The Education-Job Satisfaction Paradox in the Public Sector

Much has been written on employment in the public and private sectors and the relative attributes of each one. To begin with, there exists a vast array of literature on public–private sector wage differentials¹ (Bender, 1998; Mueller, 1998; Birch, 2006; Lucifora & Muers, 2006; Makridis, 2021; Sánchez-Sánchez & Fernández Puente, 2021), as well as job satisfaction (Ghinetti, 2007; Heywood et al., 2002). In addition, there is another area of research on employment in the public and private sectors that focuses on the returns to education that workers obtain when they are employed in both sectors. Most studies are empirical and have been done using surveys of mainly cross-sectional data for specific countries. However, within this line of research, there seems to exist a consensus that returns to education are higher in the private sector in most countries (Montenegro & Patrinos, 2022). That is to say, the return of workers in the private sector of the economy to further education is higher than for those working in the public sector (Psacharopoulos & Patrinos, 2018). According to Budria (2010), this is a worldwide phenomenon, although some authors have recently found exceptions to this stylized fact in a handful of countries such as India (Chen et al., 2022) and Turkey (Patrinos et al., 2019). Psacharopoulos (1979) explains that wages may exceed productivity in the public but not in the private sector. Furthermore, the overpayment of government workers with respect to their productivity can be mostly found at the lower end of the earnings distribution (Siminski, 2013). The public sector offers (relatively) high wages to unskilled workers and (relatively) low wages to the high-skilled (Budria, 2010). Psacharopoulos (1983) concludes that less qualified labor is better treated in the public than in the private sector.

In this paper, we analyze the self-reported levels of satisfaction with working conditions and pay of public and private sector workers who have attained different levels of education. In the context of Herzberg’s two-factor theory (Herzberg et al., 1959), our focus on satisfaction with working conditions and pay could be framed as an analysis of the hygiene or extrinsic factors of job satisfaction. In particular, we compare and rank the satisfaction with working conditions and pay of European workers who are employed in the private and public sectors. We intend to shed some new light on this topic by analyzing self-reported job satisfaction in a large sample of 43,850 workers from 35 countries, which is the last available wave of the European Working Conditions Survey (EWCS).

Since data on job satisfaction provided by surveys is based on subjective replies to an interviewer’s questions in some sort of categorical fashion, in our analysis we use the Balanced Worth (BW) procedure. The BW was developed from the concept of Net Difference (Lieberson, 1976) and incorporates the economic concepts of value and equilibrium. This methodology has been specifically designed to compare distributions of categorical data (Herrero & Villar, 2013, 2018) to analyze inequality among different population groups.

In recent years, the BW has been successfully used in different areas of the Social Sciences and has proven to be a suitable mechanism for researching happiness and well-being (Herrero & Villar, 2018). In particular, the BW procedure has been applied in studies on human capital and health (Herrero & Villar, 2013), education (Herrero et al., 2014), corporate responsibility (Gallén & Peraita, 2015), nationalist identity (Torregrosa, 2015), research (Albarrán et al., 2017), life satisfaction (Herrero & Villar, 2018), labor market (Herrero & Villar, 2019), happiness (Ravina-Ripoll et al., 2020), COVID-19 (Herrero & Villar, 2020), and income distributions (Herrero & Villar, 2021).

The paper proceeds as follows. We devote the next two sections to briefly summarizing the methodology and describing the data set. The following sections display our results regarding satisfaction with working conditions and payment in both sectors. The final section summarizes the conclusions of our empirical analysis.

Questions regarding job satisfaction are quite similar across surveys, as they ask workers to provide an overall assessment of their job satisfaction and working conditions, offering them several categorical responses. Score methods are commonly used when dealing with ordered categorical data, which consist of giving weights to each of the categories in an appropriate order and evaluating the groups according to their mean values. The problem with these methods is that the choice of weights introduces an exogenous cardinalization in the original information, leading to arbitrary results.

An alternative method is based on the concept of stochastic dominance. This procedure is robust, and only relies on categorical information, without using external weights. But this method provides neither a complete order nor cardinal information about the relative goodness of the distributions (Herrero & Villar, 2018). To amend these problems, Lieberson (1976) posed a procedure to compare any pair of distributions, based on the concept of probabilistic dominance. That is, how likely it is that an individual from one group chosen at random will belong to a higher category than an individual from any other group chosen at random.

Following Herrero and Villar (2013, 2018), let us consider the problem of comparing the distributions of g different groups over a set of k categories. The following table shows the information that we can draw from our data set, where the first row represents different categories. These categories are ordered from best to worst and the first column corresponds to the different groups and \({x}_{ij}\) denotes the relative frequency in which category j appears in group i, where \({\sum }_{j=1}^{k}{x}_{ij}=1, \forall i=\mathrm{1,2},\dots ,g.\)

Our comparison problem consists of evaluating the relative dominance of these frequencies among them. This is an extension of the concept of Net Difference introduced by Lieberson (1976) for the case of two groups. Hence, letbe the probability that a randomly chosen individual from group i belongs to a better category than a randomly chosen individual from group j. On the other hand, the probability that the randomly chosen individuals from both groups belong to the same category iswhere \({e}_{ij}={e}_{ji}\) and \({p}_{ij}+{p}_{ji}+{e}_{ij}=1\). The procedure posed by Herrero and Villar (2018) consists of choosing one individual from each group at random and comparing its category with that of a randomly chosen individual from all the other groups. In the pairwise comparison, when a randomly chosen individual of group i belongs to a higher category than a randomly chosen individual of group j, we say that the distribution of group i dominates the distribution of group j. If both individuals belong to the same category (with probability \({e}_{ij}\)), each group is declared dominant with a probability of 0.5. Hence, the probability of group i being declared not worse than group j is \({{q}_{ij}=p}_{ij}+\frac{1}{2}{e}_{ij}\). At this point, Herrero and Villar (2013, 2018) introduce the concept of the worth of group i, given by \({v}_{i}\), which could be interpreted as the value of not-being dominated by any group. To avoid problems of cycles that may arise in the pairwise comparisons between groups, Herrero and Villar (2018) propose to determine the groups’ worth by equating the expected worth of group i not being dominated by the rest of the groups with the expected worth of the rest of the groups not being dominated by group i. That is,

Extending this criterion to all the groups, we can write the following homogeneous linear system \(Mv=0\), whereand \({v}^{^{\prime}}=\left({v}_{1},{v}_{2},\dots ,{v}_{g}\right)\). Since the sum of each column of M equals zero, the vectors that conform it are linearly dependent, thus M is singular, and the homogeneous system has one degree of freedom. To amend the indetermination of the solution, Herrero and Villar (2018) add an additional equation associated with the normalization of the worth vector, say\(\sum_{1}^{g}{v}_{i}=g.\)² Herrero and Villar (2018) prove that the solution to this system exists, is unique, and is given by \({v}^{*}=\left({v}_{1}^{*},{v}_{2}^{*},\dots ,{v}_{g}^{*}\right)\) such that \(\sum_{1}^{g}{v}_{h}^{*}=g\), and

We call \({v}^{*}\) the BW vector. The main contribution of the BW is that, according to the probability of the domination criterion, it ranks the different groups using a cardinal measure. In contrast to the score methods previously used to compare categorical distributions, the BW procedure does not only rank the dominance of a group, but it also provides endogenous information about the intensity of such dominance using the cardinality of its components. The BW has other desirable properties, such as anonymity, symmetry, and monotonicity (see Herrero & Villar, 2013, 2018).

The website of the Instituto Valenciano de Investigaciones Económicas (IVIE) provides the algorithm to calculate the BWV freely (http://​www.​ivie.​es/​balanced-worth/​).

The data set that we use is the sixth wave of the European Working Conditions Survey (EWCS), which was conducted in 2015 by the European Foundation for the Improvement of Living and Working Conditions (Eurofound) and interviewed 43,850 workers whose answers provide an overview of working conditions in Europe. Using the cross-national weights given by the EWCS ensures that each country is represented in proportion to the size of its in-work population.

The EWCS is based on a questionnaire that was administered face-to-face to a random sample of 'persons in employment', both employees and self-employed, who work in the private sector, the public sector, or at other alternatives, such as joint private–public organizations or companies, the not-for-profit sector or non-governmental organizations. Among them, 68.47% of surveyed individuals declare to work in the private sector, while 23.06% declare to work in the public sector. For our purposes, we focus on the private and the public sectors, where 91.53% of workers are employed.

Regarding worker’s education, the EWCS classifies the attained educational levels of surveyed individuals according to the categories of the International Standard Classification of Education (ISCED). These categories and their percentages in the EWCS are the following: childhood education (0.58%), primary education (4.75%), lower secondary education (13.37%), upper secondary education (41.55%), post-secondary-non-tertiary education (7.02%), short-cycle tertiary education (9.42%), bachelor or equivalent (13.05%), master or equivalent (9.29%), and doctorate or equivalent (0.95%). For the sake of representativeness, we consider individuals that report levels of education from primary to master or equivalent.³ Moreover, to keep a unimodal and symmetrical distribution, we gather the groups of post-secondary-non-tertiary education and short-cycle tertiary education in one group called post-secondary-tertiary education (16.43%).

Let us first consider the educational attainment of workers employed in the private and public sectors using the BW procedure. Table 1 shows the distributions of groups for the private and the public sector according to the level of education of their workers and their corresponding Balance Worth Value (BWV).

The BW components in Table 1 indicate that a randomly chosen individual from the public sector is more likely to have a higher education level than a randomly chosen individual from the private sector. In other words, the concentration of educated workers is considerably higher in the public than in the private sector across European countries, as the frequency distributions and the subsequent BW calculation show. Notice the difference between both components of the BW vector, which suggests a high level of inequality in terms of the educational attainment of workers in both sectors. This will be our first finding.

Moreover, we have analyzed whether this result holds in each of the 35 countries of the EWCS. We have also calculated the difference between the BW components and ordered countries according to that difference. These calculations are shown in the Appendix. Figure 1 illustrates the differences in the BW components for the public and the private sectors for each country. As we can observe, in all the countries workers’ educational attainment is higher in the public than in the private sector with no exception. Nonetheless, there are differences across countries. The country with greater inequality in both sectors regarding the educational attainment of its workers is Albania, followed by Turkey and Greece. At the other end of the spectrum, we find The Netherlands, where the educational achievement of workers in the public and the private sector is more similar, remaining higher in the public than in the private sector, as in any other country in this data set.

Working in the public sector has been related to higher levels of job satisfaction (Ghinetti, 2007; Heywood et al., 2002). Ghinetti (2007) indicates that public employees differ from private employees in the way they evaluate satisfaction with job security, consideration by colleagues, and safety and health job features. Sánchez-Sánchez and Fernández Puente (2021) conclude that public sector employees are more satisfied in terms of stability but not in terms of wages. According to Clark and Postel-Vinay (2009), permanent public jobs are perceived to be by and large insulated from labor market fluctuations and this may be the reason educated workers prefer to be employed in the public sector of the economy.

We now proceed to analyze and compare the levels of satisfaction with working conditions and payment reported by workers with different levels of educational attainment employed in both sectors.

Understanding how to assess returns to education in the public and the private sectors of the economy should be crucial to implement policy decisions in public sector administration. Just as policymakers can learn from estimates of the monetary benefits of education, we argue that workers’ perceptions of returns to education in both sectors should also be considered. However, while there is an established consensus that education is more profitable -in terms of earnings- in the private than in the public sector, relatively little is known about satisfaction with working conditions and pay in both sectors. In this paper, we have analyzed both issues for public and private sector employees with different educational attainment by applying the BW procedure to a large sample of European workers. We conclude that: