Job satisfaction among US Ph.D. graduates: the effects of gender and employment sector

Abstract

In this paper we try to understand the determinants of job satisfaction. The population of US Ph.D. graduates provides a useful homogeneity - same level of education - and an interesting heterogeneity - different career outcomes, academics vs. non academics. Empirically we use the Survey of Doctorate Recipients carried out by the NSF. We estimate models on a sample of 30,000 Ph.D.s in science and engineering. Contrary to all the previous studies we find that females express themselves as less satisfied with their jobs than males. More generally, we find that job satisfaction is explained by different sets of variables respectively for males and females, and for academics and non-academics.

Full text

Page 1

Job satisfaction among US Ph.D. graduates: the
effects of gender and employment sector



Philippe Moguérou1
IREDU, CNRS-Université de Bourgogne (Dijon, France)
and SPRU, University of Sussex (Brighton, UK)


First draft, January 2002



Abstract


In this paper we try to understand the determinants of job satisfaction. The population of Ph.D.
graduates in the United States offered an interesting basis to test new factors that are likely to
influence job satisfaction. Indeed, the Ph.D. group provides a useful homogeneity - same level of
education - and an interesting heterogeneity - different career outcomes amongst them, academic vs.
non academic positions. Empirically we use the Survey of Doctorate Recipients carried out by the
National Science Foundation in 1997. We estimate various models on a sample of 30,000 Ph.D.s in
science and engineering. Contrary to all the previous studies, and more accordingly to expectations,
we find that females express themselves as less satisfied with their jobs than males, other things equal,
at least for those who work in the academic sector. We show that the number of hours worked has a
positive effect on the probability of being satisfied for males and a negative effect for females. The
absolute earnings increase the probability of being satisfied. But when a measure of comparative pay
is included in the models, the coefficient related to the absolute wage is not significant anymore. More
generally, we find that job satisfaction is explained by different sets of variables respectively for males
and females, and for academics and non-academics.
Keywords: job satisfaction, professional labor markets, Ph.D.
JEL classification: J28, J44.



1 IREDU-CNRS, Université de Bourgogne, 9 avenue Alain Savary, B.P. 47870, 21078 Dijon Cedex, France. Tel:
+33-380-395-237, Fax: +33-380-395-479, E-mail: philippe.moguerou@u-bourgogne.fr. Or, until April 2002:
SPRU, University of Sussex, Mantell Building, Falmer, Brighton, BN1 9RF, UK. Tel: +44-1273-87-7283, Fax:
+44-1273-68-5865, E-mail: p.moguerou@sussex.ac.uk.
I would like to thank the participants to the 2001 ZEW conference in Berlin on unemployment, the participants
to the 2001 IZA Summer School in Buch and my colleagues at IREDU and SPRU.

Page 2

2

Introduction
A majority of economists are still reluctant to study job satisfaction. “They view
personal judgments of satisfaction and other subjective opinions as a black box that should be
opened only by psychologists and sociologists” (Lévy-Garboua and Montmarquette 1997
p.1). But, many studies have tried to understand the determinants of job satisfaction following
the seminal papers of Hamermesh (1977), Freeman (1978) and Borjas (1979), and have
considered job satisfaction as an economic variable.2 Since then, economists have been
increasingly interested in the assessment of subjective well-being. Recently, many studies
have considered the general well-being of individuals3 or more specifically the well-being at
work, i.e. job satisfaction.4
Indeed, the satisfaction that workers derive from their jobs may be viewed as an
indication to how they react to general economic conditions. It is a useful summary measure
of numerous job characteristics. And, perhaps more interestingly, job satisfaction does also
affect these general economic outcomes. In that sense, job satisfaction is an economic variable
that is interesting to study. It can predict labour turnover, absenteeism, productivity or
different events affecting the labour force.5 Freeman (1978, p.8) noted that “subjective
variables like job satisfaction [...] contains useful information for predicting and
understanding behavior”.
Specific groups have been studied such as lawyers, nurses or academics.6 In this paper,
we will focus our analysis on individuals with a doctorate in science and engineering who are
employed in the USA. Three main elements have led us to consider the job satisfaction of this
specific population as potentially interesting to study.
Many studies have shown that higher educated workers are less satisfied than lower
educated workers.7 The common interpretation of this fact is that job satisfaction depends on
the gap between outcomes and aspirations, and that aspirations are increased by education. In
considering a unique level of education, we eliminate this type of difficulty. The homogeneity
of the group that results from the study of a unique level of education may reveal other
mechanisms that are not visible when considering different levels of education.
Other studies have shown that females are more satisfied with their jobs than males.8
The high investment in human capital necessary to obtain a Ph.D. may reverse this fact. It
may also have important consequences on the variables that explained job satisfaction.
In the USA, Ph.D.s have traditionally two really different types of careers. About half of
them are employed in the academic sector. But, the business and industrial sector recruit

2 However, and despite the “recent” interest of economists for job satisfaction, this notion remains an important
domain of the psychological and organizational behavior researches. For a general survey of this literature, not
really familiar for economists, see Spector (1997).
3 Clark (1995, 2001), Ng (1996), Kenny (1999), Frey and Stutzer (1999), Blanchflower and Oswald (1997,
2000), Easterlin (2001).
4 For example, among recent papers: Clark (1996, 1997, 1998), Brown and McIntosh (1998), Hamermesh
(1999), Sloane and Williams (2000), Clark et al. (1998), Blanchflower and Oswald (1999), Sousa-Poza and
Sousa-Poza (2000), Sloane and Ward (1999, 2001), Jürges (2001).
5 Hamermesh (1977), Freeman (1978), Akerlof et al. (1988), Tsang et al. (1991), Clark and Oswald (1996),
Lévy-Garboua et al. (1998), Clark et al. (1998), Clark (1999).
6 See respectively: Laband and Lentz (1998), Shields and Ward (2000) and Sloane and Ward (1999, 2001).
7 Clark and Oswald (1996) report that satisfaction levels are strongly declining in the level of education, other
things equal. See, among others, Clark (1995, 1997), Sloane and Williams (1996).
8 See below for more details on this topic.

Page 3

3
nearly the other half of them, and a minority (around 10%) are employed in local government
or federal administrations. In the United States, the private sector has been established as a
major employment sector for Ph.D.s for decades now. Thus, Ph.D. graduates face two
different labour market situations - academic vs. non academic positions. These two groups
may have different behaviours in terms of job satisfaction.
Thus, the homogeneity - same level of education - and the diversity - career outcomes -
of the Ph.D. group provide an interesting basis to analyse the determinants of job satisfaction.
The remaining of this paper is organized as follows.
In section I, we briefly review the job satisfaction at the theoretical level. We present
the data used and we provide some basic facts about job satisfaction for Ph.D.s.
In section II, the empirical modelling of job satisfaction is developed.
In section III, we proceed to a systematic analysis of the results from the models.
Section I. Job satisfaction: some basic facts
Theoretical background
The first model that can explain job satisfaction is related to the standard lifetime utility
microeconomic model.
We can define the utility of an individual from working as:
() h,yuU
=

where the utility increases with y, the income, and decreases with h, the number of hours
worked. In that specification, the absolute level of wage enters directly in the utility function.
Here, we give simply the static version of this model. However, the inter-temporal nature of
utility could be taken into account (Lévy-Garboua and Montmarquette 1997).
However, other phenomena are likely to influence the individual utility. The subjective
nature of income can be integrated in this simple model. Indeed, the level of utility may
depend not only on the absolute level of income but also on a reference income to which
individuals compare their earnings. This relative income captures the effect of “relative
deprivation, envy, jealousy or inequity” (Clark 1995, p.2). In the case of job satisfaction, the
annual salary is an important determinant of job satisfaction. But job satisfaction is also
affected by relative earnings. Individuals compare their earnings with other group of
individuals or have specific expectations concerning their earnings.9 But as mentioned by
Hamermesh (1999), it is not clear by which mechanism changes in earnings affect job
satisfaction.
Thus we rewrite the utility function as:
()
XhyyvU,*,,
=

where y* is the relative earnings of the individual and X a vector of individual and job
characteristics. The utility is expected to decrease with y*.

9 The initial idea was proposed by Hamermesh (1977).

Page 4

4
Presentation of the data
The Survey of Doctorate Recipients carried out by the National Science Foundation is
designed to provide information about individuals with doctoral degrees in science and
engineering fields less than 75 years old. The science and engineering fields include
individuals with doctorate in “hard” sciences but also Ph.D.s in social sciences. The Survey
we use here was carried out in April 1997 on about 35,000 individuals. It contains
information about education, work activities and history, socio-economic background...
We have selected a sample of 29114 individuals from the 1997 SDR, 21358 males and
7756 females, who are employed full-time in April 1997, from a total sample of 35189
individuals.10 Most individuals are employed in the academic sector (cf. table 1).
Table 1. Broad sectors of employment of Ph.D.s (in percentage)

Academic sector
Government
Industry
Male
46.9
11.1
42.0
Female
59.5
10.7
29.8
Total
50.2
11.0
38.8
Source: SDR 1997
Note: our sample.
Job satisfaction is described as a categorical response that underline the feelings of
individuals about their jobs. These feelings are represented by limited number of discrete
choices. In the Survey of Doctorate Recipients, the overall job satisfaction is ranged in four
categories. The exact question asked to the individuals is: “How would you rate your overall
satisfaction with the job you held during the week of April 15, 1997 ?”. Four answers are
possible: very satisfied, somewhat satisfied, somewhat dissatisfied and very dissatisfied.
Table 2. Job satisfaction by gender and sector of employment (in percentage)


All sectors
Male
52.1
37.7
7.8
2.4
Academia
Male
55.2
35.5
7.2
2.1
Government
Male
47.0
40.5
9.5
2.9
Industry
Male
49.9
39.7
7.8
2.6
Fem
48.2
38.8
9.3
3.7
Total
51.0
38.1
8.1
2.7
Fem
46.9
39.5
9.8
3.8
Fem
45.8
39.0
10.7
4.5
Fem
51.7
37.5
7.9
2.9
Very satisfied
Somewhat satisfied
Somewhat dissatisfied
Very dissatisfied
Source: SDR 1997.
Notes: our sample.
Different studies have specifically considered the impact of race or gender on job
satisfaction (Bartel 1981; Clark 1995, 1997; Groot and Brink 1998; Sloane and Williams
2000) and females are found to have higher job satisfaction than males. Contrary to these
earlier studies, in the SDR survey, women are less satisfied than men with their jobs, except
maybe for women working in the business/industry sectors. Ward and Sloane (1999) have
showed for the UK academic profession that there were no significant differences between
males and females regarding job satisfaction. But in all the other studies, women express

10 The inclusion of part-time workers in the models has raised some difficulties (particularly with the variable
umber of hours worked). In the original database, 6.7% of individuals work part-time.

Page 5

5
themselves as more satisfied with their jobs than men. The authors have tried to explain this
surprising job satisfaction11 differential by the different nature of job and personal
characteristics, by the different values or expectations among males and females or by sample
selectivity problems. Here, these basic statistics seem to indicate that the gender differential in
the job satisfaction has the opposite sign as usually.
The age has an impact on job satisfaction for both men and women An increasing
satisfaction profile with age seems to appear with these basic data.
Table 3. Job satisfaction by date of birth and gender (in percentage)
<1935 35-39 40-44
Males
54.90
36.05
6.82
2.22
Females
50.25
39.10
8.15
2.51
45-49 50-54 55-59 >1959
Very satisfied
Somewhat satisfied
Somewhat dissatisfied
Very dissatisfied
68.54
25.76
4.75
0.95
58.94
32.77
6.02
2.28
53.42
36.18
8.04
2.36
49.54
39.40
8.20
2.87
47.97
40.78
8.74
2.52
47.54
41.66
8.21
2.59
Very satisfied
Somewhat satisfied
Somewhat dissatisfied
Very dissatisfied
56.72
32.34
7.96
2.99
58.82
30.81
7.00
3.36
49.64
37.47
9.16
3.73
48.46
39.26
8.78
3.51
48.05
38.82
8.87
4.26
44.51
40.90
10.82
3.78
Source: SDR 1997.
Notes: our sample.
Section II. Empirical models for job satisfaction
To precise these basic evidences, it is necessary to estimate models to take into account
the different factors that are likely to affect job satisfaction, and to assess their relative
importance, other things equal.
Modelling the discrete but ordinal nature of job satisfaction
We will respectively denote the four job satisfaction outcomes (“very satisfied”...”very
dissatisfied”) by yi = 0, yi = 1, yi = 2 and yi = 3 for the individual i. The outcome is discrete but
of ordinal nature. So, we would like to estimate the following model:
iii
uxy
+−=
'*
β

where yi* is the independent unobserved variable, xi the vector of dependent variables, ui the
vector of error terms and β the vector of parameters to estimate.
The observed satisfaction variable yi is related to the latent variable yi* such as:

11 Surprising if we consider that females have a disadvantage positions in the labour market in terms of earnings,
promotion or job security.