Content uploaded by Arun J. Prakash
Author content
All content in this area was uploaded by Arun J. Prakash
Content may be subject to copyright.
ARE WOMEN MORE RISK-AVERSE THAN MEN?
Ann Marie Hibbert
Assistant Professor of Finance
West Virginia University
00amhibbert@mail.wvu.edu
Edward Lawrence
Assistant Professor of Finance
Florida International University
11elawrenc@fiu.edu
Arun Prakash
Knight Ridder Research Professor of Finance
Florida International University
22prakasha@fiu.edu
November 18, 2008
2
ARE WOMEN MORE RISK-AVERSE THAN MEN?
Abstract
This paper measures the gender difference in risk aversion using a sample that controls for biases
in the level of education and finance knowledge. We survey 1,382 Finance and English
professors from universities across the United States and compare their actual portfolio
allocations to that of respondents in the Federal Reserve’s Survey of Consumer Finances (SCF).
Our findings suggest that when individuals have the same level of education irrespective of their
knowledge of finance, women are no more risk averse than men. We also find that the gender-
risk aversion relation is a function of age, income, wealth, marital status, race/ethnicity and the
number of children under 18 in the household.
3
I. Introduction
There is anecdotal evidence in the finance literature that gender plays a role in the
individual/household portfolio allocation decision, where women tend to invest more in low risk
investments and hence are found to be more risk-averse than men.1 However, most studies that
find gender to be a significant determinant of risk aversion also find various proxies of the level
of education and investment knowledge to be significant. In addition, even when gender is
insignificant, risk aversion has been found to be inversely related to the level of education.2 It is
therefore not clear if there is really a gender difference in risk aversion or if the gender-risk
aversion difference is being confounded by gender biases in the level of education or in the
knowledge of finance. We address this issue by investigating if there is a gender difference in
risk aversion when individuals either (a) have attained the same level of education or (b) have
similar knowledge of finance.
To disentangle the gender bias from any education or finance knowledge bias, we use two
datasets in our empirical investigation. The first dataset is the 2004 Federal Reserve Board’s
Survey of Consumer Finances (SCF)3. The SCF has been the database often used by researchers
to investigate the individual/household portfolio allocation decision. A cursory look at the SCF
data reveals a disparity in the level of education across gender. According to the 2004 US census,
1 See for example Cohn, Lewellen, Lease and Schlarbaum (1975), Sunden and Surette (1998), Jianakoplos and
Bernasek (1998), Bajtelsmit, Bernasek and Jianakoplos (1999), Bernasek and Shwiff (2001), Dwyer, Gilkeson and
List (2002), Felton, Gibson and Sanbonmatsu (2003), Agnew, Balduzzi and Sunden (2003) and Watson and
McNaughton (2007)
2 Riley and Chow (1992) use data from the Survey of Income and Program Participation (SIPP) and conclude that
relative risk aversion is inversely related to education, but unrelated to gender.
3 The Survey of Consumer Finances (SCF) is conducted every three years by the Board of Governors of the Federal
Reserve System. The main purpose of the survey is to help the government and, ultimately, the public at large
understand the financial condition of families in the United States and study the effects of changes in the economy.
At the time of our empirical study, the 2004 results are the most current available.
4
52% of the population were female and the fraction of females who had attained at least some
college degree was only slightly less than the fraction of males; 52.6% versus 53.7%. By
comparison, of the 4519 respondents included in the SCF 2004 dataset (hereafter SCF04), 22%
were females and most of the females in the sample had “some college” education, while most of
the males had a “college degree”. This survey allows us to investigate the gender difference in
risk aversion among individuals who have attained different levels of formal education. We find
that in the SCF04 sample, women are significantly more risk-averse than men. We also find that
income and education are the most important variables in explaining the individual’s level of risk
aversion in this sample.
Our second dataset is the result of a survey of Finance and English faculty at universities
across the United States. During the fall of 2007 and the spring of 2008, we surveyed Finance
and English professors at universities across the United States and collected information on their
actual investment holdings as well as their household and other demographic information. Since
all individuals in this sample have achieved at least a graduate degree, we implicitly control for
the level of education. Using this sample, we investigate if women with the same level of
education as men are more likely to invest a larger portion of their portfolio in safe assets. Unlike
the SCF04 sample, the results of the faculty sample vary with the marital status of the individual.
We find that after controlling for education, the gender difference in risk aversion is confined to
individuals who are married or live with a partner. In other words, single women are no more
risk-averse than single men when they are both highly educated.
We then study the sub-samples of Finance and English professors to investigate the
importance of finance knowledge vis–à–vis formal education in general. The results of our Tobit
regression analysis suggest that when individuals have the same level of education, irrespective
5
of their knowledge of finance, women are no more risk-averse than men.
We make four important contributions to the body of research that investigates gender
difference in risk aversion. First, unlike prior studies that only consider a subset of the
individual/household portfolio, e.g., mutual fund holdings, (Dwyer, et al. 2002), brokerage
accounts (Cohn, et al. 1975), Defined Contribution (DC) plans (Watson and McNaughton 2007;
Sunden and Surette 1998; and others), we consider both the overall portfolio as well as the
retirement savings and find that our results are even stronger in the retirement savings.
Second, unlike most prior studies that use household data where the gender of the decision-
maker is not specified, in our survey we explicitly request information on the gender of the
financial decision-maker in the household. Lindamood, Hanna and Bi (2007) in a critique of
methodological issues with using the SCF data to perform gender analyses, highlight the fact that
since the respondent is not necessarily the household head (who may be of a different gender),
there may be incorrect inferences. Our results demonstrate that when a woman is the financial
decision-maker in the household, even if she invests jointly with her spouse, it is important to
control for her education level in gender-risk aversion studies.
Third, in our analysis we use both parametric and non-parametric tests and include an
extensive set of control variables such as race/ethnicity, number of children under 18 in the
household, non-financial asset holdings and proxies for job security. While a few studies have
included race in models that predict portfolio allocation decisions and have obtained mixed
results, to our knowledge the interaction of race and gender has only been considered by
Jianakoplos and Bernasek (1998) and they consider only Caucasians and African-Americans. We
use the entire range of race categories based on the Federal Board’s classification and find that
the gender difference in risk aversion is most significant among Caucasians. We also find that
6
the interaction between ethnicity and risk aversion is a function of the length of time that the
individual has lived in the United States.
Finally, in our survey we ask individuals to rank specific asset classes based on their
perception of the “riskiness” of each asset. Consistent with Olsen and Cox (2001) and Dwyer, et
al. (2002), who suggest that the gender difference in risk aversion is more significant for
portfolios at the extreme risk classes, we find that after controlling for education and finance
knowledge, women are less likely to have ever invested in an asset class that they rank as being
the riskiest. Further, when both men and women invest in the asset class they consider most
risky, women are more likely to invest the smallest portion of their portfolio in that asset class.
II. Literature Review
The literature on gender difference in risk aversion has proceeded mainly along two lines. The
first line of studies focuses on finding if there is indeed a gender difference in risk aversion. The
second line of research focuses on the psychological factors that would result in women being
more risk-averse than men, (see Byrnes, Miller and Schafer 1999 for an excellent review of this
area of the literature). Our study lies in the first line of research and we investigate if women are
more risk-averse than men after controlling for the level of education, knowledge of finance and
other demographic factors. We consider this to be important, since as Sunden and Surette (1998)
suggest, with the increasing trend towards defined contribution (DC) plans, if women are not
making optimal investment choices this could severely impact their accumulated wealth for
retirement. In addition, the notion of a gender difference in risk aversion has led to stereotyping.
Roszkowski and Grable (2005) analyze how accurately financial advisors gauge their male and
female clients' risk tolerance and find that financial advisors have a distorted sense of the risk
tolerance of men and the risk aversion of women. They find that advisors generally overestimate
7
the risk tolerance of men and underestimate the risk tolerance of women. They suggest that
women earn lower returns on their investments, partly because they take less risk but also
because financial planners tend to overestimate their level of risk aversion and will not even
advise certain riskier investments.
Empirical tests of gender difference in individual investment behavior have largely used
either data on actual asset holdings or questionnaires to solicit hypothetical portfolio decisions.
However, the results have varied depending on the sample studied and the control variables used.
A large number of studies have used the Federal Reserve’s Survey of Consumer Finances (SCF),
which is a triennial survey of the balance sheet, pension, income, and other demographic
characteristics of U.S. families. Sunden and Surette (1998) use data from the 1992 and 1995 SCF
to investigate individual investment in DC plans and find that women are less likely to have a
DC plan than men and both gender and marital status significantly affect the actual allocation of
assets. They also find that neither age nor education seems to affect the investment decision.
Jianakoplos and Bernasek (1998) use data from the 1989 SCF to examine the household holdings
of risky assets to determine whether there are gender differences in financial risk-taking. They
find that gender differences are influenced by wealth, age, race and the number of children in the
household. In particular, they find that single black women hold significantly more risky assets
than single white women, but the reverse is true for single men and married couples. Bajtelsmit,
Bernasek and Jianakoplos (1999) use data from the 1989 SCF survey and find that women
exhibit greater relative risk aversion in their allocation of wealth into defined contribution
pension assets. While these studies provide evidence of a gender difference, the SCF dataset does
not provide any information on the financial decision-maker in each household. In addition, the
SCF uses the “years of schooling of the head of the household” to proxy for the level of
8
education, of which on average only 25% of the respondents in each of the sample had attained
at least a college degree.
Another survey that has been used to investigate the household asset allocation decision
is the Survey of Income and Program Participation (SIPP), which provides the economic status
of U.S. households. Riley and Chow (1992) use the results of the 1985 SIPP to study individual
asset allocation and risk behavior of a random sample of US households and conclude that
relative risk aversion is inversely related to age, education, wealth and income but unrelated to
gender. In their study only 22.1% of the subjects had at least a college degree. In our analysis, we
control for the level of education, and unlike the SCF and SIPP datasets in which there is no
indication of who makes the financial decisions in the household, in our survey we specifically
elicit this information.
To investigate the influence of gender on investments, some researchers have focused on
subsets of the investor’s portfolio. Dwyer, et al. (2002) use data from a national survey of 2000
mutual fund investors to investigate whether gender is related to risk-taking as revealed in
mutual fund investment decisions. They find a positive and significant relation between risk-
taking and both income and education. When investor knowledge is excluded, the gender
variable is positive and significantly different from zero in all cases, suggesting that men take on
more risk than women when selecting mutual funds. However, when investor knowledge is
included, gender is significant for only the riskiest investment. The authors caution that care
should be taken in making conclusions about gender differences without controlling for investor
knowledge of financial markets and investments. Their study provides evidence for the need to
control for finance education before investigating gender bias, as we do in our empirical
investigation.
9
Agnew, Balduzzi and Sunden (2003) study 6,778 retirement accounts during the period April
1994-August 1998. They aim to build on the studies of Barber and Odean (2001), who find
evidence of overconfidence in discount brokerage accounts by investigating investor behavior in
401(k) investment plans. They find that men invest more in equities and trade more actively than
women. They also find that married investors hold more in equities than their single
counterparts. Additionally, investors who earn higher salaries tend to be more aggressive in their
allocations and tend to trade more often. They also report that age induces investors to allocate
less to equities and to rebalance more frequently. We extend their study and investigate if there is
a similar interaction of gender and marital status in predicting investment behavior when
financial literacy is controlled for.
Bernasek and Shwiff (2001) survey a sample of 270 faculty (42.2% females) at five
universities in Colorado in 2000 and elicit complete information on the percentage of their
contribution pension invested in stocks. They find that gender is a consistently significant factor
in explaining the percentage of an individual’s retirement fund invested in stocks, with the
percentage decreasing if the respondent is a woman. They also find that if the respondent has a
Ph.D. compared with a BA or MA or if the respondent has a degree in the liberal arts compared
to a business degree, the percentage invested in stocks decreases.
Another approach that has been used to test gender difference is simulated trading and
hypothetical portfolio allocations. Felton, Gibson and Sanbonmatsu (2003) examine the role of
gender and optimism in determining the preference for risk in investment choices of 66
undergraduate students (34 females and 32 males) in a semester-long investment contest, with
both monetary and academic incentives. They use four proxy measures of risk-taking: (1) the
number of futures and options contracts traded, (2) the overall number of transactions, (3) the
10
number of transactions on the Nasdaq and (4) the number of transactions on the NYSE. They
find that males make more risky investment choices (particularly in the case of the first two
measures) and have greater variability in their final portfolio value than females and the
difference is primarily due to the riskier choices of optimistic males. They conclude that the
well-documented gender difference in investment strategies may be due to the more specific sub-
group of optimistic males.
Grable, Lytton and O'Neill (2004) survey 421 relatively young individuals (average age of
32.03) via the internet and find that men report a higher risk tolerance score than women. The
aim of their study is to investigate if projection bias, as explained by regret theory, shape
financial risk tolerance attitudes. Even though gender difference was not the focus of their study,
they find that while both genders exhibit a form of projection bias by extrapolating recent trends
into attitudes toward taking investment risks, gender is an important factor in explaining this
attitude. However, their study does not control for other variables, such as income and age which
are known to explain risk-tolerance.
There has also been evidence of gender difference in investments outside the United Sates.
Bhandari and Deaves (2006) use a survey of 1,871 Canadian defined contribution (DC) pension
plan members from 17 pension plans to investigate the demographics of overconfidence. They
find that the average certainty of men is 46% compared to 37% for women and the difference is
statistically significant while there is no appreciable difference between the genders on
investment knowledge; which implies that men are more overconfident than women. They also
find that for both education and income, the relation with overconfidence is monotonic. They
suggest that future research should be directed at understanding why in their study the formally
educated were more at “risk” of being overconfident. They however acknowledge that formal
11
education may not have the same impact as business-related education. Degeorge, Jenter, Moel
and Tufano (2004) analyze the employees’ response to the stock offering during the partial
privatization of France Telecom in 1997. They document that even though they have no clear
hypothesis for why gender should affect the decision to participate in the stock plans, it does
have an effect. Women were about five percent more likely to participate than men. They
suggest household factors and differences in risk aversion may provide possible explanations.
More recently Watson and McNaughton (2007) investigate retirement savings in the Australian
University sector and find that after controlling for age, income and education; women choose
more conservative options than men. However, they suggest the need to include other controls
such as marital status and the number of children in the household. Our set of control variables is
motivated in large part by the findings of these earlier studies.
III. Data, Variable Description and Summary Statistics
In this paper we use data from two sources: the first is the Survey of Consumer Finances and the
second is a survey of professors in the United States. We provide a description of each of these
datasets below.
A. Survey of Consumer Finances
The Survey of Consumer Finances (SCF) is a triennial survey conducted by the Federal Reserve
Board that provides information on the balance sheet, pension, income, and other demographic
characteristics of U.S. families. We use the 2004 survey (SCF04) which is the most current
survey available at the time of our empirical investigation. The SCF04 sample includes 4,519
households, but after eliminating 233 respondents who do not have any financial assets and 70
respondents who only invest indirectly in financial assets, such as via their retirement savings or
12
life insurance, the final sample has 4,216 respondents. Of this 3,361 (79.7%) are male and 855
(20.3%) are female. In Appendix B we provide summary data of the households in the SCF04
sample.
B. Survey of professors
For our empirical study, we manually collected the names and email addresses of Finance and
English professors at universities in the United States that had a website with the email addresses
of their faculty listed. Using a questionnaire, we wanted to collect actual portfolio holdings and
demographic information from each of the individuals selected. Appendix A provides a copy of
our survey questionnaire. The questionnaire was emailed to 4,381 Finance as well as 4,447
English professors. To increase the rate of participation, an email reminder was sent to
individuals who had not responded after one week and a final follow-up email was sent to
subjects who had not responded after the second week.
We received responses from 1,430 of the Finance and 414 English professors4, a
response rate of 33% and 9% respectively. After eliminating respondents who did not answer
questions used to construct our test variable and individuals whose spouses were responsible for
financial decision making within their household, our final sample contains responses from 1,147
Finance and 235 English professors. The number of female professors in the Finance and English
samples are 180 (16%) and 98 (42%) respectively.
C. Control Variables
Since an individual’s affinity for financial risk-taking may be affected by factors other than
gender, we control for these factors in our empirical investigation. Our choice of control
4 We also received email responses from some of the subjects citing reasons why they were unable or unwilling to
participate. The reasons include being ill, being on a sabbatical, being too busy to participate, and their concern for
privacy.
13
variables is motivated by the findings of earlier empirical studies in the risk aversion literature.
The control variables we use are the individual’s age classification (Age), level of education
(Education), marital status5 (Married), race or ethnicity (Race), household income level
(Income), unsecured debt level (Debt) and the number of children under 18 in the household
(Children). The respective categories for each of these variables are shown in Table 1.
D. Summary Statistics
We provide summary statistics for the SCF04 and faculty samples in Table 2. Panel A has the
breakdown of the respondents by gender and each of the demographic factors and in Panel B we
report the sample mean, median and standard deviation for each of the control variables. To test
for differences in mean, we perform Chi-squared tests for the gender, race and marital status
variables since these are nominal and Jonckheere-Terpstra tests for the age, number of children,
income and debt variables since these are ordinal.
The Education variable is only included in the SCF04 sample since all of the respondents
in the faculty sample are within the “college degree” category (Education = 4) of the Federal
Reserve’s education classification. Table 2 shows that more than half of the males in the SCF04
sample have a college degree while just over a third of the females have a college degree.
There is a difference in the breakdown of the Race category due primarily to the design
of each survey. The “Asian/Pacific Islanders” and “Native American, Aleut or Aboriginal
Peoples” are separate categories in the faculty sample but are included in the “Other” category
of the SCF sample. We keep them separate in the faculty sample to highlight the larger
representation of other races in the faculty sample compared to the SCF04 sample. In the SCF04
sample, the total of the Other (Race) category is 4% while the Asian/Pacific Islanders as a
5 Throughout this paper, the marital status, Married is used to indicate individuals who are either married or live
with a partner.
14
separate category accounts for 11% of the faculty data. Notwithstanding the above, even though
the majority of respondents in both samples are Caucasian, there is a greater race/ethnic diversity
in the faculty sample as evidenced by the higher mean, 1.67 versus 1.37 and larger standard
deviation, 1.43 versus 0.90 of the variable Race in Panel B of Table 2.
Similar to the SCF04 sample, the mean age of the professors is 50 years and most of the
households in both samples have only one child under 18. Unsurprisingly, there is a large and
significant difference in the average income across the two datasets. The average income of the
faculty household is 2.42 which is in the $75,000 - $150,000 range, compared to an average of
1.77 which is just under $75,000 for the households in the SCF04 sample. Whereas most of the
respondents in the SCF04 sample are single (mean of 1.32)6, most of the faculty are married or
live with a partner. There is also a difference in the distribution of the marital status by gender.
In the faculty sample, a larger share of both the men and women are married, 85% and 71%
respectively. In the SCF04 sample, a large portion of the women (97%) are married or live with a
partner, but the reverse is observed for the men, where only 16% of the men are married or live
with a partner. The chi-squared test results in Panel B confirm that there is statistically
significant difference between the distributions of the Married variable between the two samples.
To ensure we are obtaining the respondents’ investment behavior and not that of their
spouse/partner, in our survey we ask respondents who are married or live with a partner, the
following two questions:
(a) Do you and your spouse invest jointly or separately?
(b)Who has greater responsibility for making savings and/or investment decisions in
your household?
6 Since the Married variable ranges from 1 (Single) to 2 (married or living with partner), a mean below 1.5 is
consistent with a greater proportion of singles in the sample.
15
In our survey, we do not collect information about the spouses of faculty, hence we exclude
those respondents who invest jointly with their spouses and whose spouses have greater
responsibility for the financial decision making within the household. In our final sample, we
include only those individuals who are either (a) single or (b) married but invest separately or (c)
married and invest jointly but have the major responsibility for financial decision making within
their household.7 In our final faculty sample 1,138 (82%) of the respondents are married or live
with a partner and of this number, the majority (69%) invest jointly with their spouse. Since it is
not possible to uniquely quantify a spouse’s influence on the respondent’s risk aversion8, in our
main empirical investigation we form sub-groups by both gender and marital status.
IV. Test Variable and Preliminary Results
Similar to Cohn, et al. (1975), we define the relative risk aversion index, RRA of the ith investor
as:
i
i
Assets Financial Total Assets FinancialRisky
1−=
i
RRA (1)
Risky Financial Assets are all financial assets other than liquid assets and certificate of deposits
(CDs). By definition, the RRA variable ranges from a minimum of zero (individuals who invest
their entire portfolio in risky assets) to a maximum of one (individuals who invest their entire
portfolio in safe assets). In Table 3 we provide summary statistics for the RRA variable in each of
the sub-categories by gender. We also perform Wilcoxon rank-sum test with the alternative
hypothesis that the RRA of females is greater than that of males in each of these sub-groups and
discuss the findings below.
7 In Section VD we investigate the impact of investing jointly with one’s spouse on the gender-risk aversion relation.
8 Bernasek and Shwiff (2001) suggest that an individual’s risk aversion is affected by the risk aversion of his/her
spouse.
16
Level of Education (Education)
Consistent with the results of Riley and Chow (1992) and Dwyer, et al. (2002), we find a
monotonic inverse relation between risk aversion and the level of education. Further, the median
RRA is one in all of the education groups except for college-educated males. This finding
suggests that individuals who are more educated are less likely to invest their entire portfolio in
safe assets. At all education levels, women are significantly more risk-averse than men.
Age
The SCF04 data indicates that both men and women become less risk-averse as they get older,
the decrease being more for men than women. Cohn, et al. (1975) and Bernasek and Shwiff
(2001) also report a negative relation between age and relative risk aversion. Except for
individuals under 30, women in the SCF04 sample are significantly more risk-averse than men.
The results of the faculty sample are very different from the SCF04 data. For women faculty,
there is a large decrease in risk aversion as the individual’s age increases from the twenties to the
thirties and from the thirties to the forties. However, for the women above forty years old there is
only a minor decrease in RRA. On the other hand, the RRA of the male professors has a U-shape,
falling up to 60 years and then increasing after 60 years. This is similar to the results of Riley and
Chow (1992) and Jianakoplos and Bernasek (1998). Unlike the SCF04 sample, here we find that
female professors are more risk-averse than their male counterparts only when they are in their
thirties and fifties.
Race/Ethnicity (Race)
In the SCF04 sample, women of all races are significantly more risk-averse than their male
counterparts. In the faculty sample, women are significantly more risk-averse than men in all
17
race categories except Black/African-American. In the case of black faculty, we find that even
though quantitatively women are less risk-averse than men, the difference is not statistically
significant. This result is similar to that of Jianakoplos and Bernasek (1998) who find that unlike
white women, black women are less risk-averse than black men.
Married or Living with a Partner (Married)
In the SCF04 sample, married men are more risk-averse than single men whereas married
women are less risk-averse than single women. In the faculty sample, both married men and
married women are less risk-averse than their single counterparts. The faculty results support the
findings of Agnew, et al. (2003) who find that married persons are less risk-averse and Sunden
and Surette (1998) who find that single women are more risk-averse than single men but
counters the findings of Cohn, et al. (1975) who use discriminant analysis and find that married
individuals are more risk-averse than singles. Table 3 also shows that in the SCF04 sample, both
married women and single women are significantly more risk-averse than their male
counterparts. However, in the faculty sample, only married women are significantly more risk-
averse than married men.
Number of Children under 18 in the Household (Children)
Jianakoplos and Bernasek (1998) find that the number of children in the household is positively
related to the risk aversion of single women, negatively related to the risk aversion of couples
and unrelated to the risk aversion of single men. Bernasek and Shwiff (2001) also document a
positive relation between the number of children under 18 in the household and the respondent’s
level of risk aversion. The results in Table 3 show that in the SCF04 sample women are
significantly more risk-averse than men irrespective of the number of children in the household.
18
However, in the faculty sample the gender difference in risk aversion is significant only at the
two extremes of the number of children under 18 in the household, i.e., when they have either
more than two children or none. In the faculty sample, women in households without any
children under 18 are more risk-averse than men in this category, but women in households that
have more than two children under 18 are significantly less risk-averse than their male
counterparts.
Household Income (Income)
Similar to previous studies (see for example Agnew, et al. 2003; Cohn, et al. 1975; Dwyer, et al.
2002; and Riley and Chow 1992), we observe a monotonic negative relation between the
household income and the relative risk aversion. As individuals earn more, there is the tendency
to allocate a larger share of their wealth to risky assets. Table 3 shows that the gender-risk
aversion relation is most significant at the extreme income groups. In the SCF04 sample, women
are significantly more risk-averse than men in the highest and lowest income groups and less
risk-averse than men in the middle income range. In the faculty sample, women are significantly
more risk-averse than men in the extreme income groups but there is no significant gender
difference in risk aversion in the middle income group.
Unsecured Debts (Debt)
Table 3 shows that for both men and women in the SCF04 sample and for the women in the
faculty sample, risk aversion initially increases with increasing levels of debt and then decreases.
For the men in the faculty sample, there is no clear pattern in the debt-risk aversion relation. In
the SCF04 sample, in all but the highest debt level group (Debt>$500,000), women are more
risk-averse than men. In the faculty sample, women are significantly more risk-averse than men
19
in two of the lowest debt level categories (Debt<$100,000). To our knowledge, none of the
previous authors who have investigated the gender-risk aversion relation have included debt as a
control variable.
Summary of Preliminary Results
The consistent finding in our preliminary test is that the persistent gender difference in the
various sub-groups of the SCF04 sample seems to weaken or disappear when the level of
education is controlled for as we do in the faculty sample. In addition, whereas the gender
difference remains among individuals married or living with a partner it seems to disappear
across educated singles.
V. Econometric Analysis
A. The Role of Education
In this section we quantify the impact of formal education on the gender-risk aversion relation. In
Panel A of Figure 1 we compare the kernel density plots of the RRA by gender and marital status
between the SCF04 sample and the faculty sample. The results of the Anderson-Darling test, the
Cramér-von Mises criterion and the Shapiro-Wilk test in Table 4 all confirm that the RRA
variable in each of the sub-categories are not normally distributed. In addition, the results of the
Levene’s test of homogeneity of variance confirms that except for married individuals in the
SCF04 sample, the variance of the RRA variable differs significantly across gender in both
samples. We therefore use two nonparametric (namely, Kolmogorov-Smirnov, and Fligner-
Policello) tests and provide the results of these tests in Panel B of Figure 1. Using Kolmogorov-
Smirnov (K-S) test, we test the null hypothesis that the distribution of the RRA variable is the
20
same across gender within each marital status group. That is, we test the following hypotheses
for each category:
single partner, with livingor married i
)(:
thatealternativ eagainst th
),(:
0
=
≤≠
⎟
⎠
⎞
⎜
⎝
⎛≤
≤=
⎟
⎠
⎞
⎜
⎝
⎛≤
a
i
men
RRAPa
i
women
RRAP
A
H
aallfora
i
men
RRAPa
i
women
RRAPH
(2)
The second test is the Fligner-Policello test 9 that the RRA of females is greater than the RRA for
males within each of the marital status group. The Fligner-Policello test can discern if the median
RRA of women is significantly greater than that of men. We test the following hypothesis:
singlepartner, a with livingor marriedi andRRA median theis where
:
1
thatealternativ eagainst th
:
0
=
>
=
θ
θθ
θθ
i
men
i
women
H
i
men
i
women
H (3)
Panel A of Figure 1 shows that both plots for the SCF04 sample have a greater portion of
the distribution closer to an RRA of 1 while the plots for the faculty have a greater portion of the
distribution in the region corresponding to an RRA less than 0.5. This supports the results of our
preliminary test that the more educated are generally less risk-averse. The results of the
Kolmogorov-Smirnov and the Fligner-Policello tests confirm that within the SCF04 sample, the
distribution of the RRA variable is significantly different between men and women whether they
are single or married. However, in the faculty sample, there is no significant difference between
the distribution of the RRA variable between single men and single women.
We also perform Tobit analysis by marital status for both the SCF04 sample and the
faculty sample. We test if the gender variable is significant in both married and single
respondents after controlling for age, race, number of children in the household, the household
9 See Fligner and Policello (1981) for a detailed description of this test.
21
income and the amount of unsecured debts. The regression model is given by10:
iii
iiiiii
eDebtIncome
ChildrenRaceAgeEducationGenderRRA
+++
+
+
+
+
+=
76
543210
αα
α
α
α
α
α
α
(4)
We use the Tobit analysis since the dependent variable (RRA) is left censored at zero and right
censored at one. For the estimation, we use the Qualitative Limited Independent Model (QLIM)
procedure which uses a non-linear optimization. The regression results provided in Table 5 are
consistent with our preliminary findings. In the SCF04 sample, income, education and age are
highly significant and negatively related to relative risk aversion. This confirms that the young
respondents who are less educated and earn the least, are the most risk-averse. Race is significant
and positive suggesting that Caucasians are less risk-averse. The number of children in the
household is significant and positively related only to the level of risk aversion when the
respondent is married or lives with a partner. This suggests that as the number of children
increases, an individual who is married or lives with a partner is less likely to invest a large share
of his/her portfolio in risky assets. After controlling for all these other factors, both single and
married females are significantly more risk-averse than their male counterparts in the SCF04
sample.
The results in Table 5 show that for the faculty sample, income and age have the most
significant and negative influence on the individual’s relative risk aversion. In addition, race is
positively related to the level of risk aversion suggesting that Caucasians are less risk-averse.
Unlike the SCF04 sample, in the faculty sample the number of children in the household is
negatively and significantly related to the level of risk aversion irrespective of whether the
respondent is single or married. Furthermore, in the faculty sample, when individuals are married
or live with a partner, the level of risk aversion is negatively related to the amount of unsecured
10 As outlined previously, the Education variable is only included in the SCF04 sample.
22
debts. Finally, as reported in Table 5, we find that while there is a gender-difference in risk
aversion among individuals who are married or live with a partner, single women are no more
risk-averse than their male counterparts after controlling for age, income, debts, race and the
number of children in the household.
B. The Role of Finance Knowledge
In this section we investigate if it is education in general or finance knowledge in particular that
is more significant in the gender-risk aversion relation. We separate our faculty sample into its
two sub-groups; namely Finance faculty and English faculty. Table 6 provides summary
statistics for each of the variables used in our analysis for the two sub-samples. Panel A has the
breakdown by gender and in Panel B we provide results of tests that the distribution of each of
the variables is the same across the Finance and English sub-samples. We use Chi-squared tests
for the nominal variables Gender, Married and Race and Jonckheere-Terpstra test for the ordinal
variables Age, Income and Debt.
Table 6 indicates that compared to the English faculty, the Finance faculty sample has a
larger percentage of males and is older. The Finance group also exhibits greater ethnic diversity
and has fewer Caucasians as evidenced by the larger standard deviation and larger mean of the
Race variable. On average, the Finance group has more respondents who are either married or
live with a partner, has households with more children and has a significantly larger mean
income. There is no significant difference in the mean age or the level of unsecured debts
between the respondents in each of the two sub-samples.
We now investigate whether the distribution of the RRA variable differs across gender
within the Finance and English faculty sub-samples. In Panel A of Figure 2 we compare the
kernel density plots of the RRA variable by gender and marital status. In Panel B we provide
23
results of the Kolmogorov-Smirnov test of the null hypothesis that the distribution of the RRA
variable is the same across gender within each marital status group. We also report the results of
the Fligner-Policello test that the median RRA of females is greater than the median RRA of
males within each of the marital status group. Unlike the stark difference observed between the
SCF and the overall faculty samples, the plots for each of the faculty sub-groups are qualitatively
similar. Further, the results in Panel B confirm that except for English faculty who are married or
live with a partner, there is no difference in risk aversion between male and female. In other
words, there is no gender difference in risk aversion among Finance and English faculty who are
single. The results for the Finance faculty also suggest that when individuals have a high level of
education and knowledge of finance, women are just as likely as men to invest a significant
portion of their portfolio in risky securities.
We also perform Tobit analysis of the RRA index on gender and the other control
variables as follows:
iii
iiiii
eDebtIncome
ChildrenRaceAgeGenderRRA
+++
+
+
+
+=
65
43210
αα
α
α
α
α
α
(5)
The results of the Qualitative Limited Independent Model estimation are given in Table 7. For
the Finance faculty, age has the most significant effect on the individual’s level of risk aversion.
In the case of English faculty, while income is significant for both married and single
individuals, it is the most important factor for single respondents. Similar to the Finance group,
age is the most important factor among English professors who are married or live with a
partner. Race and the number of children under 18 in the household are also consistently
significant factors while the level of unsecured debt in the household is consistently insignificant
in all of the gender-marital sub-groups of the faculty sample. The most important finding as
reported in Table 7 is that gender is not significant in any of the sub-groups. This provides
24
evidence that when individuals have the same level of education irrespective of their finance
knowledge, women are no more risk-averse than men.
C. Retirement Savings
In this section, we investigate whether our findings in sub-sections A. and B. for the individuals’
overall portfolio also hold for their retirement savings. As we discuss in our literature review,
most of the studies which focus on the retirement savings find different investment patterns for
men and women. We consider this important since as Sunden and Surette (1998) suggest, if
women are not making optimal investments decisions this could affect their accumulated wealth
for retirement. In addition, in the aggregate this could have severe welfare cost for the society in
general.
For each of our samples, we include only those respondents who have a retirement
account. We use the same definition for RRA as given by expression 1. In the case of the SCF04
sample we consider all non-equity retirement savings to be “non-risky”. For the faculty sample,
we ask respondents to provide the breakdown of their retirement savings based on the TIAA-
CREF®11 funds. The TIAA-CREF has six major categories for investments but we give seven
options in our question. Five of the options are the same as the five major categories of TIAA-
CREF funds, (1) Fixed Income, (2) Guaranteed Income, (3) Money Market Investments, (4)
Multi-Asset Investments and (5) Real Estate Funds. We separate the sixth TIAA-CREF category
of Equities funds into two sub-categories of (1) Equities-Domestic and (2) Equities-International.
We consider Equities-Domestic, Equities-International plus one-half of Multi-Asset Investments
to be "risky". In Table 8 we provide summary statistics by marital status and gender for the
SCF04 sample, the faculty sample and the sub-samples of Finance and English professors.
11 TIAA-CREF is the largest manager of defined contribution retirement plans for employees of educational
organizations.
25
Table 8 also provides results of the Wilcoxon rank-sum test that the mean RRA of women is
greater than that of men in each of the sub-groups.
Table 8 provides preliminary evidence that when we control for the level of education
and knowledge of finance, women are no more risk-averse than men in their retirement savings.
In fact, comparison of faculty who are married or live with a partner shows that whereas the
married female in the Finance sub-sample are more risk-averse than their male counterpart, the
reverse holds for the married female in the English sub-sample. In addition, single women are no
more risk-averse than their male counterparts in any of the faculty samples.12 These findings
suggest that a spouse’s level of risk aversion may impact the individual’s risk aversion.
In Table 9 we provide the results of the Tobit analysis of the model given by expression 4
for the retirement savings for each of our samples and sub-samples. To be consistent with our
earlier analysis, we do the estimation by marital status. Table 9 provides further support for the
need to control for education when investigating the gender-risk aversion relation. After
controlling for age, income, race, number of children in the household and the level of debt, we
find that there is no significant difference in the RRA of single women versus that of single men
in the SCF04 sample. It is important to note that similar to the overall portfolio results, education
is a highly significant (and negative) factor. Income, age and the level of debt are also significant
among single respondents in this sample, while debt and race are the other significant factors for
respondents who are married or live with a partner.
The results by marital status for the faculty sample are revealing. For the overall faculty
sample, age is the most significant factor. However, unlike the overall portfolio, in the case of
the retirement savings, age has a positive impact on the individual’s level of risk aversion. This
12 We are cautious in interpreting the singles results for the SCF04 sample since there are only 15 single female
respondents who have retirement savings.
26
suggests that individuals in the faculty sample tend to be less aggressive in their retirement
savings as they approach retirement. The importance of income also differs across disciplines.
The income level has a significantly negative effect on RRA for the English faculty but its impact
is insignificant for Finance faculty. Finally, whereas there is a gender-difference in risk aversion
among individuals who are married or live with a partner in both the English and Finance
samples; single female faculty are no more risk-averse than their male counterpart in both
disciplines.
D. Other Factors that may affect the level of Risk Aversion
Our findings so far suggest that the gender-risk aversion relation is affected by factors such as
the individual’s level of education, marital status, race and number of children. In their study,
Cohn, et al. (1975) suggest that the individual’s other non-financial holdings may influence his
level of risk aversion. In addition, Riley and Chow (1992) and Jianakoplos and Bernasek (1998)
find evidence of a negative relation between the individual’s wealth and his level of risk
aversion. We therefore hypothesize that investment in other assets such as the value of the
respondent’s home if he/she owns one (House), the value of other tangible assets such as jewelry
(Tangibles), the value of investment in real estate except home (Realestate) and if the respondent
owns a firm (Firm) should have a negative influence on the level of risk aversion of both men
and women. Similarly, proxy measures of job security such as the respondent’s number of years
at current job (JobYears) and if the respondent is a tenured professor (Tenured) should increase
the human capital and therefore should also have a negative impact on the individual’s level of
risk aversion irrespective of gender. However, we are unsure of the differential impact across
gender of these additional variables. We extend the regression model (expression 5) to include
these variables that proxy for the individual’s other assets.
27
Further, motivated by our finding that an individual’s Race influences his level of risk
aversion, we include a variable for the number of years the individual has lived in the United
States (YearsinUS) to test if the Race result is also influenced by an “immigration” factor. In
other words, we test the hypothesis that individuals who have recently migrated to the US are
more conservative in their investments. Finally, in our married sub-samples we include a dummy
variable, Joint (which is “1” if the individual invests jointly with his/her spouse) as a proxy for
the influence of the spouse on the individual’s risk aversion. A breakdown of the choices for
each of these variables is given in Table 10. Furthermore, in Figure 3 we provide a comparison
of the proportion of respondents in each of the faculty sub-samples by gender and sub-groups.
The breakdown in each category is similar in the two sub-samples. The main differences are that
in the Finance faculty there are more immigrants, the respondents have made more investment in
real estate other than their residence and the respondents’ houses are of higher value. Our
extended Tobit regression model is given by:
iiii
iiiii
iiiiiii
TenuredJobYearsFirm
ealestateRTangiblesHousetinJoinUSYears
DebtChildrenRaceIncomeAgeGenderRRA
εααα
ααααα
α
α
α
α
α
α
α
++++
+++++
+
+
+
+
++=
141312
1110987
6543210 (6)
Comparison of the results of this model as reported in Table 11 to the results in Tables 4
and 6 shows that the gender-risk aversion relation is largely unchanged, particularly among
singles, women are no more risk-averse than men when the level of education and the knowledge
of finance is controlled for. The Age variable seems to have been a proxy for the Job security
variables. The significance of Age among finance faculty is now replaced by JobYears for single
Finance faculty and Tenured among married Finance and English faculty. Consistent with our
previous findings; neither Age, JobYears nor Tenured are important factors in explaining the
relative risk aversion of single English faculty. Of the other asset holdings, only the value of the
28
respondent’s house and other tangible assets are significant factors. House is significant among
married faculty members and Tangibles is significant among single English faculty. The findings
for both the job security measures and the other assets confirm our hypothesis that these would
have a negative effect on the individual’s level of risk aversion.
The results for the YearsinUS and Joint variables suggest that these variables are highly
correlated with the race and marital status variable respectively. For the Finance faculty, the
YearsinUS variable is significant but Race is not. The relation between YearsinUS and RRA is
negative unlike the relation between Race and RRA as the majority of individuals who have lived
in the US the longest (YearsinUS =3) are more likely to be Caucasians (Race = 1). The Joint
variable is only weakly significant among married Finance faculty, which now also has a very
weak gender difference in risk aversion, providing further support for our conjecture and that of
Watson and McNaughton (2007) of the need to control for marital status when investigating the
gender-risk aversion relation.
VI. Risk Extremes and Subjective Measures of Riskiness
In this section we investigate if the gender-risk aversion difference is more pronounced for
“extreme” risk classes based on the individual’s perception of an asset’s “riskiness”. Olsen and
Cox (2001) suggest that the difference in risk aversion is most significant at the risk extremes.
We investigate if there is a gender difference in (a) the likelihood to have ever invested in a
group of assets the respondent considers most risky and (b) the likelihood that an individual will
invest the smallest share of his/her portfolio in the asset category that he/she ranks as most risky.
In our survey we ask the following supplemental questions:
How do you rank the following in order of Riskiness? (0=No/Least Risk, 6=Most Risk)
Checking, Savings, CDs, Govt. Bills and Bonds
29
Federal Agency Bonds and Municipal Bonds
Corporate Bonds
Mutual Funds
Stocks
REITs
Derivatives
We compare the ranking of each the choices above with the responses to the following two
questions:
(1) Have you ever invested in any of the following? Choose as many as apply.
(2) Please indicate what percentage of your total wealth is currently invested in each of the
following categories
In Panel A of Table 12 we provide the summary results for the ranking by gender among
the Finance faculty. The three categories that are ranked the highest by both men and women are
Derivatives, REITs and Stocks respectively. We test if women who have ranked derivatives as
the most risky are more likely than their male counterparts to have ever invested in derivatives.
We repeat this analysis for REITs and stocks. The results, provided in Panel B of Table 12, show
that among Finance faculty who rank derivatives as the riskiest; women are less likely than men
to have ever invested in these securities. In the case of REITs and stocks, women who consider
them the most risky are just as likely to invest in them as their male counterparts.
We also test if women who rank an asset class as most risky are more likely to invest the
smallest share of their portfolio in that asset class than men. The results of Panel C in Table 12
shows that when women rank both derivatives and REITs as the riskiest group of assets, they are
more likely to invest the smallest share of their portfolio in these groups than their male
counterparts. Together the results of Panels B and C suggest that after controlling for the level of
30
education and finance knowledge, there seems to be a gender-difference in risk aversion for asset
categories that are perceived as being most risky.
VII. Conclusions
In this paper we use the results of a survey of Finance and English professors to test if there is a
gender difference in risk aversion when the level of education and knowledge of finance are
controlled for. Our sample consists of 1,382 Finance and English professors from universities
across the United States. We compare their actual portfolio allocations to that of respondents in
the Federal Reserve’s Survey of Consumer Finances (SCF). Similar to most prior studies, we
find that within the SCF sample, both the level of education and gender are significant factors in
the propensity to invest in risky assets. However, most of the women in this sample have lower
education than men. Further, within the SCF sample, for those households where the respondent
is married or live with a partner, there is no indication of the gender of the financial decision-
maker.
The results of our survey of professors provide us with a rich dataset which includes
precise information on the gender of the financial decision-maker within each household.
Further, since all the respondents in this sample, have attained at least a college degree, we
implicitly control for the level of education. Results of both parametric and non-parametric tests
confirm that the gender-difference in risk aversion that is pervasive in the SCF sample is
significantly reduced when the level of education is controlled for. In particular, using cross-
sectional Tobit analysis we find that when individuals have the same level of education, after
controlling for age, income, debts, race and the number of children in the household, single
women are no more risk-averse than their male counterparts.
31
We also investigate whether it is education in general or finance knowledge in particular
that plays a greater role in reducing the gender bias in risk aversion by separating our faculty
sample into its two sub-groups of Finance and English professors. We find that when individuals
have the same level of education regardless of their knowledge of finance, there is no gender
difference in risk aversion.
Our results remain robust after including a number of additional controls such as non-
financial assets and when we consider only the individual’s retirement savings. In fact, the
results are stronger for singles when we restrict our analysis to only the retirement assets which
have been the focus of most prior studies. However, when we compare the individual actual
investment choices to their subjective rankings of the various asset classes, we find that there is a
gender difference in risk aversion at the extreme risk classes. Specifically, we find that after
controlling for the level of education and knowledge of finance, women are less likely to invest
in the asset class that they consider most risky. Furthermore, when both men and women invest a
positive amount in the asset class they consider most risky, women are significantly more likely
to invest the smallest share of their portfolio in that asset class.
32
Table 1
Variables used in Empirical Study
Gender 1 = Male Education 1 = No high school diploma/GED
2 = Female 2 = High school diploma or GED
3 = Some college
A
ge 1 = 20-29 4 = College degree
2 = 30-39
3 = 40-49 Household Income 1 = Less than $75,000
4 = 50-59 (Income) 2= $75,000 - $149,999
5 = 60 or above 3 = $150,000 and above
Marital Status (Married) 1 = Single
2 = Married or Living with partner Unsecured Debts 1 = Less than $25,000
(Debt) 2 = $25,000 - $99,999
Number of Children in Household 0 = None 3 = $100,000 - $249,999
(Children) 1 = 14 = $250,000 - $499,999
2 = 25 = $500,000 or more
3 = More than 2
Race/Ethnicity (Race) 1 = Caucasian (non Hispanic)
2 =
A
frican-
A
merican (Non-Hispanic)
3 = Latino or Hispanic
4 =
A
sian/Pacific Islanders
6 = Native American, Aleut or Aboriginal
Peoples
5 = Other
33
Table 2
Summary Statistics of the SCF 2004 Sample and Faculty Survey
Panel A presents summary statistics by gender in each category of our samples. SCF04 is the results of the 2004
Survey of Consumer Finances and Faculty is the result of our survey of professors at universities across the United
States. By definition, the Education variable is not applicable to the faculty sample. Panel B provides Chi-squared
tests of difference in means of the gender, race and marital status variable and Jonckheere-Terpstra tests of
difference in mean age, number of Children, income and debt. *, ** and *** denote significant differences between
the samples means at the 10%, 5% and 1% level respectively.
Panel A
SCF04 (N = 4216) Faculty (N = 1382)
Male (3361) Female (855) Male (1104) Female (278)
Education
No high school diploma/GED 231 7% 113 13%
High school diploma or GED 757 23% 245 29%
Some college 493 15% 201 24%
College degree 1880 56% 296 35%
Age
20-29 274 8% 114 13% 16 1% 5 2%
30-39 511 15% 118 14% 195 18% 95 34%
40-49 775 23% 173 20% 263 24% 77 28%
50-59 827 25% 172 20% 353 32% 82 29%
60 or above 974 29% 278 33% 277 25% 19 7%
Race/Ethnicity (Race)
A
frican-
A
merican (Non-Hispanic) 199 6% 177 21% 19 2% 12 4%
A
sian/Pacific Islanders 126 11% 24 9%
Caucasian (non Hispanic) 2800 83% 612 72% 886 81% 220 79%
Latino or Hispanic 223 7% 45 5% 14 1% 9 3%
Native American,
A
leut or
A
boriginal 61% 1 0%
Other 139 4% 21 2% 49 4% 11 4%
Marital Status (Married)
Single 2838 84% 25 3% 164 15% 80 29%
Married or Living with partner 523 16% 830 97% 940 85% 198 71%
Number of Children Under 18 in Household (Children)
None 1827 54% 550 64% 651 59% 171 62%
1 582 17% 182 21% 175 16% 51 18%
2 578 17% 91 11% 186 17% 40 14%
More than 2 374 11% 32 4% 92 8% 16 6%
Income
Less than $75,000 1475 44% 763 89% 70 6% 36 13%
$75,000 - $149,999 685 20% 45 5% 450 41% 126 45%
$150,000 and above 1201 36% 47 5% 572 52% 116 42%
Unsecured Debts (Debt)
Less than $25,000 1376 41% 557 65% 789 72% 193 69%
$25,000 - $99,999 580 17% 168 20% 141 13% 42 15%
$100,000 - $249,999 682 20% 107 13% 93 8% 31 11%
$250,000 - $499,999 357 11% 13 2% 48 4% 10 4%
$500,000 or more 366 11% 10 1% 29 3% 2 1%
34
Table 2 contd.
Panel B
SCF Faculty Chi-Squared
Test
Z-JT
Test
Mean Median Std Mean Median Std
Gender 1.203 1 0.402 1.201 1 0.401 0.017
Education 3.115 4 1.033
Age 3.498 4 1.302 3.503 4 1.091 1.259
Race*** 1.368 1 0.903 1.640 1 1.398 1064.241
Married*** 1.321 1 0.467 1.823 2 0.381 801.604
Children** 0.884 0 1.255 0.803 0 1.193 2.010
Income*** 1.765 1 0.879 2.425 3 0.632 -24.811
Debt*** 2.172 2 1.331 1.529 1 0.976 17.005
N 4216 1382
35
Table 3
Summary Statistics of RRA Index for SCF 2004 vs. Faculty Survey
This table presents summary statistics for the Relative Risk Aversion (RRA) Index as defined in the text. SCF04 is
the results of the 2004 Survey of Consumer Finances and Faculty is the result of our survey Finance and English
professors at universities across the United States. We report results of the Wilcoxon test that the RRA of women is
greater than that of men in each of the categories that we investigate; *, ** and *** denote significance at the 10%,
5% and 1% level respectively.
SCF04 Faculty
Mean Median Mean Median
Male Female Male Female Male Female Male Female
Education
No high school diploma/GED 0.92 0.96** 1.00 1.00
High school diploma or GED 0.82 0.88*** 1.00 1.00
Some college 0.73 0.83*** 1.00 1.00
College degree 0.52 0.77*** 0.49 1.00
Age
20-29 0.87 0.88 1.00 1.00 0.59 0.71 0.75 0.99
30-39 0.78 0.91*** 1.00 1.00 0.29 0.46*** 0.15 0.36
40-49 0.66 0.86*** 0.84 1.00 0.20 0.24 0.10 0.10
50-59 0.58 0.81*** 0.62 1.00 0.17 0.24** 0.10 0.15
60 or above 0.56 0.80*** 0.58 1.00 0.21 0.21 0.10 0.10
Race
African-American 0.89 0.95*** 1.00 1.00 0.45 0.22 0.20 0.15
Caucasian (non Hispanic) 0.60 0.80*** 0.70 1.00 0.20 0.29*** 0.10 0.15
Latino or Hispanic 0.91 0.99** 1.00 1.00 0.27 0.38** 0.12 0.30
Other 0.74 0.88** 1.00 1.00 0.28 0.45* 0.20 0.50
Married
Single 0.63 0.87*** 0.80 1.00 0.31 0.35 0.15 0.20
Married or Living with partner 0.75 0.84*** 1.00 1.00 0.20 0.31*** 0.10 0.15
Children
None 0.62 0.83*** 0.81 1.00 0.23 0.37*** 0.10 0.20
1 0.72 0.98*** 0.89 1.00 0.19 0.28 0.10 0.10
2 0.63 0.75*** 0.86 1.00 0.19 0.23 0.10 0.10
More than 2 0.67 0.88*** 0.97 1.00 0.19 0.13* 0.10 0.05
Income
Less than $75,000 0.84 0.87*** 1.00 1.00 0.43 0.66*** 0.25 0.90
$75,000 - $149,999 0.68 0.60* 0.86 0.65 0.23 0.30 0.10 0.10
$150,000 and above 0.39 0.53*** 0.24 0.45 0.18 0.24*** 0.10 0.15
Debt
Less than $25,000 0.66 0.85*** 0.95 1.00 0.22 0.32*** 0.10 0.20
$25,000 - $99,999 0.75 0.86*** 1.00 1.00 0.22 0.37 0.10 0.10
$100,000 - $249,999 0.70 0.81*** 0.90 1.00 0.18 0.35** 0.10 0.20
$250,000 - $499,999 0.56 0.74** 0.56 1.00 0.21 0.18 0.15 0.06
$500,000 or more 0.42 0.47 0.29 0.28 0.14 0.05 0.05 0.05
N 3361 855 1104 278
36
Table 4
Tests of Normality and Homogeneity of Variance of the RRA Variable
This table provides the results of three tests of normality of the Relative Risk Aversion (RRA) variable; the Shapiro-
Wilk test, the Cramer-von Mises test and the Anderson-Darling test. We provide results for each gender and marital
status sub-group within the SCF and Faculty sample. We also report results of the Levene test of the homogeneity of
the variance of the RRA variable across gender for each marital status category. Statistics in bold are significant at
the 10% level or better.
SCF04 Faculty
Single Married Single Married
Male Female Male Female Male Female Male Female
Tests of Normality
Shapiro-Wilk (W) - 0.53 0.70 0.58 0.78 0.82 0.74 0.79
Cramer-von Mises (W-Sq) 37.17 1.05 12.94 32.45 2.30 0.94 14.43 2.67
Anderson-Darling (A-Sq) 232.89 5.44 70.87 164.05 14.13 5.53 81.80 15.80
Levene's Test for
Homogeneity of Variance
(F-Value)
12.98 14.40 0.01 25.54
37
Table 5
Tobit Regression of Relative Risk Aversion (RRA) versus gender for the SCF04 Sample
and the Faculty survey
This table provides results of two-limit censored (Tobit) regression analysis for the SCF04 sample and the faculty
sample using the following model:
iii
iiiiii
eDebtIncome
ChildrenRaceAgeEducationGenderRRA
+++
+
+
+
++=
76
543210
αα
α
α
α
α
α
α
RRAi is the relative risk aversion of the ith investor as defined in the text. The Qualitative Limited Independent
Model procedure which uses a non-linear optimization is used for the estimation. The estimation is done by marital
status within each sample. The Education variable is omitted from the faculty regressions since all respondents in
this group are in the same education category. The t-values for each of the estimates are in parenthesis below the
respective estimates. Significant estimates at the 10%, 5% and 1% level are denoted by *, ** and *** respectively.
SCF04 Faculty
Single Married Single Married
Gender 0.314*** 0.083
*
0.005 0.090***
(2.58) (1.80) (0.09) (3.71)
Education -0.136*** -0.170***
(-11.12) (-7.16)
Age -0.048*** -0.055*** -0.076*** -0.046***
(-4.94) (-3.36) (-3.08) (-4.94)
Race 0.091*** 0.086*** 0.037* 0.018***
(7.87) (3.01) (1.77) (2.78)
Children 0.000 0.077*** -0.109*** -0.029***
(-0.02) (2.50) (-2.75) (-3.77)
Income -0.248*** -0.269*** -0.172*** -0.080***
(-17.34) (-7.27) (-4.04) (-5.31)
Debt 0.011 -0.034 0.019 -0.016*
(1.44) (-1.53) (0.47) (-1.79)
Intercept 1.426*** 2.025*** 0.851*** 0.485***
(10.51) (14.74) (5.40) (7.37)
Log Likelihood -1944.000 -898.068 -159.974 -394.401
N 2863 1353 244 1124
38
Table 6
Summary Statistics of the Finance and English faculty Survey
This table presents summary statistics by gender in each category of our faculty sample. Finance and English are the
results of the survey of Finance and English faculty respectively. Panel A provides the breakdown of the respective
categories and Panel B provides the results of Chi-squared tests of difference in means of the Gender, Race and
Marital Status variable and Jonckheere-Terpstra (JT) tests of difference in mean Age Income and Debt across the
samples; *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
Panel A
Finance English
Male Female Male Female
A
ge
20-29 10 1% 21% 64% 3 3%
30-39 167 17% 65 36% 28 20% 30 31%
40-49 238 25% 52 29% 25 18% 25 26%
50-59 322 33% 56 31% 31 23% 26 27%
60 or above 230 24% 53% 47 34% 14 14%
Race
A
frican-
A
merican 12 1% 53% 75% 7 7%
A
sian/Pacific Islanders 121 13% 21 12% 54% 3 3%
Caucasian (non Hispanic) 767 80% 139 78% 119 87% 81 83%
Latino or Hispanic 12 1% 63% 21% 3 3%
Native American 51% 11% 11% 0 0%
Other 46 5% 74% 32% 4 4%
Married
Single 131 14% 53 29% 33 24% 27 28%
Married/Living with partner 836 86% 127 71% 104 76% 71 72%
Children
None 545 56% 100 56% 106 77% 71 72%
1 158 16% 32 18% 17 12% 19 19%
2 174 18% 34 19% 12 9% 6 6%
More than 2 90 9% 14 8% 21% 2 2%
Income
Less than $75,000 34 4% 74% 36 26% 29 30%
$75,000 - $149,999 389 41% 78 43% 61 45% 48 49%
$150,000 and above 533 56% 95 53% 39 29% 21 21%
Debt
Less than $25,000 692 72% 125 69% 97 71% 68 69%
$25,000 - $99,999 118 12% 22 12% 23 17% 20 20%
$100,000 - $249,999 80 8% 23 13% 13 9% 8 8%
$250,000 - $499,999 44 5% 84% 43% 2 2%
$500,000 or more 29 3% 21% 00% 0 0%
39
Table 6 contd.
Panel B
Finance English Chi-
Squared
Test
Z-JT test
Mean Median Std Mean Median Std
Gender*** 1.157 1 0.364 1.417 1 0.494 82.106
Age 3.516 4 1.062 3.438 4 1.223 0.711
Race*** 1.696 1 1.451 1.370 1 1.064 35.442
Married*** 1.840 2 0.367 1.745 2 0.437 12.082
Income*** 2.517 3 0.568 1.979 2 0.732 10.628
Debt 1.548 1 1.014 1.438 1 0.762 0.241
N 1147 235
40
Table 7
Tobit Regression of Relative Risk Aversion (RRA) for Finance and English Faculty
This table provides results of two-limit censored (Tobit) regression analysis for the Finance and English faculty
samples using the following model:
iii
iiiii
eDebtIncome
ChildrenRaceAgeGenderRRA
+++
+
+
+
+=
65
43210
αα
α
α
α
α
α
RRAi is the relative risk aversion of the ith investor as defined in the text. The Qualitative Limited Independent
Model procedure which uses a non-linear optimization is used for the estimation. The estimation is done by marital
status within each sample. The t-values for each of the estimates are in parenthesis below the respective estimates.
Significant estimates at the 10%, 5% and 1% level are denoted by *, ** and *** respectively.
Finance English
Single Married Single Married
Gender -0.012 0.036 -0.095 0.050
(-0.21) (1.41) (-0.68) (0.69)
Age -0.085*** -0.026*** -0.001 -0.122***
(-3.42) (-2.87) (-0.02) (-3.43)
Race 0.035* 0.017*** 0.131* 0.077***
(1.82) (2.88) (1.71) (2.35)
Children -0.073*** -0.014** -0.522* -0.078*
(-2.13) (-1.93) (-1.88) (-1.87)
Income -0.065 -0.025 -0.306*** -0.098*
(-1.43) (-1.58) (-2.37) (-1.75)
Debt -0.004 -0.010 0.005 -0.068
(-0.09) (-1.30) (0.06) (-1.49)
Intercept 0.633*** 0.289*** 1.129*** 0.996***
(3.52) (4.26) (3.39) (4.92)
Log Likelihood -92.638 -200.700 -46.610 -122.870
N 184 950 60 174
41
Table 8
RRA by Marital Status and Gender for Retirement Savings
This table compares the Relative risk Aversion (RRA) by gender and marital status for the respondents’ retirement
savings only. Panel A provides the results for the SCF04 sample and the faculty sample and Panel B provides
similar results for the Finance and English sub-samples. Wilcoxon tests of difference between the median across
gender for each of the marital status sub-groups is also provided. Wilcoxon Z and T-statistics are provided; *, ** and
*** denote significance at the 10%, 5% and 1% level. We report Wilcoxon T- statistics for samples sizes (N) less
than 30.
Panel A
SCF04 Faculty
Single Married Single Married
Male Female Male Female Male Female Male Female
Nobs 2051 15 268 316 138 68 862 173
Mean RRA 0.430 0.486 0.429 0.528 0.229 0.260 0.277 0.280
Median RRA 0.400 0.500 0.400 0.500 0.180 0.215 0.250 0.250
Wilcoxon Z 0.218 -2.912*** 1.037 0.291
Panel B
Finance English
Single Married Single Married
Male Female Male Female Male Female Male Female
Nobs 116 50 786 122 22 18 76 51
Mean RRA 0.217 0.224 0.267 0.285 0.290 0.360 0.375 0.268
Median RRA 0.180 0.200 0.250 0.263 0.175 0.385 0.388 0.200
Wilcoxon Z/T 0.221 1.017 0.210 0.016**
42
Table 9
Tobit model of RRA by Marital Status and Gender for Retirement Savings
This table provides results of two-limit censored (Tobit) regression analysis for the SCF04 sample, the faculty sample and the Finance and English sub-samples
using the following model:
iii
iiiiii
eDebtIncome
ChildrenRaceAgeEducationGenderRRA
+++
+
+
+
+
+
=
76
543210
αα
α
α
α
α
α
α
RRAi is the relative risk aversion of the ith investor as defined in the text. The Qualitative Limited Independent Model procedure which uses a non-linear
optimization is used for the estimation. The estimation is done by marital status within each sample. The Education variable is omitted from the faculty
regressions since all respondents in this group are in the same education category. The t-values for each of the estimates are in parenthesis below the respective
estimates. Significant estimates at the 10%, 5% and 1% level are denoted by *, ** and *** respectively.
Single Married or Living with a Partner
SCF04 All Faculty Finance English SCF04 All Faculty Finance English
Gender 0.047 0.052 0.023 0.125 0.122* 0.038* 0.067*** -0.124**
(0.36) (1.18) (0.49) (1.06) (1.88) (1.62) (2.54) (-2.07)
Education -0.056***
-0.080***
(-3.85) (-2.29)
Age 0.032*** 0.037* 0.031 -0.042 0.037 0.061*** 0.068*** 0.001
(2.76) (1.85) (1.55) (-0.58) (1.49) (6.62) (6.98) (0.04)
Race 0.011 -0.015 -0.018 0.002 0.109*** -0.007 -0.008 0.017
(0.79) (-0.95) (-1.17) (0.04) (2.71) (-1.16) (-1.31) (0.66)
Children 0.007 -0.032 -0.028 0.001 0.022 -0.008 -0.004 -0.022
(0.72) (-1.13) (-1.06) (0.00) (0.52) (-1.02) (-0.52) (-0.63)
Income -0.089*** 0.011 -0.015 0.188* -0.006 -0.009 -0.017 0.095**
(-5.36) (0.34) (-0.40) (1.90) (-0.13) (-0.58) (-1.03) (1.98)
Debt -0.016* 0.022 0.007 0.050 -0.093*** -0.006 -0.005 -0.027
(-1.87) (0.73) (0.22) (0.73) (-3.12) (-0.68) (-0.59) (-0.76)
Intercept 0.649*** -0.018 0.108 -0.194 0.455*** 0.036 -0.005 0.292*
(4.34) (-0.14) (0.75) (-0.56) (2.37) (0.54) (-0.07) (1.74)
Log Likelihood -1689.311 -73.994 -46.374 -21.004 -580.711 -288.969 -223.630 -51.051
N 2066 206 166 40 584 1023 897 126
43
Table 10
Additional Control Variables used in Empirical Study
This table provides a breakdown of the variables used in our extended model for the faculty sample.
Invest Jointly with Spouse 0 = No Years lived in the US 1 = 10 years or more
(Joint) 1 = Yes (YearsinUS) 2 = Less than 10 Years
3 = Since Birth
A
pproximate Value of House 0 = I do not own a home
(House) 1 = Less than $100,000 How long have you been a 1 = Less than 3 years
2 = $100,000 - $249,999 professor/instructor 2 = More than 3 years
3 = $250,000 - $499,999 (JobYears)
4 = $500,000 or more
A
pproximate Value of 1 = Less than $50,000
A
re you a Tenured professor 1 = Less than $25,000
Tangible Assets e.g.,
j
ewelry 2 = $50,000 - $99,999 (Tenured) 2 = $25,000 - $99,999
(Tangibles) 3 = $100,000 or more 3 = $100,000 - $249,999
4 = $250,000 - $499,999
5 = $500,000 or more
A
pproximate Value of Real 0 = Not applicable
Estate other than residence 1 = $100,000 - $249,999
(Realestate) 2 = $250,000 - $499,999
3 = $500,000 or more
Do you own a firm 0 = No
(Firm) 1 = Yes
44
Table 11
Tobit Regression of Relative Risk Aversion with additional variables
This table provides results of two-limit censored (Tobit) regression analysis for the overall faculty sample and the
Finance and English sub-samples by gender and marital status using the following model:
iii
iiiiii
iiiiiii
TenuredJobYears
FirmTangiblesHouseJo
DebtChildrenRaceIncomeAgeGenderRRA
εαα
αααααα
α
α
α
α
α
α
α
+++
++++++
+
+
+
+
++=
1413
121110987
6543210
RealestateintYearsinUS
The Qualitative Limited Independent Model (qlim) procedure which uses a non-linear optimization is used for the
estimation. The t-values for each of the estimates are in parenthesis below the respective estimates. Significant
estimates at the 10%, 5% and 1% level are denoted by *, ** and *** respectively.
Single Married
All faculty Finance English All faculty Finance English
Gender 0.030 0.011 -0.018 0.084*** 0.043* 0.017
(0.52) (0.19) (-0.13) (3.43) (1.68) (0.25)
Age 0.012 0.002 0.192* -0.015 0.005 -0.051
(0.37) (0.06) (1.70) (-1.37) (0.45) (-1.27)
Income -0.131*** -0.045 -0.176 -0.056*** -0.005 -0.010
(-2.81) (-0.92) (-1.19) (-3.51) (-0.27) (-0.16)
Race 0.024 0.012 0.1458* 0.014** 0.012* 0.062**
(1.07) (0.62) (1.71) (2.05) (1.87) (1.97)
Children -0.066* -0.035 -0.408 -0.018** -0.004 -0.033
(-1.64) (-1.05) (-1.42) (-2.29) (-0.58) (-0.78)
Debt 0.034 0.017 -0.017 -0.010 -0.006 -0.049
(0.84) (0.42) (-0.19) (-1.12) (-0.73) (-1.08)
YearsinUS -0.081* -0.135** -0.088 -0.022 -0.037** -0.084
(-1.84) (-3.19) (-0.66) (-1.31) (-2.14) (-1.62)
Joint -0.016 0.035* -0.087
(-0.83) (1.78) (-1.26)
House -0.035 -0.016 -0.126** -0.039*** -0.032*** -0.088***
(-1.59) (-0.75) (-1.96) (-4.09) (-3.57) (-2.48)
Tangibles -0.036 -0.006 -0.180** 0.000 0.002 -0.002
(-1.10) (-0.20) (-2.09) (-0.03) (0.14) (-0.04)
Realestate -0.012 -0.002 -0.088 -0.004 -0.004 -0.025
(-0.35) (-0.06) (-0.74) (-0.42) (-0.44) (-0.67)
Firm -0.022 0.008 0.000 -0.002 0.016 -0.118
(-0.23) (0.10) . (-0.06) (0.65) (-0.73)
JobYears -0.183** -0.173** -0.132 -0.044 -0.043 -0.017
(-2.21) (-2.22) (-0.51) (-1.25) (-1.28) (-0.13)
Tenured -0.022 -0.044 -0.106 -0.039* -0.036* -0.214**
(-0.35) (-0.74) (-0.48) (-1.86) (-1.83) (-2.27)
Intercept 1.124*** 0.988*** 1.243** 0.605*** 0.377*** 1.268***
(5.57) (4.83) (2.14) (6.89) (4.32) (4.41)
Log Likelihood -151.570 -81.964 -41.511 -379.092 -185.606 -113.166
N 242 182 60 1120 947 173
45
Table 12
Comparison of Investment in “Riskiest” Category in Finance sample
Panel A provides summary results of how Finance professors rank various groups of security by gender. Panel B
compares the investment history in the three most risky groups by gender. Panel C provides the proportion of male
and female who invests the smallest share of their portfolio in the group that he/she ranks as most risky. Jonckheere-
Terpstra tests that women are less likely to invest in the “riskiest” group and are more likely to invest the smallest
share of their portfolio in the riskiest group are provided in Panels B and C respectively. Significant estimates at the
10%, 5% and 1% level are denoted by *, ** and *** respectively.
Panel A
Male Female All
MEAN STD MEAN STD MEAN STD
Checking, Savings, CDs, Govt. Bills and
Bonds 0.24 0.60 0.21 0.57 0.24 0.59
Federal Agency Bonds and Municipal Bonds 1.21 0.80 1.27 0.76 1.22 0.80
Corporate Bonds 2.47 0.83 2.62 0.83 2.49 0.83
Mutual Funds 3.19 0.85 3.11 0.89 3.18 0.86
Stocks 4.31 0.88 4.30 0.83 4.31 0.87
REIT's 4.21 1.03 4.48 0.98 4.25 1.03
Derivatives 5.55 1.05 5.68 0.84 5.58 1.02
N 945 201 1146
Panel B
Riskiest Investment History Male Female Z - JT Test
Derivative Invest in Derivatives 31.1% 13.6% -4.714***
Does Not Invest in Derivatives 68.9% 86.4%
REITs Invest in REITs 26.4% 16.7% -1.096
Does Not Invest in REITs 73.6% 83.3%
Stocks Invest In Stocks 97.1% 100.0% 0.641
Does Not invest in Stocks 2.9% 0.0%
Panel C
Riskiest Portfolio Male Female Z - JT Test
Derivative Derivatives is Smallest share of portfolio 92.6% 97.2% 2.229***
Derivatives is Not Smallest share of portfolio 7.4% 2.8%
REITs REITs is Smallest share of portfolio 80.2% 93.3% 1.689**
REITs is Not Smallest share of portfolio 19.8% 6.7%
Stocks Stocks is Smallest share of portfolio 35.6% 35.7% 0.010
Stocks is Not Smallest share of portfolio 64.4% 64.3%
46
Figure 1
RRA by Marital Status and Gender for SCF04 vs. Faculty
In Panel A, we provide kernel density plots by marital status and gender for each of our sample. SCF04 is the results
of the 2004 Survey of Consumer Finances and Faculty is the result of our survey Finance and English professors at
universities across the United States. The top section of each graph gives the distribution for males and the bottom
half gives the distribution for females. In Panel B we provide results of the Kolmogorov-Smirnov (KS) test that the
distribution of the RRA variable is the same across both genders and Fligner-Policello test that the RRA of Females
is greater than that for Males. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
Panel A
Panel B
EDF at maximum Dev at maximum KS_D
Fligner-
Policello
Test
Male Female Male Female
SCF Single 0.466 0.120 0.161 -1.714 0.346** -4.446***
Married 0.403 0.267 1.908 -1.514 0.136*** -4.857***
Faculty Single 0.530 0.450 0.338 -0.484 0.080 -0.928
Married 0.874 0.722 0.812 -1.770 0.152*** -3.630***
SCF04 - Married
SCF04 - Single
Faculty - Single Faculty - Married
47
Figure 2
Distribution of RRA for Finance versus English Faculty
In Panel A, we provide kernel density plots by marital status and gender for each of the faculty sub-samples. The top
section of each graph gives the distribution for Males and the bottom half gives the distribution for Females. In
Panel B we provide results of the Kolmogorov-Smirnov (KS) test that the distribution of the RRA variable is the
same across both genders and Fligner-Policello test that the RRA of Females is greater than that for Males. *, ** and
*** denote significance at the 10%, 5% and 1% level respectively.
Panel A
Panel B
EDF at maximum Dev at maximum KS_D
FLIGNER-
POLICELLO
TEST
Male Female Male Female
Finance Single 0.397 0.321 0.251 -0.395 0.076 -0.757
Married 0.604 0.520 0.322 -0.826 0.084 -1.628
English Single 0.545 0.778 -0.601 0.664 0.232 0.948
Married 0.846 0.634 0.879 -1.063 0.212** -1.765**
Finance - Single Finance - Married
English - Married English - Single
48
Figure 3
Additional Control Variables for Finance and English Faculty by Gender
In this figure, we compare the breakdown by number of years the individual has lived in the United States (YearsinUS), if the individual invests jointly with a
spouse (Joint), the value of the respondent’s home (House), the value of other tangible assets such as jewelry (Tangibles), value of Real Estate except home
(Realestate), if the respondent owns a firm (Firm), number of years at current job (JobYears) and if the respondent is a tenured professor (Tenured)
0 0.2 0.4 0.6 0.8 1
Since birth
Less than 10 years
10 years or more
Jointly
Separately
Less than $100,000
$100,000 - $249,999
$250,000 - $499,999
$500,000 or more
I do not own a home
Less than $50,000
$50,000 - $99,999
$100,000 or more
Not applicable
$100,000 - $249,999
$250,000 - $499,999
$500,000 or more
Yes
No
Less than 3 years
More than 3 years
Yes
No
YearsinUS Joint House Tangibles Realestate Firm JobYears Tenured
Finance Male Finance Female 00.20.40.60.81
Since birth
Less than 10 years
10 years or more
Jointly
Separately
Less than $100,000
$100,000 - $249,999
$250,000 - $499,999
$500,000 or more
I do not own a home
Less than $50,000
$50,000 - $99,999
$100,000 or more
Not applicable
$100,000 - $249,999
$250,000 - $499,999
$500,000 or more
Yes
No
Less than 3 years
More than 3 years
Yes
No
YearsinUS Joint House Tangible s Realestate Firm JobYears Tenure d
English Male English Female
49
Appendix A: Questionnaire
1. Gender:
Male
Female
2. Age:
20-29
30-39
40-49
50-59
60 or above
3. Racial or Ethnic Group(s):
(To which racial or ethnic group(s) do you most identify? Select more than one if applicable)
African-American (Non-Hispanic)
Asian/Pacific Islanders
Caucasian (non Hispanic)
Latino or Hispanic
Native American, Aleut or Aboriginal Peoples
Other
4. For how long have you lived in the United States?
Since birth
Less than 10 years
10 years or more
5. Marital Status
Single
Married or Living with partner
6. Do you and your spouse invest jointly or separately?
Jointly
Separately
7. Who has greater responsibility for making savings and/or investment decisions in your household?
Me
My Spouse
8. How many children under age 18 do you have in your household?
None
1
2
More than 2
9. What is the approximate value of your house?
Less than $100,000
$100,000 - $249,999
$250,000 - $499,999
$500,000 or more
50
10. What is the approximate value of your tangible assets other than your home (e.g., jewelry)?
Less than $50,000
$50,000 - $99,999
$100,000 or more
11. If you own real estate other than your personal residence, what is the combined approximate value?
Not applicable
$100,000 - $249,999
$250,000 - $499,999
$500,000 or more
I do not own a home
12. Do you currently own a firm?
Yes
No
13. What is the approximate value of your outstanding unsecured loans and other liabilities (including
credit cards)?
Less than $25,000
$25,000 - $99,999
$100,000 - $249,999
$250,000 - $499,999
$500,000 or more
14. What is your Current Household Income?
Less than $75,000
$75,000 - $149,999
$150,000 and above
15. How long have you been a Faculty/Instructor?
Less than 3 years
More than 3 years
16. Are you a tenured professor?
Yes
No
17. Which of the following expressions best describes your short-term expectations of the market?
I expect a bull period
I expect a bear period
I expect a normal (or flat) period
18. Which of the following statements best characterize your recent investment experience?
On average, I have experienced a loss on my investments
On average, my gains are just about as much as my losses
On average, I have experienced a gain on my investments.
51
19. Have you ever invested in any of the following? Choose as many as apply.
U.S. Govt. Bills and Bonds
Federal Agency Bonds or Municipal Bonds
U.S. Corp. bonds or bond mutual funds
U.S. High-yield junk bond or bond mutual funds
U.S. Large-cap stocks or stock mutual funds
U.S. Small-cap stocks or stock mutual funds
Foreign stocks/bond or foreign mutual funds
Futures and options
Managed futures or commodity pools
Private Hedge Funds
Privately managed Accounts
REIT’s
None of the Above
20. Please indicate what percentage of your total wealth is currently invested in each of the following
categories (The total must sum to 100):
Checking, Savings, CDs, Govt. Bills and Bonds
Federal Agency Bonds and Municipal Bonds
Corporate Bonds
Mutual Funds
Stocks
REIT’s
Derivatives
21. When was the last time you invested in the stock market?
I currently invest in the stock market
The last time I invested in the stock market was within the last 5 years
The last time I invested in the stock market was more than 5 years ago
I have never invested in the stock market
22. Which of the following statements best characterize your last investment in the stock market?
I made a gain on my investment
I made a loss on my investment
Not Applicable
23. How many stocks do you currently own (excluding mutual funds)?
None
Less than 5
Between 5 and 15
Greater than 15
52
24. Which of the following sectors do you own stocks in? (Choose as many as apply)
Oil & Gas
Basic Materials
Industrials
Consumer Goods
Health Care
Consumer Services
Telecommunications
Utilities
Financials
Technology
None of the Above
25. Do you currently invest in a retirement plan?
Yes
No
Plan
26. When did you start investing in your retirement plan?
Less than 5 years ago
5-20 years ago
More than 20 years ago
27. When do you anticipate withdrawing money from your plan?
Later than 25 years
10-25 years
Within 10 years
28. Please indicate the percentage of your total retirement asset invested in each of the following (Total
must sum to 100):
Equities - Domestic
Equities - International
Real Estate
Fixed Income
Money Market Accounts
Guaranteed Income
Multi-Asset Investments
29. How often do you change the composition of your retirement plan?
Never
Every 5 years or more
Every 2-5 years
At least once every 2 years
30. Excluding my primary residence, my retirement plan represents ___% of my investment holdings.
less than 20 percent
20 percent to 49 percent
50 percent to 79 percent
80 percent or more
53
31. How do you rank the following in order of Riskiness? (0=No/Least Risk, 6=Most Risk)
Checking, Savings, CDs, Govt. Bills and Bonds
Federal Agency Bonds and Municipal Bonds
Corporate Bonds
Mutual Funds
Stocks
REIT’s
Derivatives
54
Appendix B: SCF Data, 2004 - Descriptive Statistics
Income, Equity and Financial Assets are in Thousands, Gender is 1 for male, 2 for female; Age is in Years, Educ is
years of formal education, Children is number of children under 18, Race is a categorical variable from 1 to 5 for
each of the 5 Federal classification, 1 is White
All Households
VARIABLE NMIN MAX MEAN STD
GENDER 4519 1 2 1.22 0.41
AGE 4519 18 95 50.74 15.7
MARRIED 4519 1 2 1.34 0.47
EDUC 4519 117 13.97 2.89
CHILDREN 4519 0 8 0.84 1.16
RACE 4519 1 5 1.41 0.91
INCOME 4519 0105069.9 792.14 3914.71
EQUITY 4519 0202657.4 1891.1 10824.7
FIN 4519 0577788 3382.7 17512.6
Households with no Financial Assets
VARIABLE NMIN MAX MEAN STD
GENDER 233 1 2 1.425 0.495
AGE 233 18 91 42.176 14.403
MARRIED 233 1 2 1.597 0.492
EDUC 233 117 9.936 3.3
CHILDREN 233 0 6 1.116 1.438
RACE 233 1 5 2.039 0.939
INCOME 233 0129.39 17.185 15.112
EQUITY 233 0 0 0 0
FIN 233 0 0 0 0
Households with Financial Assets
VARIABLE NMIN MAX MEAN STD
GENDER 4286 1 2 1.2 0.4
AGE 4286 18 95 51.2 15.63
MARRIED 4286 1 2 1.33 0.47
EDUC 4286 117 14.19 2.69
CHILDREN 4286 0 8 0.83 1.14
RACE 4286 1 5 1.37 0.9
INCOME 4286 0105070 834.3 4015.45
EQUITY 4286 0202657 1994 11105.85
FIN 4286 0577788 3567 17964.16
55
References
Agnew, J., P. Balduzzi, and A. Sunden. 2003. "Portfolio Choice and Trading in a Large 401 (K)
Plan." American Economic Review 93: 193-215.
Bajtelsmit, V. L., A. Bernasek, and N. A. Jianakoplos. 1999. "Gender Differences in Defined
Contribution Pension Decisions." Financial Services Review 8: 1-10.
Barber, B. M., and T. Odean. 2001. "Boys Will Be Boys: Gender, Overconfidence, and Common
Stock Investment." The Quarterly Journal of Economics 116: 261-92.
Bernasek, A., and S. Shwiff. 2001. "Gender, Risk, and Retirement." Journal of Economic Issues
35: 345-56.
Bhandari, G., and R. Deaves. 2006. "The Demographics of Overconfidence." Journal of
Behavioral Finance 7: 5-11.
Byrnes, J. P., D. C. Miller, and W. D. Schafer. 1999. "Gender Differences in Risk Taking: A
Meta-Analysis." Psychological bulletin 125: 367-83.
Cohn, R. A., W. G. Lewellen, R. C. Lease, and G. G. Schlarbaum. 1975. "Individual Investor
Risk Aversion and Investment Portfolio Composition." The Journal of Finance 30: 605-
20.
Degeorge, F., D. Jenter, A. Moel, and P. Tufano. 2004. "Selling Company Shares to Reluctant
Employees: France Telecom's Experience." Journal of Financial Economics 71: 169-202.
Dwyer, P. D., J. H. Gilkeson, and J. A. List. 2002. "Gender Differences in Revealed Risk
Taking: Evidence from Mutual Fund Investors." Economics Letters 76: 151-58.
Felton, J., B. Gibson, and D. M. Sanbonmatsu. 2003. "Preference for Risk in Investing as a
Function of Trait Optimism and Gender." Journal of Behavioral Finance 4: 33-40.
Fligner, Michael A., and George E. Policello II. 1981. "Robust Rank Procedures for the Behrens-
Fisher Problem." Journal of the American Statistical Association 76: 162-68.
Grable, J., R. Lytton, and B. O'Neill. 2004. "Projection Bias and Financial Risk Tolerance."
Journal of Behavioral Finance 5: 142-47.
Jianakoplos, Nancy Ammon, and Alexandra Bernasek. 1998. "Are Women More Risk Averse?"
Economic Inquiry 36: 620-30.
Lindamood, S., S. D. Hanna, and L. Bi. 2007. "Using the Survey of Consumer Finances: Some
Methodological Considerations and Issues." Journal of Consumer Affairs 41: 195-214.
56
Olsen, R. A., and C. M. Cox. 2001. "The Influence of Gender on the Perception and Response to
Investment Risk: The Case of Professional Investors." The Journal of Psychology and
Financial Markets 2: 29-36.
Riley, W. B., and K. V. Chow. 1992. "Asset Allocation and Individual Risk Aversion."
Financial Analysts Journal 48: 32-37.
Roszkowski, M. J., and J. Grable. 2005. "Gender Stereotypes in Advisors' Clinical Judgments of
Financial Risk Tolerance: Objects in the Mirror Are Closer Than They Appear." Journal
of Behavioral Finance 6: 181-91.
Sunden, Annika E., and Brian J. Surette. 1998. "Gender Differences in the Allocation of Assets
in Retirement Savings Plans." American Economic Review 88: 207-11.
Watson, John, and Mark McNaughton. 2007. "Gender Differences in Risk Aversion and
Expected Retirement Benefits." Financial Analysts Journal 63: 52-62.
