Está en la página 1de 20

The Relationship between Further Training

and Job Satisfaction

Claudia Burgard
Ruhr Graduate School in Economics (RGS Econ) and RWI Essen

Katja Gorlitz
RWI Essen

-PRELIMINARY-
Do not cite. Comments are very welcome.

May 14, 2010

Abstract. This paper analyzes the relationship between participation in further training
courses and satisfaction using data from the German Socio-Economic Panel (GSOEP). In
order to do that, we estimate a binary Logit model as well as a so-called Probit-adapted
OLS (POLS) model introduced by van Praag and Ferrer-i-Carbonell (2004). Our prelim-
inary results indicate a significant positive correlation between attending training courses
and employees job satisfaction as well as their life satisfaction. Separate estimations by
gender show that this result only holds for males while the training coefficient for fe-
males is insignificant. Controlling for time invariant unobserved heterogeneity scales up
the coefficients. We also find for male workers that the association between training and
job satisfaction seems to intensify with the duration of courses. According to our esti-
mation results there seems to be a larger increase in work satisfaction for men the longer
the courses are. In addition to that we examine differences with respect to the financing
of training courses. Participants who received financial support for all courses appear to
be more satisfied compared to those who had to finance at least one course completely
themselves. Moreover, we differentiate between firm-specific and (partly) general training
courses. Entirely firm-specific courses seem to have a negative correlation with job satis-
faction for men.

JEL-Classification: I29, J24, J28, M53, P36

Keywords: Training, Satisfaction, Logit, POLS, fixed effects

Financial support by the Ruhr Graduate School in Economics is gratefully acknowledged.


All correspondence to Claudia Burgard, RGS Econ, Hohenzollernstr. 1-3, 45128 Essen,
Germany, e-mail: claudia.burgard@rwi-essen.de.
1 Introduction
As employees working lives are nowadays often characterized by rapidly changing requirements
and a rising demand for skilled personnel, the role of on-the-job training becomes increasingly
important. The improvement of access to continuous training programs is one way to satisfy the
need for lifelong learning, especially in a society with an aging work-force.

Following Beckers human capital theory (Becker, 1964), participation in training might lead to
a higher productivity of workers and to a higher income. Beside this, it might be that further train-
ing influences job satisfaction as well. We find the analysis of this relation an interesting question
because job satisfaction as a subjective valuation of ones work measures different aspects of a
job than monetary outcomes do, and might be of importance for individuals and firms as well.
The relation of training and life satisfaction might be worth being examined as in addition to the
often analyzed productive part of education the consumption part of education might also play a
role in training decisions. In Alstadster and Sievertsen (2009), the consumption value of higher
education is defined as the private, intended, non-pecuniary returns to higher education. This con-
sumption value of education or training might be captured by changes in life satisfaction.

While there is a broad literature focusing on monetary outcomes of on-the-job training, only a
few studies investigate the non-monetary returns like (job) satisfaction. Job satisfaction contributes
essentially to total life satisfaction of employees and it might also affect firms performance through
its potential impact on productivity (Jones et al., 2008). In his study Economic Aspects of Job
Satisfaction, Hamermesh (1977) hypothesizes a positive relation between training opportunities
and job satisfaction.

This paper investigates the association between participation in formal training courses and job sat-
isfaction of employees using data from the GSOEP. Estimating linear models and discrete choice
models combined with panel estimation methods, our preliminary results suggest a significant pos-
itive relationship between participation in training courses and work satisfaction. Separate analyses
for men and women reveal that this positive correlation seems to be driven solely by male workers.
Furthermore, the estimations indicate that the association between training and job satisfaction
might depend on course duration, financing of the courses and the type of content; i.e. firm-specific
or general courses.

The outline of the paper is as follows: Section 2 gives information about the theoretical back-
ground and existing literature. Section 3 introduces the data and provides some descriptive results.
Section 3 describes the empirical strategy and section 4 reports (preliminary) estimation results.
Finally, Section 5 concludes the paper and discusses possible approaches for future research.

2
2 Theory and Literature
According to the human capital theory (Becker, 1964), training is often seen as an investment in
human capital. This means that there are costs and returns involved by the decision to invest in
training. Costs can be of direct or indirect nature, e.g. participation fees, travelling costs, foregone
income or foregone leisure time as well as psychological costs etc., on which this study will not
focus further. Adressing potential outcomes of a training investment, many economists tried to
measure monetary returns, like wage or earnings effects of training (see e.g. Buchel and Pannen-
berg (2004); Jurges and Schneider (2005); Kuckulenz and Zwick (2003); Leuven and Oosterbeek
(2004); Muehler et al. (2007); Pischke (2001)). Examples for non-monetary returns mainly include
promotion prospects, job security or employment probability. However, workers utility in terms
of satisfaction with their job or with their life might as well be affected by further training. Con-
sider the following function for utility from working which was introduced by Clark and Oswald
(1996)):

= (, , , ),

where is income, are working hours, contains individual characteristics and comprises job
characteristics. We assume that includes participation in training courses, firm characteristics
and other working characteristics. Job satisfaction is a subjective measure reflecting workers as-
sessment of different aspects of the job, i.e. it provides information about individuals valuation of
their utility from working.
To our knowledge, there are only few studies dealing with training and job satisfaction. Gazioglu
and Tansel (2002) find a significant positive relation of training and job satisfaction in Britain using
ordered Probit estimation. (Schmidt, 2007) investigates the relationship between training satisfac-
tion and overall job satisfaction when analyzing a sample of 552 customer and technical service
employees from the United States and Canada. He finds a significant positive relationship in terms
of a bivariate regression coefficient. Oreopoulos and Salvanes (2009) examine the effect of school-
ing on lifetime well-being and find positive effects using twins and siblings data. Georgellis and
Lange (2007) investigate the relationship of on-the-job training and job satisfaction. They use
three waves of the GSOEP where the latest one is from 2000. Estimating an ordered Probit model
leads them to a significant positive correlation of employer-financed training and job satisfaction
for men. DAddio et al. (2007) analyze determinants of job satisfaction where they also include
information about training participation. They use data from Denmark and estimate ordered Logit
fixed effects model which was proposed by Ferrer-i-Carbonell and Frijters (2004) and Das and van
Soest (1999). The coefficient of their training variable is significantly positive for men. For women
it is the same when using the Das and van Soest (1999) estimation but leads to an insignificant
coefficient with the Ferrer-i-Carbonell and Frijters (2004) estimator.

3
3 Data and Descriptive Results
We use data from the GSOEP1 which is provided by the DIW Berlin (German Institute of Economic
Research). The GSOEP is a representative longitudinal data set which started in 1984 and conducts
annual surveys. The most recent wave available, 2008, comprises more than 20,000 persons living
in about 11,000 households. The data include information about demographic and socioeconomic
characteristics, educational background and the occupational situation of the respondents. The
GSOEP asks interviewed persons to rate their job satisfaction on a 0 (low) to 10 (high) scale every
year. As training information is collected only in some years and the questionnaire was modified
frequently over time, we use three waves of the GSOEP data, i.e. 2000, 2004 and 2008. Only these
waves provide comparable information on further training which each refer to the past three years.
Following the questionnaire, (further) training is defined as participation in professionally oriented
courses. In addition to that, the data contain information on the course duration of the last three
courses for those who attended at least one course. Moreover, respondents are asked about whether
they received financial support from their employer, from the Employment Office, or from another
source. Information on firm-specific or general-type training is given as well.
Our sample consists of full- and part-time employed persons aged between 18 and 64 years, exclud-
ing apprentices, public servants, self-employed persons, and people from the high income sample
generated by the GSOEP. These subgroups might exhibit a strongly differing behavior regarding
further training. The remaining sample, after dropping missing values, contains 21,710 observa-
tions which are used for the estimation since they provide information on all variables needed.

As can be seen from Table 1, mean training participation in the sample is 31.1% where it is on
average slightly higher for women than for men.

Table 1: Participation in further training


2000 2004 2008 mean
All 30.7% 30.2% 32.7% 31.1%
Male 29.2% 30.6% 31.3% 30.2%
Female 32.5% 29.7% 34.3% 32.1%

Table 2 presents means of job satisfaction for training participants and non-participants separated
by gender. Overall average job satisfaction is about 7. For male workers, there is a significant
difference in the average job satisfaction between participants and non-participants whereas this is
not the case for females. Figure 1 shows the development of the values of mean job satisfaction
1
The dataset was extracted using PanelWhiz. See Haisken-DeNew (2007) and Haisken-DeNew and Hahn
(2006). For more information on the GSOEP, please see http://www.diw.de/soep. Any data or computational
errors in this paper are our own.

4
Figure 1: Average job satisfaction over time

over time for the above groups of male and female participants and non-participants. An overall
drop in the employees valuation of their work satisfaction can be observed. More interestingly,
there is a large difference in the satisfaction level of men over the whole time span (upper and lower
line). This gap is not distinctive for women (middle two lines). Table A.2 (see Appendix) shows
means and standard deviations of all explanatory variables for participants and non-participants as
well as by gender.
The following section presents the model specification and the empirical strategy used for our
analysis.

Table 2: Means of job satisfaction


Participants Non-Participants
Mean (SD) Mean (SD) N
1
All 7.023 (0.037) 6.9071 (0.026) 23,900
2
Male 7.047 (0.051) 6.8992 (0.035) 12,874
Female 7.010 (0.055) 6.917 (0.038) 11,026
Note: 1 ,2 Difference statistically significant at 1%-level.

5
4 Empirical Strategy
In our empirical framework, the model specification we use is derived from the utility function
from working (see Section 2). More precisely, we estimate the following model:

= 0 + 1 + 2 ( ) + 3 + 4 + 5 + 6 + , (1)

where the subscripts and denote the ith individual and the year t in the panel dataset, respec-
tively. Job satisfaction is modeled as a function of a binary training indicator , (the natural log-
arithm of) income (()), weekly hours of overtime (), and other individual/socio-demographic
(), job-related () and firm-specific characteristics ( ). In detail, we control for income, hours
of overtime per week, nationality, residence in West or East Germany, gender, age, marital status,
children living in the household, years of education, having a part- or full-time working contract,
being a blue- or white-collar worker, tenure of being employed at the current firm, job change dur-
ing last year, worries about job security, firm size and industry.
All estimations are conducted for men and women separately in order to figure out potential differ-
ences between the two groups. The Chow Test supports this decision.
Due to the ordinal nature of our dependent variable, the estimations are carried out by discrete
choice methods. Firstly, the original scale measuring job satisfaction which ranges from zero to
ten is collapsed into a binary variable2 in order to conduct Logit and binary OLS estimations.
To address potential correlation of the training variable with unobserved factors, we apply fixed ef-
fects estimations. By doing so we are able to account for time invariant unobserved factors, which
could affect the training decision as well as job satisfaction. Problems of reverse causality are re-
garded as less severe for our estimation because of the time lag between reporting job satisfaction
and training participation.
Since the conversion of the dependent variable into an indicator variable implies a loss of informa-
tion (and in case of the Logit fixed effects estimation also a loss of observations), we additionally
run so-called Probit adapted OLS (POLS) fixed effects estimations. Therefore, secondly, the job
satisfaction measure is converted into a standardized variable according to van Praag and Ferrer-
i-Carbonell (2004). The standardized variable retains the original eleven categories. The POLS
method uses the implicit cardinalization implied by the ordered Probit model. Once the dependent
variable is standardized, usual OLS estimation can be applied since the values of the cardinalized
outcome are not limited to certain bounds, e.g. zero and ten, anymore (see Appendix B for a de-
scription of the calculation of the standardized variable).

In a second step of our analysis, we investigate possible heterogeneity with respect to the follow-
2
This binary variable equals one if reported job satisfaction is equal to or larger than the individual mean
of job satisfaction over the years 1997 until 2008; and zero otherwise. For details, see Kassenbohmer and
Haisken-DeNew (2008).

6
ing characteristics: course length3 , financial support from employers, and type of human capital
aquired in the course (specific or general). These estimations are conducted by using the POLS
method as well since this is our preferred technique.
The next section provides the empirical results of our estimations.

5 (Preliminary) Results
The correlation between participation in further training courses and job satisfaction is presented
in Table 3. It shows results from binary Logit, binary OLS and POLS regressions for men and
women separately. It can be seen that there is a significant positive association between training
participation and work satisfaction for men but not for women.4
Since there might likely be the problem of unobserved heterogeneity in our model, in a next step
we control for fixed effects. This rules out the influence of time invariant factors that possibly bias
our estimated coefficients. The results of the fixed effects estimations are shown in Table 4. The
same pattern as in table 3 can be observed: There is indication for a positive association between
participation in continuous training and job satisfaction. As before, this result seems to hold only
for males. For female workers, the estimated coefficients of the training variable are insignificant
again.

If we compare the Logit, OLS and POLS results, we see that the estimations of the training
coefficients are robust across these methods. The signs and significance levels of the estimated
coefficients are equal. Furthermore, it can be noticed that including fixed effects leads to larger
coefficients across all three methods.
The observation that the correlations between training and job satisfaction tend to become stronger
and more significant when implementing fixed effects estimations indicates underestimation in the
simple regressions without fixed effects. This might be due to unobserved time invariant factors
influencing the training participation as well as satisfaction. Examples for such factors are motiva-
tion or ability. They probably affect the training decision in a positive way but job satisfaction in
a negative way. The latter relation might not be intuitively clear, however, there are studies which
find that persons with higher ability or with a higher educational degree seem to be less satisfied
with their work (Ganzach, 2003; Barrett et al., 1980).
As presented above, our results suggest a positive relationship between training and job satisfaction
for men but not for women. One possible explanation for this could be that men are more likely
to get a better position in their job after having attended e.g. a career-oriented training course.
3
The course length used here is an approximate measure constructed by the authors. In case the person
took part in more than one course, it is the sum of the lengths of the last two or three courses.
4
All estimations have been as well conducted excluding income from the set of control variables (See
Appendix A.3, A.4, A.9). The results show no considerable differences compared to our baseline specifica-
tion.

7
Maybe they experience a higher self-esteem by taking more responsibility in their job or a higher
reputation by their boss which makes them more satisfied.

Estimation results regarding training and life satisfaction can be found in Table 5 and Table 6.
Since, like described above, unobserved heterogeneity seems to play a role we present here the
fixed effects results.5 They are quite similar with the job satisfaction estimations when maintaining
the same model specification (Table 5). However, when controlling for job satisfaction in the life
satisfaction equation, the significant correlations for men disappear. Thus, there seems to be no
additional change in life satisfaction once job satisfaction is controlled for. This could mean that
there is no support for the assumption of an consumption value of training activities.

The relationship of attending training courses and satisfaction might not only differ by gender
but depend on other factors as well. It is supposable that training duration, financing and contents
of the courses play a role. In order to examine this we conduct additional estimations where we in-
clude information about the course length, financial support from the employer and firm-specificity
of the knowledge aquired, respectively, into the model. Regarding the latter two, we only consider
participants where for the former the control group are non-participants. We focus on POLS fixed
effects estimations since this is our preferred method. Results are reported in Table 7. Evidence
suggests that, for men, the longer the course is the higher is their job satisfaction. This coincides
with the intuitive consideration that more training yields higher returns. That this seems not to be
true for women requires another explanation. It could for example be that opportunity costs for
femals are higher than for males.
If courses are (at least partly) financially supported there seems to be a significant positive correla-
tion with job satisfaction for both males and females. This is immediately comprehensible.
Participation in courses of firm-specific nature is negatively associated with job satisfaction for the
male sample. A possible explanation for this could be that females value the security of their job
at their current firm more than males do.

6 Conclusion and Outlook


This paper investigates the association between further training and satisfaction. Our results sug-
gest a significant positive relation between participation in professional oriented training courses
and job satisfaction. This result seems to hold only for male employees but not for female workers.
The same result is obtained with respect to life satisfaction. This changes, however, if we addi-
tionally control for job satisfaction. Furthermore, we find evidence that this association depends
5
For comparison purposes, the results from estimations without fixed effects are shown in the Appendix
(Table A.5 and Table A.6).

8
Table 3: Job satisfaction estimations
Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.034*** 0.136*** 0.060*** -0.007 -0.027 -0.026
(0.011) (0.046) (0.022) (0.012) (0.049) (0.024)
Overtime 0.001 0.006 -0.006** 0.002 0.008 -0.010***
(0.001) (0.005) (0.003) (0.002) (0.008) (0.004)
German -0.001 -0.003 0.025 -0.021 -0.085 0.012
(0.017) (0.067) (0.036) (0.021) (0.087) (0.045)
West -0.042*** -0.170*** 0.024 -0.042*** -0.173*** 0.000
(0.013) (0.053) (0.027) (0.013) (0.052) (0.028)
Age 0.018*** 0.075*** -0.027*** 0.022*** 0.088*** -0.008
(0.004) (0.017) (0.008) (0.004) (0.017) (0.009)
Age2 /100 -0.021*** -0.087*** 0.021** -0.026*** -0.106*** 0.005
(0.005) (0.020) (0.010) (0.005) (0.021) (0.011)
Married 0.007 0.027 0.014 0.014 0.059 0.101***
(0.013) (0.053) (0.027) (0.012) (0.049) (0.027)
Children 0.008 0.032 0.012 0.034*** 0.139*** 0.090***
(0.012) (0.049) (0.024) (0.013) (0.051) (0.027)
Years of education -0.006** -0.025** -0.015*** -0.003 -0.014 -0.015***
(0.003) (0.010) (0.005) (0.003) (0.011) (0.006)
Part time -0.004 -0.016 -0.100 0.001 0.003 0.019
(0.034) (0.138) (0.075) (0.014) (0.059) (0.031)
Blue collar -0.003 -0.012 0.002 -0.030** -0.121** -0.155***
(0.013) (0.053) (0.027) (0.015) (0.061) (0.033)
Tenure -0.008*** -0.033*** -0.004 -0.009*** -0.036*** -0.018***
(0.002) (0.008) (0.004) (0.002) (0.009) (0.004)
Tenure2 /100 0.017*** 0.070*** 0.009 0.025*** 0.104*** 0.043***
(0.005) (0.021) (0.010) (0.006) (0.026) (0.013)
Job change 0.047*** 0.192*** 0.072** 0.042** 0.170** 0.020
(0.016) (0.067) (0.031) (0.017) (0.068) (0.034)
ln(Income) 0.018 0.072 0.232*** 0.018 0.074 0.124***
(0.017) (0.068) (0.036) (0.014) (0.059) (0.030)
Worries (job) -0.073*** -0.295*** -0.410*** -0.072*** -0.293*** -0.390***
(0.010) (0.043) (0.021) (0.011) (0.045) (0.023)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
Year effects Yes Yes Yes Yes Yes Yes
R2 /Pseudo-R2 0.017 0.012 0.071 0.018 0.013 0.057
F-value 6.9 - 26.9 6.4 - 17.9
Log-Likelihood - -7,276 - - -6,253 -
Obs. 10,631 10,631 10,631 9,143 9,143 9,143
Note: Coefficients show marginal effects; Standard errors in parentheses.
Significance levels: * < 0.10, ** < 0.05, *** < 0.01.
Firm size: 3 dummy variables; Industry: 6 dummy variables.

9
Table 4: Job satisfaction estimations (fixed effects)
Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.067*** 0.277*** 0.124*** 0.009 0.042 -0.032
(0.021) (0.086) (0.030) (0.023) (0.090) (0.034)
Overtime 0.003 0.015 0.004 0.002 0.011 -0.001
(0.002) (0.010) (0.003) (0.004) (0.014) (0.005)
Age 0.024* 0.088* 0.006 0.023* 0.103** 0.000
(0.013) (0.051) (0.017) (0.013) (0.052) (0.018)
Age2 /100 -0.012 -0.044 -0.003 -0.017 -0.071 -0.008
(0.014) (0.056) (0.020) (0.015) (0.059) (0.022)
Married -0.000 -0.015 0.030 0.001 -0.002 -0.050
(0.034) (0.133) (0.049) (0.037) (0.147) (0.055)
Children -0.032 -0.119 -0.040 0.036 0.153 0.084**
(0.025) (0.102) (0.037) (0.028) (0.114) (0.042)
Years of education 0.007 0.022 -0.001 -0.010 -0.047 -0.052
(0.046) (0.172) (0.067) (0.048) (0.199) (0.072)
Part time -0.005 0.059 -0.153 0.055 0.264* 0.055
(0.076) (0.294) (0.111) (0.035) (0.140) (0.052)
Blue collar -0.112*** -0.431*** -0.196*** -0.075 -0.307 -0.131*
(0.036) (0.151) (0.052) (0.047) (0.192) (0.071)
Tenure -0.016*** -0.064*** -0.027*** -0.020*** -0.081*** -0.035***
(0.004) (0.018) (0.006) (0.005) (0.020) (0.008)
Tenure2 /100 0.009 0.039 0.017 0.039** 0.153** 0.053**
(0.014) (0.058) (0.020) (0.017) (0.067) (0.025)
Job change 0.034 0.149 0.079* 0.037 0.133 0.067
(0.028) (0.116) (0.041) (0.030) (0.118) (0.045)
ln(Income) 0.165*** 0.603*** 0.281*** 0.119*** 0.526*** 0.147**
(0.045) (0.186) (0.066) (0.039) (0.156) (0.058)
Worries (job) -0.137*** -0.553*** -0.218*** -0.156*** -0.627*** -0.287***
(0.020) (0.083) (0.029) (0.022) (0.090) (0.033)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
Year effects Yes Yes Yes Yes Yes Yes
R2 /Pseudo-R2 0.047 0.058 0.062 0.045 0.057 0.065
F-value 8.9 - 12.1 7.0 - 10.3
Log-Likelihood - -1,648 - - -1,386 -
Obs. 10,631 4,873 10,631 9,143 4,101 9,143
Note: Coefficients show marginal effects; Standard errors in parentheses.
Significance levels: * < 0.10, ** < 0.05, *** < 0.01.
Firm size: 3 dummy variables; Industry: 6 dummy variables.

on certain characteristics of the course as well. Our estimations indicate a positive relation be-
tween course duration and satisfaction for male training participants. Receiving financial support
for training courses seems to correlate positively with job satisfaction, for both men and women.

10
Table 5: Life satisfaction estimations (fixed effects)
Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.068*** 0.336*** 0.050* -0.010 -0.035 0.042
(0.018) (0.094) (0.027) (0.019) (0.105) (0.032)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 /Pseudo-R2 0.047 0.075 0.043 0.044 0.076 0.041
F-value 9.4 - 8.5 7.2 - 6.7
Log-Likelihood - -1,313 - - -1,038 -
Obs. 10,983 3,950 10,983 9,410 3,130 9,410
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 4). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

Table 6: Life satisfaction estimations (fixed effects, control for js)


Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.014 -0.129 -0.002 0.001 -0.229 0.050
(0.013) (1.5e+07) (0.027) (0.014) (2.1e+07) (0.031)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 /Pseudo-R2 0.540 1.000 0.122 0.536 1.000 0.091
F-value 205.6 - 24.4 165.4 - 14.4
Log-Likelihood - -0 - - -0 -
Obs. 10,621 3,712 10,621 9,134 2,975 9,134
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 4). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

11
Table 7: POLS estimations (fixed effects)
Male Female
duration finance specific duration finance specific
Training 0.125 - - -0.138* - -
(< 1 day) (0.086) - - (0.079) - -
Training 0.089 - - -0.054 - -
(1 day - 1 week) (0.064) - - (0.075) - -
Training 0.189*** - - -0.062 - -
(1 week - 1 month) (0.070) - - (0.095) - -
Training 0.278*** - - -0.149 - -
(> 1 month) (0.106) - - (0.126) - -
Training finan- - 0.141** - - 0.128* -
cially supported - (0.064) - - (0.073) -
Firm-specific - - -0.194** - - 0.005
training - - (0.095) - - (0.134)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
2
R 0.056 0.111 0.126 0.062 0.110 0.160
F-value 7.1 3.9 2.4 6.2 3.1 2.7
Obs. 8,875 2,895 2,021 7,432 2,532 1,888
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 4). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

Finally, firm-specific training leads to a negative association with satisfaction for men.

Unfortunately, using data from the GSOEP only allows us controlling for a small set of firm char-
acteristics (e.g. industry and size). However, changes at the firm level such as the introduction
of new technological equipment or organizational changes could have occurred simultaneously
within the period under investigation (Bauer, 2004). Therefore, it is not clear whether the observed
correlations also capture changes in employees work environment. Hence, our results have to be
interpreted cautiously and our estimated coefficients measure correlations rather than causal ef-
fects. Further research on the identification of the causal effect of training participation on (job)
satisfaction is needed. This could be examined by applying instrumental variable regressions. For
the implementation there is a strong and valid instrument needed which accounts for unobserved
heterogeneity in a more appropriate way.

In addition to the analyses we conducted so far, it would be interesting to take into account the
heterogeneity of training courses with respect to contents, aims and other properties even more. In
order to investigate such issues, of course adequate data is necessary.

12
References
Alstadster, Annette and Hans Henrik Sievertsen, The Consumption Value of Higher
Education, Working Paper 2871, CESIFO 2009.
Barrett, Gerald V., J. Benjamin Forbes, Edward J. OConnor, and Ralph A. Alexander,
Ability-Satisfaction Relationships: Field and Laboratory Studies, Academy of Manage-
ment Journal, 1980, 23 (3), 550555.
Bauer, Thomas K., High Performance Workplace Practices and Job Satisfaction: Evidence
from Europe, IZA Discussion Paper 1265, Institute for the Study of Labor (IZA) 2004.
Buchel, Felix and Markus Pannenberg, Berufliche Weiterbildung in West- und Ost-
deutschland - Teilnehmer, Struktur und individueller Ertrag, Zeitschrift fur Arbeitsmark-
tforschung, 2004, 37, 73126.
Becker, Gary S., Human Capital: A Theoretical and Empirical Analysis with Special Refer-
ence to Education, Chicago: University of Chicago Press, 1964.
Clark, Andrew and Andrew Oswald, Satisfaction and Comparison Income, Journal of
Public Economics, 1996, 63 (1), 359381.
Cornelien, Thomas, The Interaction of Job Satisfaction, Job Search, and Job Changes. An
Empirical Investigation with German Panel Data, Journal of Happiness Studies, 2009,
10, 367384.
DAddio, Anna Cristina, Tor Eriksson, and Paul Frijters, An Analysis of the Deter-
minants of Job Satisfaction when Individuals Baseline Satisfaction Levels May Differ,
Applied Economics, 2007, 39 (19), 24132423.
Das, Marcel and Arthur van Soest, A panel data model for subjective information on
household income growth, Journal of Economic Behavior and Organization, 1999, 40,
409426.
Ferrer-i-Carbonell and Frijters, How Important is Methodology for the Estimates of the
Determinants of Happiness?, Economic Journal, 2004, 114 (497), 641659.
Ganzach, Yoav, Job satisfaction as an economic variable, Work and Occupations, 2003, 30
(1), 97122.
Gazioglu, Saziye and Aysit Tansel, Job Satisfaction in Britain: Individual and Job Related
Factors, ERC Working Papers 0303, ERC - Economic Research Center, Middle East
Technical University 2002.
Georgellis, Yannis and Thomas W.K. Lange, Participation in Continuous, On-the-Job
Training and the Impact on Job Satisfaction: Longitudinal Evidence from the German
Labour Market , International Journal of Human Resource Management, 2007, 18 (6),
969985.

13
Haisken-DeNew, John P., PanelWhiz: A Stata interface for large scale panel datasets,
German Stata Users Group Meetings 2007 08, Stata Users Group April 2007.
and Markus Hahn, PanelWhiz: A Flexible Modularized
Stata Interface for Accessing Large Scale Panel Data Sets, 2006.
http://www.panelwhiz.eu/docs/PanelWhiz Introduction.pdf (Zugriffsdatum: 30. August
2008).
Hamermesh, Daniel S., Economic Aspects of Job Satisfaction, in Orley C. Ashenfelter
and Wallace E. Oates, eds., Essays in Labor Market Analysis, J. Wiley Sons, 1977.
Jones, Melanie K., Richard J. Jones, Paul L. Latreille, and Peter J. Sloane, Training,
Job Satisfaction and Workplace Performance in Britain: Evidence from WERS 2004,
IZA Discussion Papers 3677, Institut zur Zukunft der Arbeit (IZA) 2008.
Jurges, Hendrik and Kerstin Schneider, Dynamische Lohneffekte beruflicher Weiterbil-
dung, MEA Discussion Paper 5092, Mannheim Research Institute for the Economics of
Aging, Universitat Mannheim 2005.
Kassenbohmer, Sonja C. and John P. Haisken-DeNew, Youre Fired! The Causal Nega-
tive Effect of Unemployment on Life Satisfaction, Ruhr Economic Papers No. 63, RWI
Essen 2008.
Kuckulenz, Anja and Thomas Zwick, The Impact of Training on Earnings - Differences
between Participant Groups and Training Forms, ZEW Discussion Papers 03-57, Zen-
trum fur Europaische Wirtschaftsforschung (ZEW) Mannheim 2003.
Leuven, Edwin and Hessel Oosterbeek, Evaluating the Effect of Tax Deductions on Train-
ing, Journal of Labor Economics, 2004, 22 (1), 461488.
Muehler, Grit, Michael Beckman, and Bernd Schauenberg, Returns to Continuous Train-
ing in Germany: New Evidence from Propensity Score Matching Estimators, WWZ
Discussion Paper 04/07, Wirtschaftswissenschaftliches Zentrum (WWZ) der Universitat
Basel 2007.
Oreopoulos, Philip and Kjell G. Salvanes, How large are returns to schooling? hint:
money isnt everything, NBER Working Paper 15339, National Bureau of Economic
Research 2009.
Pischke, Jorn-Steffen, Continuous Training in Germany, Journal of Population Eco-
nomics, 2001, 14 (3), 523548.
Schmidt, Steven W., The relationship between satisfaction with workplace training and
Overall Job Satisfaction, Human Resource Development Quarterly, 2007, 18 (4), 481
498.
van Praag and Ferrer-i-Carbonell, Happiness quantified: A satisfaction calculus approach,
Oxford: Oxford University Press, 2004.

14
A Appendix - Tables

Table A.1: Definition of Variables


Variable Description
a. Dependent variables
Job satisfaction (0-10) Reported job satisfaction measured on a scale from 0 to 10
Job sat., standardized Standardized job satisfaction
Job satisfaction (0/1) Reported job satisfaction recoded into a binary variable
Life satisfaction (0-10) Reported life satisfaction measured on a scale from 0 to 10
Life sat., standardised Standardized life satisfaction
Life satisfaction (0/1) Reported life satisfaction recoded into a binary variable
b. Explanatory variables
Training 0/1-variable; 1 if respondent participated in formal training
course during the previous three years; 0 otherwise
Training (< 1 day) 0/1-variable; 1 if training course length less than one day; 0 oth-
erwise
Training (1 day - 1 week) 0/1-variable; 1 if training course length is between one day and
one week; 0 otherwise
Training (1 week - 1 month) 0/1-variable; 1 if training course length is between one week and
one month; 0 otherwise
Training (> 1 month) 0/1-variable; 1 if training course length more than one month; 0
otherwise
Firm-specific training 0/1-variable; 1 if acquired skills of all attended courses are firm-
specific; 0 otherwise
Training financially supported 0/1-variable; 1 if all attended courses are (partly) financially sup-
ported; 0 otherwise
Demographics
Age Age in years
Nationality 0/1-variable; 1 if nationality is German; 0 otherwise
East 0/1-variable; 1 if respondent living in east Germany; 0 for West
Germany
Marital status 0/1-variable; 1 if married; 0 otherwise
Children 0/1-variable; 1 if children living in respondents household; 0
otherwise
Education/Employment
ln(Income) Natural Logarithm of gross income last month
Years of education Length of education in years
Tenure Firm tenure in years
Overtime Overtime hours per week
Part/full time worker 0/1-variable; 1 if respondent works part time; 0 if respondent
works full time
Blue-/white-collar worker 0/1-variable; 1 if respondent is a blue-collar worker; 0 if respon-
dent is a white-collar worker
Job change during last year 0/1-variable; 1 if respondent changed his job during the last year;
0 otherwise
Worries about job 0/1-variable; 1 if respondent is (somewhat or very) concerned
about his job security; 0 otherwise
Firm characteristics
Firm size dummies Firm size in categories: <20, 20-199, 200-1999, >2000
Industry dummies Categories: agriculture, energy, mining; manufacturing; con-
struction; trade; transport; bank, insurance; services

15
Table A.2: Summary statistics
Male Participants Female Participants Male Non-Participants Female Non-Participants
Variable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.
Job satisfaction (0-10) 7.16 1.86 7.05 1.94 6.95 1.99 6.98 2.04
Job sat., standardized 0.06 0.93 0.02 0.97 -0.03 0.98 -0.01 1.01
Job satisfaction (0/1) 0.501 0.500 0.492 0.500 0.478 0.500 0.488 0.500
Life satisfaction (0-10) 7.19 1.49 7.12 1.62 7.01 1.61 6.98 1.68
Life sat., standardized 0.08 0.91 0.05 0.99 -0.03 0.97 -0.03 1.01
Life satisfaction (0/1) 0.535 0.499 0.531 0.499 0.525 0.499 0.526 0.499
Training 1 0 1 0 0 0 0 0
Training (< 1 day) 0.182 0.386 0.306 0.461 0.000 0.000 0.000 0.000
Training (1 day - 1 week) 0.356 0.479 0.341 0.474 0.000 0.000 0.000 0.000
Training (1 week - 1 month) 0.343 0.475 0.230 0.421 0.000 0.000 0.000 0.000
Training (> 1 month) 0.116 0.320 0.115 0.319 0.000 0.000 0.000 0.000
Firm-specific training 0.404 0.491 0.399 0.490 - - - -
Training financially supported 0.626 0.484 0.554 0.497 - - - -
German 0.957 0.203 0.973 0.163 0.877 0.329 0.907 0.290
West 0.771 0.420 0.677 0.468 0.769 0.421 0.764 0.424
Male 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000
Age 40.48 9.51 40.15 9.85 41.58 10.66 41.75 10.69
Age2 /100 17.29 7.94 17.09 7.99 18.43 8.99 18.58 8.88
Married 0.660 0.474 0.548 0.498 0.656 0.475 0.623 0.485

16
Children 0.457 0.498 0.356 0.479 0.430 0.495 0.371 0.483
ln(Income) 8.23 0.48 7.79 0.55 8.02 0.51 7.49 0.64
Years of education 13.03 2.63 13.06 2.41 11.70 2.33 11.73 2.20
Overtime 3.55 4.35 2.24 3.13 2.54 3.93 1.42 2.72
Part time 0.023 0.149 0.350 0.477 0.032 0.176 0.447 0.497
Blue collar 0.297 0.457 0.058 0.234 0.619 0.486 0.293 0.455
Tenure 10.97 9.53 9.72 8.79 10.86 9.91 9.29 8.77
Tenure2 /100 2.11 3.19 1.72 2.81 2.16 3.42 1.63 2.76
Job change 0.147 0.354 0.173 0.378 0.165 0.371 0.190 0.392
Worries (job) 0.584 0.493 0.585 0.493 0.663 0.473 0.606 0.489
Firm size <20 0.143 0.350 0.249 0.433 0.238 0.426 0.306 0.461
Firm size 20-199 0.264 0.441 0.276 0.447 0.340 0.474 0.317 0.465
Firm size 200-1999 0.264 0.441 0.253 0.435 0.232 0.422 0.207 0.405
Firm size >2000 0.329 0.470 0.222 0.416 0.190 0.392 0.170 0.375
Agricul., energy, mining 0.049 0.215 0.018 0.135 0.036 0.187 0.014 0.116
Manufacturing 0.235 0.424 0.086 0.280 0.305 0.460 0.164 0.370
Construction 0.194 0.395 0.037 0.190 0.255 0.436 0.053 0.225
Trade 0.104 0.305 0.116 0.321 0.121 0.326 0.244 0.429
Transport 0.064 0.244 0.034 0.182 0.076 0.265 0.034 0.182
Bank, insurance 0.071 0.257 0.067 0.251 0.028 0.164 0.043 0.203
Services 0.284 0.451 0.641 0.480 0.179 0.384 0.448 0.497
Number of observations 3,933 3,578 9,015 7,464
Table A.3: Job satisfaction estimations without income
Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.028*** 0.114*** 0.057*** -0.006 -0.025 -0.010
(0.011) (0.043) (0.021) (0.011) (0.046) (0.023)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 /Pseudo-R2 0.016 0.011 0.063 0.016 0.012 0.054
F-value 7.4 - 27.2 6.6 - 19.3
Log-Likelihood - -8,004 - - -6,865 -
Obs. 11,686 11,686 11,686 10,024 10,024 10,024
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 3). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

Table A.4: Job satisfaction estimations without income (fixed effects)


Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.063*** 0.263*** 0.118*** 0.013 0.056 -0.023
(0.019) (0.080) (0.029) (0.021) (0.084) (0.032)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 /Pseudo-R2 0.044 0.056 0.058 0.042 0.053 0.062
F-value 10.0 - 13.3 7.8 - 11.6
Log-Likelihood - -1,884 - - -1,574 -
Obs. 11,686 5,551 11,686 10,024 4,636 10,024
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 4). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

Table A.5: Life satisfaction estimations


Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.036*** 0.148*** 0.019 -0.010 -0.042 0.038
(0.012) (0.048) (0.021) (0.012) (0.050) (0.024)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 /Pseudo-R2 0.028 0.020 0.094 0.027 0.020 0.072
F-value 11.2 - 35.0 9.2 - 24.2
Log-Likelihood - -7,439 - - -6,381 -
Obs. 10,983 10,983 10,983 9,410 9,410 9,410
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 3). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

17
Table A.6: Life satisfaction estimations (control for js)
Male Female
lin. PM Logit POLS lin. PM Logit POLS
Training 0.008 0.059 -0.007 -0.008 -0.047 0.048**
(0.009) (0.065) (0.019) (0.010) (0.068) (0.022)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
2 2
R /Pseudo-R 0.392 0.384 0.267 0.397 0.386 0.203
F-value 316.4 - 105.3 272.5 - 64.7
Log-Likelihood - -4,530 - - -3,883 -
Obs. 10,621 10,621 10,621 9,134 9,134 9,134
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 3). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

Table A.7: POLS FE, with income, LS


Male Female
duration finance specific duration finance specific
Training 0.152* - - 0.012 - -
(< 1 day) (0.078) - - (0.073) - -
Training 0.003 - - 0.017 - -
(1 day - 1 week) (0.059) - - (0.069) - -
Training 0.082 - - 0.052 - -
(1 week - 1 month) (0.064) - - (0.087) - -
Training 0.073 - - -0.010 - -
(> 1 month) (0.099) - - (0.116) - -
Training finan- - 0.082 - - -0.067 -
cially supported - (0.059) - - (0.062) -
Firm-specific - - -0.125 - - 0.182
training - - (0.098) - - (0.120)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 0.046 0.052 0.063 0.043 0.081 0.107
F-value 5.9 1.8 1.1 4.3 2.4 1.7
Obs. 9,139 3,021 2,027 7,630 2,623 1,891
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 4). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

18
Table A.8: POLS FE, with income, LS with JS
Male Female
duration finance specific duration finance specific
Training 0.049 - - 0.052 - -
(< 1 day) (0.079) - - (0.072) - -
Training -0.045 - - 0.021 - -
(1 day - 1 week) (0.058) - - (0.068) - -
Training 0.011 - - 0.046 - -
(1 week - 1 month) (0.064) - - (0.086) - -
Training -0.012 - - 0.064 - -
(> 1 month) (0.096) - - (0.115) - -
Training finan- - 0.035 - - -0.104* -
cially supported - (0.058) - - (0.063) -
Firm-specific - - -0.089 - - 0.179
training - - (0.096) - - (0.119)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 0.117 0.155 0.099 0.092 0.134 0.143
F-value 15.2 5.5 1.8 9.2 3.8 2.3
Obs. 8,865 2,893 2,019 7,425 2,531 1,887
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 4). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

Table A.9: POLS FE, without income, JS


Male Female
duration finance specific duration finance specific
Training 0.144* - - -0.117 - -
(< 1 day) (0.081) - - (0.076) - -
Training 0.080 - - -0.005 - -
(1 day - 1 week) (0.062) - - (0.072) - -
Training 0.201*** - - -0.069 - -
(1 week - 1 month) (0.067) - - (0.091) - -
Training 0.247** - - -0.069 - -
(> 1 month) (0.100) - - (0.119) - -
Training finan- - 0.124** - - 0.114* -
cially supported - (0.060) - - (0.068) -
Firm-specific - - -0.186* - - 0.033
training - - (0.097) - - (0.132)
Firm size Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes
R2 0.053 0.100 0.107 0.055 0.104 0.156
F-value 7.8 4.1 2.4 6.5 3.4 3.1
Obs. 9,732 3,173 2,215 8,158 2,736 2,028
Note: Coefficients show marginal effects; Standard errors in parentheses. Control variables included
(compare Table 4). Significance levels: * < 0.10, ** < 0.05, *** < 0.01.

19
B Appendix - POLS method
In order to apply the POLS estimation, the categorical dependent variable is rescaled by deriving Z-
values of the standard normal distribution that correspond to cumulative frequencies of the original
categories. The cardinalized dependent variable is then the expectation of a standard normally
distributed variable under the condition that the value of the original variable lies between two cut-
off points. The calculation of this cardinalized variable works as follows (compare Cornelien
(2009)):

= (1 < < 2 ) = [(1 ) (2 )]/[(2 ) (1 )]

where is a standard normal random variable, is the standard normal probability density func-
tion, and is the standard normal cumulative density function. In our analysis, the standardized
values for the job satisfaction (, ) and life satisfaction variable (, ), respectively, are then as
follows:




2.92 if = 0

3.23 if = 0



2.44 if = 1 2.79 if = 1









2.09 if = 2 2.42 if = 2







1.72 if = 3 2.02 if = 3









1.41 if = 4 1.67 if = 4







, = 1.02 if = 5 , = 1.21 if = 5



0.62 if = 6 0.74 if = 6









0.20 if = 7 0.22 if = 7










0.42 if = 8






0.55 if = 8






1.10 if = 9

1.40 if = 9




1.88 if = 10,
2.23
if = 10

20

También podría gustarte