Sung, J. & Hanna, S. (1996). Factors Related to Risk Tolerance, Financial Counseling and Planning, 7, 11-20.
Copyright 1996 by AFCPE
Factors Related To Risk ToleranceJaimie Sung,(1) The Ohio State University
Sherman Hanna,(2) The Ohio State University
Effects of financial and demographic variables on risk tolerance were estimated for households with
an employed respondent in the 1992 Survey of Consumer Finances. Logistic regression analysis
showed that female headed households were less likely to be risk tolerant than otherwise similar
households with a male head or a married couple. Differences in risk tolerance by gender/marital
status, ethnic group and education could be due to differences in understanding of the nature of risk.
KEY WORDS: risk tolerance, individual investors, Survey of Consumer Finances
Risk tolerance plays an important role in each household’s optimal portfolio decisions. It may also be
an important factor in determining many government policies related to consumer risks regarding
financial decisions. An investor’s ability to handle risks may be related to individual characteristics
such as age, time horizon, liquidity needs, portfolio size, income, investment knowledge, and attitude
toward price fluctuations (Fredman, 1996). It has been widely perceived that, for financial planners,
it is essential to make an effort to determine every investor’s risk tolerance level using a subjective
measure (Mittra, 1995). However, there may be objective as well as subjective aspects of risk tolerance.
Malkiel (1996, p. 401) stated that “The risks you can afford to take depend on your total financial
situation, including the types and sources of your income exclusive of investment income.” In their
study, Hanna and Chen used an expected utility and simulation approach to derive optimal portfolios,
based on risk aversion and the ratio of a household’s financial investment portfolio to total wealth,
including human wealth. Hanna and Chen (1995) demonstrated that the ratio of financial assets to total
wealth (including human wealth) was an important determining what level of volatility was optimal for
a portfolio, and that ratio would tend to be related to such objective factors as years until retirement.
Based on plausible assumptions about risk aversion and the actual distribution of the ratio of financial
assets to total wealth in the United States (Lee & Hanna, 1995b) Hanna and Chen concluded that it
would be rational for most households to have only stocks in portfolios intended for long run goals such
as retirement. For younger workers investing for retirement, willingness to accept some risk (volatility)
would lead to substantially greater wealth at retirement (Chen & Hanna, 1996).
The purpose of this paper is to investigate effects of financial variables and individual characteristics
on risk tolerance, with the most recent appropriate dataset, the 1992 Survey of Consumer Finances
(SCF.) Because retired households face very different portfolio issues from those who are not retired,
only working respondents aged between 16 and 70 were included in the analysis. The results have
implications for financial counselors and financial planners in providing portfolio advice to their clients.
A number of articles have analyzed factors related to risk tolerance. Using the 1983 Survey of
Consumer Finances, Hawley and Fujii (1993) employed ordered logit models to investigate effects of
net worth and individual characteristics on risk tolerance. The study included economically active
respondents aged 25-62. Education, income and debt were positively related to risk tolerance. Married
couples and households headed by a single male were more risk tolerant than otherwise similar
households headed by a single female. Age was not statistically significant in the analysis. The Hawley
and Fujii (1993) are consistent with results from Warner and Cramer (1995) and Lee and Hanna
(1995a). Using 1983 SCF data on risk tolerance, Lee and Hanna (1995a) derived the distribution of
dichotomous risk tolerance level by demographic groups. Of 2,691 respondents in the sample, 60%
were willing to take financial risks. Predicted risk tolerance was approximately the same for all ages
under 55, then decreased with age. Predicted risk tolerance increased with education.
Using the 1983 SCF risk tolerance data, Sung and Hanna (1996) employed an ordered probit model of
a 3-level dependent variable to analyze effects of income and demographic variables on risk tolerance.
They found that income and education were positively related to risk The general pattern from the
dummy variables for age was that risk tolerance decreased with age after 45. Self-employed and farmers
were significantly likely willing to take financial risks than their counterparts.
This article is different from previous studies on risk tolerance in use of the 1992 Survey of Consumer,
and in specifying factors that should be logically related to the household’s ability to tolerate risk, such
as years to retirement. This article is also unique in extensive use of graphs to illustrate patterns of risk
Hanna and Chen (1995) used an expected utility approach to demonstrate that it is optimal for almost
all households with an investment horizon of at least five years to invest in stocks, despite higher
volatility. Hanna and Chen assumed that the expected utility of a household is based on the total wealth
of the household, including human wealth. For young households, the investment portfolio represents
such a small proportion of total wealth that even those who are very risk averse should invest in the
asset category with the highest expected return, small stocks. As households approach, human wealth
typically decreases and financial wealth typically increases. The investment horizon becomes shorter,
ultimately becoming less than a year for households depending on investment income for ordinary
living expenses. Therefore, the number of years until expected retirement should be related to a
household’s risk tolerance. Having short-term goals should also be related to risk tolerance. For
instance, if a household has not yet accumulated its desired level of emergency funds, its investment
horizon may be very short, in that it cannot tolerate much volatility in investments until emergency
funds have been accumulated. Therefore, the premise in this article is that only factors related to having
important short-term goals should be important in not being willing to take some risk in obtaining a
higher return on financial investments. Households who do not have adequate financial assets to cover
emergencies or perhaps even normal month-to-month transactions may not be in a position to invest
in stocks or other risky assets.a Those who might have other short-term goals, such as saving for a down
payment for a home, also might not be in a position to invest in risky assets. We assume that for those
with long investment horizons being unwilling to take some chances to obtain a higher return on
investments indicates a lack of information, as it is not rational to be unwilling to take some risk for
Data and Sample
The dataset for this study is the 1992 Survey of Consumer Finances (SCF). The Survey is sponsored
by the Federal Reserve Board in cooperation with the Department of the Treasury and conducted by the
National Opinion Research Center at the University of Chicago. The SCF was primarily designed as
an instrument for the study of assets and liabilities.
In this study, respondents who were working, were aged between 16 and 70, and had positive non-investment income were included, resulting in a sample of 2,659 respondents.
The 1992 SCF had a question on financial risk tolerance. The possible responses and the distribution
of responses for the present sample are shown in Table 1. Although it would seem reasonable to
analyze the distribution of all four response levels to the risk tolerance question, the substantial risk
category is so small that meaningful analysis of it is not appropriate for multivariate analysis with many
variables such as education, race, age and income. By combining the substantial and above average
risk categories, multivariate analysis may be appropriate, as 18% of the respondents in this sample were
in the combined category. Appendix Table 2 presents mean income and the distribution of the risk
tolerance categories of no risk, substantial and above average categories, for demographic variables.
There is no consistent pattern of risk tolerance categories. For instance, there was little difference
between the proportion of married couples and male-headed households in the average risk category,
yet male-headed households were much more likely to be in the above average/substantial risk
tolerance category than were married couples. Therefore, two levels of risk — no risk and
average/above average/substantial risk tolerance (referred to simply as risk tolerant) are used in the
dependent variable. In the sample, 60% of households were risk tolerant and 40% were not risk
The independent variables are non-investment income and dummy variables representing whether liquid
assets were equal to at least 3 months of non-investment income, whether non-liquid assets were equal
to at least 6 months of non-investment income, household size, age, number of years until expected
retirement, education, race/ ethnicity, occupation, self-employed, marital status and gender. Definition
of variables, their measurement, and their sample statistics are shown in Appendix Table 1.
Distribution of Answers to Risk Tolerance Question In Sample From 1992 Survey of Consumer
When you save or make investment, would you take substantial financial risks expecting to earn substantial returns. 3.7%
When you save or make investment, would you take above-average financial risks expecting to earn above-average returns. 14.6%
When you save or make investment, would you take average financial risks expecting to earn average returns. 42.1%
When you save or make investment, would you take no financial risks. 39.6%
*Sample used consisted of respondents who were working, age between 16 and 70, and had positive non-investment income, n= 2659
For descriptive purposes, the results were weighted to reflect the general population of households.b
Chi square statistics were calculated to test for significant bivariate risk tolerance differences in sets of
variables. A logit model was employed in order to identify effects of variables on risk tolerance.c
Table 2 shows differences in risk tolerance for the independent variables. All variables except age and
years to retirement were significantly related to risk tolerance. Logit results are shown in Table 3. Most
of the sets of independent variables had significant effects, except for household size, occupation, and
homeownership status. The patterns with some significant differences in the logit are illustrated in
Figures 1 through 7 as the predicted graphs, and for comparison, the bivariate patterns are shown as the
actual graphs. The actual patterns may be useful to someone interested in inferring risk tolerance from
one characteristic, such as marital status. The predicted patterns provide insight into the effect of a
variable after controlling for the effects of other variables.
Effect of Non-investment Income on Risk Tolerance.
Predicted based on Table3 (other variables at mean values.) Actual based on logit of risk tolerance on log of non-investment income only.
Effects of Non-investment Income The level of non-investment income had a positive effect on risk
tolerance. Figure 1 shows the effect of income on predicted risk tolerance, calculated with other
variables at their mean values.d The predicted probability of being risk tolerant increased with non-investment income, reaching 60% at a level of $50,000. For comparison, a logistic regression of risk
tolerance as a function of only non-investment income was run, and the “actual” risk tolerance by
income was calculated and shown in Figure 1.
Effects of Liquid and Non-liquid Financial Assets. Households with liquid assets greater than or equal
to 3 months of non-investment income had a predicted risk tolerance of 70%, compared to a predicted
level of 58% for otherwise similar households who did not meet the 3 month guideline. Households
with non- liquid assets greater than or equal to 6 months of non-investment income had a predicted risk
tolerance of 73%, compared to a predicted level of 58% for otherwise similar households who did.
Bivariate (Actual) Risk Tolerance Patterns
|Liquid assets3 months income||77.2||61.0***|
|Liquid assets<3 months income||57.1|
|Non-liquid financial assets6 months income||74.1||96.2***|
|Non-liquid financial assets<6 months income||54.0|
|Years until expected retirement||6.6|
|Retire in 0-9 years||52.6|
|Retire in 10-19 years||59.5|
|Retire in 20-29 years||61.9|
|Retire in 30 and over||61.1|
|Ages less than 25||57.4|
|Age 55 and over||57.3|
|Less than high school||32.7|
|High school graduate||52.1|
|More than college||76.3|
|Race or Ethnicity (%)||80.5***|
|Household size (%)||17.4**|
|Size 5 and more||51.3|
|Marital status (“||64.6***|
|Renting or other alternatives to owning||54.9|
|Own without a mortgage||53.8|
|Own with a mortgage||66.2|
Source: The Survey of Consumer Finances, 1992. (N=2,659) All estimates are weighted. * p<.05 ** p<.01 *** for p<.001.
Effects of the Number of Years to Retirement Figure 4 shows differences in risk tolerance among categories of number
of year until expected retirement.e Those who were 30 years or more away from retirement had significantly higher risk
tolerance than otherwise similar respondents whose expected retirement was closer.
Race/Ethnic Group Non-Hispanic whites had higher predicted risk tolerance than Hispanics or those in other racial/ethnic
groups other than Blacks (Figure 6). (The predicted level for non-Hispanic Blacks is lower than the level for whites, but
the difference is not significant at the 0.05 level.) Non-Hispanic Blacks had the lowest actual risk tolerance level (Figure
6), but the predicted level (holding other variables at the sample mean values) was second only to non-Hispanic whites.
Effect of Having Liquid Assets Equal To At Least 3 Months on Risk Tolerance.
Predicted based on Table 3 (other variables at mean values.) Actual based on Table 2.
Effect of Having Non-liquid Assets Equal To At Least 6 Months on Risk Tolerance.
Predicted based on Table 3 (other variables at mean values.) Actual based on Table 2.>