2005). Judgemental Overconfidence, Self-
Monitoring, and Trading Performance in an Experimental Financial Market, Review of
Economic Studies. 72(2): 287-312.
Blinder, Alan S. (2000). Central-Bank Credibility: Why Do We Care? How Do We Build It?,
American Economic Review. 90(5): 1421-1431.
Brown, Keith C., W. Van Harlow and Laura T. Starks (1996). Of Tournaments and Tempta-
tions: An Analysis of Managerial Incentives in the Mutual Fund Industry, Journal of Fi-
nance. 51(1): 85-110.
Busse, Jeffrey A. (2001). Another Look at Mutual Fund Tournaments, Journal of Financial
and Quantitative Analysis. 36(1): 53-73.
Byrnes, James P., David C. Miller and William D. Schafer (1999). Gender Differences in
Risk-Taking: A Meta-Analysis, Psychological Bulletin. 125(3): 367-383.
Campbell III, Carl M. and Kunal S. Kamlani (1997). The Reasons for Wage Rigidity: Evi-
dence from a Survey of Firms, Quarterly Journal of Economics. 112(3): 759-789.
Chevalier, Judith und Glenn Ellison (1999). Career Concerns of Mutual Fund Managers,
Quarterly Journal of Economics. 114(2): 389-432.
Clement, Michael B. and Senyo Y. Tse (2005). Financial Analyst Characteristics and Herding
Behavior in Forecasting, Journal of Finance. 60(1): 307-341.
Dhar, Ravi and Ning Zhu (2006). Up, Close and Personal: An Individual Level Analysis of
the Disposition Effect, Management Science. 52(5): 726-740.
Deaves, Richard, Erik Lüders, and Guo Ying Luo (2004). An Experimental Test of the Impact
of Overconfidence and Gender on Trading Activity, Working Paper, Center for Euro-
pean Economic Research (ZEW) and Rutgers University.
Dohmen, Thomas and Armin Falk (2006). Performance, Pay and Multi-Dimensional Sorting:
Productivity, Preferences and Gender, IZA Discussion Paper No. 2001, Bonn.
Dohmen, Thomas, Armin Falk, David Huffman, Uwe Sunde, Jürgen Schupp, Gert G. Wagner
(2005). Individual Risk Attitudes, New Evidence from Large, Representative, Experi-
mentally-Validated Survey, IZA Discussion Paper, No. 1730, Bonn.
Donkers, Bas, Bertrand Melenberg, and Arthur van Soest (2001). Estimating Risk Attitudes
Using Lotteries: A Large Sample Approach, Journal of Risk and Uncertainty. 22(2):
165-195.
Dwyer, Peggy D., James H. Gilkeson, and John A. List (2002). Gender Differences in Re-
vealed Risk Taking: Evidence from Mutual Fund Investors, Economics Letters. 76(2):
151-158.
Eckel, Catherine C. and Philip J. Grossman (2006). Men, Women and Risk Aversion: Ex-
perimental Evidence, in: Plott, Charles R., and Vernon L. Smith (eds.), Handbook of
Experimental Economics Results. New York, forthcoming.
Estes, Ralph and Jinoos Hosseini (1988). The Gender Gap on Wall Street: An Empirical
Analysis of Confidence in Investment Decision Making, Journal of Psychology. 122(6):
577-590.
 16

Gervais, Simon and Terrance Odean (2001). Learning to be Overconfident, Review of Finan-
cial Studies. 14(1): 1-27.
Glaser, Markus, Thomas Langer and Martin Weber (2005). Overconfidence of Professionals
and Lay Men: Individual Differences Within and Between Tasks, Sonderforschungs-
bereich 504 Publications No. 05-25, University of Mannheim.
Gneezy, Uri, Muriel Niederle and Aldo Rustichini (2003). Performance in Competitive Envi-
ronments: Gender Differences, Quarterly Journal of Economics. 118(3): 1049-1074.
Gneezy, Uri and Aldo Rustichini (2004). Gender and Competition at a Young Age, American
Economic Review. 89(2): 386-391.
Goriaev, Alexei P., Theo E. Nijman and Bas J. M. Werker (2005). Yet Another Look at Mu-
tual Fund Tournaments, Journal of Empirical Finance. 12(1): 127-137.
Graham, John R. (1999). Herding Among Investment Newsletters: Theory and Evidence,
Journal of Finance. 54(1): 237-268.
Graham, John R., Campbell R. Harvey and Hai Huang (2006). Investor Competence, Trading
Frequency, and Home Bias, AFA 2006 Boston Meetings Paper.
Green, T. Clifton, Narasimhan Jegadeesh and Yue Tang (2007). Gender and Job Performance:
Evidence from Wall Street, NBER Working Paper 12897.
Gupta, Nabanita D., Anders Poulsen and Marie-Claire Villeval (2005). Male and Female
Competitive Behavior – Experimental Evidence, CNRS Working Paper No. 05-12,
Ecully.
Heath, Chip and Amos Tversky (1991). Preferences and Belief: Ambiguity and Competence
in Choice under Uncertainty, Journal of Risk and Uncertainty. 4(1): 5-28.
Jianakoplos, Nancy A. and Alexandra Bernasek (1998). Are Women More Risk Averse?,
Economic Inquiry. 36(4): 620-630.
Kahneman, Daniel and Amos Tversky (1979). Prospect Theory: An Analysis of Decision Un-
der Risk, Econometrica. 47(2): 263-291.
Kapteyn, Arie and Federica Teppa (2002). Subjective Measures of Risk Aversion and Portfo-
lio Choice, Center for Economic Research Discussion Paper No. 11, Tilburg University.
Kumar, Alok (2007). Do Social Biases Influence Market’s Interpretation of New Public In-
formation?, Working Paper, University of Texas at Austin.
Lakonishok, Josef, Andrei Shleifer and Robert W. Vishny (1992). The Structure and Perform-
ance of the Money Management Industry, Brookings Papers on Economic Activity, Mi-
croeconomics. 339-391.
Lenney, Ellen (1977). Women’s Self-Confidence in Achievement Settings, Psychological
Bulletin. 84(1): 1-13.
Levine, David I. (1993). Fairness, Markets, and Ability to Pay: Evidence from Compensation
Executives, American Economic Review. 83(5): 1241-1259.
Locke, Peter R. and Steven C. Mann (2001). House Money and Overconfidence on the Trad-
ing Floor, Working Paper, George Washington University.
Lundeberg, Mary A., Paul W. Fox and Judith Punccohar (1992). Highly Confident but
Wrong, Gender Differences and Similarities in Confidence Judgments, Journal of Edu-
cational Psychology. 86(1): 114-121.
Menkhoff, Lukas (2001). Short-Term Horizons in Foreign Exchange? Survey Evidence from
Dealers and Fund Managers, Kyklos. 54(1): 27-47.
Menkhoff, Lukas, Ulrich Schmidt and Torsten Brozynski (2006). The Impact of Experience
on Risk Taking, Overconfidence, and Herding of Fund Managers: Complementary Sur-
vey Evidence, European Economic Review. 50(7): 1753-1766.
Niederle, Muriel, and Lise Vesterlund (2007). Do Women Shy Away from Competition? Do
Men Comepete Too Much?, Quarterly Journal of Economics. 122(3): 1067-1101.
Niessen, Alexandra and Stefan Ruenzi (2006). Sex Matters: Gender and Mutual Funds, CFR
Working Paper No. 06-01, University of Cologne
 17

Oberlechner, Thomas and Carol Osler (2006). Overconfidence in Currency Markets, Working
Paper, Brandeis University.
Olsen, Robert A. and Constance M. Cox (2001). The Influence of Gender on the Perception
and Response to Investment Risk: The Case of Professional Investors, Journal of Behav-
ioral Finance. 2(1): 29-36.
Paarsch, Harry J. and Bruce S. Shearer (2007). Do Women React Differently to Incentives?
Evidence from Experimental Data and Payroll Records, European Economic Review.
51(7): 1682-1707.
Schmidt, Ulrich and Stefan Traub (2002). An Experimental Test of Loss Aversion, Journal of
Risk and Uncertainty. 25(3): 233-249.
Schubert, Renate, Martin Brown, Matthias Gysler and Hans Wolfgang Brachinger (1999).
Financial Decision-Making: Are Women Really More Risk-Averse?, American Eco-
nomic Review. 89(2): 381-385.
Shefrin, Hersh and Statman, Meir (1985). The Disposition to Sell Winners Too Early and
Ride Losers Too Long: Theory and Evidence, Journal of Finance. 40(3): 777-790.
Sirri, Erik R. and Peter Tufano (1998). Costly Search and Mutual Fund Flows, Journal of Fi-
nance. 53(5): 1589-1622.
Slovic, Paul, B. Fischhoff, and S. Lichtenstein (2000). Facts and Fears: Understanding Per-
ceived Risk, in: Slovic, Paul (ed.), The Perception of Risk. London: Earthscan/James &
James: 137-153.
Sunden, Annika E. and Brian J. Surette (1998). Gender Differences in the Allocation of As-
sets in Retirement Savings Plans, American Economic Review. 88(2): 207-211.
Weber, Martin and Frank Welfens (2006). An Individual Level Analysis of the Disposition
Effect: Empirical and Experimental Evidence, Working Paper, University of Mannheim.
TABLE 1. Comparison of the sample with the industry's structure1
Structure of the asset management industry
in relation to respective country sub sample
(assets under management)
USA GER ITA THA
H0: no difference2 -1.213 -0.669 -0.403 -0.136
(0.225) (0.503) (0.687) (0.892)
Firm participation rate 29.6% 77.3% 58.2% 93.5%
Number of questionnaires 148 263 112 126
Share of female participants 11.4% 10.0% 21.1% 46.4%
1
The U.S. market data is taken from the 'Pensions & Investments' money managers directory 2003
(www.pionline.com). For Germany, we use market data from the annual report 2003 of the BVI. The Italian
market data for 2004/2005 is taken from the Italian Investment Management Association’s Website “Asso-
gestioni”. Thailand data for 2004 emanates from a market share datasheet provided by the Thai Association
of Investment Management Companies “AIMC” as well as selected companies’ information.
2
The table gives the z-value of the respective Mann-Whitney U test with the p-value in parentheses.
TABLE 2. Descriptive statistics of fund managers’ characteristics
Characteristics H0: no gender difference1
split by country
USA GER ITA THA
and gender
♂ ♀ ♂ ♀ ♂ ♀ ♂ ♀

Demographic characteristics
2
Age in years ~42 ~40 ~36 ~34 ~37 ~32 ~35 ~35
-1.263 (0.206) -2.002***(0.045) -3.115*** (0.002) -0.527 (0.598)
Experience2 ~13 ~12 ~7 ~6 ~9 ~6 ~6 ~8
in years -1.008 (0.313) -1.386 (0.166) -2.098** (0.036) -2.110**(0.035)
Married3 in % 82.5 82.4 57.9 37.5 57.1 36.4 47.9 38.6
-0.039 (0.969) -1.816* (0.069) -2.173** (0.030) -0.783 (0.434)
Academic 80.3 93.8 86.5 88.0 91.7 100.0 95.5 96.6
education3 in % -1.308 (0.191) -0.206 (0.837) -1.394 (0.163) -0.292 (0.771)

Career characteristics
4
Position 2.43 2.24 2.01 1.46 2.63 2.00 1.82 1.84
-0.636 (0.525) -3.172***(0.002) -2.398** (0.017) -0.536 (0.592)
Personal AuM5 4.00 3.60 3.52 2.45 3.56 2.15 3.16 3.18
-0.843 (0.399) -3.332***(0.001) -4.170*** (0.000) -0.070 (0.944)
Working hours6 ~52 ~51 ~50 ~47 ~49 ~47 ~46 ~45
-1.097 (0.273) -1.811* (0.070) -2.035** (0.042) 0.928 (0.354)

Fund characteristics
7
Type of fund 2.16 1.82 2.23 1.88 1.50 1.17 2.00 2.11
-1.555 (0.120) -1.883* (0.060) - 1.501 (0.133) -0.750 (0.453)
Investment 1.75 1.59 1.56 1.48 2.02 1.65 2.02 2.20
Segment8 -0.667 (0.505) -0.664 (0.507) -1.567 (0.117) -1.083 (0.279)
Allowed 2.74 2.41 2.73 2.70 2.81 2.35 3.12 3.12
tracking error9 -0.946 (0.344) -0.271 (0.786) -1.731* (0.083) -0.413 (0.679)
1
The table gives the mean value for male and female asset managers in comparison as well as the z-value of
the respective Mann-Whitney U test regarding gender specific differences in the four displayed countries.
The p-value is given in parentheses. Asterisks refer to level of significance: * 10%, ** 5%, *** 1%.
2
Mean age and experience are derived from six response categories that range from 1=<31 years old, and
1=<4 years of experience, respectively, to 6=>older than 50 years, and 6=more than 15 years of experience.
3
Marital status and education are binary variables. While marital status is classified by 1=single and
3=married, education is divided into 1=non-academic and 3=academic.
4
Position ranges from 1=junior asset manager, over 2=senior, 3=head of team to 4=CIO/CEO.
5
Assets under personal responsibility / management (AuM) are classified into six response categories from 1
(lowest) to 6 (highest) with amounts that are country specifically adopted.
6
Working hours are assessed by six response categories ranging from 1=<41 hours to 6=>60 hours.
7
Type of fund is classified by 1=mutual fund, 2=both mutual and other funds, 3=other funds, i.e. pension/
provident, restricted/private fund.
8
For the investment segment we split equity fund management from the one of bond and money market
funds. Codings are 1=equities, 2=both equities and bonds/money market, 3=bonds and money market.
9
The allowed tracking error is assessed as follows: “How actively (i.e. high tracking error) can you manage
your portfolio at most?”. Response categories range from 1=high tracking error to 6=indexing.
TABLE 3. Risk behavior and gender
[A] “In respect of professional investment decisions, I mostly act…”; six response categories ran-
ging from 1=very risk averse to 6=little risk averse.
[B] “In case of loss positions in my portfolio I generally wait for a price rebound instead of selling
those securities.” Six response categories, ranging from 1=complete approval to 6=complete
contradiction.
[C] “I prefer to take profits instead of cutting losses, when I am confronted with unexpected liquid-
dity demands.” Six response categories, ranging from 1=complete approval to 6=complete con-
tradiction.
Asset managers’ Ho: no gender difference1
risk behavior
USA GER ITA THA
♂ ♀ ♂ ♀ ♂ ♀ ♂ ♀

[A] Risk taking 3.55 3.59 3.50 3.50 3.63 3.25 3.30 3.05
-0.139 (0.890) -0.070 (0.944) -1.671* (0.095) -1.451 (0.147)
[B] Loss aversion 4.18 4.47 4.58 4.39 4.29 4.05 3.83 3.45
-0.925 (0.355) -0.790 (0.430) -0.640 (0.522) -1.953* (0.051)
[C] Disposition 3.82 3.53 4.22 3.54 3.96 3.35 3.31 2.90
effect -0.856 (0.392) -2.043** (0.041) -1.667* (0.095) - 1.716* (0.086)
1
The table gives the mean value for male and female asset managers in comparison as well as the z-value of
the respective Mann-Whitney U test regarding gender specific differences in the four displayed countries.
The p-value is given in parentheses. Asterisks refer to level of significance: * 10%, ** 5%, *** 1%.

TABLE 4. Risk behavior in a multivariate framework

Lower Lower Smaller
risk aversion loss aversion disposition effect
Female -0.127* (0.074) -0.062 (0.376) -0.172** (0.014)
Older -0.104 (0.133) -0.010 (0.881) 0.031 (0.643)
More experienced -0.001 (0.990) -0.046 (0.413) -0.042 (0.458)
Married 0.004 (0.941) 0.186*** (0.002) 0.080 (0.178)
Better educated -0.055 (0.535) 0.055 (0.524) 0.093 (0.280)
Higher position 0.158* (0.061) 0.078 (0.359) 0.216** (0.010)
More personal AuM -0.011 (0.802) 0.144*** (0.001) 0.122*** (0.005)
More working hours 0.011 (0.836) 0.018 (0.731) 0.039 (0.445)
Other than mutual fund type 0.031 (0.618) -0.023 (0.704) -0.122** (0.047)
Rather bond investment fund -0.127** (0.034) -0.070 (0.237) -0.219*** (0.000)
Lower tracking error allowed -0.219*** (0.000) 0.036 (0.430) 0.023 (0.611)
Dummy GER -0.182 (0.227) 0.358** (0.015) 0.340** (0.021)
Dummy ITA -0.029 (0.890) 0.053 (0.796) 0.066 (0.746)
Dummy THA -0.098 (0.600) -0.303 (0.101) -0.117 (0.524)
LR statistic 41.66*** 63.61*** 83.92***
(Pseudo-)R2 0.04 0.05 0.06
1
The table gives the coefficients of the ordered PROBIT regressions with p-values in parentheses. Asterisks
refer to level of significance: * 10%, ** 5%, *** 1%. Explanations of the three dependent variables can be
found in Table 3, information about the control variables is given in Table 2. Country dummies for Ger-
many, Italy and Thailand are binary coded.
TABLE 5. Overconfidence and gender
[A] Unrealistically positive self-evaluation is tested as follows: “How do you assess your achievement in asset
management – compared to other asset managers’ achievement in the same investment segment?” Re-
sponse categories range from 1=much better to 7=much worse.
[B] Control illusion is assessed by: “Most of the published business news does not surprise me at all.” Re-
sponse categories cover 1=complete approval to 6=complete contradiction.
[C] Miscalibration is tested by 90% probability performance forecasts of home indices: “Please estimate the
development of the S&P 500 (EuroSTOXX 50 / SET, respectively) within the next month.”

Evidence for male Ho: no gender difference1

overconfidence? USA GER ITA THA
♂ ♀ ♂ ♀ ♂ ♀ ♂ ♀
[A] Management 2.67 2.75 3.13 3.00 2.81 3.30 3.09 3.22
performance -0.046 (0.964) -0.587 (0.557) -2.000** (0.046) -0.879 (0.380)
[B] News novelty 3.08 3.53 3.23 3.42 3.41 3.71 3.06 3.19
-1.418 (0.156) -1.338 (0.181) -1.142 (0.253) -0.855 (0.392)
[C] Home index 17.6 21.5 22.2 37.4 11.8 8.4 15.6 16.3
range in %2 -1.331 (0.183) -1.934* (0.053) -0.894 (0.371) -0.251 (0.802)
1
The table gives the mean value for male and female asset managers in comparison as well as the z-value of
the respective Mann-Whitney U test regarding gender specific differences in the four displayed countries.
The p-value is given in parentheses. Asterisks refer to level of significance: * 10%, ** 5%, *** 1%.
2
Index ranges are given in percentages; calculated by relating the difference of maximum and minimum level
estimations to the actual level
TABLE 6. Overconfidence in a multivariate framework
Overconfidence controlled for Ordered PROBIT / TOBIT regressions1
personal characteristics Worse assessment Higher surprise Broader
of management about business home index
performance news estimation
Female 0.068 (0.347) 0.077 (0.271) -0.399 (0.699)
Older 0.076 (0.273) -0.113 (0.100) 0.465 (0.645)
More experienced -0.102* (0.074) 0.031 (0.578) -0.640 (0.420)
Married 0.027 (0.653) -0.019 (0.755) 1.158 (0.164)
Better educated 0.023 (0.799) 0.044 (0.601) 1.591 (0.207)
Higher position -0.061 (0.467) -0.009 (0.916) -0.808 (0.510)
More personal AuM 0.011 (0.802) -0.017 (0.690) 1.082* (0.087)
More working hours -0.147*** (0.005) 0.049 (0.343) 1.051 (0.157)
Other than mutual fund type 0.021 (0.730) 0.093 (0.128) 0.885 (0.298)
Rather bond investment fund 0.007 (0.904) 0.032 (0.593) -0.345 (0.688)
Lower tracking error allowed 0.064 (0.170) 0.030 (0.508) -0.359 (0.580)
Dummy GER 0.447*** (0.003) -0.271* (0.066) 3.888* (0.087)
Dummy ITA 0.157 (0.447) -0.018 (0.930) -6.035* (0.055)
Dummy THA 0.223 (0.236) -0.607*** (0.001) 0.778 (0.777)
Constant - - 5.279 (0.413)
LR statistic 42.35*** 20.92 -
(Pseudo-)R2 0.04 0.02 0.10
1
The table shows the coefficients of the ordered PROBIT regressions for the first two dependant variables,
testing for unrealistically positive self-evaluation and control illusion, and a TOBIT regression for the home
index estimation range that addresses miscalibration, respectively. P-values are given in parentheses. Aster-
isks refer to level of significance: * 10%, ** 5%, *** 1%. Further explanations of the three dependent vari-
ables can be found in Table 5, information about the control variables is given in Table 2. Country dummies
for Germany, Italy and Thailand are binary coded.
TABLE 7. Tournament behavior and gender
Tournament behavior is assessed by requesting the surveyed asset managers to imagine that – apart from any
fund's restriction – their portfolio's performance differs from its benchmark near the end of the period.
We then address two scenarios as follows: “If my portfolio has outperformed (underperformed, respectively in
the second scenario) its benchmark so far, I would...
[A] not change my strategy,
[B] increase the relative risk level to become a top performer,
[C] decrease the relative risk level to lock in the performance.”

Gender differences Ho: no gender difference1

in tournament be- USA GER ITA THA
havior ♂ ♀ ♂ ♀ ♂ ♀ ♂ ♀
Scenario 1: If the portfolio has outperformed its benchmark so far…
[A] Change instead
of keeping 20.0 5.9 33.6 45.8 61.4 81.8 38.8 60.4
strategy -1.409 (0.159) -1.191 (0.234) -1.781* (0.075) -2.393** (0.017)
[B] Increase risk 0.8 0.0 0.9 0.0 4.8 0.0 7.5 5.2
-0.390 (0.696) -0.415 (0.678) -0.694 (0.488) 0.09 (0.93)
[C] Decrease risk 19.2 5.9 32.7 45.8 56.6 81.8 31.3 55.2
1.363 (0.173) -1.253 (0.210) -1.924* (0.054) -2.625*** (0.009)
Scenario 2: If the portfolio has underperformed its benchmark so far …
[A] Change instead
of keeping 15.4 24.6 30.4 37.5 41.0 50.0 47.0 74.1
strategy -0.852 (0.394) 0.714 (0.475) -0.758 (0.449) -3.065*** (0.002)
[B] Increase risk 12.3 11.8 7.9 20.8 18.1 36.4 25.8 58.6
-0.069 (0.945) -1.940* (0.052) -1.587 (0.113) -3.668*** (0.000)
[C] Decrease risk 3.1 11.8 22.5 16.7 22.9 13.6 21.2 15.5
-1.692* (0.091) -0.326 (0.744) -0.498 (0.619) -0.766 (0.444)
1
The table gives mean percentages for male and female asset managers in comparison as well as the z-value
of the respective Mann-Whitney U test regarding gender specific differences in the four displayed countries.
The p-value is given in parentheses. Asterisks refer to level of significance: * 10%, ** 5%, *** 1%.
TABLE 8. Tournament behavior in a multivariate framework

Tournament behavior Binary PROBIT regressions1

controlled for personal
characteristics Scenario 1: Outperformance Scenario 2: Underperformance
Change Decrease risk Change Increase Decrease
strategy strategy risk risk
Female 0.181** 0.199** 0.196** 0.294*** 0.058
(0.043) (0.028) (0.027) (0.004) (0.632)
Older -0.135 -0.131 -0.122 -0.102 -0.087
(0.147) (0.163) (0.196) (0.397) (0.451)
More experienced 0.079 0.088 0.035 0.032 0.080
(0.288) (0.243) (0.647) (0.743) (0.396)
Married -0.006 -0.025 -0.049 -0.091 -0.029
(0.937) (0.742) (0.531) (0.343) (0.769)
Better educated -0.075 -0.085 -0.138 -0.282** 0.142
(0.502) (0.450) (0.226) (0.031) (0.396)
Higher position -0.068 -0.087 0.028 -0.059 0.096
(0.534) (0.439) (0.801) (0.666) (0.495)
More personal AuM -0.081 -0.082 -0.119** -0.145** -0.109
(0.155) (0.156) (0.040) (0.042) (0.150)
More working hours 0.034 0.013 0.080 0.175** -0.028
(0.612) (0.853) (0.235) (0.039) (0.742)
Other than mutual fund -0.013 -0.030 -0.096 -0.061 -0.069
type (0.868) (0.704) (0.227) (0.535) (0.490)
Rather bond investment 0.181 0.193** 0.072 0.088 0.054
fund (0.017) (0.012) (0.350) (0.341) (0.598)
Lower tracking 0.060 0.049 0.080 0.118 0.021
error allowed (0.304) (0.409) (0.179) (0.104) (0.772)
Dummy GER 0.590*** 0.589*** 0.301 -0.131 0.899***
(0.003) (0.003) (0.141) (0.595) (0.004)
Dummy ITA 1.259*** 1.202*** 0.561** 0.568* 0.662*
(0.000) (0.000) (0.035) (0.063) (0.079)
Dummy THA 0.699*** 0.518** 1.056*** 1.070*** 1.170***
(0.004) (0.036) (0.000) (0.000) (0.009)
Constant -0.968* -0.801 -0.495 -0.693 -1.852**
(0.099) (0.177) (0.404) (0.323) (0.026)
LR statistic 65.11*** 61.65*** 69.66*** 84.60*** 26.48**
(McFadden) R2 0.12 0.11 0.13 0.22 0.08
1
The table gives the coefficients of the binary PROBIT regressions with p-values in parentheses (dependent
Variables are binary coded). Asterisks refer to level of significance: * 10%, ** 5%, *** 1%. Further expla-
nations of the five binary dependent variables can be found in Table 7, information about the control vari-
ables is given in Table 2. Country dummies for Germany, Italy and Thailand are binary coded.