Upload
georgetown
View
0
Download
0
Embed Size (px)
Citation preview
Electronic copy available at: http://ssrn.com/abstract=2439593
1
EXTRINSIC MOTIVATION, PSM AND LABOUR MARKET CHARACTERISTICS:
A MULTILEVEL MODEL OF PUBLIC SECTOR EMPLOYMENT PREFERENCE
IN 26 COUNTRIES1
Forthcoming in
International Review of Administrative Sciences
Steven Van de Walle Bram Steijn Sebastian Jilke
[email protected] [email protected] [email protected]
Department of Public Administration
Erasmus University Rotterdam
1 We would like to thank Dion Curry, Niels Schenk and Sandra Groeneveld for their valuable assistance. We also
would like to thank the University of Bergen, department of administration and organizational science, for the
hospitality given to one of the authors who wrote part of this article during his sabbatical there; The research
leading to these results has received funding from the European Union’s Seventh Framework Programme under
grant agreement No. 266887 (Project COCOPS), Socio-economic Sciences and Humanities.
Electronic copy available at: http://ssrn.com/abstract=2439593
2
Abstract
Research findings have been contradictory with respect to the determinants of why people
choose a public sector job. In this paper we use an internationally comparative design with
data from 26 countries to explain public sector employment preference. The study shows that
on the individual level public service motivation and extrinsic motivation both are important
drivers for this preference. Intrinsic motivation, in turn, is negatively related with people's
inclination to work for the public sector. Moreover, having a lower income and lower
education is associated with a greater preference for public sector employment. This suggests
that working for the public sector is seen as a good and safe career option. Our results
furthermore show that variation in this preference can only partly be explained by country
differences. Nevertheless, in countries with a career- rather than position-based system of
public employment people are more likely to prefer public employment.
Points for practitioners
Attracting the best and brightest to work for the public sector requires an insight into why
people prefer public over private sector employment. This paper looks at what makes people
prefer public sector employment in 26 countries. Findings reveal that public service
motivation (helping other people, being useful to society) and extrinsic motives (job security,
a high income, opportunities for advancement) play an important role in this preference. Still,
there are considerable differences between countries. In countries with a career-based system
of public employment, working in the public sector is seen as more attractive.
Keywords: Public service motivation, work motives, work values, public sector sorting, public
employment
3
Introduction
‘Who wants to work for the government?’ is a provocative question asked by Lewis and
Frank (2002). According to their answer ‘becoming a public sector employee involves both
choice and chance’ (2002: 395). Like Lewis and Frank (2002) our focus in this paper is on the
question why people prefer a public sector job. We argue that the choice for public
employment is affected by both individual preferences as well as by country dependent
structural labour market characteristics. Choice is in our view thus dependent on preferences
and opportunities. As only a few studies have looked at the effect of the opportunity structure
on public employment preference, this will be our main focal point.
Nevertheless, two mechanisms situated on the level of individual job attribute
preferences should be mentioned when explaining why people prefer a public sector job over
one in the private sector. The first mechanism, popular in recent public administration
literature, points to the importance of public service motivation (PSM) (Perry & Wise, 1990;
Vandenabeele, 2008; Ritz & Waldner, 2011). According to Perry and Wise (1990: 368) PSM
‘may be understood as an individual’s predisposition to motives grounded primarily or
uniquely in public institutions and organizations’. In other words, people are attracted to a
public sector job because they want to contribute to society or to the public good.
The second mechanism is put forward by economists and is quite different. In this
view, public sector workers are shirkers, self-interested and not interested in performing for
the public good (Niskanen, 1971; Downs, 1967). They are mainly attracted to the public
sector by good employment conditions, high job security and low (and difficult) monitoring
of individual performance (Delfgaauw & Dur, 2008).
At first glance, these two explanatory mechanisms for public sector attractiveness are
contradictory. In this respect, Delfgaauw and Dur (2008) suggest that two types of workers
4
are attracted to the public sector: ‘dedicated’ workers driven by PSM, and ‘lazy’ workers who
may crowd out the dedicated ones.
However, preferences based on job attributes (or work values)1 are not the only
determinants of career choice. Lewis and Frank (2002) observe that the preference for
working in the public sector declines with each birth cohort. Others refer to economic
conditions. For instance, Llorens and Stazyk (2011: 119), although addressing turnover rather
than sector preference, observe that ‘in lean economic times, one generally expects public
sector employees to be less likely to separate because of the decreased likelihood of
alternative employment opportunities’.
Thus a preference for public sector employment is not constant, but varies over time
and space. Few studies have looked at this preference from a comparative perspective; one of
the few exceptions (Norris 2003) showed that public sector employment preference varies
widely among countries.
Norris (2003) suggested three possible factors to explain this variation: public
management reform, cultural differences and structural patterns in the labour force. With
respect to the latter, we will argue that both characteristics of the internal (public) labour
market (career structure) and external labour market conditions (wage differences,
unemployment and economic growth) are relevant.
In order to look at the effect of these internal and external labour market conditions on
the preference for public sector employment, we will use an international comparative
perspective using data for 26 countries taken from the International Social Survey Programme
(ISSP 2005b). Our research question is: what is the effect of internal and external labour
market conditions on public sector employment preference?
In the next section we first hypothesize on the effect of labour market conditions on
public sector employment preference, resulting in three hypotheses. Following this, we will
5
briefly discuss the effect of work values, leading to a further three hypotheses. We will then
describe the data. Multilevel statistics are used to explain public sector employment
preference in 26 countries. We end by discussing the limitations and implications of our
findings.
Theoretical considerations
Differences in employment sector preference between countries: the importance of the labour
market
Norris (2003) reported large differences between countries with respect to the preference for
working within the public sector. According to her data (based on ISSP, 1997), there is a
strong preference for public employment in post-Communist and developing countries, and a
lower desire in Anglo-American and Scandinavian countries.
It is not difficult to imagine a possible effect of labour market conditions on this
preference. The idea of a ‘rational man’ is a cornerstone of traditional economic theory
(Simon, 1955). According to this theory people will, confronted with several possible courses
of action, choose the option which will be most profitable for them. Applied to sector of work
preference, people will thus be attracted to a sector that offers favourable conditions.
With respect to the external labour market, we will in this respect look at two aspects:
a) the differences in wages between the public and private sector (the wage premium), and b)
economic conditions.
Wage premium
Classical economic theory predicts that people will behave rationally. Applied to wages, this
implies that they will prefer jobs with a higher wage over lower paid jobs. Various studies
6
have shown that, in this respect, a ‘wage premium’ makes the public sector a more attractive
employer (Gunderson, 1979; Tansel, 2005). As the public sector will, in general, not be able
to set wages in the private sector the wage premium can be seen as an external labour market
condition. Boudarbat (2008) claims that this mechanism explains why university graduates in
Morocco prefer public to private employment. Although Llorens and Stazyk (2011) did not
find that public sector wage premiums explained differences in employee turnover between
US states, several studies have shown that, in some countries, the public sector does pay
relatively higher wages (Bender, 1998; Gornick & Jacobs, 1998), and that such a wage
premium affects employment sector preferences (Adamchik & Bedi, 2000; Tansel, 2005).
Therefore, we will test the following hypothesis:
H1. The higher the public sector’s wage premium in a country, the stronger the preference for
working in the public sector.
Economic conditions
Boudarbat (2008) notes that ‘employment in the public sector is generally desired for its
stability, which attracts risk-adverse workers’. On a country level, this argument becomes
especially valid during times of economic upheaval: job security will be more important
during difficult economic times.
Several studies suggested that this is indeed the case. According to Groeneveld et al.
(2009), economic circumstances affect public sector attractiveness. They found that people
looking for job security will prefer public sector employment during an economic crisis, but
might switch to the private sector during an economic boom since job security is then not an
issue. A similar argument was developed by Llorens and Stazyk (2011) in an analysis of
differences in turnover between US states. We will therefore test the following hypothesis:
7
H2. Weak economic conditions and levels of unemployment in a country are associated with a
stronger preference for working in the public sector.
Characteristics of the internal labour market
So far, we have looked at the external labour market to explain differences in sector
employment preferences. However, characteristics of the internal labour market will also be
important.
Doeringer and Piore (1971) highlighted the importance of the internal labour market,
defined as ‘an administrative unit where the levels of employment and wages are determined
by a set of internal administrative rules and procedures’. Within the public sector, two types
of internal market can be discerned: career-based and position-based systems (OECD, 2005;
Hammerschmid et al., 2007). In career-based systems, employees are expected to spend their
whole working life in public service. After initial post-education entry, their careers depend to
a large part on decisions made by the organization. Possibilities to enter public employment in
later life are limited (OECD, 2005: 164). In contrast, within position-based systems, the best-
suited candidate is recruited to fill a vacancy and external recruitment is relatively common.
Such a system provides less security for public sector employees as they have to compete with
others to achieve a career move. We expect that the career-based system is more attractive to
prospective public sector employees as it provides greater security. Thus, we will test the
following hypothesis:
H3. The preference for working for the public sector is higher in countries with a career-
based, rather than a position-based, internal job market.
8
Work values
Although our focus is on explaining differences between countries in employment sector
preference, individual work values will also matter. In this respect, Norris (2003) noted that
two distinct ‘motivational values’ contributed to explaining public sector employment
differences in nearly all geographical regions. Firstly, she noted that those who want a job that
is useful for society prefer public sector employment and, secondly, that the same holds for
people with a preference for job security.
This brings us back to the mechanisms outlined in the introduction for explaining a
preference for public sector employment. Perry and Wise (1990) have put the concept of
public service motivation on the academic agenda. They formulated the hypothesis that a
higher level of PSM is related to greater interest in public sector employment.
Since then, many studies have confirmed their initial hypothesis (Vandenabeele, 2008;
Steijn, 2008, Wright & Christensen, 2010). Others, however, have argued that it is job type
and type of occupation rather than just sector that matters (Christensen & Wright, 2011;
Houston, 2011; Kjeldsen & Jacobsen, 2013). The fourth hypothesis is therefore formulated as
follows:
H4. A higher level of PSM is associated with a higher preference for working in the public
sector.
Several authors have suggested that ‘selfish’ motives rather than PSM may attract people to
the public sector. Hypotheses 1-3 already dealt with the importance of these motives for
sector of employment preference on a country level. Here we are interested in effects at the
individual level as people will differ in the degree to which selfish motives are important to
them.
9
The classic distinction between extrinsic and intrinsic motivations in Herzberg’s
(1966) theory, although controversial (Latham, 2007: 102), can be used here. Behaviour that
is intrinsically motivated is undertaken because of the inherent satisfaction derived from a
task (Houston, 2011). In contrast, extrinsic motivational factors lie outside the job: people
work ‘because they have to’ - it provides them with money or security (Groeneveld et al.
2009). In other words, ‘the task is undertaken for instrumental reasons and therefore satisfies
personal needs indirectly’ (Houston, 2011: 762).
Economic theorists tend to stress the importance of extrinsic motivation as an
important determinant of public sector employment preference (Downs, 1967; Delfgaauw &
Dur, 2008). However, the same argument is made within the public administration literature.
For instance, Perry and Hondeghem (2008: 3) note that the ‘public sector has traditionally
offered some strong extrinsic motivators, such as security of tenure, the career system, and
pension systems’. It is generally assumed that public sector employees value job security
more than private sector workers do (Lewis & Frank, 2002; Norris, 2003; Houston, 2011)
although some studies disagree (Crewson, 1997). With respect to pay, another important
extrinsic motivator, the empirical evidence is less clear. Some studies have found that public
servants are less motivated by pay (Buelens & Van den Broeck, 2007; Houston, 2011), but the
finding of Lewis and Frank (2002), that a high income was relatively more important for
those wanting to work for the US government, contradict this. In a comparative study,
Houston (2011) found that, especially in Anglo-Saxon countries, ‘controlled extrinsic
motives’ (that include high income and job security) are relatively important. Although the
empirical evidence on pay as an extrinsic motivator for public sector preference is thus
somewhat inconclusive, the overall evidence on all aspects of extrinsic motivation leads to the
following hypothesis.
10
H5. A higher extrinsic work orientation is associated with a stronger preference for working
in the public sector.
Several studies have found that public sector workers are more intrinsically motivated
(Crewson, 1997; Houston, 2000). However, some recent studies report a different conclusion.
Buelens and Van den Broeck (2007) for instance found that Flemish civil servants were less
motivated by self-development and responsibility; Norris (2003) found that people who
preferred to work in business gave greater priority to autonomy and the ability to work
independently. Houston (2011) did not find differences regarding the sector and whether or
not one finds ‘an interesting job’ a very important work motive. He does, however, find
significant differences with respect to the importance adhered to ‘help other people’ and ‘a
useful job to society’. This finding raises the question whether the previous finding that public
sector workers are more intrinsically motivated has in fact only to do with PSM as a special
dimension of intrinsic motivation. If the latter is indeed the case, we can formulate the
following hypothesis.
H6. A higher intrinsic work orientation is associated with a lower preference to work in the
public sector.
Data
We used data from the Work Orientations III survey (2005) that was undertaken within the
International Social Survey Programme (ISSP 2005b). This survey used a multistage stratified
random sample of respondents aged 18 and above2. Data were collected using a standardized
questionnaire, but with a variety of methods (self-completion by mail, CAPI etc.) depending
on local conditions.
11
The results from this survey have frequently been used to study work values. We add
to this work by looking at a large number of countries. We expand Norris’ (2003) approach,
using the ISSP 1997, by using a multilevel design. In the next section we will explain why
such a design is preferable. Houston (2011) used the 2005 survey to look at work motives
within a subset of eleven countries but did not relate these motives to public sector
employment preferences. Taylor and Westover (2011; see also Westover & Taylor, 2010)
used both the 1997 and 2005 ISSP data to relate work motives to job satisfaction within a
subset of seven countries. Their study does not address a possible preference for public sector
employment and also does not use a multilevel design. Taylor and Taylor (2011) also used a
subsample of 15 countries to look at the relationships among efficiency wages, PSM and
effort, but did also not use a multilevel design. The same was true for Jin (2013a, 2013b) who
looked at PSM and sector choice among unemployed people in several countries.
We use a subset of the data consisting of respondents up to 65 years of age – the
official upper limit retirement age in many countries (notwithstanding differences across
countries and between male and female retirement ages). Further, the analysis is restricted to
respondents who were currently in paid employment because we assume jobless respondents
will be interested in any job, whether public or private. We focus on those 26 countries in the
dataset that are members of the EU or OECD: Australia, Belgium (Flanders only), Bulgaria,
Canada, Cyprus, Czech Republic, Denmark, Finland, France, Germany, Hungary, Ireland,
Israel, Japan, Korea, Latvia, Mexico, New Zealand, Norway, Portugal, Slovenia, Spain,
Sweden, Switzerland, United States, and United Kingdom (Great Britain only). This resulted
in a total dataset of 19,271 respondents, with a country-level sample size ranging from 465
(Hungary) to 1,202 (Denmark). Although response rates varied across countries (see Scholz et
al., 2008), stringent checks and fieldwork controls by the ISSP mean that the sampling can
generally be considered as representative. After deleting cases with missing values, the
12
analysis was carried out with a dataset where N=15,222. The income variable accounted for
the largest number of missing data (13.9% of the original sample). Respondents who did not
report income were slightly more likely to prefer public sector employment than other
respondents. Also, people currently working in the private sector are less likely to report their
income, but both groups do not differ in a statistically significant way in terms of age and
gender. Moreover, non-response on income items tends to be concentrated at the tails of the
wider income distribution and is also relatively random (Riphahn & Serfling, 2005).
Variables and statistical model
Modelling strategy
Considering both individual-level and country-level factors, we seek to explain preferences
for public sector employment in a set of 26 countries. We use a multilevel model because the
data is nested and, therefore, there could potentially be clustering effects in the data due to
country-specific public sector or labour market related variables (Guo & Zhao, 2000). Further,
using a multilevel model makes it possible to assess which country-level factors explain
differences between countries. Using country dummies instead would only allow the
researcher to observe and control for country-level effects, not to explain them. A danger is
that small level-2 sample sizes (country-level) could lead to biased estimates of standard
errors. However, a sample size of 26 (countries) is considered sufficient for random intercept
models with binary outcomes to produce sufficiently unbiased estimates (Stegmueller, 2013).
Dependent variable: public sector employment preference
A preference for public sector employment was measured as follows: ‘Suppose you were
working and could choose between different kinds of jobs, which of the following would you
personally choose?’ Respondents could choose between: ‘I would choose working in a private
13
business’ and ‘I would choose working for the government or civil service’. In theory, there
was a third spontaneous response - ‘no preference’ - but respondents were not provided with
this answer category. However, there was a ‘cannot choose’ option and its usage differed
substantially across countries, and in some cases exceeded 20%. These differences suggest
that there were country differences in fieldwork protocols. We decided to combine the ‘cannot
choose’ answers with ‘choose working in a private business’3. Thus, the category ‘choosing to
work for the public sector’ reflects a clear and deliberate choice.
Figure 1 shows how sectoral preference varies across countries: with over 80% in
Cyprus and over 60% in Slovenia opting for the public sector, but below 20% in New
Zealand, Sweden and Denmark. This is in line with Norris’s (2003) findings based on the
ISSP 1997 survey.
Figure 1: Sector preference, by country (N=15,222)
Independent variables at the individual level
Socio-demographic variables. Earlier studies have shown that work values are related to
gender (Tolbert & Moen, 1998), age (Kooij et al., 2011) and educational level (Rosseel,
14
1985). Age was recoded into three categories. Education was measured as the number of
years of continuous full-time education undertaken by the respondent. In the case of
respondents still in education (e.g. students who also work for pay at the same time), this
value was recoded as the mean years of education of respondents in that country who had
completed a similar education level.
Sector of employment. The analysis was limited to respondents currently in paid employment.
Sector of employment was recoded as a dichotomous public/private variable. Respondents
who indicated that they worked for the government or for a publicly owned company or
nationalized industry4 were coded as working in the public sector; those working for a private
firm and the self-employed were coded as working in the private sector.
Income was measured by asking for the respondents’ average net monthly income. Within
each country, we divided the incomes into three roughly equal-sized groups (low, medium
and high incomes), to establish the respondent’s position relative to the overall income
distribution in the country. These within-country categorizations were then transformed into
cross-national income dummies (low, medium and high incomes). This means that a
respondent’s income category is relative to the income distribution in their country. Income
was not corrected for family situation or working hours.
Work values were measured using eight items on a five-point Likert scale (ranging from very
important to not important at all). These were of the form, ‘For each of the following, please
tick one box to show how important you personally think it is in a job’. A factor analysis with
oblimin rotation (to allow the factors to correlate) returns three factors which we label
extrinsic work values, intrinsic work values and public service motivation, in line with
15
Westover and Taylor’s analysis of workplace attributes and job satisfaction using the same
data (2011). Factor scores were used in the analysis.
Table 1: Factor analysis on work values items (Pattern matrix)
Personally important in job Intrinsic work values Extrinsic work values Public service motivation
Job security -.227 .732 -.165
High income .105 .831 .178
Opportunities for advancement .237 .634 -.082
An interesting job .557 .098 -.178
Work independently .756 -.072 -.128
Help other people .097 -.025 -.875
A job useful to society .007 .040 -.888
Decide time of work .751 .035 .105
Oblimin rotation with Kaiser Normalization. The three dimensions account for 63.3% of the common variance in the initial
correlation matrix
Country-level independent variables
At the country level, we distinguish three variables: public sector wage premium (Hypothesis
1), economic conditions (measured by two variables Hypothesis 2) and internal labour market
(Hypothesis 3).
Public sector wage premium. Unfortunately, there are no good, comparable cross-national
data that could be used to measure the extent of the wage gap in all countries included in our
sample. Consequently, we constructed our own public sector wage premium indicator to
measure the gap between public and private sector wages. We used data from ISSP 2005 to
compare mean respondent incomes, using ANOVA to control for age, education and gender.
Taylor and Taylor (2011) used a similar method to estimate the ‘efficiency wages’ of 15
countries included in the ISSP 2005 survey (but see also Llorens and Stazyk’s (2011)
approach). We limited our analysis to those respondents currently in paid employment who
had finished their education. This resulted in an indicator comparing mean public salaries to
mean private salaries in a country, as shown in Figure 2.
16
Figure 2: Mean country public sector wage as ratio of mean private sector wage
Figure 2 shows that public wages tend to be lower than private ones in most Central and
Eastern European countries, and in Scandinavian countries. However, there is a public sector
wage premium in many Southern European countries. The largest public sector wage
premiums found were in Germany and Japan.
We use two indicators for the economic condition of a country having hypothesized that
adverse economic conditions will increase preferences for public sector employment:
GDP per capita in 2005 obtained from World Bank National Accounts Data and based on
purchasing power parity in international $. A high GDP is seen as indicating good economic
conditions.
Unemployment rate. High unemployment is indicative of adverse economic conditions. We
use 2005 unemployment rates as provided by the International Labour Organization.
17
As a final variable, we add a variable reflecting whether a country has a career-based or a
position-based system of public employment. This distinction is ideal typical. In reality, one
will see various systems for different categories of public sector workers. Top public officials
may, for instance, be appointed on short-term contracts, whereas general civil servants are
recruited into a career-based system. We therefore categorized countries as having a career-
based, a position-based or a hybrid system of public sector employment. This categorization
was based on a combination of five existing studies that focus on various subsets of countries.
The main source for constructing this variable is the OECD’s Government at a Glance 2009,
in which selected countries were categorized by the type of recruitment system used in central
government in 2005. Other sources were: Hammerschmid et al.’s study of HR systems in the
EU27 (2007); a World Bank study on the Senior Public Service (2005); the Quality of
Government Institute’s Expert Survey with a variable on the closedness of recruitment
(Teorell et al., 2011); and a EUPAN-commissioned study on senior civil servants in Europe
(Kuperus & Rode, 2008). In general, there was considerable consensus among the sources. If
no consensus emerged, countries were categorized as having a hybrid system.
Analysis
A multilevel logistic model is appropriate because the dependent variable is binary. We use
Stata for our analysis of unit-specific models and consequently report odds ratios. All Level-2
variables have been mean centred. Error! Reference source not found. reports the
descriptive statistics. Correlations between country-level variables are all lower than .45.
Table 2: Descriptive statistics
N Mean SD Min Max
Dependent variable
Sector preference (0=private; 1=public) 15,222 0.37 0.48 0.00 1.00
18
Level-1 Individual Variables
Sex (0=male; 1=female) 15,222 0.49 0.50 0.00 1.00
Age low (under 35 years) 15,222 0.32 0.47 0.00 1.00
Age medium (35-54 years) 15,222 0.53 0.50 0.00 1.00
Age high (55 years and older) 15,222 0.15 0.36
Education (years) 15,222 13.14 3.61 0.00 44.00
Current work sector (0=private; 1=public) 15,222 0.31 0.46 0.00 1.00
Income (low) 15,222 0.19 0.39 0.00 1.00
Income (medium) 15,222 0.36 0.48 0.00 1.00
Income (high) 15,222 0.45 0.50 0.00 1.00
Extrinsic work values 15,222 -0.03 1.00 -5.48 1.71
Intrinsic work values 15,222 -0.02 1.00 -5.31 2.58
Public service motivation 15,222 -0.01 1.00 -4.77 2.24
Level-2 Country Level Variables
GDP per capita 26 27,886 9,084 9,818 47,305
Unemployment5 26 6.59 2.23 3.50 11.10
Public sector wage premium 26 1.01 0.12 0.80 1.21
Career-based system 26 0.50 0.51 0.00 1.00
Position-based system 26 0.35 0.49 0.00 1.00
Hybrid system 26 0.15 0.37 0.00 1.00
Unconditional model
We first run the intercept-only, or unconditional, model (Model 0), and subsequently estimate
the extent of the total variance explained on the country level. We find that the chi square
value is significant in the intercept-only model (χ2 = 1499.38, df = 26, p <0.001), meaning
that between-group (country) variance is significantly different from zero, and that therefore
there are cluster effects in the data. In other words, individuals living in the same country are
more similar in terms of public sector preference than individuals are across countries. Having
19
determined this, multilevel techniques are required to explain employment sector preferences
across individuals and countries. The intraclass correlation (ICC) is 0.137, indicating that
almost 14% of the total variance is located on the second (country) level. This proportion is
similar to that reported by Houston (2011) in his multilevel analysis of public service
motivation across 11 nations. Respondents' odds of preferring public sector employment is
0.623. As we have seen in Figure 1, this proportion differs considerably across countries.
We then run a number of models, each looking at a different set of variables. Model 1
adds the basic socio-demographic variables, while Model 2 looks at work values.
Subsequently, we explore country-level predictors: first, the external labour market through
economic factors (unemployment and GDP) in Models 3 and 4 respectively; and then the
public sector wage premium (in Model 5). Then we add the internal labour market
characteristics (in Model 6). In Model 7 we did only incorporate those level-2 predictors that
made a significant contribution to our model fit. This parsimonious approach gives us
sufficient degrees of freedom at level-2 to estimate the effects of country characteristics on
individuals' sector preferences. The findings are presented in the form of odds ratios. Also,
standard errors are reported.
20
Model 0 Model 1-
Socio-
demographics
Model 2:
Socio-
demographics & motivation
Model 3:
GDP per
capita
Model 4:
Unemployment
Model 5:
Public sector
wage premium
Model 6:
Career- vs.
position-based system
Model 7:
Career- vs.
position-based & GDP per
capita
Intercept 0.623***
(0.089)
0.546***
(0.107)
0.511***
(0.095)
0.486***
(0.083)
0.508***
(0.094)
0.504***
(0.093)
0.722***
(0.147)
0.673*
(0.132) Individual-level
Sex (0=male; 1= female) 1.277***
(0.053)
1.241***
(0.052)
1.240***
(0.052)
1.241***
(0.052)
1.241***
(0.052)
1.241***
(0.052)
1.240***
(0.052)
Age low (under 35 years) 0.835*** (0.051)
0.822** (0.051)
0.821** (0.051)
0.822** (0.051)
0.822** (0.051)
0.822** (0.051)
0.821** (0.051)
Age medium (35-54 years) 0.892*
(0.050)
0.900
(0.050)
0.900
(0.050)
0.900
(0.050)
0.900
(0.050)
0.900
(0.050)
0.900
(0.050) Age high (55 years and older) (reference
category)
Low Income 1.290*** (0.075)
1.297*** (0.076)
1.300*** (0.076)
1.300*** (0.076)
1.297*** (0.076)
1.300*** (0.076)
1.300*** (0.076)
Medium Income 1.238***
(0.056)
1.218***
(0.056)
1.220***
(0.056)
1.218***
(0.056)
1.218***
(0.056)
1.219***
(0.056)
1.220***
(0.056) High income (reference category)
Education (years) 0.960***
(0.006)
0.966***
(0.006)
0.966***
(0.006)
0.966***
(0.006)
0.966***
(0.006)
0.966***
(0.006)
0.966***
(0.006) Current employment (0= private; 1=public) 4.844***
(0.213)
4.585***
(0.204)
4.582***
(0.204)
4.584***
(0.204)
4.588***
(0.204)
4.591***
(0.204)
4.589***
(0.204) Extrinsic work values 1.213***
(0.028)
1.211***
(0.028)
1.213***
(0.028)
1.213***
(0.028)
1.211***
(0.028)
1.209***
(0.028)
Intrinsic work values 0.848*** (0.018)
0.848*** (0.018)
0.848*** (0.018)
0.848*** (0.018)
0.848*** (0.018)
0.848*** (0.018)
Public service motivation 1.224***
(0.027)
1.225***
(0.027)
1.225***
(0.027)
1.224***
(0.027)
1.223***
(0.027)
1.223***
(0.027)
Country-level
GDP per capita 0.9999549** (0.000)
0.999971 (0.000)
Unemployment 1.064
(0.078)
Public sector wage premium 3.831
(5.194)
Position-based system 0.354*** (0.101)
0.418** (0.116)
Hybrid system 1.082 0.956
21
(0.407) (0.341)
Career-based system (reference category)
Model fit -2logL 18,470.90 (2) 16,829.03 16,644.98 16,637.97 16,644.26 16,644.02 16,632.93 16,629.34
Change in -2logL - -1,633.87***
(compared to Model 0)
-192.05***
(compared to Model 1)
-7.01**
(compared to Model 2)
+0.72
(compared to Model 2)
+0.96
(compared to Model 2)
-12.05**
(compared to Model 2)
-15.64** / -
3.59 (not sig.) (compared to
Model 2 / 6)
AIC 18,474.90 16,855.03 16,668.98 16,663.97 16,670.26 16,670.02 16,660.93 16,659.34
BIC 18,490.16 16.923.70 16,760.55 16,763.17 16,769.46 16,769.21 16,767.75
16,773.8
Interclass correlation 0.137 0.189 0.170 0.135 0.166 0.165 0.113 0.100
Significance level: *p<0.05, **p<0.01, ***p<0.001; N = 15,222; Number of countries: 26
22
Evaluation of the model fit
Determining the overall traction and strength of a model in logistic multilevel modelling is
not straightforward. Traditional measures such as R2 are ambiguous (Snijders & Bosker,
1999). Further, the intercept variance will change as other variables are added to the model
and, therefore, we cannot use changes in variance to indicate the explanatory power of the
model. What we can do, however, is to look at the decrease in -2 log Likelihood across
models (last row in the table), and evaluate the decrease using the chi-square statistic. If this
decrease is statistically significant across models, then the model improves fit. Adding socio-
demographic variables (in Model 1) and work values (Model 2) to the unconditional model
significantly improves the results. At the country-level, adding GDP per capita and whether a
position-based or a career-based system is in place, leads to significant improvements.
Unemployment and the wage premium variable do not improve the model. Combining the
significant country-level variables in a single model does not improve model fit (the change in
-2logL between models 6 and 7 is not significant). This makes model 6 our final model. The
decline in -2logL is our primary indicator of model fit. This is confirmed by a decrease in
AIC. However, the second information-based criterion, BIC, increases slightly. Nevertheless,
our main evaluation criteria (change in -2logL and AIC) suggest that model 6 is the model
that best fits the data.
23
Findings
As the individual level accounts for most of the variance in employment sector preference, we
will first report the findings at the individual level (level-1), and then those at the country
level (level-2).
Individual-level findings
Table 3 shows odds ratios, which are typically used in logistic regression. Odds ratios
represent the odds that an outcome (preferring public sector employment) will occur under
certain conditions, compared to that outcome happening in the absence of that condition. An
odds ratio of 1 means that both categories have the same odds. Odds ratios that diverge from 1
indicate stronger effects. These effects have to be interpreted though in relation to the unit
used to measure the independent variable.
Odds for females to prefer public sector employment are 1.241. This means that
female respondents are more likely to prefer public sector employment than men do. More
specifically, the odds for females to prefer public sector employment are higher than those for
men. There is also a positive effect of age, meaning that older people are more likely than
younger people to choose public employment. This supports the finding of Lewis and Frank
(2002) for the US that younger birth cohorts are less attracted to the public sector. Lower and
medium income groups also have a stronger preference for public employment than higher
income groups. The more highly educated respondents are also slightly less likely to opt for
public employment. A person’s current sector of employment has, not surprisingly, in all
models6 a very strong relationship with sector preference: respondents working in the public
sector are considerably more likely to prefer public sector employment.
24
Work values also have a significant effect, and adding them to the model improves
model fit significantly. Here, it is found that being extrinsically motivated is associated with a
preference for public employment, which supports Hypothesis 5. Respondents who would opt
for public employment highly value job characteristics such as job security, high income and
good opportunities for advancement. In line with Hypothesis 4, public service motivation is
also associated with favouring public sector employment: respondents who opt for public
employment consider it important to have a job that allows them to help others and that is
useful to society. However, intrinsic work values are related to a preference for working in the
private sector, which is in accordance with Hypothesis 6. In other words, respondents tend not
to associate public sector work with being allowed to work independently, choosing one’s
own working times or having an interesting job.
Country-level findings
The country-level variables we added to the model contribute little to the explanatory power
of the model. Nevertheless, some findings are worthy of note.
One is that a public sector wage premium is not significant in explaining sector choice when
added as a country-level variable. This contradicts Hypothesis 1 and is surprising because it is
at odds with the expectations derived from the economic literature discussed earlier. It is also
unexpected given the importance of the extrinsic work values and income effects found at the
individual level.
The nature of the public internal labour market does have a significant effect. Results
are in accordance with Hypothesis 3: in countries with a position-based system, public sector
employment is seen as less attractive than in countries with a career-based system.
25
There is mixed evidence for Hypothesis 2. Firstly, GDP has a significant effect, and its
direction suggests that respondents are more likely to prefer working for the government or
the civil service in countries with a low GDP7. To test the robustness of this finding, GDP has
been replaced with GDP growth (3 year average). This variable, however, was not significant.
The effect of unemployment level, although in the expected direction, is not significant. We
thus have only limited support for the link between adverse economic conditions and a
preference for public sector employment. When all the significant country-level variables are
added together, the effect of GDP per capita (just) disappears while the effect of the internal
labour market remains significant. On adding GDP per capita, the effect of the internal labour
market decreases, but model 7 does not fit better than model 6 (which only has the internal
labour market variable) No evidence was found for the existence of significant cross-level
interactions. Because labour market conditions and extrinsic motives may interact (e.g.
extrinsic motives may become more important under bad economic conditions), we have
explicitly tested cross-level interactions between extrinsic motives and GDP, unemployment,
and internal labour market. None of these interactions terms were significant (not included in
the table).
Limitations
This study has a number of data limitations. First, it assumes cross-national measurement
equivalence of the three work-value factors used in the analysis (see Jilke, Meuleman and Van
de Walle, 2013). Although this equivalence was not tested explicitly in this paper, other
researchers using the ISSP 1997 data found modest equivalence across nations for the
intrinsic and extrinsic motivation scales (Huang & van de Vliert, 2003). As in other studies,
and supported by our factor analysis, we included both income and security within the
extrinsic dimension. However, it could be that the effects on sector preference of both these
26
aspects of extrinsic motivation differ across countries. Second, the analysis is based on the
work values scale used in the ISSP. Unlike most research, which measures PSM with an
elaborate and multidimensional scale, the ISSP does not measure PSM as such. Using such a
scale may return a more differentiated view of the impact of PSM. Third, the public sector
wage premium variable is rather crude, lacking comprehensive (sub)sector comparisons
across countries. Fourth, interactions between the country-level variables may exist but cannot
be properly tested statistically because of the relatively small number of countries in the
model. Nevertheless, we have seen that only 14% of the variance in the data is attributable to
the country level. Fifth, the dependent variable is a single item construct and not a scale.
Further concerns are raised by the country-level variation in the proportion of ‘no preference’
answers. These limitations are not uncommon in international comparative research. Finally,
there may exist endogeneity because the current sector of employment may also affect other
preferences and values. A split sample estimation of the final model for public and private
sector workers separately did not indicate substantial differences for work values between
these groups, or with regard to the impact of career versus position system. But there are some
differences regarding income, age, and country level GDP.
Conclusion
This paper started with the observation by Lewis and Frank (2002) that becoming a public
employee involves both choice and chance. We have argued in this paper that public sector
choice depends on both individual characteristics as well as country dependent opportunities
provided by internal and external labour market conditions.
Conceptualized in this way, individual characteristics is shown to be a more important
determinant of public sector employment preference than structural opportunities. According
to our analysis, only 14% of the variance of public sector preference resides on the country
27
level. Moreover, the labour market variables included in our multilevel model were only
partly successful in explaining this variance. This leads to the conclusion that individual
characteristics are a more important determinant of public sector employment than country
level labour market conditions.
Within these individual characteristics, work values are important. Our three
hypotheses in this respect are confirmed, which supports the observation of Delfgaauw and
Dur (2008) that two types of workers (‘dedicated’ and ‘lazy’ ones) are especially attracted to
the public sector. Although these findings are not new, we want to stress them because current
research on PSM seems to underestimate the importance of extrinsic motivation for public
sector employment. Therefore, future research should take into account the fact that people do
not only want to work in the public sector to serve the public good, but that factors such as
money or job security also play a role.
Individual characteristics are also relevant. Older employees, women and the less
highly educated have a stronger preference for public sector employment. . However, we see
that respondents in higher income groups tend to prefer private sector employment. This
probably has to do with a lower wage spread in the public sector and may be associated with
lower probabilities of gaining highly paid public employment (Borjas, 2002). It must be
stressed that these effects are independent of country-level characteristics and are thus
independent from characteristics of the internal and external labour market. Nevertheless, on
the individual level the opportunity structure could still play an important role in determining
public sector employment preference. For instance, it could be that people are not so much
affected by the general labour market conditions in a country, but by specific economic
conditions which for instance only affect the lower educated or younger workers. Further
research into this relation between specific economic conditions and individual characteristics
is therefore needed. Although the effect of country-level characteristics on public sector
28
employment preference is limited, there nevertheless is such an effect. In accordance with
hypothesis 2 a country’s GDP has a minor effect (but level of unemployment does not), but
this effect disappears when internal labour market characteristics are added. In contrast to
hypothesis 1, we did not find an effect of the public sector wage premium on sector
preference. This rather limited support for the effect of external labour market conditions is
surprising.
The most interesting finding at the country level is the confirmation of hypothesis 3,
which states that the public sector is a more preferred sector of employment when it has a
career-based system of employment. This is possibly the result of the fact that the career-
based system is perceived as more attractive because it provides greater security. A relevant
next research question is what happens if the public sector internal labour market changes
from a career-based into a position-based system. Based on our findings, one would expect
that public sector employment preference will decline in such a case.
As outlined above, country level characteristics play only a limited role in the
explanation of public sector employment preference. Moreover, the variables we have used to
explain this country effect have only a limited explanatory power. Obviously, further research
is needed to interpret this country effect more fully. Institutional variables (as Houston, 2011
used to look at welfare regimes) and cultural variables at the country-level are obvious
candidates.
Our paper has extended current research on public sector employment preferences by
including a large number of countries in its analysis and by including individual- as well as
country-level variables. In spite of the discipline’s current focus on PSM, it re-emphasizes the
importance of also looking at extrinsic factors when analyzing public sector employment, and
stresses the need to also consider structural factors which affect the opportunities people have.
29
References
Adamchik, V. & A.S. Bedi (2000). Wage Differentials between the Public and the Private Sectors: Evidence
from an Economy in Transition. Labor Economics, 7(2), 203-224.
Bender, K. A. (1998). The Central Government-Private Sector Wage Differential. Journal of Economic Surveys,
12(2), 177–220.
Borjas, G.J. (2002). The Wage Structure and the Sorting of Workers into the Public Sector. NBER Working
Paper 9313. Cambridge: National Bureau of Economic Research.
Boudarbat, B. (2008). Job-Search Strategies and the Unemployment of University Graduates in Morocco. 14, 15-
33
Buelens, M. & H. Van den Broeck (2007). An Analysis of Differences in Work Motivation between Public and
Private Sector Organizations. Public Administration Review, 67(1): 65–74.
Christensen, R. K. and B. E. Wright (2011). The Effects of Public Service Motivation on Job Choice Decisions:
Disentangling the Contributions of Person-Organization Fit and Person-Job Fit. Journal of Public
Administration Research and Theory, 21 (4): 723-743
Crewson, P.E. (1997). Public-Service Motivation: Building Empirical Evidence of Incidence and Effect. Journal
of Public Administration Research and Theory, 7(4): 499 – 518 .
Delfgaauw, J. & R. Dur (2008). Incentives and Workers’ Motivation in the Public Sector. The Economic
Journal, 118 (525): 171–191
Doeringer, P. & M.J. Piore (1971). Internal Labor Markets and Manpower Adjustment. New York: D.C. Heath
and Company.
Downs, A. (1967). Inside Bureaucracy. Boston: Little Brown.
Gornick, J.C, & J.A. Jacobs (1998). Gender, the welfare state, and public employment: a comparative study of
seven industrialized countries. American Sociological Review, 63(5), 688-710
Groeneveld, S.M., B. Steijn, & P. Van der Parre (2009). Joining the Dutch Civil Service. Influencing Motives in
a Changing Economic Context. Public Management Review, 11(2), 173-189.
Gunderson, M. (1979). Earnings Differentials between the Public and Private Sectors. The Canadian Journal of
Economics, 12(2): 228-242.
Guo, G, & H. Zhao (2000). Multilevel Modeling for Binary Data. Annual Review of Sociology, 6: 441–62
30
Hammerschmid, G., R.E. Meyer, & C. Demmke (2007). ‘Public Administration Modernization: Common
Reform Trends or Different Paths and National Understandings in the EU Countries’. In: K. Schedler &
I. Proeller, I. (eds) Cultural Aspects of Public Management Reform, Emerald Group Publishing Limited,
pp. 145 – 169
Herzberg, F. (1966). Work and the Nature of Man. Ohio: World Publishing
Houston, D. J. (2000). Public-service motivation: A multivariate test. Journal of public administration research
and theory, 10(4), 713-728.
Houston, D.J. (2011). Implications of Occupational Locus and Focus for Public Service Motivation: Attitudes
Toward Work Motives Across Nations. Public Administration Review, 71(5): 761-771.
Huang, X. & E. van de Vliert (2003). Where Intrinsic Job Satisfaction Fails to Work: National Moderators of
Intrinsic Motivation. Journal of Organizational Behavior, 24(2), 159-179.
ISSP - International Social Survey Programme (2005a). Study Descriptions for Participating countries,
Mannheim: GESIS.
ISSP - International Social Survey Programme (2005b). Work Orientation III dataset, Mannheim: GESIS.
Jilke, S., Meuleman, B. & S. Van de Walle (2013). We need to compare, but how? Measurement equivalence in
comparative public administration. Paper presented at the PMR Conference, Madison, Wisconsin.
Jin, M. (2013a). Comparing Employed and Unemployed Workers' Job Motivations for Sector Choice in East
Asia: Does Employment Status Matter? International Journal of Public Administration, 36(5), 305-319.
Jin, M. (2013b). Public Service Motivation: A Cross-Country Study. International Journal of Public
Administration, 36(5), 331-343.
Kjeldsen, A. M. and C. B. Jacobsen (2013). Public Service Motivation and Employment Sector: Attraction or
Socialization? Journal of Public Administration Research & Theory, 23(4): 899-926
Konrad, A.M., J.E. Ritchie, P. Lieb and E. Corrigal (2000). Sex Differences and Similarities in Job Attribute
Preferences. A Meta-Analysis. Psychological Bulletin, 126(4), 593-641.
Kooij, D. T. A. M., A. H. De Lange, P. G. W Jansen,. R. Kanfer. & J. S. E. Dikkers, (2011). Age and Work-
Related Motives: Results of a Meta-analysis. Journal of Organizational Behaviour, 32(2), 197–225.
Kuperus, H. & A. Rode (2008). Top public managers in Europe: Management and working conditions of the
senior civil servants in the European Union member states. Study commissioned by the French EU
presidency. Maastricht: EIPA.
Latham, G.P. (2007) Work Motivation. History, Theory, Research and Practice. Thousand Oaks: Sage.
31
Lewis, G.B. & S.A. Frank (2002). Who Wants to Work for the Government? Public Administration Review
62(4), 395-404.
Llorens, J.J. & E.C. Stazyk (2011). How Important Are Competitive Wages? Exploring the Impact of Relative
Wage Rates on Employee Turnover in State Government. Review of Public Personnel Administration
31(2), 111-127.
Niskanen, W. A. (1971). Bureaucracy and Representative Government, Chicago: Aldine-Atherton.
Norris, P. (2003). ‘Is There Still a Public Service Ethos? Work Values, Experience, and Job Satisfaction among
Government Workers’. In J. D. Donahue and J. S. Nye, Jr (eds) For the People: Can We Fix Public
Service? Washington, DC: Brookings Institution Press.
OECD, (2005). Modernising Government. The Way Forward. Paris
OECD (2009). Government at a Glance. Paris: OECD
Perry, J. L., & A. Hondeghem (Eds.). (2008). Motivation in Public Management. The Call of Public Service.
Oxford: Oxford University Press.
Perry, J.L. & L.R. Wise (1990). The Motivational Bases of Public Service. Public Administration Review,
May/June: 367-373.
Riphahn, R.T. & Serfling, O. (2005). Item Non-response on Income and Wealth Questions. Empirical
Economics, 30(2), 521-538
Ritz, A. & C. Waldner (2011). Competing for Future Leaders. A Study of Attractiveness of Public Sector
Organizations to Potential Job Applicants. Review of Public Personnel Administration, 31(3), 291-316
Rosseel, E. (1985). Work Ethic and Orientations to Work of the Young Generations: The Impact of Educational
Level. Social Indicators Research, 17(2), 171-87.
Scholz, E., J. Harkness & T. Faaß (2008). ISSP study monitoring 2005, report to the ISSP General Assembly on
monitoring work undertaken for the ISSP. GESIS-Methodenberichte, 2008/04, Mannheim: GESIS
Simon, H. A. (1955). A behavioral model of rational choice. The Quarterly Journal of Economics, 69(1), 99-
118.
Snijders, T.A.B. & R.J. Bosker (1999). Multilevel Analysis: An Introduction to Basic and Advanced Multilevel
Modeling, London: Sage.
Stegmueller, D. (2013). How Many Countries for Multilevel Modeling? A Monte Carlo Study Comparing
Bayesian and Frequentist Approaches. American Journal of Political Science, 57(3), 748-761.
32
Steijn, B. (2008). Person Environment Fit and Public Service Motivation. International Public Management
Journal, 11(1): 13-27.
Tansel, A. (2005). Public-Private Employment Choice, Wage Differentials, and Gender in Turkey. Economic
Development and Cultural Change, 53(2), 453-477.
Taylor, J. & R. Taylor (2011). Working Hard for More Money or Working Hard to Make a Difference?
Efficiency Wages, Public Service Motivation, and Effort. Review of Public Personnel Administration,
31(1), 67-86
Taylor, J., & J. H Westover (2011). Job satisfaction in the public service. Public Management Review, 8(1): 1-
21.
Teorell, J., C. Dahlström & S. Dahlberg (2011). The QoG Expert Survey Dataset. University of Gothenburg: The
Quality of Government Institute.
Tolbert, P.S., & P. Moen (1998). Men's and Women's Definitions of "Good" Jobs: Similarities and Differences
by Age and Across Time. Work and Occupations, 25(2), 168-194.
Vandenabeele, W. (2008). Government calling: Public Service Motivation as an Element in Selecting
Government as an Employer of Choice. Public Administration, 86(4), 1089-1105
Westover, J.H. & J. Taylor (2010). International Differences in Job Satisfaction: The Effects of Public Service
Motivation, Rewards and Work Relations. International Journal of Productivity and Performance
Management, 59(8): 811 – 828
World Bank (2005). Senior Public Service: High Performing Managers of Government. Washington D.C.:
World Bank,
Wright, B.E. & R.K. Christensen (2010). Public Service Motivation: A Test of the Job Attraction-Selection-
Attrition Model. International Public Management Journal, 13(2), 155-76.
33
1 Some authors use the concept of job attributes preference (Lewis & Frank, 2002), whereas others use work
values (Norris, 2003) or work motives (Groeneveld et al., 2009). These all reflect ‘the extent to which people
desire a variety of specific qualities and outcomes for paid work’ (Konrad et al., 2000). Consequently, we use
‘work values’ in our argument.
2 With the exceptions of Finland and Japan who used 15 and 16 years respectively as the cut-off.
3 A preliminary test with country dummies and with two models, one binomial and one multinomial, revealed no
differences in terms of changes in effect signs. We therefore decided not to run a multinomial multilevel model.
4 The latter category was not included in countries where such terminology is not used. An example is the US
where only ‘working for government’ was used.
5 The reported unemployment rates for Mexico appear to be rather low. This is probably more an influence of the
social security system than the labour market.
6 When running the models without current employment sector, both education and low income cease to be
significant, but age and gender effects strengthen. Not surprisingly, public service motivation is also stronger in
the model without current sector of employment.
7GPD per capita has the perhaps surprising value of 0.999. This is a direct result of the high absolute values of
GDP per capita (see Table 2).