Center For Economic Research
Reviewed Paper
Gender Differences in Expected Risk and Earnings:
The Case of UW-River Falls*
Authors:
John R. Walker
Associate Professor
Department of Economics
University of Wisconsin-River Falls
And
Hamid Tabesh
Associate Professor
Department of Economics
University of Wisconsin-River Falls
* Direct all correspondence to John R. Walker or Hamid Tabesh, Department
of Economics, UW-River Falls, 410 S. 3rd Street, River Falls, WI 54022 (email:
john.r.walker@uwrf.edu or hamid.tabesh@uwrf.edu). We thank the students who
participated in the study and Alexander Ansah for excellent research assistance.
Thanks also to Mary Ellen Benedict and Jim Peach who provided very helpful
comments. Any remaining inadequacies are the responsibility of the authors.
Abstract
Objectives. We test the hypothesis that women (due to greater risk aversion)
anticipate a higher wage premium for expected earnings uncertainty than men.
Also examined is the human capital hypothesis that women expect a flatter earnings
profile due to fewer years of planned full-time work.
Methods. Regression analysis is used to estimate expected earnings models
using data measuring the labor market expectations of college seniors and juniors
in business and education.
Results. In business, women anticipate a higher risk premium than men in early
expected salary estimates and a lower one after twenty years. Planned years
worked full-time are negative in expected salary estimates after ten and twenty
years in business and are positive in all pooled regressions in education.
Conclusion. The evidence does not support the hypothesis of greater risk aversion
for women. The higher expected risk premium for women in business may reflect
a lower anticipated wage elasticity of demand for leisure. The results also
do not support the human capital hypothesis that women, who plan fewer years
of full-time work, expect a flatter earnings profile than men.
Gender Differences in Expected Risk and Earnings: The Case of UW-River Falls
Introduction
This study examines the impact of expected risk in the determination of gender
differences in earnings expectations. The analysis employs detailed data that
measure uncertainty about future earnings, expected earnings, labor force plans,
and attitudes towards job characteristics of college seniors and juniors (majoring
in business and education) at the University of Wisconsin-River Falls (
1 ).
Previous studies that examined gender differences in earnings expectations
(Blau and Ferber, 1991; Walker, 1998) did not include a risk measure. In addition,
recent work by McGoldrick (1995)-which found that women received a higher risk
premium for wage uncertainty when compared to men-did not control for the effects
of labor force participation or individual attitudes towards job characteristics.
In the following, we first provide a brief literature review. We then describe
the data, model, descriptive statistics, and present the results. The paper
concludes with a summary of the significant findings and a discussion of the
potential impact of student error in predicting future earnings.
Background
Human capital analysis has ignored the role which wage uncertainty plays in
the determination of occupational and earnings differences among women and
men. This could be an important omission in light of evidence that suggests
men tend towards greater risk taking in labor markets than women (Subich, 1989)
( 2 ). If that is the case, women (even though they are career oriented) may
avoid higher paying jobs with greater wage uncertainty. This could explain
the previous findings of Corcoran et. al (1984) and England (1982) that indicated
women, who worked full time, still tended to be concentrated in lower paid
occupations. Also, more recently, Chauvin and Ash (1994) (in a survey of business
school graduates) found career oriented women selected into occupations that
paid lower salaries with less variable (or contingent) pay than do men.
Gronau (1988) suggested the reliance of empirical studies on outcomes based
data may be the reason why measures of labor force participation do not explain
the lower earnings of women. This is due to the possibility that women may
plan lower levels of labor force participation than they actually work (Shaw
and Shapiro, 1987). Since job selection is based on future plans, women therefore
still choose lower paying jobs compared to men. Blau and Ferber (1991) and
Walker (1998), following up on Gronau's point, tested the human capital model
using detailed data that included labor force plans and earnings expectations
of college seniors. In both studies, labor force plans did not explain the
expected earnings of the students. However, neither study controlled for possible
gender differences in expected risk.
McGoldrick (1995), contributing to a small literature examining the relationship
between earnings and risk, found that women received a significantly greater
risk premium for earnings uncertainty than men (
3 ). McGoldrick's earnings
model, however, did not control for labor force participation. In addition,
McGoldrick's analysis did not include the subjective attitudes of individuals
towards job attributes. Thus, it is not clear if that result reflects a higher
degree of risk aversion among women that contributes to their lower overall
earnings ( 4 ).
In the analysis below we estimate a more complete model that investigates
the impact of labor force plans, attitudes towards jobs, and anticipated earnings
uncertainty on the expected earnings differences between women and men. This
allows us to make stronger inferences regarding whether or not gender differences
in expected returns to uncertainty reflect differing attitudes between women
and men towards risk. By controlling for uncertainty the analysis also provides
a stronger test of the human capital model that emphasizes the role of labor
force plans in the determination of expected earnings differences between women
and men. The specific hypotheses we examine are:
1. Risk averse seniors and juniors require a positive wage premium to enter
occupations in which earnings are uncertain (Johnson, 1977; Feinberg, 1981;
King, 1974; Leigh, 1983).
2. Risk averse seniors and juniors are willing to forgo earnings by selecting
occupations in which there is a small chance of making a high salary (positively
skewed earnings distributions) (King, 1974; McGoldrick, 1995).
3. Due to a greater degree of risk aversion, women anticipate a higher wage
premium than men for a given level of wage uncertainty.
4. Women, who plan labor force withdrawal, expect a lifetime earnings profile
which is flatter compared to men who work full time (Polachek 1976a; 1981b).
Also of interest (given our sample), is the degree to which seniors and juniors
in education anticipate a higher wage premium compared to their counterparts
in business. This would be consistent with the analysis of Bellante and Link
(1981) which suggests that public sector workers are more risk averse than
workers in the private sector. In addition, since our analysis estimates regression
models for expected beginning salaries and expected salaries after ten and
twenty years, we are interested in the degree to which the expected wage premium
for uncertainty declines in the later expected salary estimates. This would
be consistent with the hypothesis of decreasing absolute risk aversion which
suggests a decreasing risk premium for individuals who move into higher income
(wealth) categories (Arrow, 1971).
The Data
The data were collected during Fall semester 1997. The target population was
753 seniors and juniors majoring in Business Administration, Accounting, Agricultural
Business, Elementary and Secondary Education at the University of Wisconsin-River
Falls. Surveys were administered to students in upper division courses and
through their academic advisors. In the end, 570 surveys were returned for
a response rate of 76 percent (
5 ).
Information obtained from the survey included age, gender, cumulative GPA,
major, and class standing (whether a senior or junior) of the students. Labor
force plans were measured by asking students to estimate the number of years
over their working lives they planned to work full-time, part-time while raising
children, part-time for other reasons, or be a full-time homemaker. A seven
point Likert scale was used to obtain student attitudes towards fifteen different
job characteristics (including making a decent salary, earnings prestige, having
pleasant co-workers, or making a contribution to society).
To measure wage expectations, we first asked students to indicate their two
most preferred occupations. Following that, students provided point estimates
(in 1997 dollars) of the expected beginning salaries and expected salaries
after ten and twenty years (for continuous employment) in those occupations.
In addition, students provided high and low values for each salary estimate
to measure their degree of uncertainty regarding those future wages (Kodde,
1986; Manski ,1996) (
6 ).
The Model and Descriptive Statistics
We estimate OLS regression models on expected beginning salaries and expected
salaries after ten and twenty years. The three models are estimated within
business and education on both pooled and separate samples of women and men
in these fields. The basic equation in these estimations is:
(1) _______ Ln Expected Salary = b0 + b1X1 + b2X2 + b3X3 + b4X4 + e
where the dependent variable is the natural log of expected salary in job
of first choice. Among the independent variables, X1 is a vector that includes
labor force plans and other productivity variables. The vector X2 accounts
for the effects of subjective tastes on earnings expectations (Daymont and
Andrisiani, 1984; Filer, 1985). Demographic characteristics such as college
major and gender are included in X3, while X4 measures the impact of risk and
skewness on future wages. Interaction terms are added to the pooled estimates
within business and education to test for gender differences in risk and skewness
(7).
Definitions of the variables (along with their anticipated signs) are presented
in Table 1 (click here to see Table 1). As indicated, we anticipate the coefficient
on uncertainty to be positive and skewness negative for each of the models.
The gender-uncertainty interaction term is expected to be negative reflecting
a higher degree of risk aversion among women. We do not have an a priori hypothesis
regarding the sign of the gender-skewness interaction term. However, we note
McGoldrick (1995) found women gave up significantly greater earnings for a
given amount of positive earnings skewness than men did.
Among the human capital variables, the coefficient on planned years worked
full-time (our measure of labor force plans) should be negative in estimates
of expected beginning salaries and positive after ten and twenty years. This
is consistent with the human capital view that those planning more full time
work choose occupations with steeper expected earnings profiles. Cumulative
GPA should be positive indicating higher expected earnings for those with greater
ability. We anticipate age to be negative as older students would be less inclined
to invest in additional human capital to augment expected earnings.
In terms of job preferences, those variables associated with an ambitious
attitude toward the labor market (i.e., decent salary, opportunity to advance,
earnings prestige) should be positive in each of the estimates. Those variables
associated with compensating differentials (i.e., close to family, pleasant
work environment, flexible hours) should be negative in the regression models.
The mean values of the dependent and independent variables (by gender) within
business and education are presented in Table 2. Also indicated are t-tests
of the difference between means of women and men in those fields. The descriptive
data suggest that men tend to take on greater earnings uncertainty than women.
As indicated, within education, men anticipate significantly greater uncertainty
than women in the three salary estimates. In business, men anticipate significantly
greater uncertainty in expected salary estimates after twenty years. Men in
business also expect significantly greater skewness in expected earnings after
ten and twenty years than women in that field (
8 ).
The earnings expectations and labor force plans of these students are consistent
with human capital assumptions (
9 ). Women in both fields expect flatter earnings
profiles than their male counterparts. However, the expected earnings gap is
wider after twenty years in business (80 percent) than education (92 percent).
At the same time, women in both business and education plan significantly lower
levels of full time work compared to men in those fields.
________________________ [click here to see Table 2 ]
The significant gender differences in job preferences are (for the most part)
consistent with a traditional view regarding the attitudes of women and men
towards jobs. For example, women in business place greater emphasis on the
job attributes close to family, pleasant work environment, and contribution
to society. Men in both business and education place greater emphasis on earnings
prestige. However, contrary to this view, women in business place significantly
greater emphasis on intellectual challenge than men (
10 ).
Between the fields, differences with regard to decent salary, opportunity
for advancement, pleasant work environment, contribution to society, earnings
prestige, and job security are consistent with assumed priority differences
between business and education majors. However, women and men in business place
greater emphasis on jobs that are not too demanding and have flexible work
hours ( 11 ). This may reflect less anticipated job flexibility by these seniors
and juniors compared to their counterparts in education.
The Regression Results
The OLS estimates of expected earnings within business and education are presented,
respectively, in Tables 3 and 4. To simplify these tables, we exclude the results
for job preferences, major, and class standing. The estimates of these coefficients
(for business and education) are reported separately in Tables A1 and A2 in
Appendix A.
The results (in Tables 3 and 4) support the risk and skewness hypotheses.
As indicated, the risk and skewness coefficients are statistically significant,
with the anticipated signs, in all estimations within business and education.
[click here to see Table 3]
[ Table 4 ]
The models provide mixed support for the hypothesis that women are more risk
averse than men. The gender-uncertainty interaction terms do not suggest statistically
significant differences in expected risk between women and men in the education
field ( 12 ). Within business, the separate regressions for expected beginning
salaries indicate women anticipate a 1.2 percent increase in expected salary
(per thousand dollar increase in risk) compared to .89 percent for men. After
ten years the risk premium is .66 percent for women compared to .3 percent
for men. That women anticipate higher risk premiums in these estimates is consistent
with the hypothesis of greater risk aversion for women (
13 ). However, after
twenty years, the results suggest men expect a significantly greater risk premium
(.55 percent compared to .36 percent for women).
The regression results fail to support the human capital hypothesis that those
planning more full-time work expect steeper earnings profiles. Each of the
total estimates for education majors indicate a positive (and statistically
significant) effect for planned years worked full time. This suggests students
in education (who plan more full time work) expect higher earnings over their
entire lifecycle. Within business, planned years worked full time is not significant
in any of the estimates of expected beginning salaries. Moreover, in the estimates
of expected salaries after ten and twenty years, the coefficient is negative
suggesting those who plan fewer years of full time work expect higher earnings
( 14 ).
In terms of the other risk hypotheses, the results indicate that male education
majors (compared to male business majors) expect higher risk premiums in all
the expected salary estimates. In addition, higher risk premiums are evident
in women's estimates (and the total) after ten and twenty years in education
compared to business. These results lend support to the analysis of Bellante
and Link (1981) suggesting public sector workers are more risk averse than
those in the private sector (
15 ). The tendency of the risk premium to decline
in the later expected salary estimates within business (particularly for women)
is consistent with the hypothesis of decreasing absolute risk aversion. Within
education, focusing on the total estimates, the expected risk premium is relatively
constant in each of the expected salary regressions.
Among other results reported in Tables 3 and 4, the positive effects of age
(contrary to our hypothesis) in expected beginning salaries within business
and after ten and twenty years in education may reflect anticipated productivity
effects due to maturation (Lazear, 1976). The results in business which indicate
that after ten years women anticipate giving up a greater percentage of expected
earnings per thousand dollar increase in positive skewness (.7 percent compared
to .25 percent for men) is consistent with the findings of McGoldrick (1995).
However, in estimates of expected salaries after twenty years, men in business
anticipate giving up a greater percentage of expected earnings (.48 percent)
than women (.3 percent). We also note the significance of the gender coefficient
in business indicating (all else equal) men anticipate higher earnings than
women in each of the total expected salary estimates.
That gender differences in returns to expected risk changes in the expected
salary estimates after twenty years within business is puzzling. However, this
may reflect the fact that planned years worked full time does not capture the
timing of labor force withdrawal. Thus, employing the logic of De Meza (1984),
women in business may expect a higher wage premium in beginning salaries and
after ten years (due to a lower wage elasticity of demand for leisure) because
they plan to have children during this earlier stage of their career. However,
later in their careers, women in business may plan to become more involved
in the labor market (reflected in a higher wage elasticity of demand for leisure)
which results in a significantly lower risk premium compared to men in the
estimates of expected salary after twenty years.
McGoldrick and Robst (1996) provide a complementary explanation with their
finding that workers with greater job mobility (who may engage in job shopping)
receive lower returns to risk. Thus, it might also be that women in business
(who may plan families early in their careers) expect less labor market mobility
than men resulting in a higher risk premium in the early salary estimates.
However, later in their careers these women may anticipate greater mobility
(as they devote more time to the labor market) and expect a lower risk premium.
Also puzzling is the negative effect of planned years worked full-time in
expected salary estimates after ten and twenty years within business. These
results, however, may reflect England's (1982) point that women (or men) who
plan labor market withdrawal may not confine themselves to flat earnings profiles
in order to maximize income over their life-cycle. A woman (or man) who expects
to withdraw from the labor market could also plan to re-enter on a career path
that provides a steeper earnings profile. Indeed, the size and statistical
significance of the cumulative GPA variable (after ten and twenty years) within
business suggests students in this field are highly confident (all else equal)
that the market will reward their ability.
Summary and Conclusion
This study has analyzed gender differences in expected risk and earnings using
a sample of college seniors and juniors majoring in business and education
at the University of Wisconsin-River Falls. Results of this study indicate
women and men in both fields expect a positive wage premium to compensate for
earnings uncertainty. At the same time, women and men in both fields anticipate
lower earnings because of plans to enter occupations with positively skewed
earnings distributions.
Results do not support the hypothesis that women in the sample are more risk
averse than men. Earnings estimations within business suggest women expect
a higher risk premium than men in estimates of expected beginning salaries
and after ten years. However, in estimates of expected salaries after twenty
years, men anticipate a significantly greater risk premium than women. This
is consistent with an increase in women's wage elasticity for leisure demand
due to reduced household responsibilities. It is also consistent with the hypothesis
that women may expect greater job mobility as they devote more attention to
the labor market later in their careers.
Results also suggest education majors may be more risk averse than business
majors due to significantly larger uncertainty coefficients especially in expected
salary estimates after ten and twenty years. In addition, that the risk coefficients
in business (particularly for women) tend to decline in the later salary estimates
suggests the influence of decreasing absolute risk aversion.
Regression results do not support the human capital hypothesis of gender differences
in expected earnings. Results in business suggest women and men planning fewer
years of full time work expect significantly higher salaries after ten and
twenty years. The positive and significant effects of planned years worked
full time in the total expected salary estimates within education is not consistent
with the human capital hypothesis that those planning to work more expect steeper
earnings profiles.
The results presented above should be qualified in light of research by Betts
(1996) indicating students may not be fully informed about wages in the labor
market. Since our sample includes seniors and juniors (who are estimating salaries
within their fields), these students most likely have a good knowledge of pay
rates in the labor market. However, the degree to which there is error (based
on lack of student knowledge) suggests our uncertainty variable could overstate
the impact of risk on earnings. Also, Zarkin (1985) suggested that women (planning
less labor market participation) may tend towards greater error as they have
less incentive to gain information about wages. Future studies, using expectations
data, might address this issue by focusing on fields that have data available
on both actual salaries and labor force participation. This would allow for
a comparison with expected salaries and labor force plans to estimate error
and incorporate its effects into the analysis.
Return to home page
|
