Credential and Skill Mismatches Among Tertiary Graduates: The effect of labour market institutions on the differences between fields of study in 18 countries

Starting with Freeman's pioneering work (1976), in recent decades, studies of school-to-work transition have increasingly focused on analysing the determinants and consequences of overeducation: this phenomenon is considered a relevant indicator of an inefficient allocation of skilled labour force on labour market. General agreement exists on the costly consequences of overeducation which are borne not only by individuals (Duncan and Hoffman 1981; Hartog 2000), but also by firms (Tsang 1987) and by the whole economy (Maier et al.2003; McGuinness 2006).

Following educational expansion in developed countries, recent research on overeducation has mainly focused on the incidence of the phenomenon among tertiary graduates: the risk of overeducation is found to be smaller among graduates from technical and scientific fields than among graduates from the Humanities and Social Sciences (Barone and Ortiz 2011). Another well-established result concerns the association between graduates’ risk of overeducation and cyclical and structural dimensions of both the demand and the supply of skilled workers (Hannan et al.1997; Ortiz 2006; Verhaest and van der Velden 2013). What has been seldom investigated, however, is whether these factors affect not only the overall incidence of the phenomenon among tertiary graduates, but also its distribution among different fields of study.

The main purpose of this work is to fill this gap by analysing the influence of labour market institutions on the differential risk of overeducation among graduates from different fields of study. In particular, data from two comparative surveys – REFLEX and HEGESCO – are employed to assess the impact of three elements that are likely to affect not only the incidence but especially the distribution of overeducation among graduates: the level of employment protection, the regulation of access to the so-called liberal professions, and the propensity of welfare states to hire skilled workers.

Before presenting the theoretical framework and the hypotheses that drive the subsequent analyses, in the next section a few words are spent to discuss how overeducation can be defined and measured, and the rationale of distinguishing between credential and skill mismatches is addressed. Indeed, as outlined in the following, this distinction has important theoretical and methodological implications, widely discussed in the literature, and the empirical results that will be presented confirm its heuristic value.

A widespread discussion around the definition and measurement of overeducation has been developed in last decades. Generally speaking, individuals are defined as overeducated if their education levels exceed job requirements. However, while measuring education levels is quite straightforward, assessing job requirements is complex. A review of the existing literature suggests three feasible means of measuring them and, consequently, overeducation. The first is a self-assessment (SA) of the match between the attained level of education and that required to obtain (Green and Zhu 2010) or perform (Brynin et al.2006) the job. The second approach determines job requirements on the basis of surveys conducted among job analysts (JA) and then compares them with the education levels of job incumbents (Rumberger 1987). The third solution exploits information on realised matches (RM): required levels of education are inferred from the mean years of schooling completed by workers in each occupational category, and individuals are considered overeducated if their levels of education are one standard deviation or more above this average (Verdugo and Verdugo, 1989).

The pros and cons of these three measurement methods are well established. Therefore, here, we provide only a brief summary of the main critiques raised in the literature.¹ SA strongly depends on the formulation of questions (Green et al.1999) and on the contingent factors considered by individuals answering them, which may pertain to the formal requirements to obtain a job and/or to the level of education needed to perform it (Dolton and Vignoles 2000). JA is static in nature, because dictionaries of occupations are seldom updated, it tends to ignore the variability of skill requirements within occupational categories (van der Velden and van Smoorenburg 1997), and it is only available for a small number of countries. Finally, RM is likely to be significantly affected by the presence of outliers and by the inflation of educational credentials (Mendes de Oliveira et al.2000).

A preference for a given method typically depends on data availability, and it is common practice to use these measures interchangeably in order to bypass data constraints. However, it has been suggested that different indicators of overeducation may measure different – though potentially correlated – phenomena (Allen and de Weert 2007). Indeed, an increasing number of studies report that using different indicators leads to significant discrepancies in the estimates of the incidence of and economic returns to overeducation, a weak correlation among these estimates, and different patterns of determinants (Groot and Maassen van den Brink 2000; Rubb 2003).

This study embraces the idea that it is more informative to use different indicators to catch distinct features of the broader phenomenon of overeducation. Thus, in the following analyses, a distinction is drawn between credential and skill mismatches: the first term refers to the mismatch between formal educational credentials and job requirements, and the second term refers to the mismatch between the competencies acquired through education and the skills needed to perform a job (Allen and van der Velden 2001). This distinction is not minor. Credential mismatch does not necessarily indicate a lack of demand for graduate skills per se. Increasing demand for advanced competencies may be the result of technological and organisational change that require more advanced skills to perform a job.

In the following, both credential and skill mismatches are defined on the basis of SA, since, in a comparative perspective, this is the only feasible way to distinguish between the two phenomena. However, the possibility given by the data sources to differentiate between credential and skill requirements of occupations is fully exploited. In the first case, respondents are asked which type of education is most appropriate for their job while, in the second case, they are asked to what extent they use their knowledge and skills in performing their job.

It has been widely documented that field of study matters in determining graduates’ occupational opportunities (Kim and Kim 2003; Reimer et al. 2008). Graduates from more generalist fields of study are exposed to the highest risks of mismatch, whereas overeducation is less likely in job-oriented technical and scientific fields (Wolbers 2003; Ortiz and Kucel 2008). These variations in the incidence of overeducation have been interpreted in light of theoretical approaches that differ with respect to the mechanisms they assume connecting education to labour market opportunities (van de Werfhorst 2011).

First, in a human capital framework, education is considered as an indicator of productive skills (Becker 1964). If we assume that a worker's productivity is a function of both the amount and type of skills acquired through education, and that graduates from occupation-specific fields are generally more productive, we can expect that overeducation is less widespread in technical and scientific fields than in the generalist fields of the Humanities and Social Sciences (Reimer et al. 2008).

Second, education may be considered as a positional good, i.e. a signal of motivation and ability used by employers to predict the productivity of potential workers in contexts of imperfect information (Spence 1973; Stiglitz 1975). According to this approach, employers prefer graduates from technical and scientific fields because these fields are perceived as more selective.

Third, education may function as a means for social closure. Educational qualifications may serve as instruments of exclusion, as they strictly regulate the access to specific occupations (Collins 1979; Parkin 1979). According to this approach, graduates from several fields – such as Health, Architecture, Engineering, and Law – are expected to have better labour market chances, since the corresponding interest groups are more capable than others of keeping the supply of graduates low. However, as discussed in the following, social closure may also happen through the regulation of the demand for professional services (Weeden 2002), which might counterbalance the positive effect of the requirement of educational credentials to access these professions.

Although most literature on overeducation has assessed which approach best fits the empirical evidence, recent studies have suggested that overeducation is unlikely to be entirely explained by only one of these three models. Instead, scholars tend to focus on the macro conditions under which one mechanism tends to prevail over the others. The institutional context is likely to affect not only the strength of the educational effect on labour market opportunities for job seekers, but also the mechanism through which individual education affects occupational outcomes (van de Werfhorst 2011).

This study focuses on three labour market features that are likely to affect both the overall incidence of credential and skill mismatches among graduates and the patterns of differences between fields of study: the level of employment protection, the extent to which access to some professions is regulated by the requirement of specific credentials, and the role of the welfare state as an employer. These factors may affect the occupational opportunities of graduates, by influencing both the number and the quality of job vacancies and the job screening process.

Scholars have mainly focused on the influence of employment protection on unemployment rates, with ambiguous findings. The stricter the employment protection, the more difficult it is for employers to dismiss workers, but also the lower is employers’ willingness to hire workers (Breen 2005). It remains unclear whether the strictness of employment protection also affects the incidence of overeducation. Although some studies indicate a positive association (Di Pietro 2002), others do not find any significant effect (Verhaest and Van der Velden 2013). It may be expected that employment protection affects credential mismatches among graduates from different fields of study, but not the distribution of skill mismatches (Hypothesis 1). The stricter the employment protection, the more fields of study do affect the risk of credential mismatch. Since dismissing workers in these contexts implies higher costs, employers have a greater incentive to select graduates from fields of study that send more powerful signals of applicants’ abilities and motivations, i.e. scientific and technical fields. In other words, employment protection enhances the signalling value of education rather than its skill content.

The second institutional factor taken into account is the regulation of liberal professions, usually promoted by professional orders. This kind of regulation takes the form of restrictions to the access both to fields of study and, in a subsequent stage, to the corresponding professions. These two kinds of restrictions may have opposite effects on graduates’ labour market prospects (Barone and Schindler 2014). The former (e.g. numerus clausus for access to tertiary education or very high tuition fees) reduces the number of students from the matching fields, thus lowering the risk of oversupply of graduates. Since overeducation, in terms of both credential and skill mismatches, is primarily due to a gap between demand and supply of graduates, reducing the number of graduates could improve their labour market prospects, in terms of both the analysed phenomena. On the contrary, the latter type of regulation (e.g. long periods of postgraduate training or additional selective entry examinations to the professions) constrains the access to these professions also for graduates from the matching fields, thus possibly lowering the connected occupational rewards. Graduates who do not succeed in accessing the corresponding professions (e.g. graduates in law that not succeed in becoming lawyers) are likely to end up in jobs that do not require nor their tertiary degree nor the skills they have acquired at university. Summing up, we can expect the regulation of liberal professions to affect the risk of both credential and skill mismatches for graduates from the corresponding fields of study (Hypothesis 2). However, the direction of these associations strongly depends on the prevalence of one of the two aforementioned types of restrictions.

Finally, the existing literature suggests that the match between workers and jobs may be affected by the capacity of the welfare state to employ skilled workers. As some studies suggest, the risk of overeducation is lower in the public than in the private sector (Wolbers 2003). In the developed countries, substantial shares of graduates are employed in the public sector, particularly in services (i.e. health, social services, and education) and public administration, which means that the public sector can absorb graduates who are at greater risk of overeducation, such as those from the Humanities and Social Sciences. Therefore, the higher the capacity of the public sector to absorb graduates, the smaller the differentials between fields of study will be in terms of both credential and skill mismatches (Hypothesis 3). Because a higher level of education is formally required for several managerial positions in public bureaucracies, we expect that the larger the public sector, the more graduates from generalist fields – such as the Humanities and Social Sciences – will improve their labour market prospects. The public sector generally comprises professions requiring specific competencies, such as education, social and health services. Therefore, we hypothesise that graduates from these fields of study will be at less risk of skill mismatch in countries with a higher rate of public employment.

The analyses are based on data from two comparative surveys conducted on representative samples of graduates, interviewed approximately 5 years after graduation. These studies report detailed comparable information on individuals’ educational careers, occupational status, job experience before and after the conclusion of tertiary education, and family background. The first dataset, Research into Employment and Professional Flexibility (REFLEX), includes 34,347 individuals from 13 European countries (Italy, Spain, France, Austria, Germany, the Netherlands, the United Kingdom, Finland, Norway, the Czech Republic, Portugal, Belgium, and Estonia) and Japan who graduated between 2000 and 2001 and were interviewed in 2005. The second study, Higher Education as a Generator of Strategic Competencies (HEGESCO), was based on the same methodology developed in the REFLEX study and was conducted approximately 3 years later. It includes 8742 graduates of the 2002–2003 academic year from Slovenia, Turkey, Lithuania, Poland, and Hungary. Lithuania is excluded from the analysis, because not all aggregate indicators that are used in this study are available for it. Consequently, this study reports findings for 17 European countries plus Japan. The analysis is restricted to individuals who were employed at the time of the interview. A more detailed description of the employed data is provided in appendix.

Credential and skill mismatches are defined on the basis of an SA of the match between workers’ education and jobs held 5 years after graduation. The indicator of credential mismatch is based on responses to the question ‘What type of education do you feel is most appropriate for this work?’. The possible responses are (a) Ph.D.; (b) other postgraduate qualification; (c) master; (d) bachelor; and (e) lower than higher education. Credential mismatch occurs when individuals indicate that a level lower than higher education is most appropriate for their work. The indicator of skill mismatch is based on responses from graduates to the question: ‘To what extent are your knowledge and skills utilised in your current work?’. The possible answers range from 1 (not at all) to 5 (to a very high extent). Skill mismatch occurs when the response is 2 or below. It is worth noting that, when answering, individuals might refer to skills not acquired through education. However, because the respondents are new entrants in the labour market, their job experience is assumed to play a less relevant role than the competencies acquired in Higher Education.

where y is the dependent variable of interest, X is a vector of characteristics of individual i, C is a vector of characteristics of country c, and U and V are random error terms. Markov Chain Monte Carlo (MCMC) methods are used to compute estimation (Browne and Draper 2006) that provide more accurate results than those derived by classical likelihood-based frequentist methods.²

The baseline model (Model 1) consists solely of an intercept and additional random effects at the country level. In Model 2, the covariates measured at the individual level are added: the distribution of these variables is summarized in Table 1.

The main independent variable is field of study, classified as follows: Education; Art and Humanities; Social Sciences; Business and Administration; Law; Natural and Applied Sciences; Mathematics and Statistics; Engineering and Architecture; and Health. Education is chosen as the reference category since it stays midway between generalist fields (Humanities and Social Sciences) and the other scientific, technical, or highly job-oriented fields. The other included covariates are all the antecedent variables relevant for the outcomes of interest, which are known to affect both the choice of field of study and labour market opportunities. These are gender, age, country of birth, parents’ level of education, type of upper secondary education completed, final examination grade at the end of secondary education, type of tertiary study programme completed (whether it provides or not direct access to doctorate), involvement in further education after graduation, and work experience before and during higher education.

Model 3 includes the variables measured at the country level: their distribution is reported in Table 2.

The first variable is the OECD index ‘Overall Strictness of Employment Protection’, which combines 21 items in three main categories: (i) protection of regular workers against dismissal, (ii) regulation of temporary employment, and (iii) specific requirements for collective dismissals (see Venn 2009 for further details). The possible values range from 0 (least stringent) to 6 (most restrictive). The second variable is the OECD index ‘Professional Services Regulation’, which indicates the extent to which access to some professions is regulated by strict credential requirements; the scale ranges from 0 to 6 (from least to most restrictive). This indicator measures regulatory conditions in professional services, covering entry and conduct regulation in the legal, accounting, engineering, and architectural professions³ (see Conway and Nicoletti 2006 for information about its composition and its advantages and disadvantages).⁴ The reference year for both indexes is the year in which the survey was conducted in each country. Finally, the third variable is employment in general government as a percentage of the labour force, provided by OECD and referring to 2008.

Moreover, some country-level covariates are added to control for both the economic structure and cycle, namely GDP per capita and unemployment rates (World Bank). Finally, we control for tertiary graduation rates (UNESCO). To account for the structure of the educational system, we first included different indicators of the level of stratification (e.g. the age of first selection into educational tracks) as control variables in the models. However, the inclusion of this variable did not affect the pattern of results concerning labour market institutions, nor were its coefficients statistically significant.⁵ Finally, Model 4 includes interactions between country-level variables and field of study to test whether different institutional settings affect the distribution of credential and skill mismatches among graduates from different fields.⁶

The average incidence of both credential and skill mismatches 5 years after graduation across the 18 countries studied is approximately 10%. However, Figure 1 indicates that significant differences exist in the incidence of the phenomena among countries and that the correlation among the two is far from perfect. The highest incidence of credential mismatch is observed in Japan (18.9), followed by Spain (16.8), Hungary (15.4), and UK (14.7); the lowest incidence is recorded in Finland (5.6), Poland (3.9), and Estonia (1.7). The diffusion of skill mismatch differs by country to a significant extent. The highest value is again observed in Japan (21.4), but Poland and Estonia score higher than before (13.8 and 7.1, respectively); the lowest incidence of skill mismatch is observed in Norway (4.5) and Portugal (2.8). The correlation between the two measures at the country level is 0.69. However, their average correlation at the individual level is only 0.36, with great variability among countries, spanning from 0.07 in Estonia to 0.50 in Japan (see Table A2 in appendix). These descriptive results provide prima facie evidence that credential and skill mismatches refer to different situations.

Significant differences are observed among fields of study in the incidence of both credential and skill mismatches. A substantial risk of credential mismatch is faced by graduates from Art and Humanities (15.51) and Social Sciences (13.13). The incidence of this phenomenon is lower among graduates from technical and scientific fields, such as Engineering and Architecture (7.23), Mathematics and Statistics (6.4), and Health (3.81). The distribution of skill mismatches is partially different. The highest values are observed among graduates from Natural and Applied Sciences (14.25), followed by graduates from Art and Humanities (13.62). The lowest incidence of skill mismatches is found among graduates from Health (3.84). Additionally, when considering different fields of study separately, the correlation between credential and skill mismatches is far from perfect (see Table A3 in appendix).

The results on the risk of credential and skill mismatches 5 years after graduation are reported in Tables 3 and 4, respectively. Due to space limitations, we report only an extract of the full results here. The full estimation results are reported in appendix (Tables A5 and A6)

These results are obtained by controlling for the main socio-demographic variables and for some indicators of educational and working experience of graduates, as previously indicated. The coefficients of these covariates are statistically significant and consistent with results of previous studies, so they will not be discussed here. Consistent with the previous descriptive analysis, the results demonstrate that field of study matters. Model 2 suggests that Humanities graduates are the worst performers, whereas those from technical and scientific fields enjoy much better labour market prospects. Interestingly, comparing the results for credential and skill mismatches reveals that the relative position of different fields changes. Credential mismatches seem to be distributed according to the distinction between more-or-less specific and job-oriented fields of study (Table 3). However, the distribution of skill mismatches presents some interesting exceptions to this rule (Table 4). For instance, graduates from Engineering and Architecture and from Mathematics and Statistics are among those at least risk of credential mismatch, but they do not perform as well in terms of skill mismatches. These findings may reflect the gap between the signalling value of degrees, which would lead employers to prefer hiring graduates from these fields of study, and the actual demand for their specific skills in the labour market. If comparable repeated cross-sectional data were available, it would be of real interest to assess how credential and skill mismatches do react to changes in graduation rates from different fields of study.

Model 3 includes country-level covariates. Consistent with the descriptive results reported in appendix (see Figure A4), the coefficients for the level of employment protection and for the rate of public employment are negative, while the one for the regulation of professional services is positive. However, the association between these covariates and the overall risks of both credential and skill mismatches never appears to be statistically significant. Moreover, the inclusion of country-level variables does not change the significance nor the magnitude of the individual level coefficients. Because we were interested in assessing whether these country-level variables affect field of study differentials in overeducation, in Model 4 we add interaction terms between these variables and field of study.

Our first hypothesis concerning the effects of employment protection is confirmed. Model 4 in Table 3 suggests that fields of study can be divided into three groups, according to graduates’ risk of credential mismatch. On the one hand, we have technical and scientific fields, which are associated with a lower risk of mismatch: these are Health, Engineering and Architecture, Mathematics and Statistics. On the other hand, we find fields with a higher risk of credential mismatch: Arts and Humanities, Social Sciences, Business and Administration and Natural Sciences. Education, which is the reference category, and Law stay in between the two categories. The level of employment protection seems not to be statistically associated with the overall risk of credential mismatch, but it influences the distribution of this risk across fields of study. A stricter employment protection, indeed, seems to further reduce the already low risk of credential mismatches for graduates from fields of study that prepare students for specific professions, such as Engineering and Architecture and Health, and from scientific fields, particularly Mathematics and Statistics. At the same time, it leaves unaltered the risk for graduates from the other fields of study. In short, the stricter the employment protection, the greater the gap between graduates in terms of the risk of credential mismatch. Conversely, consistent with our hypothesis, we find that strictness of employment protection does not significantly affect the overall risk of skill mismatch, nor its distribution among graduates from different fields (Table 4).

The empirical evidence partially supports also the second hypothesis, concerning the regulation of professional services. Model 4 in Table 3 suggests that a higher level of regulation is associated with an overall higher risk of credential mismatch. Moreover, the interaction terms indicate that in contexts with a stricter professional regulation we find a lower risk of credential mismatch for graduates who do not have to comply with these additional requirements to access the profession, such as graduates from the Humanities or Social Sciences. At the same time, we find no alteration of the risk of credential mismatch for graduates who are subjected to this kind of regulation, such as graduates from Engineering, Architecture and Law (Table 3). Thus, we end up with a reduction of the comparative advantage for graduates from such fields of study. The only relevant difference is the one of graduates from Business and Administration, whose risk of credential mismatch is found to be lower in countries with a stricter regulation of professional services. On the other hand, results reported in Table 4 suggest that the regulation of professional services does not influence the distribution of the risk of skill mismatch among graduates: the main effect is positive and statistically significant, but it is counterbalanced by the interaction terms, which all appear negative and significant.

The findings on the role of the public sector as an employer partially support our hypothesis. As expected, the higher the rate of public employment, the smaller the gap between more and less job-oriented fields of study with respect to credential mismatches. The results reported in Table 3 suggest, indeed, that a higher rate of public employment reduces the relative advantage of more occupationally oriented fields, such as Engineering, Architecture, and Health. We find, however, public employment not to significantly affect the risk of skill mismatch (Table 4). Even though the main effect of the rate of public employment is negative and statistically significant, the interaction terms are not significant, suggesting that this factor is not capable of affecting the differential risk of graduates from different fields of study.

In this paper, we have deepened the existing knowledge around the association between some labour market institutions and the diffusion of overeducation among tertiary graduates, by distinguishing between two forms of suboptimal allocation of workers in the labour market: credential and skill mismatches. In a context of widespread educational expansion, this distinction is not of minor importance: individuals may possess formal qualifications higher than those required by their jobs, but actually perform jobs for which their skills are well suited. Consistent with previous findings, we have found important differences in the overall incidence of the two phenomena and a weak correlation between them at the individual level. Also, the distributions of credential and skill mismatches among graduates from different fields have been shown to differ to a significant extent, even though both phenomena seem to be more widespread among graduates from generalist fields, such as the Humanities and Social Sciences.

The analyses presented in this study have mainly focused on the variation across countries of fields of study differentials in terms of both credential and skill mismatches. While the literature has mainly explained differences across countries as the result of different educational systems, we have investigated the role of labour market and welfare institutions. In particular, we have analysed three institutional factors that are likely to affect the number and the quality of the available vacancies: the strictness of employment protection, the level of professional services regulation, and the capacity of the welfare state to hire skilled workers.

The first result is that these institutional factors do not affect credential and skill mismatches similarly: they have been shown to mainly affect the match between individual educational credentials and job, rather than the actual utilisation of skills on the workplace. This finding, together with the descriptive evidence that they happen to be weakly correlated at the individual level and across fields, further supports the claim that we need to trace a distinction between the two phenomena.

The way in which these institutional factors affect fields of study differential is far from homogenous. The level of employment protection – is found to augment the differential in terms of credential mismatch, by further reducing the already low risk for graduates from more job-oriented fields of study. Employment protection, on the contrary, seems not to affect the match between individual skills and those required on-the-job. In other words, this factor seems to enhance the marketability of some credentials – the ‘stronger’ ones, in fact – without altering the capacity of employers to fully recognize and use graduates’ competencies.

The second analysed factor – the regulation of professional services – is found to create rigidities to the individual-job match. Graduates who have to comply with these additional requirements, such as those from Architecture, Engineering, Law or Health, seem in fact not to be advantaged at all by this form of closure: on the contrary, in contexts with a stricter regulation of professional services, it appears to be harder for graduates from the matching fields to access these professions.

Finally, the level of public employment seems to play a beneficial effect on the gap between fields of study, at least in terms of credential mismatch. In contexts with a higher rate of public employment, indeed, we have found a reduction of the differential between generalist fields of study and those more job-oriented, probably due to the welfare state being able to employ graduates with less marketable credentials, such as those from the Humanities or Social Sciences.

Because of data constraints, this study has only employed subjective indicators. Future research might use other indicators of overeducation, in order to strengthen these conclusions. In particular, skill mismatches might be analysed by referring to dictionaries of occupations, and credential mismatches could be measured by indicators based on RM. These alternative solutions would offer a chance to test the robustness of our findings with different indicators of the two concepts.

No potential conflict of interest was reported by the author.

The data collection for the REFLEX and HEGESCO projects took place, respectively, in 2005 and 2008. In each country, in both the REFLEX and HEGESCO Projects, a representative sample was drawn of graduates from ISCED 5A (bachelors and masters or equivalent) who got their degree 5 years prior to the time of the survey (in most REFLEX countries this was the academic year 1999/2000, in the HEGESCO countries the academic year 2002/2003).

The size of the sample varied according to the anticipated response rate and the targeted number of respondents in each country. To increase the efficiency of the sample, stratified sampling was used. The strata used were dependent on the national context, but usually comprised type and field of higher education, and in some countries also region and gender.

The mail questionnaire focuses on educational experiences before and during higher education, the transition to the labour market, characteristics of the first job, characteristics of the occupational and labour market career up to the present, characteristics of the current job, characteristics of the current organisation, assessment of required and acquired skills, evaluation of the educational programme, work orientations, and some socio-biographical information.

Table A1 contains an overview of the number of available respondents and the response percentage per country.

Giulia Assirelli is a Ph.D. candidate in Sociology and Social Research at the University of Trento, Italy. Her research interests include transitions from education to labour market, comparative sociology, and quantitative methods of data analysis.