29
Revista de Econom?a Aplicada N?mero 61 (vol. XXI), 2013, p?gs. 29 a 55EA
JOB SATISFACTION
OF UNIVERSITY GRADUATES*
ALEKSANDER KUCEL
Escola Universit?ria del Maresme (UPF and Tecnocampus)
MONTSERRAT VILALTA-BUF?
Universitat de Barcelona and CREB
This paper investigates the determinants of job satisfaction of university
graduates in Spain. We base our analysis on Locke?s discrepancy theory
[Locke (1969)] and decompose subjective evaluation of job characteris-
tics into surplus and deficit levels. We also study the importance of over-
education and over-skilling on job satisfaction. We use REFLEX data, a
survey of university graduates. We conclude that job satisfaction is most-
ly determined by the subjective evaluation of intrinsic job characteristics,
with an asymmetric impact of surpluses and deficits. Over-skilling is
much more important than over-education in explaining the job satisfac-
tion of university graduates, although the latter is also significant.
Key words: job satisfaction, discrepancy theory, over-education, over-
skilling.
JEL Classification: J28.
T his paper investigates the determinants of the job satisfaction of Spanishuniversity graduates. We base our analysis on Locke?s discrepancy theory[Locke (1969)] by which individuals? job satisfaction is the result of theirsubjective evaluation of the existing discrepancy between what they wantfrom their job and what they perceive they get from the job. We comple-
ment it with the basic assumption of prospect theory whereby economic agents
value losses more than gains [Kahneman and Tversky (1979)].
We bring two important contributions to the literature on job satisfaction.
Firstly, we decompose job characteristics into surplus and deficit levels in the
worker?s perception. We demonstrate that deficits and surpluses of certain desired
job characteristics (such as work autonomy or learning opportunities) have an
asymmetric impact on job satisfaction. Moreover, we show that omitting this dis-
tinction between deficits and surpluses might prevent the researcher from identi-
(*) Montserrat Vilalta-Buf? is grateful to the Spanish Ministry of Science and Innovation for finan-
cial support through the grant ECO2012-34046 and to the Government of Catalonia through the
Barcelona GSE Research Network and grant 2009SGR1051. The authors also wish to thank the
REFLEX project for granting access to their data.
fying the effect of some job characteristics on job satisfaction, as is the case for
challenging jobs and time for leisure in this paper.
Secondly, we contribute to the debate on job satisfaction of over-educated
workers. While some research shows that having more education than required in
the job makes people unsatisfied [Green and McIntosh (2007); Maynard, Joseph,
and Maynard (2006); Pollmann-Schult and Buchel (2004); Vaisey (2006); Vila,
Garcia-Aracil, and Mora (2007)], there is evidence showing no relationship be-
tween being over-educated and the level of job satisfaction [Allen and van der
Velden (2001); Glenn and Weaver (1982)]. Our findings support the former view
that over-education in the workplace diminishes the job satisfaction of workers.
We demonstrate, however, that having more skills than necessary to perform the
current job has a much stronger negative effect than over-education.
The paper is organized as follows. Section 1 presents the theoretical back-
ground necessary to understand our empirical analysis. Section 2 describes the
data and econometric framework employed in this paper. In Section 3, we present
the descriptive statistics relevant to our study. Econometric results are presented
in Section 4 and Section 5 concludes the paper.
1. THEORETICAL BACKGROUND
Job satisfaction has long been considered an important economic [Freeman
(1978)], sociological [Kalleberg (1977)] and psychological variable [Locke
(1969)]. Since the seventies, the literature on job satisfaction has grown to a con-
siderable size. One can distinguish three major streams of research on job satisfac-
tion. Firstly, gender differences in job satisfaction have been thoroughly investigat-
ed. The main finding is that women are usually more satisfied with their jobs than
men, chiefly due to their lower expectations [Bender, Donohue, and Heywood (2005);
Clark (1997); Hanson and Sloane (1992); Mora and Ferrer-i-Carbonell (2009); Sloane
and Williams (2000); Sousa-Poza and Sousa-Poza (2000); Sousa-Poza and Sousa-
Poza (2003); Sousa-Poza and Sousa-Poza (2007); Usui (2008)].
A second line of research investigates the relationship between job satisfaction
and wages [Allen and van der Velden (2001); Bowles, Gintis and Osborne (2001);
Clark (1996); Clark (1999); Clark and Oswald (1996); Gazioglu and Tansel (2006);
Lydon and Chevalier (2002); Sloane and Williams (2000)]. In most of these stud-
ies, measures of relative wage, changes in income and future wage expectations
tend to affect job satisfaction more than the level of pay itself.
The third aspect of studies on job satisfaction is education. The effects of edu-
cation on job satisfaction tend to be positive when job characteristics are not con-
trolled for [see Glenn and Weaver (1982)], while they become negative when job
characteristics are included in the analysis [see Clark (1996); Clark and Oswald
(1996); Clark, Oswald, and Warr (1996)]. As Fabra and Camis?n (2009) point out,
the positive effect is most likely due to better quality jobs achieved by higher edu-
cated individuals, which make them more satisfied. When controlling for job char-
acteristics, the effect of the level of education on job satisfaction becomes negative,
probably due to the higher expectations of better educated individuals.
Revista de Econom?a Aplicada
30
We analyze the determinants of job satisfaction of Spanish university gradu-
ates. Our empirical strategy is based on Locke?s discrepancy theory [Locke (1969)].
Locke distinguishes three elements affecting job satisfaction: 1) the worker?s per-
ception of some aspect of the job (not the objective description of it), 2) an im-
plicit or explicit value standard (what one wants from the job), and 3) a conscious
or subconscious judgment of the relationship between one?s perception(s) and
one?s value(s) (perceived discrepancy). According to Locke, job satisfaction and
dissatisfaction are a function of the perceived discrepancy between what one
wants from the job (value standard), and what one perceives one gets from the
job. Moreover, he admits the possibility of different functional forms for this rela-
tionship depending on the job characteristic. He discusses examples of two func-
tional forms (Graphic 1). For some job characteristics (e.g. hours of work), the in-
dividual prefers some optimal amount and any deviations from this optimal
amount (due to excess or deficit) diminish job satisfaction. This creates a bell-
shaped function between discrepancies in these factors and job satisfaction (Panel
a, Graphic 1). For some other job characteristics (e.g. pay) the function is found to
be linear (Panel b, Graphic 1). Individuals compute their ideal pay (value) based on
what similar others earn and the amount that would fulfill their economic wants.
In this case, though, discrepancies in deficit decrease their job satisfaction, but
discrepancies in excess make them more satisfied with their job.
Job satisfaction of university graduates
31
Graphic 1: TWO POSSIBLE FUNCTIONAL FORMS OF THE RELATIONSHIP BETWEEN JOB
SATISFACTION AND DISCREPANCY IN PERCEIVED JOB CHARACTERISTIC AND
THE IDEAL VALUE OF THIS JOB CHARACTERISTIC
Source: Own elaboration.
It is therefore important to recognize the possibility of a bell-shaped function-
al form between perceived discrepancy and job satisfaction for the empirical analy-
sis. One has to distinguish between excess and deficit of a job characteristic to be
able to identify any effect of this job characteristic on job satisfaction. The standard
assumption of a linear relationship would mistakenly show no relationship when
the underlying functional form is bell-shaped. The empirical strategy we follow is,
thus, to distinguish between excess and deficit discrepancy in order to capture non-
linear functional forms in the influence of job characteristics on job satisfaction.
Another central element in the theory of Locke is the intensity of values. The
effect of a discrepancy on one?s job satisfaction depends on the importance of the
job characteristic for the individual. For instance, the effect of a pay increase on
job satisfaction will be larger for someone who puts more value on money than
for someone who considers it less important. Therefore, ?satisfaction with some
object or situation is a function not only of the amount of discrepancy between
percept and value, but of the importance of that value to the individual? [Locke
(1969), pag. 330]. The importance of a value is reflected in the slope of the func-
tional form. The more important the value for the individual, the more it affects
his/her satisfaction level and, therefore, the steeper the function. We expect to ob-
tain a larger coefficient of one job characteristic for an individual who reports
high importance of this job characteristic than for an individual who does not at-
tach such importance to the job characteristic. Although Locke?s theory has found
large empirical support in the psychological literature [see Jackson and Corr
(2002); Mobley and Locke (1970); Rice, Gentile, and McFarlin (1991)], it has not
received much attention in the fields of economics and sociology except for
Kalleberg (1977). If different functional forms and different intensities in the rela-
tionship between job characteristics and job satisfaction are common, then the
consequent model misspecification might pose serious limitations to some of the
existing literature on job satisfaction.
We complement discrepancy theory with the basic assumption of prospect
theory whereby people value losses more than gains [see Kahneman and Tversky
(1979)]. Therefore, we consider the possibility of different intensities in the rela-
tionship between job satisfaction and perceived discrepancy depending on the lat-
ter being deficit or surplus. If the assumption holds true in our context, we should
obtain a stronger effect on job satisfaction for deficit discrepancies than for excess
discrepancies. The REFLEX survey data provide information on the individual
subjective evaluation of the following job characteristics: work autonomy, job se-
curity, learning opportunities, high earnings, new challenges, good career
prospects, enough time for leisure, social status, useful for society and time for
family. We introduce them all decomposing deficits and surpluses in our analysis.
Finally, in addition to analyzing the effects of perceived discrepancies in the
job characteristics mentioned above, we also introduce educational mismatches as
potential explanatory variables of job satisfaction.
Based on the previous theoretical discussion, we derive the following hy-
potheses:
Hypothesis 1: A perceived discrepancy in any job characteristic better explains
the level of job satisfaction than the objective measure of the job characteristic.
Most papers on job satisfaction use objective/subjective job characteristics
indistinctively depending on data availability. Locke?s theory argues that job satis-
faction depends on the worker?s perception of some aspect of the job relative to
Revista de Econom?a Aplicada
32
the value standard (perceived discrepancy). We want to test whether objective
measures of job characteristics matter when subjective measures are also present
in the equation. Only a few studies include both measures to check which one is
relevant in explaining job satisfaction, and, when they do, they focus on a single
job characteristic. Origo and Pagani (2009) study the effect of perceived and actu-
al job stability on job satisfaction across several European countries. Their results
show that perceived job stability, rather than actual stability, matters for job satis-
faction. Sloane and Williams (2000) study the gender gap as well as the effect of
objective and subjective comparative pay measures on job satisfaction. Once again,
the subjective measure turns out to be significant. We want to investigate whether
subjective or objective evaluations of wage, job stability, hours of work and job
training are relevant for job satisfaction. We expect objective measures to become
insignificant once we introduce their subjective counterparts.
Hypothesis 2: The relationship between perceived discrepancies and job sat-
isfaction depends on two aspects: whether discrepancies are in surplus or deficit
and on the importance that the individual gives to the job characteristic.
The two aspects should be taken into account. First, the relationship between
perceived discrepancies and job satisfaction may be bell-shaped or linear1. There-
fore, we distinguish between discrepancies in deficit and excess to allow for other
than linear functional forms. Second, the importance an individual gives to each
aspect of the job matters for the intensity of the relationship. We allow for differ-
ent effects of each discrepancy depending on the level of importance given to the
aspect of the job. This distinction, furthermore, permits us to test whether deficit
discrepancies have a stronger effect than excess discrepancies on job satisfaction,
as the assumption of prospect theory would predict. As far as we are aware, this has
not been analyzed so far in the literature on job satisfaction. If the asymmetric valu-
ation of gains and losses in job satisfaction is confirmed, optimal policies to im-
prove workers? well-being will be different from those under a symmetric valuation.
Hypothesis 3: Being over-educated and over-skilled reduces the job satisfac-
tion of individuals.
While some research shows that having more education than required in the
job makes people unsatisfied [Green and McIntosh (2007); Maynard et al. (2006);
Vaisey (2006)], there is evidence showing no relationship between being over-ed-
ucated and the level of job satisfaction [see Glenn and Weaver (1982)]. Many
show that it is skill mismatch and not over-education that matters for job satisfac-
tion [see Allen and van der Velden (2001); McGuinness and Sloane (2011)]. We
contribute to this debate by introducing over-education and over-skilling into the
equation on job satisfaction.
Job satisfaction of university graduates
33
(1) The bell-shaped and linear functional forms are just two possibilities. One could imagine other
functional forms that are also allowed with our estimation strategy.
2. DATA AND METHODOLOGY
We use REFLEX survey data2. It consists of information on a representative
group of individuals who graduated in 1999/2000 and were interviewed five years
later in 2005. The survey covers several European countries and Japan. In this
paper, we analyze the Spanish case.
The REFLEX survey provides information on the individuals graduate stud-
ies, personal background, first job and current job. We restrict our sample to indi-
viduals who are currently working at least 20 hours per week and are not self-em-
ployed3. Our final sample consists of 2,216 individuals.
Our dependent variable is the level of job satisfaction. Respondents were
asked how satisfied they are with their current job on a 5-point Likert scale (1-
very dissatisfied, 5-very satisfied). We construct a dichotomous variable, which
takes value 1 if the respondent replied 4 or 5, and zero otherwise4. We use the lo-
gistic model throughout the analysis.
Respondents were asked to indicate how important the following job charac-
teristics were to them personally, and to what extent they actually applied to their
current job situation. In both cases, they could choose from among five categories
between 1 (not at all) and 5 (to a very high extent). The job characteristics under
evaluation were the following: work autonomy, job security, opportunity to learn
new things, high earnings, new challenges, good career prospects, enough time
for leisure activities, social status, chance of doing something useful for society,
and good chance to combine work with family tasks.
The psychological literature considers that the report of how much each job
characteristic applies to the current job is not affected by how satisfied the re-
spondent is with it [see Locke and Latham (1990)]. It is, therefore, a good de-
scriptive measure of perceived discrepancy between what one would like to ob-
tain from a job and what one perceives she or he gets from the job. The same
applies to the measure of the importance of each job characteristic [see Rice et al.
(1991)].
Based on the replies to the questions mentioned above, we compute two sets of
dummy variables for each job characteristic (Table 1). In the first set of variables,
we distinguish between reporting a surplus (jobfacetS: if the respondent reported 4
or 5) and a deficit (jobfacetD: if the respondent reported 1 or 2) in the question
?how much the job characteristic applies to your current job?. Clearly, we set the
reference category to be when the respondent reported 3. In the second set of vari-
ables, the surpluses and deficits are decomposed by the level of importance assigned
Revista de Econom?a Aplicada
34
(2) A detailed description of the REFLEX survey is available at
http://www.fdewb.unimaas.nl/roa/reflex/
(3) We exclude self-employed and part-time workers because they form entirely distinct groups
and do not permit the homogeneity of individuals in key job characteristics necessary to assess the
influences of job characteristics on job satisfaction.
(4) In the ordered logit model, the cut-off parameters were not statistically significantly different
from each other until we collapsed the categories to two.
to each job characteristic. We consider the job facet is important for the respondent
when s/he reported 4 or 5, and non-important otherwise. Therefore, we obtain four
dummy variables for each job characteristic in the second decomposition.
Job satisfaction of university graduates
35
Table 1: DUMMY VARIABLES ON PERCEIVED DISCREPANCY,
DISTINGUISHING BY REPORTED IMPORTANCE
Perceived Perceived How much it Importance of the
discrepancy discrepancy by applies to the job characteristic
importance level current job for the respondent
DEFICIT (D) Low importance &
Deficit discrepancy (DU) 1-2 1-3
High importance &
Deficit discrepancy (DI) 1-2 4-5
SURPLUS (S) Low importance &
Surplus discrepancy (SU) 4-5 1-3
High importance &
Surplus discrepancy (SI) 4-5 4-5
Note: Reference group: Those who report no discrepancy (?How much it applies to your job?= 3).
Source: Own elaboration.
The dummy jobfacetSI (surplus, important) takes value 1 if the individual re-
sponded 4 or 5 to both questions referring to this job facet; zero otherwise. The
dummy jobfacetSU (surplus, unimportant) takes value 1 if the individual respond-
ed that the job facet has importance 1 to 3 and that the job facet applies to current
job 4 or 5; zero otherwise. The dummy jobfacetDI (deficit, important) takes value
1 if the individual responded 4 or 5 on the importance of the job facet and 1 or 2 on
how much it applies to the job; zero otherwise. Finally, the dummy jobfacetDU
(deficit, unimportant) takes value 1 if the individual responded 1 or 2 in both mea-
sures; zero otherwise. The reference category is the same as previously, i.e. the re-
spondent reported no discrepancy (the job characteristic applies with level 3, no
matter how much importance the individual attaches to the job characteristic).
These variables give us enough flexibility to test hypothesis 2, that is, to test
whether surplus and deficit discrepancies have asymmetric effects and whether
the reported importance of job characteristic matters.
In order to test hypothesis 1, whereby the subjective evaluation of job char-
acteristics is better at determining job satisfaction than the objective job charac-
teristics, we use objective variables on wage (standardized gross hourly wage),
job stability (permanent contract), hours of work (standardized total working
hours in current job per week) and job training (participated in job-related train-
ing in the last 12 months). The subjective counterparts of these job characteristics
Figure 3 shows the perception of deficit and surpluses decomposed by the
importance of each job characteristic. SI (SU) indicates the percentage of individ-
uals who report having surplus in the job characteristic and (do not) consider it
important. DI (DU) indicates the percentage of individuals who report having
deficit in the job characteristic and (do not) consider it important.
The first observation is that most individuals report having a surplus or
deficit of a job characteristic when they care about it. This is especially true for
are the discrepancies reported in high earnings, job security, time for leisure and
family tasks and learning, respectively.
We introduce subjective measures of over-education and over-skilling to test the
third hypothesis. Over-education is a dummy variable that takes value 1 if the re-
spondent felt that the appropriate level of education for the current job is lower than
the one attained, zero otherwise. We consider over-skilled the individual who report-
ed 1 or 2 from a 5-point scale (1-not at all; 5-to a very high extent) to the question:
?to what extent are your knowledge and skills utilized in your current work?? Sub-
jective measures are broadly used in the studies on educational mismatches.
3. DESCRIPTIVE STATISTICS
Here, we present the descriptive statistics of the main variables. Figure 1 pre-
sents the percentage of individuals who perceive that a job characteristic applies
to his/her job with surplus (responded 4 or 5 on a 5-point scale). Figure 2 shows
the same information for perceptions of deficit of each job characteristic (re-
sponded 1 or 2 on a 5-point scale). Most respondents perceive having a job with
work autonomy, security and learning opportunities (Figure 1), while high earn-
ings, status, usefulness for the society and time for family and leisure are the job
characteristics respondents miss the most (Figure 2).
Revista de Econom?a Aplicada
36
Figure 1: PERCEPTION OF SURPLUS IN JOB CHARACTERISTICS
Source: Own elaboration.
Job satisfaction of university graduates
37
Figure 2: PERCEPTION OF DEFICIT IN JOB CHARACTERISTICS
Source: Own computations.
Figure 3: PERCEPTION OF DISCREPANCIES IN JOB CHARACTERISTICS PER IMPORTANCE
Source: Own elaboration.
surpluses. In contrast, those who consider a job characteristic unimportant tend to
perceive neither deficit nor surplus of that job characteristic in their current job.
The main exceptions occur in the evaluation of high earnings, challenging job, so-
cial status and usefulness of the job for the society. Those reporting surplus of a
job characteristic which they do not care about are few (less than 10% in most
cases). In terms of deficit of particular job facets, deficit in social status and use-
fulness of the job for the society are the most common.
Figure 4 shows the incidence of over-education and over-skilling, the per-
centage of individuals with permanent contracts and those who received job-relat-
ed training during the 12 months before the interview. The incidence of over-edu-
cation is double the incidence of over-skilling for Spanish university graduates;
nearly 63% of graduates have a permanent job and 71% of individuals report hav-
ing attended some job-related training during last 12 months.
Revista de Econom?a Aplicada
38
Figure 4: INCIDENCE OF OVER-EDUCATION, OVER-SKILLING,
JOB PERMANENCY AND JOB TRAINING
Source: Own elaboration.
4. RESULTS
In this section, we present and discuss the results of the econometric estima-
tion. We estimate logistic regressions to explain the probability of being satisfied
with the current job. We report coefficient results in Table A in the Appendix. In
all the models, we control for education and skill mismatch (over-education and
over-skilling), objective job characteristics (permanent, wage, hours of work, job
training, tenure, public) and individual characteristics (gender, marital status, age,
child, education level and field of study).
Model A presents the estimation results when the evaluation of how much
each job characteristic applies to the current job is introduced. Recall that these
variables are constructed on a 5-point Likert scale. In Model B, we decompose
each job characteristic evaluation into two dummy variables, one for reported sur-
plus and the other for reported deficit as explained above. Model C extends the
analysis by additionally distinguishing between four possibilities: reported sur-
plus and important, reported surplus and unimportant, reported deficit and impor-
tant and reported deficit and unimportant.
62.7%
71.5%
14.8%
30.3%
4.1. Testing Hypothesis 1
To test hypothesis 1, whereby objective measures of job characteristics are
not important in explaining job satisfaction, we test whether the coefficients of the
variables job permanency (permanent), wage (stwage), job training (jtrain) and
hours worked (sthrswrk) are all equal zero. We perform a log-likelihood ratio test
for all three models, which rejects the null hypothesis in all cases5. From the table
of coefficients in the Appendix (Table A), it is clear that wage has a strongly sig-
nificant coefficient in all equations. Therefore, we can conclude that, in Spain, the
wage level has a significant positive effect in explaining job satisfaction, even
after controlling for the subjective evaluation of high earnings in the job. Notwith-
standing, subjective measures of earnings also matter for job satisfaction as we
discuss below. One explanation for the significance of the objective wage level
might be that its subjective counterpart refers to having ?high earnings? instead of
?fair earnings?, which would reflect the wage level of the respondent much more.
4.2. Testing Hypothesis 2
Hypothesis 2 states that subjective discrepancies on job characteristics should
be decomposed into deficit and surplus to allow for both linear and other non-lin-
ear functional forms in their effect on job satisfaction. Furthermore, it argues that
the reported importance of each job characteristic for the respondent should also
be taken into account as it influences the strength of the relationship between the
existing discrepancies and job satisfaction.
We observe that, for some job characteristics, it is important to decompose
the deficit and surplus in order to identify their non-linear shape relationship with
job satisfaction. Model A fails to identify a significant effect of having a challeng-
ing job, while Model B reveals that having positive discrepancy (surplus) in the
amount of challenge in the job makes people more satisfied with their jobs. Simi-
larly, in Model A, having time for leisure activities does not seem to affect job sat-
isfaction. However, when decomposing leisure into surplus and deficit, results re-
veal that having less leisure time than one would desire significantly decreases
satisfaction with the job. We can, therefore, conclude that our results support hy-
pothesis 2.
Model C reveals that further decomposing the perceived discrepancies by job
facet importance gives new information. First of all, job satisfaction is mostly af-
fected by those job facets that the respondent considers important (learning, earn-
ing, career prospects, leisure time, usefulness and family time)6. We observe,
however, an interesting result for the characteristic ?challenging job?. Graduates
who report deficit in this job facet and do not care about it are actually happy to
get a non-challenging job.
Table C in the Appendix shows several specification and goodness of fit tests
of the three models (A, B and C). Although the two information criteria (AIC and
Job satisfaction of university graduates
39
(5) Results are reported in Table B of the Appendix.
(6) Notice, though, that most coefficients of the job facets reported as unimportant for the respondent
(those finishing with SU or DU) have larger standard errors due to these variables having many zeros.
BIC) generally support Model A, Model C turns out to be a slightly better predic-
tor of job satisfaction (larger adjusted count R-squared). Moreover, distinguishing
between surplus, deficits and importance level (Model C) is the only specification
that is not rejected by the link test. Therefore, our preferred specification is Model
C, which confirms hypothesis 2.
4.3. Determinants of job satisfaction and prospect theory
We proceed now to investigate the importance of each job characteristic in ex-
plaining job satisfaction based on our preferred specification (Model C). Figures 5
and 6 show the predicted discrete change in probability of being satisfied with the
job when job characteristics are important and unimportant, respectively. They are
computed for a 30 year-old male with a permanent job, average tenure (nearly 3.5
years) and zero perceived discrepancies in all job facets. Table F in the Appendix
reports the same results plus their level of significance. As expected, the strongest
effects occur when there is a perceived discrepancy in a job characteristic that is
important to the respondent (Figure 5). Moreover, surplus discrepancies tend to in-
crease job satisfaction when one cares about the job characteristic and deficit dis-
crepancies tend to decrease job satisfaction when the job characteristic is important
to the respondent. In contrast, discrepancies when the job characteristic is unim-
portant (Figure 6) do not generally affect job satisfaction and their effect is more
mixed. While a surplus in learning opportunities and good career prospects in-
crease the likelihood of job satisfaction, a surplus in job autonomy, job security
and usefulness for society may decrease job satisfaction if the respondent does not
find these job characteristics important. A deficit discrepancy in work autonomy
decreases the probability of job satisfaction when the respondent reports these job
facets as unimportant. Only having a deficit discrepancy in the degree of challenge
of the job and learning possibilities increases the likelihood of being satisfied with
the job when the respondent does not regard these job characteristics as important.
We find some support for prospect theory in this context. Figure 5 reveals that,
for some job facets (work autonomy, high earnings, leisure time and usefulness for
the society), the deficit discrepancies have a stronger effect on job satisfaction than
the surplus discrepancies, as would be predicted by prospect theory. However,
learning possibilities, challenging job, good career prospects and time for family
show stronger positive effects of surpluses than the negative effects of deficits.
Summarizing, the most important job characteristics that lead to high job sat-
isfaction are having learning opportunities, having a challenging job, good career
prospects and having time for family tasks. In contrast, low work autonomy, little
time for leisure and performing a job not useful for the society are the main factors
that decrease job satisfaction. Extrinsic job characteristics, such as job security and
social status, do not seem to matter strongly for being satisfied with the job.
Revista de Econom?a Aplicada
40
Job satisfaction of university graduates
41
Figure 5: INCIDENCE OF OVER-EDUCATION, OVER-SKILLING,
JOB PERMANENCY AND JOB TRAINING
Source: Own elaboration.
Revista de Econom?a Aplicada
42
Figure 6: CHANGE IN PREDICTED PROBABILITIES DUE TO DISCREPANCIES
WHEN JOB CHARACTERISTIC IS UNIMPORTANT
Source: Own elaboration.
4.4. Testing Hypothesis 3
Table 2 reports the predicted change in the probability of being satisfied with
the job caused by being over-educated and over-skilled. Results reveal that being
over-skilled has a larger effect on job satisfaction than being over-educated. An
over-skilled worker has as much as 25% less probability of being satisfied with the
job than someone non-over-skilled. Nonetheless, being over-educated decreases the
probability of job satisfaction by 5%. Therefore, our results are consistent with pre-
vious studies that find a significant negative effect of over-education on job satisfac-
tion, although we show that mismatch in skills has a negative effect five times larg-
er. This latter result is in line with the work of Allen and van der Velden (2001).
5. CONCLUSIONS
In this paper, we have explored the determinants of job satisfaction of Span-
ish university graduates. We have based our analysis on Locke?s discrepancy theo-
ry whereby job satisfaction depends on the subjective evaluation of the discrepan-
cy between what one perceives he gets from the job and what he wants to obtain
from it. We have decomposed these discrepancies in surplus and deficit as well as
distinguishing by the importance level that the respondent associates with each
job characteristic. These decompositions have proven to be relevant when estimat-
ing the effect of several job characteristics on job satisfaction since it allows for
other than monotone relationships.
Our results show that, in general, subjective reports on the discrepancies in
job characteristics are more important in explaining job satisfaction than the re-
ported objective job characteristics. This was true for job training, hours of work
and permanency of the contract. This result did not hold, however, for wage,
which was significant as an objective measure and also as a subjective discrepan-
cy measure. Moreover, we found that intrinsic job characteristics (work autono-
my, learning opportunities, good career prospects and challenging job) are the
most important ones in explaining the likelihood of job satisfaction, while extrin-
sic job characteristics (social status, job security) have small effects. We obtain
support for prospect theory in some job facets (work autonomy, high earnings,
leisure time and usefulness for the society), but not for the rest.
We also study the impact of over-education and over-skilling on job satisfac-
tion for Spanish university graduates. Results reveal that not using all one?s skills
in the job reduces job satisfaction five times more than working in a job that re-
quires a lower education level than the one acquired.
Our analysis shows that Locke?s discrepancy theory should be considered
when studying the effects of job characteristics on job satisfaction. First of all, the
perceived importance of each facet by the individual strongly influences the effect
of this facet on job satisfaction. Secondly, we found a non-linear functional rela-
tionship between two job characteristics (having a challenging job and having
time for leisure) and job satisfaction, which would go undetected if one did not
distinguish between the deficits and surpluses of the job characteristics.
Job satisfaction of university graduates
43
Table 2: DISCRETE CHANGE IN PROBABILITY OF JOB SATISFACTION
(BASED ON MODEL C)
overed -0.055*
overskill -0.256***
Note: permanent = 1 tenure = 41.2 age = 30 and rest at zero.
Source: Own elaboration.
Revista de Econom?a Aplicada
44
APPENDIX
Data description: variables
Variable Definition
dsatis Dummy: =1 if reported 4-5 level of job satisfaction, =0 otherwise.
j1autona To what extent work autonomy applies to your current work situa-
tion (1-not at all, 5-to a very high extent).
j1secura To what extent job security applies to your current work situation
(1-not at all, 5-to a very high extent).
j1learna To what extent opportunity to learn new things applies to your cur-
rent work situation (1-not at all, 5-to a very high extent).
j1earnia To what extent high earnings applies to your current work situation
(1-not at all, 5-to a very high extent).
j1challa To what extent new challenges applies to your current work situa-
tion (1-not at all, 5-to a very high extent).
j1careea To what extent good career prospects applies to your current work
situation (1-not at all, 5-to a very high extent).
j1leisua To what extent enough time for leisure applies to your current work
situation (1-not at all, 5-to a very high extent).
j1statua To what extent social status applies to your current work situation
(1-not at all, 5-to a very high extent).
j1usefua To what extent chance of doing something useful for society applies
to your current work situation (1-not at all, 5-to a very high extent).
j1famila To what extent good chance to combine work with family tasks ap-
plies to your current work situation (1-not at all, 5-to a very high ex-
tent).
j1autoni how important is work autonomy to you personally (1-not at all, 5-
very important).
j1secure how important is job security to you personally (1-not at all, 5-very
important).
j1learni how important is opportunity to learn new things to you personally
(1-not at all, 5-very important).
j1earni1 how important is high earnings to you personally (1-not at all, 5-
very important).
j1challi how important is new challenges to you personally (1-not at all, 5-
very important).
j1careei how important is good career prospects to you personally (1-not at
all, 5-very important).
j1leisui how important is enough time for leisure to you personally (1-not at
all, 5-very important).
j1statui how important is social status to you personally (1-not at all, 5-very
important).
Job satisfaction of university graduates
45
Variable Definition
j1usefui how important is chance of doing something useful for society to
you personally (1-not at all, 5-very important).
j1famili how important is good chance to combine work with family tasks to
you personally (1-not at all, 5-very important).
overed Dummy: =1 if reported appropriate education for the work lower
than the level acquired, =0 otherwise.
overskill Dummy: =1 if reported that knowledge and skills are not utilized in
current work, =0 otherwise.
permanent Dummy: =1 if permanent contract, =0 if fixed-term.
stwage Standardized gross hourly wage.
jtrain Dummy: =1 if had job-related training in the past 12 months, =0
otherwise.
sthrswrk Total working hours in the present job (standardized).
female Dummy: =1 if female, =0 otherwise.
married Dummy: =1 if living with a partner at present, =0 Otherwise.
child Dummy: =1 if having childern, =0 otherwise.
edulvlG_2 Current level of education: dummy variable= 1 if isced 5a long pro-
gramme providing direct access to doctorate or above, 0 otherwise
(isced5a not providing direct access to doctorate).
public Dummy for working in the public sector.
tenure Number of months working with the current employer.
age Age of the respondent when interviewed.
Education Field of study: Education
Humanities Field of study: Humanities and arts
Social Field of study: Social Sciences, Business and Law.
Science Field of study: Science, Mathematics and Computing.
Engineer Field of study: Engineering, Manufacturing and Construction.
Agricult Field of study: Agriculture and Veterinary
Health Field of study: Health and Welfare
Services Field of study: Services
Revista de Econom?a Aplicada
46
Summary statistics
Variable Mean sd Min Max
dsatis 0.62 0.48 0 1
j1autona 3.75 1.01 1 5
j1secura 3.73 1.26 1 5
j1learna 3.65 1.13 1 5
j1earnia 3.10 1.11 1 5
j1challa 3.41 1.14 1 5
j1careea 3.25 1.17 1 5
j1leisua 3.32 1.24 1 5
j1statua 3.05 1.14 1 5
j1usefua 3.22 1.29 1 5
j1famila 3.30 1.24 1 5
j1autoni 4.26 0.75 1 5
j1secure 4.67 0.62 1 5
j1learni 4.61 0.62 1 5
j1earni1 4.14 0.81 1 5
j1challi 4.17 0.80 1 5
j1careei 4.39 0.73 1 5
j1leisui 4.50 0.74 1 5
j1statui 3.57 1.04 1 5
j1usefui 4.05 0.94 1 5
j1famili 4.51 0.77 1 5
Source: Own elaboration.
Job satisfaction of university graduates
47
Variable Mean sd Min Max
dovered 0.30 0.46 0 1
overskill 0.15 0.36 0 1
permanent 0.63 0.48 0 1
stwage 0.00 1.00 -2.16 4.15
jtrain 0.71 0.45 0 1
sthrswrk 0.00 1.00 -2.23 9.05
female 0.62 0.48 0 1
married 0.45 0.50 0 1
child 0.09 0.29 0 1
edulvlG_2 0.69 0.46 0 1
public 0.33 0.47 0 1
tenure 37.76 34.22 0 401
age 29.86 3.20 26 54
Education 0.11 0.32 0 1
Humanities 0.07 0.25 0 1
Social 0.35 0.48 0 1
Science 0.15 0.35 0 1
Engineer 0.15 0.36 0 1
Agricult 0.04 0.19 0 1
Health 0.13 0.33 0 1
Services 0.01 0.08 0 1
Source: Own elaboration.
Revista de Econom?a Aplicada
48
Econometric results: tables of coefficients
Table A: TABLE OF COEFFICIENTS
Model A Model B Model C
j1autona autonS autonSI 0.156*** 0.165 0.172
(0.060) (0.126) (0.130)
autonSU -0.407
(0.313)
autonD autonDI -0.561*** -0.533**
(0.205) (0.242)
autonDU -0.647**
(0.326)
j1secura securS securSI -0.00671 0.0750 0.113
(0.054) (0.151) (0.154)
securSU -0.729*
(0.416)
securD securDI -0.120 -0.121
(0.181) (0.187)
securDU -0.00565
(0.494)
j1learna learnS learnSI 0.396*** 0.814*** 0.795***
(0.067) (0.135) (0.138)
learnSU 0.677
(0.507)
learnD learnDI -0.108 -0.118
(0.179) (0.188)
learnDU 0.315
(0.529)
j1earnia earnS earnSI 0.139** 0.0237 0.0381
(0.064) (0.142) (0.148)
earnSU 0.0474
(0.345)
earnD earnDI -0.355** -0.408***
(0.142) (0.155)
earnDU -0.222
(0.227)
Job satisfaction of university graduates
49
Table A: TABLE OF COEFFICIENTS (continuation)
Model A Model B Model C
j1challa challS challSI 0.0783 0.308** 0.322**
(0.072) (0.140) (0.145)
challSU 0.411
(0.308)
challD challDI 0.143 0.0177
(0.172) (0.197)
challDU 0.419*
(0.249)
j1careea careerS careerSI 0.387*** 0.514*** 0.514***
(0.068) (0.144) (0.148)
careerSU 0.629
(0.505)
careerD careerDI -0.442*** -0.432**
(0.156) (0.169)
careerDU -0.297
(0.289)
j1leisua leisurS leisurSI 0.0762 -0.0909 -0.126
(0.065) (0.156) (0.159)
leisurSU -0.00392
(0.470)
leisurD leisurDI -0.343** -0.449***
(0.163) (0.172)
leisurDU 0.204
(0.297)
j1statua statuS statuSI -0.0599 -0.0209 -0.101
(0.062) (0.143) (0.161)
statuSU 0.281
(0.224)
statuD statuDI -0.0629 -0.0364
(0.142) (0.194)
statuDU -0.0454
(0.164)
j1usefua usefulS usefulSI 0.226*** 0.164 0.209
(0.054) (0.145) (0.150)
usefulSU -0.415
(0.332)
Revista de Econom?a Aplicada
50
Table A: TABLE OF COEFFICIENTS (continuation)
Model A Model B Model C
usefulD usefulDI -0.423*** -0.568***
(0.147) (0.174)
usefulDU -0.285
(0.187)
j1famila familyS familySI 0.113* 0.286* 0.297*
(0.066) (0.155) (0.159)
familySU 0.175
(0.477)
familyD familyDI -0.0203 -0.0247
(0.159) (0.168)
familyDU -0.0323
(0.309)
overed -0.262** -0.266** -0.225*
(0.133) (0.134) (0.136)
overskill -1.018*** -1.056*** -1.053***
(0.168) (0.168) (0.171)
permanent 0.294** 0.219 0.221
(0.143) (0.144) (0.145)
stwage 0.282*** 0.299*** 0.322***
(0.071) (0.072) (0.073)
jtrain 0.00963 0.0235 0.0287
(0.124) (0.125) (0.127)
sthrswrk -0.00548 0.00358 0.0288
(0.060) (0.060) (0.061)
female -0.0516 -0.0248 -0.00813
(0.121) (0.122) (0.124)
married 0.164 0.210* 0.227*
(0.117) (0.118) (0.121)
child -0.228 -0.260 -0.301
(0.213) (0.214) (0.217)
edulvlG_2 -0.0571 -0.0682 -0.104
(0.138) (0.140) (0.142)
public 0.622*** 0.674*** 0.665***
(0.145) (0.146) (0.149)
tenure -0.000479 -0.000836 -0.00121
(0.002) (0.002) (0.002)
Job satisfaction of university graduates
51
Table A: TABLE OF COEFFICIENTS (continuation)
Model A Model B Model C
age -0.0138 -0.0160 -0.0177
(0.022) (0.022) (0.022)
Education -0.0549 0.0134 0.00590
(0.229) (0.227) (0.229)
Humanities -0.171 -0.154 -0.180
(0.239) (0.240) (0.242)
Science -0.403** -0.420** -0.401**
(0.165) (0.167) (0.170)
Engineering -0.415** -0.456** -0.497***
(0.176) (0.179) (0.181)
Agric. & Vet -0.619** -0.660** -0.666**
(0.271) (0.274) (0.278)
Health -0.291 -0.306 -0.314
(0.199) (0.201) (0.206)
Services 0.671 0.485 0.498
(0.651) (0.655) (0.667)
_cons -3.927*** 0.640 0.707
(0.733) (0.680) (0.689)
N 2216 2216 2216
pseudo R2 0.277 0.283 0.293
AIC 2185.4 2186.9 2199.0
BIC 2362.2 2420.8 2546.9
chi2 812.7 831.2 859.1
Dependent variable: Job satisfaction. Logistic estimation. Standard errors in parentheses. Referen-
ce field of study: Social Sciences.
* p < 0.10, ** p < 0.05, *** p < 0.01.
Source: Own elaboration.
Revista de Econom?a Aplicada
52
Econometric results: post-estimation tests
Table B: LIKELIHOOD-RATIO TESTS ON THE OBJECTIVE JOB CHARACTERISTICS
Model A Model B Model C
LR chi2(4) 24.31 23.55 25.76
Prob > chi2 0.0001 0.0001 0.00
Table C: SPECIFICATION AND GOODNESS OF FIT TESTS
Model A Model B Model C
AIC 2185.4 2186.9 2199
BIC 2362.2 2420.8 2546.9
R2-count 77.66% 77.89% 78.47%
Adj. R2-count 0.407 0.413 0.429
HL Chi2 11.54 9.35 12.94
HL p-value 0.173 0.314 0.114
Link test (p-value) 0.016 0.085 0.288
AIC: Akaike?s information criterion; BIC: Bayesian information criterion; R2-count: Proportion of
correct predictions; Adj. R2-count: Adjusted proportion of correct predictions; HL: Hosmer-Le-
meshow specification test with 10 groups (null hypothesis: model well specified); Link test: the
null hypothesis is that conditional mean is correctly specified.
Source: Own elaboration.
REFERENCES
Allen, Jim and van der Velden, Rolf (2001): ?Educational Mismatches vs. Skill Mismatch-
es: Effects of Wages, Job Satisfaction and On-the-job Search?, Oxford Economic Pa-
pers, vol. 3, pp. 434-452.
Bender, Keith A., Donohue, Susan M. and Heywood, John S. (2005): ?Job satisfaction and
gender segregation?, Oxford Economic Papers, vol. 57, pp. 479-496.
Bowles, Samuel, Gintis, Herbert and Osborne, Melissa (2001): ?The Determinants of Earn-
ings: A Behavioral Approach?, Journal of Economic Literature, vol. 39, pp. 1137-1176.
Clark, Andrew E. (1996): ?Job Satisfaction in Britain?, British Journal of Industrial Rela-
tions, vol. 34 (2), pp. 189-217.
Clark, Andrew E. (1997): ?Job satisfaction and gender: Why are women so happy at work??,
Labour Economics, vol. 4, pp. 341-372.
Job satisfaction of university graduates
53
Predicted probabilities
Table D: DISCRETE CHANGE IN PROBABILITIES OF JOB SATISFACTION (MODEL C)
autonSI 0.041
autonSU -0.101
autonDI -0.132**
autonDU -0.160**
securSI 0.027
securSU -0.180*
securDI -0.030
securDU -0.001
learnSI 0.172***
learnSU 0.150
learnDI -0.029
learnDU 0.074
earnSI 0.009
earnSU 0.011
earnDI -0.101***
earnDU -0.055
challSI 0.075**
challSU 0.095
challDI 0.004
challDU 0.097*
Note: permanent =1 tenure = 41.2 age = 30 and rest at zero, * p < 0.10, ** p < 0.05, *** p < 0.01.
Source: Own elaboration.
careerSI 0.117***
careerSU 0.140
careerDI -0.107**
careerDU -0.073
leisurSI -0.031
leisurSU -0.001
leisurDI -0.111***
leisurDU 0.048
statuSI -0.025
statuSU 0.066
statuDI -0.009
statuDU -0.011
usefulSI 0.050
usefulSU -0.103
usefulDI -0.141***
usefulDU -0.070
familySI 0.070*
familySU 0.042
familyDI -0.006
familyDU -0.008
E
A
Clark, Andrew E. (1999): ?Are wages habit-forming? Evidence from micro data?, Journal
of Economic Behavior & Organization, vol. 39, pp. 179-200.
Clark, Andrew E. and Oswald, Andrew J. (1996): ?Satisfaction and comparison income?,
Journal of Public Economics, vol. 61, pp. 359-381.
Clark, Andrew E., Oswald, Andrew J. and Warr, Peter (1996): ?Is job satisfaction U-shaped
in age??, Journal of Occupational and Organizational Psychology, vol. 69, pp. 57-81.
Fabra, M. Eugenia and Camis?n, Cesar (2009): ?Direct and indirect effects of education on
job satisfaction: A structural equation model for the Spanish case?, Economics of Edu-
cation Review, vol. 28, pp. 600-610.
Freeman, Richard (1978): ?Job satisfaction as an economic variable?, American Economic
Review, vol. 68, pp. 135-141.
Gazioglu, Saziye and Tansel, Aysit (2006): ?Job satisfaction in Britain: individual and job
related factors?, Applied Economics, vol. 38, pp. 1163-1171.
Glenn, Norval D. and Weaver, Charles N. (1982): ?Further Evidence on Education and Job
Satisfaction?, Social Forces, vol. 61 (1), pp. 46-55.
Green, Fancis and McIntosh, Steven (2007): ?Is there a genuine under-utilization of skills
amongst the over-qualified??, Applied Economics, vol. 39, pp. 427-439.
Hanson, Sandra L. and Sloane, Douglas M. (1992): ?Young Children and Job Satisfac-
tion?, Journal of Marriage and the Family, vol. 54 (4), pp. 799-811.
Jackson, Chris and Corr, Philip (2002): ?Global Job Satisfaction and Facet Description: the
moderating role of facet importance?, European Journal of Psychological Assessment,
vol. 18 (1), pp. 1-8.
Kahneman, Daniel and Tversky, Amos (1979): ?Prospect Theory: An Analysis of Decision
under Risk?, Econometrica, vol. 47 (2), pp. 263-292.
Kalleberg, Arne (1977): ?Work Values and Job Rewards: A Theory of Job Satisfaction?,
American Sociological Review, vol. 42 (1), pp. 124-143.
Locke, Edwin A. (1969): ?What is Job Satisfaction??, Organizational Behavior and Human
Performance, vol. 4, pp. 309-336.
Locke, Edwin A. and Latham, G.P. (1990): ?A theory of goal setting and task performan -
ce?: Englewood Cliffs, NJ: Prentice Hall.
Lydon, Reamonn and Chevalier, Arnaud (2002): ?Estimates of the Effect of Wages on Job
Satisfaction?, Publication of the Centre for Economic Performance.
Maynard, Douglas C., Joseph, Todd Allen and Maynard, Amanda M. (2006): ?Underem-
ployment, job attitudes and turnover intentions?, Journal of Organizational Behavior,
vol. 27, pp. 509-536.
McGuinness, Seamus and Sloane, Peter J. (2011): ?Labour Market Mismatch Among UK
Graduates: An Analysis Using REFLEX Data?, Economics of Education Review, vol.
30 (1), pp. 130-145.
Mobley, William H. and Locke, Edwin A. (1970): ?The relationship of value importance to
satisfaction?, Organizational Behavior and Human Performance, vol. 5, pp. 463-483.
Mora, Toni and Ferrer-i-Carbonell, Ada (2009): ?The job satisfaction gender gap among
young recent university graduates: Evidence from Catalonia?, The Journal of Socio-
Economics, vol. 38, pp. 581-589.
Origo, Federica and Pagani, Laura (2009): ?Flexicurity and job satisfaction in Europe: The
importance of perceived and actual job stability for well-being at work?, Labour Eco-
nomics, vol. 16, pp. 547-555.
Pollmann-Schult, Matthias and Buchel, Felix (2004): ?Career Prospects of Overeducated
Workers in West Germany?, European Sociological Review, vol. 20 (4), pp. 321-331.
Revista de Econom?a Aplicada
54
Rice, Robert, Gentile, Douglas and McFarlin, Dean (1991): ?Facet importance and job sat-
isfaction?, Journal of Applied Psychology, vol. 76 (1), pp. 31-39.
Sloane, Peter J. and Williams, H. (2000): ?Job Satisfaction, Comparison Earnings, and
Gender?, Labour, vol. 14 (3), pp. 473-502.
Sousa-Poza, Alfonso and Sousa-Poza, Andr?s A. (2000): ?Taking Another Look at the
Gender Job-Satisfaction Paradox?, Kyklos, vol. 53, pp. 135-152.
Sousa-Poza, Alfonso and Sousa-Poza, Andr?s A. (2003): ?Gender differences in job satis-
faction in Great Britain, 1991-2000: permanent or transitory??, Applied Economics
Letters, vol. 10 (11), pp. 691-694.
Sousa-Poza, Alfonso and Sousa-Poza, Andr?s A. (2007): ?The effect of job satisfaction on
labor turnover by gender: An analysis for Switzerland?, The Journal of Socio-Econom-
ics, vol. 36, pp. 895-913.
Usui, Emiko (2008): ?Job satisfaction and the gender composition of jobs?, Economics
Letters, vol. 99, pp. 23-26.
Vaisey, Stephen (2006): ?Education and its Discontents: Overqualification in America
1972-2002?, Social Forces, vol. 85 (2), pp. 835-864.
Vila, Luis E., Garcia-Aracil, Adela and Mora, Jose-Gines (2007): ?The Distribution of Job
Satisfaction Among Young European Graduates: Does the Choice of Study Field Mat-
ter??, The Journal of Higher Education, vol. 78 (1), pp. 97-118.
Fecha de recepci?n del original: noviembre, 2011
Versi?n final: noviembre, 2011
RESUMEN
Este art?culo investiga los determinantes de la satisfacci?n laboral de los
graduados universitarios en Espa?a. Basamos el an?lisis en la teor?a de la
discrepancia de Locke [Locke (1969)] y descomponemos la evaluaci?n
subjetiva de las caracter?sticas del trabajo en niveles de exceso y d?ficit.
Tambi?n estudiamos la importancia de la sobre-educaci?n y la sobre-ca-
pacitaci?n en la satisfacci?n laboral. Utilizamos los datos de la encuesta
REFLEX. Nuestros resultados muestran que la satisfacci?n laboral es
b?sicamente determinada por la evaluaci?n subjetiva de las caracter?sti-
cas intr?nsecas del trabajo, con un impacto asim?trico de excesos y d?fi-
cits. Otro resultado es que la sobre-capacitaci?n es mucho m?s relevante
que la sobre-educaci?n para explicar la satisfacci?n laboral.
Palabras clave: satisfacci?n laboral, teor?a de la discrepancia, sobre-edu-
caci?n, sobre-capacitaci?n.
Clasificaci?n JEL: J28.
Job satisfaction of university graduates
55