You are on page 1of 4

Physical appearance and earnings:

further evidence
MICHAEL T. FRENCH
University of Miami (D93), 1801 NW 9th Avenue, Third Floor, Miami, Florida 33136,
USA. E-mail: mfrench@miami.edu
The literature contains numerous studies on earnings diVerentials based on age, race,
and gender. Comparatively few studies have examined diVerences in labour market
success related to physical appearance. Using three waves of data collected at two
organizations, this paper tested for earnings diVerentials among workers based on
their self-reported appearance. Signicant earnings premiums for attractiveness were
found for women, but not for men.
I. INTRODUCTION
Discrimination in the labour market can take many forms.
Researchers have long explored the relationships between
gender, race, physical disabilities, and rate of pay (e.g.
Taubman, 1975; Luft, 1975; Cain, 1986; Blau and Beller,
1992; Blau and Kahn, 1997). While earnings diVerentials
based on physical appearance may occur in many
companies and across occupations, data limitations often
preclude a careful statistical analysis. Economists have
begun to explore relationships between the physical
characteristics of employees and labour market success.
However, numerous questions remain and no study has
analysed these relationships with a measure of self-assessed
physical appearance.
II . BACKGROUND
Frieze et al. (1991) considered the relationship between
facial attractiveness and income. Using longitudinal data
on 737 MBA graduates, they discovered that attractive
males were able to secure higher starting salaries and the
earnings diVerentials remained over time. The most
attractive female graduates did not earn higher starting
salaries, but they did earn more income later in their
careers.
A recent article by Hamermesh and Biddle (1994)
examined diVerentials in earnings among men and
women who were rated on their attractiveness by inter-
viewers from several diVerent surveys. The primary
nding of this ambitious study was that plain people
generally earned less than average-looking people (i.e. 5
10% diVerential) and average-looking people earned less
than good-looking people (i.e. slightly less than the plain-
ness penalty).
Biddle and Hamermesh (1998) extended their earlier
study by analysing longitudinal data on a large sample of
graduates from one law school. Earnings were self-reported
by graduates on follow-up questionnaires from the school
and a measure of physical appearance was developed by
rating matriculation photographs . Results showed that
physical appearance was positively related to earnings for
attorneys who graduated in the 1970s and the diVerential
grew with experience. In addition, they found that
attorneys in the private sector had higher beauty ratings
than attorneys in the pubic sector.
The empirical literature on `self-assessed appearance
and labour market measures is scant, with no apparent
research with regards to earnings per se. However, a
theoretical framework on this issue was oVered by
Jackson (1992). In her model, both the sociobiological
(reproductive potential) and sociocultural (cultural values)
perspectives predict that physical attractiveness has greater
implications for females than for males. The empirical
literature supports these perspectives as women feel that
their body appearance is more important than men do,
women engage in more appearance-enhancing behaviour
than men, and females are more dissatised with their
appearance during the life cycle (Jackson, 1992).
Applied Economics ISSN 00036846 print/ISSN 14664283 online # 2002 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/0003684001002756 8
Applied Economics, 2002, 34, 569572
569
II I. METHODS
Following the Hamermesh and Biddle convention of using
average appearance as the index category, the present
analysis determined whether a plainness penalty existed
for below average appearance (i.e. lower wages and
income) and whether a beauty premium existed for above
average appearance (i.e. higher wages and income). For
comparison purposes, the data exclusion restrictions were
dened in the same manner as in Hamermesh and Biddle
(1994). Furthermore, the response categories for the
physical appearance question were almost identical to the
ones used in their study.
1
The basic empirical specication took the following
form:
E
i

0

j
X
ij

s
S
i

y
Y
i

u
U
i

a
A
i
"
i
1
where E
i
is a measure of earnings (i.e. natural log of hourly
wage and natural log of annual personal income), for
worker i, X represents demographic, job-related, and per-
sonal characteristics that inuence wages, S is an
indicator variable for one of the two worksites, Y is an
indicator variable for survey year, U is an indicator
variable for an unattractive individual, A is an indicator
variable for an attractive individual, the s are coecients
to estimate, and " is a random error term. Gender-specic
regression models of wages and income were estimated
with ordinary least squares (OLS).
IV. SAMPLE AND DATA
The data for this analysis were collected in three annual
and independent cross-sections beginning in 1995.
Approximatel y 600 employees were randomly selected
each year (1995, 1996, and 1997) from two worksites in a
Midwestern community to complete a condential
questionnaire on workplace policies, personal
characteristics, job performance and earnings, and their
use of alcohol and illicit drugs. The questionnaires were
administered by a survey research professional that was
unaliated with either worksite. All respondents were
guaranteed anonymity. The survey took about 30 minutes
to complete. Participation was voluntary, each respondent
was given a small gift certicate, and they were eligible for
a US$500 cash lottery. The overall response rate was
greater than 80% at both worksites.
Worksite 1 is a non-prot hospital with approximatel y
2600 full- and part-time workers. Some employees work at
various satellite clinics, but most employees are located at
the main hospital. Worksite 2 is a relatively large school
district with about 3800 full- and part-time employees.
School district employees are scattered around the city
and county in various schools, administrative facilities,
and maintenance facilities. The school district employs a
wide range of workers, including teachers, administrators,
bus drivers, and clerical staV.
To allow comparison with the ndings in Hamermesh
and Biddle (1994), the same sample restrictions were
imposed. For example, 1702 observations had complete
data. The sample was then narrowed to those individuals
between the ages of 18 and 64, with an hourly wage
above US$2, and a self-reported health status of `fair
570 M. T. French
1
The following question from the survey instrument was used to construct a measure of self-assessed physical appearance: `How would
you rate your own physical appearance, compared to other people who are the same age and sex? Please consider your overall
appearance, including your height, weight, and body type, as well as your facial features. Response categories include: (1) very
handsome or beautiful; (2) above average; (3) average; (4) below average; (5) very unattractive.
Table 1. Variable means by gender
Females Males
Variable (n 1310) (n 382)
Demographics
Non-white
a
0.1344 0.2147
Hispanic 0.0244 0.0432
Married 0.7120 0.7454
Education
a
14.8169 15.6623
Enrolled in school 0.1658 0.1658
Occupations
Manager/Professional
a
0.5331 0.6111
Technical/Administrative support
a
0.2334 0.0847
Service 0.0948 0.1138
Transportation 0.0316 0.0714
Other specied occupations
b
0.1071 0.1190
Survey measures
1995 survey
b
0.2087 0.2120
1996 survey 0.4151 0.3979
1997 survey 0.3762 0.3901
Worksite 1
a
0.6047 0.3168
Workplace measures
Experience
a;c
19.2456 20.7513
Experience squared
a;c
472.9893 509.2487
Working full-time
a
0.7565 0.9526
Ln(hourly wage)
a
2.6213 2.8860
Ln(annual personal income)
a
9.9879 10.4812
Physical appearance
Above average
a
0.3339 0.4654
Average
a;b
0.6109 0.4947
Below average 0.0553 0.0399
Note:
a
Statistically signicant diVerences in variable means by
gender, p 40:05 (KruskalWallis rank-order test).
b
Excluded category in the regression analyses.
or higher.
2
A full justication of these restrictions can be
found in Hamermesh and Biddle (1994) . The nal sample
size for the present study was 1692 observations, includ-
ing 1310 females and 382 males.
Table 1 reports variable means for demographics, by gen-
der. Approximately 33.4% (46.5) of the women (men) rated
their appearance as `above average, 61.1% (49.5) rated
themselves as `average, and 5.5% (4.0) considered their
appearance `below average.
3
The majority of the sample
was female, white, and married, with an average age of 40
for women and 42 for men. Twelve variables in Table 1 had
statistically signicant mean diVerences across gender,
which further supports gender-speci c regression models.
V. RESULTS
Table 2 presents the coecient estimates for all variables
in Equation 1. Looking rst at the results for females,
the coecient estimates for above average appearance
were positive, signicant, and remarkably similar. The
quantitative interpretation is that females with above
average appearance had about 8% higher earnings
relative to females with average appearance. The
coecient estimates for below average appearance were
small and insignicant for females. Although none of the
appearance coecients in the earnings regressions were
signicant for males, the smaller sample size for men
reduced the power to nd signicant coecient estimates
(Kraemer and Thiemann, 1987).
The coecient estimates for other variables in
Table 2 generally displayed predictable signs and mag-
nitude. Earnings was a concave function of experience
(age-education-6) and education was positive and sig-
nicant. Married individuals had higher income and
full-time workers earned higher wages than part-time
employees.
Physical appearance and earnings 571
Table 2. Regression statistics and coecient estimates
Females Males
Ln(personal Ln(personal
Independent variable Ln(wage) income) Ln(wage) income)
Below average appearance 0.0268 0.0296 0.0616 0.1450
Above average appearance 0.0800** 0.0784* 0.0327 0.0321
Nonwhite 0.0021 0.0160 0.0194 0.0298
Hispanic 0.0407 0.0211 0.4057** 0.1945
Married 0.0480 0.0127 0.1858** 0.1746**
Education 0.0687** 0.0889** 0.0824** 0.0818**
Enrolled in school 0.0719** 0.0446 0.1101* 0.0055
Manager/Professional 0.4861** 0.4715** 0.0907 0.1280
Technical/Admin. support 0.0922* 0.1315 0.0636 0.1549
Service 0.1229** 0.0537 0.2195** 0.1420
Transportation 0.1573* 0.3352** 0.1391 0.3791**
Experience 0.0197** 0.0326** 0.0238** 0.0638**
Experience squared 0.0003** 0.0005** 0.0004 0.0012**
Working full-time 0.1192** 0.6470** 0.3190** 0.0518
1996 survey 0.0644* 0.0072 0.1072 -0.0407
1997 survey 0.0229 0.0287 0.0070 0.0348
Worksite 1 0.0406 0.3231** 0.0252 0.2215**
Constant 0.8788** 7.2753** 0.9075** 8.2394**
* Signicantly diVerent from zero, p 40:10:
** Signicantly diVerent from zero, p 40:05:
2
Hamermesh and Biddle (1994) chose a wage cut-oV at US$1 per hour for the databases that they analysed. Since their data were
collected in the early 1970s, it was felt that a cut-oV at US$2 per hour was more appropriate for the current study. However, this decision
had little eVect on the results.
3
Compared to the three surveys analysed by Hamermesh and Biddle (1994), a slightly greater percentage of individuals in the worksite-
based survey rated their appearance as `above average and a slightly lower percentage rated their appearance as `below average. These
statistics suggest the possibility of a sample selection problem if `attractiveness was a criterion for employment at these organizations. If
attractive candidates had a higher probability of receiving employment oVers relative to less attractive candidates, then the worksites
would be disproportionately represented by above average looking workers. It was not possible to control for this type of sample
selection with a sample of employed individuals, but the likely impact was to lower the odds of detecting signicant diVerences in
earnings across the appearance categories.
VI. CONCLUSION
The main purpose of this paper was to determine whether
the ndings from earlier studies on physical appearance
and earnings could be replicated using a worksite-based
dataset and a self-assessed measure of physical appearance.
The results of these exercises suggests that wage diVeren-
tials were present in the bivariate analysis for attractive and
unattractive employees relative to their average-looking
counterparts. However, the regression analysis yielded sig-
nicant earnings diVerentials for attractive women only.
The regression results were consistent with those in earlier
studies for women (e.g. Frieze et al. 1991; Hamermesh and
Biddle, 1994; Biddle and Hamermesh, 1998), suggesting
that appearance was important at these two worksites,
but the eVects were clearly gender-specic.
It should be noted that characteristics of the research
design played an important role in the empirical results
and comparability with other studies. First, the present
study used a self-reported measure of physical appearance
whereas other studies analysed data with interviewer-
assessed measures or independent ratings of photographs.
It is unclear whether individuals can answer in a consistent
and objective manner when rating their own appearance,
whether the relationships were causal, or whether self-
assessed ratings were better or worse than interviewer-
assessed ratings. Alternatively, a self-reported question
may be more representative of a persons self-esteem rather
than their physical appearance. While it is dicult to
obtain a completely objective measure of physical
appearance, the self-assessed measure used in this study
oVers a new perspective relative to previous research.
Second, worksite-specic data provided a diVerent,
albeit narrow, angle for analysing appearance and earn-
ings, which may limit the comparability of these ndings
to those in other studies. In addition, it could be argued
that earnings diVerentials on the basis of appearance may
be less likely among a technically skilled workforce in a
hospital and a school district relative to a service oriented
workforce, like in a department store or a bank.
Nevertheless, this is an empirical issue that requires
additional research.
In summary, the ndings reported in this paper were not
necessarily intended to corroborat e or refute the careful
and important studies on physical appearance and earnings
(e.g. Register and Williams, 1990; Frieze et al. 1991; Loh,
1993; Hamermesh and Biddle, 1994; Averett and Koreman,
1996; Biddle and Hamermesh, 1998). Rather, the intent
was to oVer additional information on this little-studied
topic through analysis of a diVerent and more recent data-
base. While the weight of statistical evidence in the litera-
ture strongly suggests that wage diVerentials are present for
certain groups of employees (e.g. Taubman, 1975; Cain,
1986; Blau and Kahn, 1997) beauty premiums and plain-
ness penalties require additional empirical examination.
ACKNOWLEDGEMENTS
Financial assistance for this research was provided by the
National Institute on Alcohol Abuse and Alcoholism
(grant No. AA10318). The author is thankful for adminis-
trative support from Carmen Martinez and Silvana Zavala.
In addition, helpful suggestions from Jeremy Bray, Laura
Dunlap, Robert Martin, Kerry Anne McGeary, Jeanne
Salome , Sara Solnick, Gary Zarkin, and seminar
participants at the University of Miami are gratefully
appreciated.
REFERENCES
Averett, S. and Korenman, S. (1996) The economic reality of The
Beauty Myth, The Journal of Human Resources, XXXI, 304
30.
Biddle, J. E. and Hamermesh, D. S. (1998) Beauty, productivity,
and discrimination: lawyers looks and lucre, Journal of
Labour Economics, 16, 172201.
Blau, F. D. and Beller, A. (1992) Black-White earnings over the
1970s and 1980s: gender diVerences and trends, Review of
Economics and Statistics, 74, 27686.
Blau, F. D. and Kahn, L. M. (1997) Swimming upstream: trends
in the gender wage diVerential in the 1980s, Journal of Labour
Economics, 15, 142.
Cain, G. (1986) The economic analysis of labour market discrimi-
nation: a survey, in Handbook of Labour Economics (Eds)
O. Ashenfelter and R. Layard, Amsterdam, North
Holland, pp. 693785.
Frieze, I. H., Olson, J. E. and Russell, J. (1991) Atractiveness and
income for men and women in management, Journal of
Applied Social Psychology, 21, 103957.
Hamermesh, D. S. and Biddle, J. E. (1994) Beauty and the labour
market, American Economic Review, 84, 117494.
Jackson, L. A. (1992) Physical Appearance and Gender:
Sociobiological and Sociocultural Perspectives, State
University of New York Press, Albany, NY.
Kraemer, H. C. and Thiemann, S. (1987) How Many Subjects?:
Statistical Power Analysis in Research, Sage Publications,
Newbury Park, CA.
Loh, E. S. (1993) The economic eVects of physical appearance,
Social Science Quarterly, 74, 42038.
Luft, H. S. (1975) The impact of poor health on earnings, Review
of Economics and Statistics, 57, 4357.
Register, C. A. and Williams, D. R. (1990) Wage eVects of obesity
among young workers, Social Science Quarterly, 71, 13041.
Taubman, P. (1975) Sources of Inequality in Earnings, North
Holland, Amsterdam.
572 M. T. French

You might also like