## Jacqueline O'Reilly, Janine Leschke, Renate Ortlieb, Martin Seeleib-Kaiser, and Paola Villa

Print publication date: 2018

Print ISBN-13: 9780190864798

Published to Oxford Scholarship Online: January 2019

DOI: 10.1093/oso/9780190864798.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy).date: 24 May 2019

# Youth Overeducation in Europe

## Is there scope for a common policy approach?

Chapter:
(p.530) 18 Youth Overeducation in Europe
Source:
Youth Labor in Transition
Publisher:
Oxford University Press
DOI:10.1093/oso/9780190864798.003.0018

# Abstract and Keywords

Less favorable outcomes—such as overeducation—early in the careers of younger workers may impact negatively on future labor market success, so it is important to understand the incidence of youth overeducation, its evolution over time, and the drivers of youth labor market mismatch. Most research has focused on examining the incidence and impacts of overeducation. This chapter represents one of the few attempts to examine patterns of overeducation within countries, while the adoption of a time-series approach enables the identification of common trends across Europe. Overeducation rates in Europe are converging upward over time, and the general pattern of overeducation is linked across many countries, suggesting that the phenomenon responds in a similar way to external shocks and, consequently, may react in similar ways to appropriate policy interventions. This chapter finds that youth overeducation is driven by the composition of education provision, aggregate labor demand, and labor market flexibility.

# 18.1. Introduction

Overeducation describes the situation in which individuals are employed in jobs for which the level of education required to either get or do the jobs in question is below the level of schooling held by the workers. Overeducation has become an increasingly important issue for discussion both within national governments and at the European and Organization for Economic Co-operation and Development (OECD) levels, and policymakers have become ever more concerned about the apparent inability of large shares of new labor market entrants to acquire jobs that are commensurate with their levels of education. Overeducation is costly at an individual level, with mismatched workers typically earning 15% less than their well-matched counterparts with similar levels of education. Furthermore, overeducation tends to reduce levels of job satisfaction and increase rates of job mobility (for a review of the evidence, see Quintini 2011). At the firm level, although there is some evidence that overeducated workers raise productivity levels somewhat,1 higher rates of job mobility imply that overeducation can impose additional hiring costs on firms. At the macroeconomic level, total output will be lower as a consequence of a significant proportion of the workforce operating below their full potential productivity, while public finances are adversely affected as a consequence of lower income tax receipts and suboptimal investments in educational provision. Given the various (p.531) impacts of overeducation, it is extremely important to assess the evolution of its rates over time (both within and between countries) so as to develop our understanding of the phenomenon and ascertain the extent to which policies combating overeducation can be coordinated at a European level or whether country-specific responses are likely to be more appropriate.

Currently, almost all of the research on labor market mismatch, measured in terms of either overeducation or overskilling, has relied on country-specific, cross-sectional, or panel data sets. To date, the research has focused on identifying the individual- or firm-level determinants of mismatch and/or the impact of mismatch on individual outcomes such as income or job satisfaction. Although such insights are crucial to understanding mismatch, it is only by studying the phenomenon at a more aggregate level that we can come to an understanding of the macroeconomic, demographic, and institutional forces that drive it. In this study, we use the European Union Labour Force Survey (EU-LFS) to construct quarterly time series of both youth and adult overeducation rates between 1997 and 2012 for 29 European countries. This chapter has a number of objectives, including (1) providing a descriptive assessment of trends in overeducation in European countries over time, (2) assessing the extent to which the rate of overeducation among youth and adult cohorts moves together within countries, (3) measuring the degree of interdependence and convergence in the evolution of overeducation between countries over time, and (4) identifying some of the underlying drivers of youth overeducation.

From a policy perspective, the extent to which overeducation could be suitable for a common policy approach, at either a European or a national level, will largely depend on the similarities in the evolution of overeducation over time both between and within countries. In this chapter, we adopt advanced econometric techniques that can confirm if two time series are driven by a common underlying economic relationship, as opposed to merely trending together in a spurious, noncausal manner. If overeducation has evolved in different directions at different rates across countries, this will provide a strong indication that it is driven by a range of factors that will vary in terms of both their magnitude and their significance across countries. Conversely, if movements in overeducation are confirmed through econometric testing to be driven by the same underlying causal factors over time, this would be supportive of a centralized policy approach aimed at targeting the common underlying causal influences driving both series. We consider a range of potential drivers relating to labor market demand, labor market supply, the structure of education systems, and macroeconomic factors. The potential for a future common policy approach to overeducation, at either a national or a pan-European level, is consequently assessed on the basis of this analysis.

## (p.532) 18.1.1. Existing evidence on international variations in overeducation

Although the general literature on overeducation has expanded rapidly, particularly during the past two decades (for reviews, see McGuinness 2006; Quintini 2011; McGuinness et al., 2018), there has been little assessment of overeducation from an aggregate country-level perspective; nevertheless, some exceptions do exist. Pouliakas (2013), also using data from the EU-LFS and analyzing the average rate of overeducation between 2001 and 2011, demonstrates the existence of considerable variation in overeducation rates across European countries. Pouliakas further concludes that although the average level of overeducation among EU25 member states exhibited a relatively stable time series between 2001 and 2009, there was substantial credentialism present in the labor market, with the growth in overeducation being largely subdued by higher occupational entry requirements.2 Despite the relatively constant trend, the Pouliakas study does indicate that during the financial crisis, the average rate of overeducation in Europe increased during the years 2008 and 2009, implying that levels of overeducation may vary with the business cycle. In support of this view, Mavromaras and McGuinness (2012) argue that there are grounds to expect the rate of mismatch to vary with macroeconomic conditions, on the basis that fluctuations in the economy will change the composition in the demand for labor and, consequently, how workers are utilized within firms. Ex ante, we might reasonably expect rates of overeducation to rise during times of recession and to fall during periods of economic growth. However, it is also reasonable to suppose that business-cycle impacts will be more heavily felt among newly qualified younger workers and that variations in the overall rate of overeducation are likely to be less affected by variations in aggregate output. These hypotheses will be further explored in Sections 18.3 and 18.5.

With respect to the potential drivers of overeducation at the macroeconomic level, there is limited research primarily because of the paucity of cross-country data sets. A number of possible effects could potentially explain the existence and persistence of overeducation at a national level. Overeducation could arise when the supply of educated labor outstrips demand, primarily as a result of the tendency of governments in developed economies to continually seek to raise the proportion of individuals with third-level qualifications. Alternatively, it may be that the quantity of educated labor does not exceed supply but that there are imbalances in composition; in other words, individuals are being educated in areas in which there is little demand, leading to people from certain fields of study being particularly prone to overeducation.3 Furthermore, labor demand and supply might be perfectly synchronized yet overeducation might still arise because of frictions deriving from asymmetric information, institutional factors that prevent labor market clearance, or variations in individual preferences related to either job mobility or work–life balance.

(p.533) Applying a multilevel model to a cross-country graduate cohort database, Verhaest and van der Velden (2012) derive a number of variables from the individual-level data to explain cross-country differences in the incidence of overeducation. Explanatory variables in the Verhaest and van der Velden study include measures for the composition of higher education supply in terms of vocational versus academic orientation and field of study, proxies for educational quality,4 measures of the output and unemployment gaps,5 indicators of employment protection legislation within each country, and the level of education oversupply. In their study, Verhaest and van der Velden calculate the share of graduates in the population older than age 25 years and gross expenditure on research and development (R&D). Graduate oversupply is then taken as the difference in the standardized values of these two variables. Verhaest and van der Velden find that cross-country differences in overeducation are related to their measures, which, they argue, capture variations in quality and orientation (general vs. specific) of the education system, business-cycle effects, and the relative oversupply of highly skilled labor.

Davia, McGuinness, and O’Connell (2017) attempt a similar exercise using EU-SILC data. Similar to Verhaest and van der Velden (2012), Davia et al. find evidence to support the notion that overeducation is more prevalent in regions where the level of educated labor supply exceeds demand and where university enrolment levels are highest.6 These authors also report that the overeducation rate is positively related to the share of migrants in the labor market and is lower for females in regions with strong employment protection. Thus, although some concerns may be raised regarding the quality of some of the indicator variables derived in studies relying on cross-sectional international data, the studies by Verhaest and van der Velden and Davia et al. demonstrate the potential importance of aggregate-level variables in explaining overeducation, with both studies pointing toward education oversupply as an important driving force. Recently, McGuinness and Pouliakas (2017), using cross-country European data, have attempted to assess the relative importance of the various explanations for overeducation in terms of the proportion of the overeducation pay penalty that can be attributed to them. McGuinness and Pouliakas argue that there is merit to the view that overeducation is related to differences in the human capital of overeducated and matched workers; however, differences in job conditions and skill requirements were also important. Furthermore, McGuinness and Pouliakas suggest that the quality of information that workers acquire before accepting a job is also an important component in explaining the impact of overeducation among European graduates.

# 18.2. Data and methods

To our knowledge, there are no reliable time-series data on overeducation that would allow a systematic cross-country comparison across time, and the (p.534) data-development aspect is a key contribution of the current study. The data used in this study are the quarterly anonymized country-level files of the EU-LFS for the period up to the fourth quarter of 2012. Because there is no subjective question within the EU-LFS related to the level of schooling necessary to get, or undertake, a person’s current job, overeducation is measured objectively. There are essentially three standard methods of measuring overeducation. The subjective measure is based on individual responses comparing attained education levels with perceived job-entry requirements; the occupational-dictionary approach compares individual-level education with the required level of schooling detailed for specific occupations in the documentation accompanying occupational classification systems; finally, the objective approach compares individual levels of schooling with either the mean or the mode level of schooling of the respective occupation. The goal of this chapter is to examine overeducation over time across a large number of EU countries. In this regard, the EU-LFS is one of the only data sets that enables this type of analysis; however, using this data set means that the only measure of overeducation we can exploit is the objective approach. Existing studies indicate that although the correlation between the various definitional approaches tends not to be particularly high, they generate very similar results with respect to both the incidence and the impacts of overeducation (for review, see McGuinness 2006).

For each country, in each quarter, overeducation is defined as the proportion of employees in employment whose International Standard Classification of Education (ISCED) level of schooling lies one level or more above the occupational mode. The occupational modal level of education is the most common qualification possessed by workers in each two-digit International Standard Classification of Occupations (ISCO) occupation group. Overeducation is calculated within two-digit occupational codes and using five ISCED categories of <2, 3, 4, 5B, and 5A + 6. Thus, if the modal level of schooling in a particular two-digit occupation were measured at ISCED-3, then all individuals educated to ISCED levels 4 and above would be deemed to be overeducated in our approach. We calculate the overall rate of overeducation in each country for each quarter, and we also calculate the rates for individuals aged 15–24 and 25–64 years. Given that we are dealing with a large number of countries, for the purposes of our analysis we group these into three categories on the basis of an initial inspection of patterns in the data. Moreover, the selected groupings are likely to have common linkages in terms of geographical proximity, levels of economic development, and access to the single market. The first category is composed of the countries that acceded to the EU from 2004—which include Bulgaria, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, the Slovak Republic, and Slovenia—and are referred to as the “Eastern” states. The second category refers to Portugal, Ireland, Italy, Greece, and Spain—the traditional “Periphery” of the EU. The third group (“Central”) is made up of the remaining countries located in Central and Northern Europe and includes Austria, Belgium, Denmark, (p.535) Finland, France, Iceland, Luxembourg, the Netherlands, Norway, Sweden, and the United Kingdom.7 Generally, we found that the average rate of overeducation is lowest in the Eastern European countries, highest in the Periphery, and somewhere in between in the Central European countries (see descriptive evidence in Section 18.3).

In terms of the empirical approach, we are interested in determining the extent to which youth and adult overeducation move together within countries and also the degree to which long-term relationships in the rates of overeducation exist between countries. We classify these long-term equilibrium relationships as “completed convergence” on the grounds that, if detected, they indicate that certain series are sufficiently correlated that overeducation is likely to be driven by a common set of macroeconomic and/or institutional factors. We might expect a link between youth and general overeducation within countries on the grounds that they are likely to be driven by a common set of macroeconomic variables related to, for instance, the nature of labor market demand, labor supply, or wage-setting institutions. The overall overeducation rate is closely related to a stock measure that will react more slowly to major changes in determining factors. However, the youth overeducation rate is more of a flow measure that may react with more volatility to changes in labor market conditions. This raises uncertainties related to the extent to which the two series will be highly synced even if they do share common determinants. Regarding intercountry completed convergence, there are grounds to believe that convergence could prevail within an EU context. This could happen, for example, when cross-country differences in key labor market variables such as unemployment and, possibly, overeducation are reduced by the free movement of workers. Conversely, completed convergence (a tendency for the overeducation rates across countries to equalize over time) may be limited for Eastern European countries or between countries where language or other noneconomic barriers prevent equalizing labor flows.

In this chapter, we are dealing with time-series data, which should not be approached using a traditional regression methodology. Historically, econometricians have tended to assume that most time-series data are “nonstationary” and, crucially, this had no impact on their empirical analysis. Time-series data tend to increase or decrease over time and, therefore, do not have a constant “stationary” mean and variance. Running regressions on data of this nature (nonstationary) can give rise to misleading results and essentially lead to erroneous conclusions about the existence of a relationship between variables where one may not in fact exist; this is commonly known as the spurious regression problem. Spurious regressions occur when two variables are statistically related to each other but no causal relationship exists, meaning they are related purely by coincidence or they are both influenced by another external variable. For example, in examining ice cream sales, we may find that sales are highest when the rate of drowning is highest. To imply that ice cream sales cause drowning or vice versa is an example of a spurious relationship. In (p.536) fact, a contemporaneous increase in these two variables could be caused by a heat wave. Consequently, when we test for common underlying trends between the overeducation series, we take account of the spurious regression problem. In order to overcome such issues, we adopt a cointegration estimation approach. Two nonstationary variables are said to be cointegrated when they move together in a similar manner over time—for example, variables such as household income and expenditure—and, in this case, the regression results are meaningful.

We begin by establishing whether each respective series is stationary or nonstationary by applying standard Phillips–Perron unit root tests (Phillips 1987; Phillips and Perron 1988).8 The Phillips–Perron test is written formally for a time series yt in Eq. (18.1), where t is a time trend. The null hypothesis of the Phillips–Perron test is that there is a unit root or that the series is nonstationary; that is, $β1=0$:

$Display mathematics$
(18.1)

If we establish that two overeducation time series are nonstationary, then we adopt the Phillips–Ouliaris test for a cointegrating relationship. If both series are stationary, we perform ordinary least squares (OLS) on the basis that spurious regressions are no longer an issue. Finally, if one series is stationary and the other nonstationary, we do not undertake any further tests for an underlying relationship.

The Phillips–Ouliaris test is a residual-based test for cointegration involving a two-step estimation approach. In the first stage, Eq. (18.2) is estimated:

$Display mathematics$
(18.2)

$β^$ is a cointegrating vector if $ut=Xit−α−βiYit$, and the second stage of the procedure tests whether the regression residuals from Eq. (18.2) are stationary using the Phillips–Perron test.

In addition to testing for long-term relationships in overeducation rates both within and between countries, we also examine the extent to which overeducation rates in Europe have been converging or diverging over time by estimating a Barro regression (Eq. 18.3; Barro 1997). This investigates the relationship between a country’s initial level of overeducation and how the rate has evolved over time. In instances in which completed convergence has not been achieved (where overeducation rates across countries have not equalized over time), overeducation rates may converge as workers from saturated graduate labor markets relocate to areas with greater levels of job opportunity and lower levels of overeducation (see Akgüç and Beblavý, this volume). For example, the lack of convergence could arise from some countries remaining outside of the monetary union. Under these circumstances, the consequence of labor market inflows would be to raise overeducation levels in areas of oversupply. At the same (p.537) time, labor market outflows, in the form of outmigration, would tend to reduce overeducation rates in highly saturated labor markets.

The application of the Barro model involves examining the relationship between the growth rate of overeducation and the initial level of overeducation using a regression model. If a country with a lower initial level of overeducation tended to have a higher growth in overeducation over time, then the estimate of the coefficient of interest—β‎1 in Eq. (18.3)—would be negative and significant. This implies that this country’s overeducation rate would converge to the average prevailing in other countries. Therefore, disparities in rates across countries over time would tend to dissipate. In contrast, a positive coefficient would point toward divergence in overeducation rates across countries. In addition to the Barro regression, we also check for convergence by plotting the cross-country variance in overeducation rates for specific groups of countries:

$Display mathematics$
(18.3)

Finally, we examine the determinants of youth overeducation for countries with a stationary series. Twenty-one of the 28 youth overeducation series were found to be stationary in nature, suggesting that the application of standard OLS is appropriate.9 For the stationary series, we estimate the following model for all countries initially and then for our three country groupings:

$Display mathematics$
(18.4)

where $yit$ is the dependent variable observed for country i in time t, $yit−1$ is the lagged dependent variable, $Xijt$ represents a number of j independent variables with $βj$ the associated coefficients, $αi$ is the unobserved time-invariant country effect that allows us to control for institutional factors (fixed effect), and $εit$ is the error term. Using this fixed-effect approach allows us to model the determinants of youth overeducation, but we cannot exclude the possibility that some variables may be endogenous and, in further analysis, we plan to build on this approach using panel data in a dynamic framework.10

# 18.3. Descriptive evidence

The average levels of overeducation, based on quarterly data for the period 2001–2011, are reported in the first column of Table 18.1. Our sample is restricted to employees in full-time employment and so will largely exclude the student population but include paid apprenticeships and traineeships. The estimated rate of overeducation varies from 8% in the Czech and Slovak Republics to 30% or greater in Ireland, Cyprus, and Spain. In general, we observe the estimated (p.538) incidence of overeducation to be lowest in the Eastern countries (e.g., the Czech Republic, Slovenia, and the Slovak Republic) and highest in the Peripheral countries (e.g., Spain and Ireland), with the Central countries lying somewhere in the middle. There are, however, some exceptions to this general pattern; for instance, overeducation rates were relatively high in Lithuania and Estonia, whereas overeducation in Portugal was well below the level observed in other (p.539) Peripheral countries. The second column of Table 18.1 provides a comparison with a number of estimates for 2014 generated by Flisi et al. (2014), who applied a comparable approach to the OECD Programme for the International Assessment of Adult Competencies (PIAAC) data. In general, our overeducation estimates match closely with those from the PIAAC-based study, with the exception of the estimate for Denmark, where a relatively large discrepancy exists.

Table 18.1 Overeducation rates: Comparison of estimates from EU-LFS data averaged over 2001–2011 and estimates based on PIAAC data for 2014

Country

(1)

(2)

Estimates based on EU-LFS (2001–2011 average)

Estimates based on PIAAC (2014)

Austria

0.19

0.23

Belgium

0.26

0.24

Bulgaria

0.11

Cyprus

0.31

0.31

Czech Republic

0.08

0.12

Germany

0.18

0.22

Denmark

0.18

0.31

Estonia

0.24

0.26

Spain

0.30

0.34

Finland

0.14

0.17

France

0.17

0.17

Greece

0.28

Hungary

0.13

Ireland

0.33

0.33

Italy

0.24

0.24

Lithuania

0.25

Luxembourg

0.17

Latvia

0.19

Netherlands

0.22

0.22

Poland

0.11

0.11

Portugal

0.18

Romania

0.10

Sweden

0.14

0.19

Slovenia

0.09

Slovak Republic

0.08

0.10

United Kingdom

0.21

0.20

Sources: Column (1), authors’ calculations based on EU-LFS data; column (2), Flisi et al. (2014).

We plot the country rates for total overeducation and for the 15- to 24-year-old and 25- to 64-year-old age groups for each country in Figure 18.1.11 The length of the time series varies depending on data availability. There is a high (p.540) (p.541) (p.542) degree of cross-country variation in terms of the level of overeducation, the general direction of the trend over time, and the relationship between youth and adult overeducation within countries.

Figure 18.1 Quarterly overeducation rates (restricted to full-time employees) for each country plotted for the time periods available from Q1/1998 to Q4/2010.

For slightly less than half of the countries, overeducation appears to be trending upward over time. However, although the rate of increase seems quite (p.543) slight, a much steeper slope is observed for most countries in the Peripheral group (Spain, Greece, Portugal, and Italy) and also in Poland. Furthermore, overeducation appears not to have risen in any observable way in 12 countries, including Austria, Belgium, Denmark, Germany, Iceland, Ireland, and Luxembourg, whereas it has fallen over time in Cyprus, Croatia, Lithuania, and Latvia.

With respect to youth overeducation, the pattern appears much more volatile relative to adult overeducation. Youth overeducation lies below the average in the vast majority of countries; however, it has been consistently above the average in the Peripheral group and in Belgium, Cyprus, France, and Poland. It may be the case that the consistently high levels of youth overeducation in countries in the Peripheral group are also contributing to the observed trend increase in total overeducation over time. For example, this may happen as a consequence of higher proportions of consecutive generations of young people failing to achieve an appropriate labor market match. The main characteristics of the country-level overeducation series are summarized in Table 18.2.

Table 18.2 Key characteristics of country-level overeducation series based on estimates from EU-LFS data, 2001–2011

Country

Positive trend

Negative trend

No trend

Austria

X

X

Belgium

X

X

Bulgaria

X

X

Cyprus

X

X

Czech Republic

X

X

Germany

X

X

Denmark

X

X

Estonia

X

X

Spain

X

X

Finland

X

X

France

X

X

Greece

X

X

Croatia

X

X

Hungary

X

X

Ireland

X

X

Iceland

X

X

Italy

X

X

Lithuania

X

X

Luxembourg

X

X

Latvia

X

X

Netherlands

X

X

Norway

X

X

Poland

X

X

Portugal

X

X

Romania

X

X

Sweden

X

X

Slovenia

X

X

Slovak Republic

X

X

United Kingdom

X

X

# 18.4. Have overeducation rates converged or are they continuing to converge?

To investigate the existence of a long-term relationship between overeducation rates across countries, we adopt the Phillips–Ouliaris approach (described in Section 18.2) and perform pairwise analysis of overeducation rates. Cointegration tests should reveal whether overeducation rates move together over a longer time period. A finding of a common trend in the rates across countries may signify that an international policy approach to overeducation is appropriate. Even if there is no finding of cointegration across countries, overeducation may still respond to the same underlying processes, which we explore in Section 18.5.

For each country, the tests for stationarity are performed either with or without a time trend. The decision to include a time trend or not depends on the evolution of the overall overeducation rate over time in each country. The null hypothesis (that the series is nonstationary) is the presence of a unit root. We conclude that we cannot reject the null hypothesis of a unit root for any series where the test statistic is below the critical value at the 10% level of significance. These countries are then included in the cointegration analysis to ascertain if the overeducation rates move together over time in an equilibrium manner. We perform pairwise OLS on the other countries where we conclude that the overeducation rate is stationary and include a time trend depending on the nature of the stationarity. For example, a series is trend stationary if the underlying series is stationary after removing the time trend.

The finding of nonstationarity means that the overeducation rate has a nonconstant mean and/or variance, suggesting that the phenomenon is somewhat unstable over time. Conversely, a finding of stationarity implies relative (p.544) (p.545) stability, suggesting that overeducation rates are generally stable in the sense that they are constant over time or increase/decrease at a constant rate with no volatility. Table 18.3 shows that for the majority of countries, overeducation is stationary, meaning the average rates are stable over time. The null hypothesis of nonstationarity could not be rejected for Cyprus, Hungary, Poland, Romania, the Slovak Republic, Norway, the United Kingdom, Greece, Italy, Ireland, and Spain, (p.546) indicating that these series are nonstationary and therefore should be included in the pairwise cointegration analysis. In the sense that the tests suggest that the development of overeducation is somewhat unpredictable, it appears more likely to be erratic in most countries in the Peripheral group, which could reflect their greater exposure to macroeconomic shocks.

Table 18.3 Country-level Phillips–Perron stationarity tests

Country

Phillips–Perron test statistic

Trend

Austria

–4.066***

No

Belgium

–4.302***

No

Bulgaria

–5.161***

No

Cyprus

–3.098

Yes

Czech Republic

–3.468*

Yes

Germany

–2.824*

No

Denmark

–4.842***

No

Estonia

–4.937***

No

Spain

–3.032

Yes

Finland

–4.189***

Yes

France

–2.836*

No

Greece

–1.962

No

Hungary

–2.063

Yes

Ireland

–2.594

No

Iceland

–3.899***

No

Italy

–2.177

Yes

Lithuania

–5.368***

Yes

Luxembourg

–2.985**

No

Latvia

–3.485*

Yes

Netherlands

–2.704*

No

Norway

–2.573

Yes

Poland

–2.006

Yes

Portugal

–5.670***

Yes

Romania

–2.367

Yes

Sweden

–5.548***

Yes

Slovenia

–3.749**

Yes

Slovak Republic

–3.078

Yes

United Kingdom

–2.272

Yes

(*) p < .10.

(**) p < .05.

(***) p < .01.

Source: EU-LFS.

Table 18.4 shows the test results from the cointegration analysis. Although the patterns are not clear-cut, the table provides some evidence of cointegrating relationships within the Peripheral group, indicating completed convergence. For example, Greece, Italy, Ireland, and Spain are all bilaterally cointegrated at varying levels of statistical significance. This implies that overeducation in these countries responds in a similar manner to external shocks; in other words, there is some evidence of a long-term relationship in overeducation rates between these countries. This arguably suggests that they should be subject to a particular policy response. Outside of this, the table indicates no clear pattern, with some evidence of cointegration between countries in the Central, Eastern, and Peripheral groups.12 The pairwise OLS results, presented in Table 18.5, reveal similar patterns. These findings of long-term relationships between several of the Central group countries and also between the Central and Eastern groups indicate that there are similarities in the general evolution of overeducation across certain countries, and they may justify a common policy approach for these countries. However, in a minority of countries, overeducation series were found not to be heavily correlated with those of other European countries; examples are Austria, Portugal, and Sweden, which exhibit little or no commonality in their overeducation series. This finding suggests that a common policy approach may not be appropriate for these countries.

Table 18.4 Phillips–Ouliaris cointegration statistics testing the existence of a long-term relationship (null hypothesis of no cointegration between paired countries against alternative hypothesis of stable cointegration relationship), 2001–2011

Country

Hungary

Poland

Romania

Slovak Republic

Norway

United Kingdom

Greece

Italy

Ireland

Spain

Cyprus

–3.242

–3.208

–3.292

–3.026

–3.921*

–3.585

–3.189

–3.346

–3.171

–3.613

Hungary

–3.122

–2.635

–2.401

–4.326**

–3.674*

–2.779

–4.951***

–2.221

–5.050***

Poland

–2.642

–2.846

–2.313

–3.111

–3.142

–3.451

–2.167

–3.674*

Romania

–3.161

–2.861

–2.594

–3.190

–2.978

–3.037

–3.660*

Slovak Republic

–4.108**

–3.793*

–5.204***

–4.674***

–4.683***

–4.463**

Norway

–3.280

–2.160

–4.651***

–2.810

–5.659***

United Kingdom

–2.824

–3.108

–2.558

–3.387

Greece

–4.348**

–3.814*

–3.912*

Italy

–1.976

–6.471***

Ireland

–3.903*

(*) p < .10.

(**) p < .05.

(***) p < .01.

Source: EU-LFS.

Table 18.5 Pairwise OLS estimates to examine the existence of long-term relationships between pairs of countries

Country

AT

BE

DE

DK

FI

FR

NL

SE

IS

LU

PT

BG

CZ

EE

LT

LV

SI

AT

–0.065

0.458***

0.056

0.002

0.385**

0.025

0.026

0.008

–0.017

0.101

–0.130

0.294

0.076

0.051

0.046

–0.045

BE

–0.468

–0.070

–0.447***

0.063

0.336***

–0.348*

–0.162*

0.113

–0.242

–0.303

–1.352***

0.066

0.245*

0.202**

–0.170

DE

0.175**

0.107

0.129

0.335**

0.014

0.111**

0.105

0.024

–0.148

0.725*

0.003

0.067

0.145*

–0.212*

DK

–0.319**

0.255

0.270**

–0.103

0.156

–0.090

–0.531***

–0.704***

–0.466

–0.090

0.039

0.121

–0.084

FI

–0.496***

–0.087

–0.516*

0.038

–0.054

–0.039

0.153

1.250***

–0.044

0.254**

–0.144**

–0.029

FR

0.050

0.306***

–0.056

–0.017

0.213**

0.011

0.346

–0.053

–0.259***

–0.092

0.142

NL

–0.709***

0.064

–0.032

–0.578***

–0.684**

–1.185***

0.311***

0.433***

0.442***

–0.579***

SE

–0.031

0.063**

0.074

–0.020

–0.418**

–0.041

–0.086

0.050

–0.088

IS

0.265

–0.854**

–0.436

–0.010

0.294

0.571**

0.315

–0.270

LU

–0.678*

0.462

–3.373***

0.514**

0.647***

0.699***

–1.286***

PT

0.144

–0.294

–0.078

0.062

–0.090

–0.050

BG

0.139

–0.023

–0.037

–0.079

–0.030

CZ

–0.015

0.062

–0.034

0.028

EE

0.792***

0.478***

–1.009***

LT

0.033

–0.080

LV

–0.316

SI

Country

AT

BE

DE

DK

FI

FR

NL

SE

IS

LU

PT

BG

CZ

EE

LT

LV

SI

AT

BE

–0.222

DE

0.439***

–0.112

DK

0.234

–0.084

0.811**

FI

–0.064

–0.186***

0.143

–0.047

FR

0.302**

0.014

0.128

0.048

0.057

NL

0.044

0.402***

0.493**

0.307**

–0.461***

0.112

SE

–0.032

0.081**

0.094

0.031

–0.138*

0.106

0.105***

IS

0.073

–0.452*

1.601**

0.361

–0.451

–0.689

0.351

–0.851**

LU

–0.159

0.301

0.752

–0.198

–1.363***

–0.198

–0.165

–1.135***

0.252

PT

0.142

0.095

0.119

–0.097

–0.042

0.160

–0.061

0.331

–0.095

0.035

BG

–0.184

–0.105

–0.128

–0.232***

0.116

0.021

–0.188**

0.085

–0.064

0.071

0.167

CZ

0.073

–0.088***

0.165**

0.053**

0.154***

–0.090

–0.012

–0.213**

0.054***

–0.053**

–0.034

–0.016

EE

0.388

0.142

0.020

–0.184

–0.389**

–0.348

0.561***

–1.142***

0.157

0.288**

–0.687**

–0.112

–1.108**

LT

0.280

0.041

0.261

–0.031

0.566**

–0.724***

0.056

–0.680

0.092

–0.020

0.104

–0.030

0.898

0.212**

LV

0.348

0.229

0.670**

0.218

–0.690**

0.317

0.725***

0.844

–0.018

0.050

–0.322

–0.531**

–1.066

0.294*

0.070

SI

–0.195

0.043

–0.470**

0.092

–0.075

–0.155

–0.277**

–0.390

0.062

–0.189***

–0.121

–0.214

0.594

–0.326***

–0.042

–0.077

(*) p < .10.

(**) p < .05.

(***) p < .01.

Source: EU-LFS.

In summary, the completed convergence evidence suggests that overeducation in Europe is likely to respond to a coordinated policy approach. However, overeducation in the Peripheral group appears to behave somewhat differently from the rest of Europe, suggesting that a separate policy response is likely to be required for this block of countries.

Although there is some evidence of completed convergence within and between the Central group countries and some Eastern group countries, it is still possible that the countries in our study are converging to a common overeducation rate. Ongoing convergence is feasible given that many countries were found to be stationary with a common trend, suggesting that they continue to rise or fall over time, whereas others were found to follow no discernible pattern or trend.

Next, we test for the presence of ongoing convergence over the period first quarter Q1/2003 to Q1/2010. This time period was chosen so as to maximize the number of countries that could be included in the model; nevertheless, the results remained unchanged when the model was tested on a longer time series including fewer countries. (p.547) (p.548) (p.549)

(p.550) Ongoing convergence would imply that overeducation increased at a faster rate between 2003 and 2010 in countries that had a lower initial overeducation rate in 2003. This is equivalent to a negative and significant β‎1 coefficient in the Barro regression from Eq. (18.3). Conversely, a positive and significant coefficient would be indicative of divergence. The coefficients from the Barro models are presented in Table 18.6 and indicate that ongoing convergence was a feature of the time period. The results suggest that there is a tendency for countries to converge toward a common overeducation rate over time for all measures of overeducation.

Table 18.6 Barro regression results: Time period Q1/2003–Q1/2012 for 26 countries

Overeducation shares

Coefficients

Total overeducation

–0.033***(0.009)

Female overeducation

–0.036**(0.011)

Male overeducation

–0.032***(0.008)

* p < .10.

(**) p < .05.

(***) p < .01.

It may be the case that the degree of ongoing convergence varies among groups of countries with common structural, geographical, and historical features. It is not possible to estimate Barro regressions separately for our three groups because the sample size is too small. In order to overcome this difficulty, we assess the rate of ongoing convergence by plotting the variance of overeducation rates across countries, on the grounds that ongoing convergence would be consistent with a falling variance over time. Plotting the variance across all countries confirms the results from Table 18.6 that ongoing convergence did occur over the time period (Figures 18.2–18.4). However, the aggregate picture appears to conceal substantial variation because it is apparent that ongoing convergence was more modest in the Central group relative to the Eastern and Peripheral groups (Figures 18.5–18.7).

Figure 18.2 Variance in total overeducation across countries from Q1/2003 to Q1/2012 (26 countries).

Figure 18.3 Variance in adult overeducation across countries from Q1/2003 to Q1/2012 (26 countries).

Figure 18.4 Variance in youth overeducation across countries from Q1/2003 to Q1/2012 (26 countries).

Figure 18.5 Variance in total overeducation across Central group countries from Q1/2003 to Q1/2012 (Austria, Belgium, Denmark, Finland, France, Iceland, Luxembourg, Netherlands, Norway, Sweden, and United Kingdom).

Figure 18.6 Variance in total overeducation across Eastern group countries from Q1/2003 to Q1/2012 (Bulgaria, Czech Republic, Estonia, Hungary, Romania, Slovak Republic, and Slovenia).

Figure 18.7 Variance in total overeducation across Peripheral group countries from Q1/2003 to Q1/2012 (Greece, Ireland, Italy, Portugal, and Spain).

# 18.5. Determinants of youth overeducation

We now bring the analysis full circle by using the EU-LFS data to calculate a number of additional variables that can potentially explain movements in youth overeducation within countries. Specifically, for each country for each quarter, we compute variables measuring the labor force shares of migrants, the employment shares of workers who are part-time and workers who are temporary, the shares of workers employed in various sectors (administration, sales, and (p.551) (p.552) manufacturing), the unemployment rate, and the participation rate. We also compute a number of variables related to relative educational supply, specifically (1) the ratio of workers employed in professional occupations to graduates in employment and (2) the ratio of workers employed in professional occupations to workers in low-skilled occupations. Whereas the first variable is designed as a straightforward measure of graduate oversupply, the second is intended to pick up (p.553) the effects of skill-biased technological change, which is generally associated with a shift in relative demand away from high-skilled and toward low-skilled labor and in many countries with a general hollowing out of mid-skilled occupations. In addition to the variables calculated from the individual labor force surveys, we also derive some indicators from external data sources, and where necessary, annual data are interpolated to quarterly data series. Information on gross domestic product (GDP) per capita and R&D spending was sourced from Eurostat and the OECD.13 Information on the number of students enrolled in tertiary and vocational programs was sourced from the OECD and standardized by age cohort using the EU-LFS data.14

A number of patterns are present in the results shown in Table 18.7. In the model that combines the data across all countries, the results suggest that overeducation declines with an increase in part-time employment, labor force participation, and manufacturing employment. Conversely, overeducation was found to rise with increases in the share of temporary workers and in employment in the sales and hotel sectors. The results are difficult to interpret because, on the one hand, the finding with respect to part-time workers suggests that overeducation tends to be lower in more flexible labor markets, whereas on the other hand, the finding related to temporary workers suggests the opposite. The estimates suggest that the higher the overall participation rate and GDP per capita, the lower the youth overeducation rate. To the extent that a rise in the participation rate is generally accompanied by increases in wage rates and general labor demand, the results suggest that youth overeducation will tend to (p.554) (p.555) decline as general labor market conditions tighten. In the context of the model, the participation rate and GDP per capita tend to capture changing labor market demand more effectively compared to the unemployment rate. The measure relating to skill-biased technological change is positive, suggesting that youth overeducation is increasing as a consequence of declining relative demand for unskilled labor. This suggests that as the labor market restructures, jobs that were traditionally occupied by poorly educated workers are now being occupied by workers with higher levels of schooling. The results suggest that higher R&D spending has a positive effect on the youth overeducation rate. At first glance, this result seems counterintuitive because one would expect countries with higher R&D spending to have more high-skilled jobs so that, all else being equal, this would have a negative impact on overeducation. However, it could be the case that this does not apply to the youth cohort given that a certain level of experience may be required for such jobs. Finally, the aggregate model provides consistent support for the view that overeducation will be higher in countries with comprehensive-based education systems and lower in countries providing viable vocational alternatives.

Table 18.7 Determinants of youth overeducation for countries with stationary series (fixed-effects model)

Dependent variable: Youth overeducation

(1)

(2)

(3)

(4)

Variable

All countries

Central group

Eastern group

Peripheral group

Lagged youth overeducation

0.45***

0.35***

0.17***

0.35***

(0.030)

(0.041)

(0.063)

(0.101)

% Migrants in labor force

–0.03

0.09

–0.11**

–0.26

(0.042)

(0.077)

(0.056)

(0.455)

% Temporary workers

0.13**

–0.02

0.20

0.46*

(0.060)

(0.091)

(0.152)

(0.269)

Overall unemployment rate

0.01

0.18

–0.07

0.06

(0.052)

(0.144)

(0.078)

(0.249)

% Part-time workers

–0.33***

–0.38***

0.04

–0.78**

(0.070)

(0.091)

(0.176)

(0.352)

–0.14

0.52

–0.40

0.62

(0.244)

(0.365)

(0.376)

(1.023)

% Employed in sales and hotels

0.44***

0.69***

–0.03

–0.66

(0.149)

(0.237)

(0.227)

(0.648)

Overall participation rate

–0.21***

–0.22

–0.02

0.79

(0.078)

(0.134)

(0.118)

(0.511)

Ratio of employed in occupations 2, 3 to grads in employment

–0.02

–0.01

–0.07***

–0.00

(0.011)

(0.021)

(0.019)

(0.055)

Ratio of workers in high (2, 3) to low (7, 8, 9) ISCO

0.03**

0.02

0.03

0.06

(0.010)

(0.012)

(0.032)

(0.086)

Share of manufacturing

–0.20*

–0.31**

–0.26

0.16

(0.107)

(0.150)

(0.167)

(0.694)

Ratio of tertiary students to population (aged 20–24 years)

0.06***

0.05**

–0.15**

3.41**

(0.022)

(0.028)

(0.072)

(1.327)

Ratio of vocational students to population (aged 15–19 years)

–0.04**

–0.04*

0.03

–2.35**

(0.016)

(0.021)

(0.048)

(1.121)

Ln GDP per capita

–0.04***

–0.09**

–0.06***

0.03

(0.016)

(0.036)

(0.021)

(0.086)

R&D expenditure

0.02***

0.02***

–0.01

–0.02

(0.005)

(0.008)

(0.010)

(0.030)

Constant

0.57***

0.99***

0.80***

–0.55

(0.140)

(0.354)

(0.173)

(0.664)

No. of observations

903

491

284

128

R2

0.32

0.31

0.24

0.76

No. of countries

21

11

7

3

Prob > F

0.00

0.00

0.00

0.00

Note: Standard errors in parentheses. Ln = Natural Log.

(*) p < .10.

(**) p < .05.

(***) p < .01.

When the model is estimated separately for country groupings, we find that many of the results hold, although some variations exist. For example, within the Eastern group, the relative balance between vocational and comprehensive-based education appears less important, whereas overeducation was found to decrease along with an increase in the availability of graduate-level jobs and in migrants in the labor force. Within the Central and Peripheral groups, the share of part-time employment was found to have a strong negative effect, but no significant effect was found for the Eastern group. The positive temporary worker effect observed within the aggregate model was only evident for the Peripheral group.

# (p.556) 18.6. Conclusions

Overeducation is known to be costly to workers, and it also has negative implications for firms and the wider macroeconomy. To date, the vast body of research in the area has focused on examining the incidence and impacts of overeducation within countries. This chapter represents one of the few existing attempts to examine patterns of overeducation within countries, while the adoption of a time-series approach enables the identification of common trends across Europe. The evidence suggests that although overeducation rates in Europe are converging upward over time, the general pattern of overeducation is linked across many countries, suggesting that the phenomenon responds in a similar way to external shocks and, consequently, is likely to react in similar ways to appropriate policy interventions. However, the research indicates that overeducation within the Peripheral group is evolving somewhat differently compared to the rest of Europe, suggesting that a separate policy response is likely to be appropriate.

Although the overall model results are complex for the determinants of youth overeducation, a number of impacts are consistently present for all or most country groupings. Specifically, youth overeducation is highly driven by the composition of education provision and will tend to be lower in countries with more developed vocational pathways. Furthermore, youth overeducation tends to be heavily related to the level of aggregate labor demand, proxied in the model by variations in the participation rate and GDP per capita. Finally, youth overeducation tends to be lower the higher the employment share of part-time workers, suggesting that the phenomenon may be partly driven by labor market flexibility.

So what form are appropriate policy interventions likely to take? Although much remains unknown with respect to the drivers of overeducation, a number of recent studies have identified some key factors that influence overeducation across countries. The research by Verhaest and van der Velden (2012) and by Davia et al. (2017) suggests that overeducation is, at least to some degree, related to an excess supply of university graduates, implying that education policy should take closer account of the demand for graduate labor before agreeing to further increases in the number of university places. However, responsible education expansion is likely to be only part of the policy response, given that the study by McGuinness and Pouliakas (2017) identified a number of policy areas likely to be effective in tackling the problem of overeducation. Overeducation is partly related to inferior human capital, suggesting that policies aimed at improving the job readiness of students will help alleviate the problem (McGuinness and Pouliakas 2017; McGuinness, Whelan, and Bergin 2016). For example, increasing the practical aspects of degree programs, irrespective of the field of study, was found to reduce the incidence of initial mismatch for graduates (McGuinness et al. 2016). Job conditions are also part of (p.557) the problem, with the research suggesting that policies targeted at improving job quality and flexibility will also make a positive contribution (McGuinness and Pouliakas 2017). Finally, the quality of information that individuals acquire about a potential job before deciding to accept the post is also important, as is the method of job search that is undertaken (McGuinness and Pouliakas 2017; McGuinness et al. 2016), leading to the conclusion that policy initiatives that facilitate a smoother and more informed route into the labor market should also be pursued. For example, higher education work placements with the potential to develop into permanent posts and the provision of higher education job-placement assistance were found to have substantial impacts in reducing the incidence of graduate mismatch (McGuinness et al. 2016). Therefore, there are many initiatives that have the potential to lessen the impact of overeducation, and the research presented here suggests that many of these can be facilitated and coordinated at a central European level.

References

Bibliography references:

Barro, Robert J. 1997. Determinants of Economic Growth: A Cross-Country Empirical Study. Cambridge, MA: MIT Press.

Davia, M. A., McGuinness, S., and O’Connell, P. J. 2017. “Determinants of regional differences in rates of overeducation in Europe.” Social Science Research 63: 67–80.

Flisi, Sara, Valentina Goglio, Elena Meroni, Margarida Rodrigues, and Esperanza Vera-Toscano. 2014. “Occupational Mismatch in Europe: Understanding Overeducation and Overskilling for Policy Making.” JRC Science and Policy Report. Luxembourg: European Commission.

Mavromaras, Kostas, and Seamus McGuinness. 2012. “Overskilling Dynamics and Education Pathways.” Economics of Education Review 31 (5): 619–28.

McGuinness, Seamus. 2006. “Overeducation in the Labour Market.” Journal of Economic Surveys 20 (3): 387–418.

McGuinness, Seamus, Adele Whelan, and Adele Bergin. 2016. “Is There a Role for Higher Education Institutions in Improving the Quality of First Employment?” BE Journal of Economic Analysis and Policy 16 (4): 12–23.

(p.559) McGuinness, Seamus, and Pouliakas, K. 2017. “Deconstructing theories of overeducation in Europe: a wage decomposition approach.” In S. Polachek, K. Pouliakas, G. Russo, and K. Tatsiramos (eds) Skill Mismatch in Labor Markets (Research in Labor Economics, Volume 45), Emerald. Publishing Limited, Bingley, UK, 81–127.

McGuinness, S., Pouliakas, K., and Redmond, P. 2018. “Skills mismatch: concepts, measurement and policy approaches.” Journal of Economic Surveys. doi:10.1111/joes.12254

Phillips, Peter C. B. 1987. “Time Series Regression with a Unit Root.” Econometrica 55: 227–301.

Phillips, Peter C. B., and Pierre Perron. 1988. “Testing for Unit Roots in Time Series Regression.” Biometrika 75: 335–46.

Pouliakas, Konstantinos. 2012. “The skill mismatch challenge in Europe.” In European Commission (ed.). Employment and social developments in Europe. Luxembourg: Publications Office, 351–394.

Quintini, Glenda. 2011. “Over-Qualified or Under-Skilled: A Review of the Existing Literature.” OECD Social, Employment and Migration Working Paper 121. Paris: Organization for Economic Co-operation and Development.

Verhaest, Dieter, and Rolf van der Velden. 2012. “Cross-Country Differences in Graduate Overeducation.” European Sociological Review 29 (3): 642–53.

## Notes:

(1) Although their earnings are penalized relative to matched workers with similar levels of schooling, overeducated workers enjoy a wage premium relative to workers with lower levels of education doing the same job (McGuinness 2006).

(2) Pouliakas (2013) measured overeducation subjectively by comparing individual levels of education with the modal level of education in the chosen occupation. The study demonstrates that overeducation in the EU25 would have increased much more rapidly between 2001 and 2009 had occupational entry requirements remained at their 2001 levels.

(3) There is ample evidence in the literature of a higher prevalence of overeducation among graduates from fields such as Arts and Social Sciences.

(4) Derived from factor analyses carried out on subjective variables.

(5) Deviations of the observed rate from the natural rate.

(6) Measured by the ratio between the share of workers with ISCED-5 educational attainment and the share of workers in professional-directive occupations—that is, ISCO groups I and II, which consist of legislators; senior officials and managers; corporate managers; managers of small enterprises; physical, mathematical, and engineering science professionals; life science and health professionals; teaching professionals; and other professionals.

(7) The descriptive analysis and the tests for long-term relationships also include Cyprus, Croatia, and Germany. These countries are excluded from later analysis because of missing or incomplete data.

(8) The augmented Dickey–Fuller (ADF) test is the most commonly used test for this purpose, but it can behave poorly, especially in the presence of serial correlation. Dickey and Fuller correct for serial correlation by including (p.558) lagged difference terms in the regression; however, the size and power of the ADF test are sensitive to the number of these terms. The nonparametric test developed by Phillips (1987) and Phillips and Perron (1988) allows for both heteroskedasticity and serial correlation in the error term.

(9) For the remaining seven countries, where overeducation was found to be more volatile, OLS can only be applied after each series is differenced a sufficient number of times to induce stationarity.

(10) Our dependent variable runs from 0 to 1, and a standard panel regression may generate predicted values that lie outside the 0 to 1 interval. However, the incidence of overeducation typically lies in the range of 10%–30%. This implies that there is no clustering around the extreme values of 0 or 1 and suggests that the use of a fractional outcome variable is not highly problematic in this instance.

(11) The 15- to 24-year-old age group was chosen on the basis that it allowed us to observe overeducation among young people across all levels of educational attainment.

(12) The results for the Slovak Republic are somewhat implausible and should be treated with caution because a visual inspection of the data suggests that the series is stationary, contrary to the test statistic result.

(13) Gross domestic expenditure on R&D from the OECD was used.

(14) Some existing research has indicated that overeducation tends to be lower in countries with more developed vocational pathways (Mavromaras and McGuinness 2012).