1. Introduction

A substantial corpus of literature has emerged on the wage effects of educational mismatch, the bulk of it suggesting that overeducated (undereducated) workers tend to receive a pay penalty (premium) with respect to other individuals who, with the same level of schooling, are adequately educated for the jobs that they do.¹ The traditional explanation for the wage penalty associated with overeducation is based on the idea that overeducation implies an underutilization of workers’ competencies. Accordingly, skill mismatch is seen as the counterpart of educational mismatch. The assumption behind this interpretation is that achieving a certain level of education implies attaining the corresponding skill level. However, several studies published over the last two decades have questioned this hypothesis, pointing out that individuals with similar schooling may have very different skill levels (Green, McIntosh and Vignoles 1999 and 2002; Quintini 2011).

Different lines of research, responding to the heterogeneity in knowledge among workers with the same level of education, have converged to suggest that overeducation does not necessarily imply an underutilization of skills. Some overeducated workers could actually have lower skill levels than expected given their schooling. Accordingly, traditional measures of overeducation would overstate the actual level of educational mismatch in the labour market (Chevalier 2003).

The idea behind this reasoning is that individuals could compensate for a lower skill level with surplus education. The level of knowledge acquired over a specified number of years of attained schooling, or for a specific academic degree, will be higher if the innate abilities or the motivation of the individuals are higher. Conversely, individuals with less ability might need more years of schooling to achieve a specific skill level. In such cases, the individuals would be considered to be overeducated in terms of schooling, but since the surplus schooling is counterbalanced by a deficit in ability, the years of effective overeducation could be overestimated.

In this context, the aim of this article is to improve the accuracy of the traditional measure of educational mismatch – which focuses only on the quantity of schooling attained by workers, thus ignoring its quality in terms of skills – and hence contribute to a better understanding of the phenomenon. To do this, we propose a novel definition of schooling that takes into account the different levels of skills acquired by individuals with the same level of educational attainment, and compensation between schooling and skills. Although other contributions have mentioned the correlation between overeducation and underskilling, the literature has not, to the best of our knowledge, analysed the substitution between educational and skill mismatches. Moreover, skill mismatch measurement is challenging since data on workers’ real skill levels are rather sparse. We draw on data gathered by the Programme for the International Assessment of Adult Competencies (PIAAC) to assess whether individuals’ observed skills differ from expected skills, given their human capital. Assuming that skill mismatch can be explained by the differences between observed and expected skills, we estimate the rate of compensation between educational and skill mismatches in terms of wages. According to this estimation, we propose the concept of “effective schooling”, taking into account both the years of attained schooling and the skill mismatch. Therefore, effective schooling indicates the number of years of education adjusted in terms of skills. Lastly, we define the “effective educational mismatch” as effective schooling minus the years of schooling required by the job and assess the level of effective educational mismatch in a set of member countries of the Organisation for Economic Co-operation and Development (OECD). We conclude that overeducation figures tend to drop after accounting for workers’ skills, with the difference between observed and effective educational mismatch varying by country.

Since the publication of the PIAAC data, various papers have proposed improvements to the skill mismatch measures existing in the literature – see Allen, Levels and van der Velden (2013) and Pellizzari and Fichen (2017). These papers classify individuals as well-matched, overskilled or underskilled, depending on their proficiency compared to the skill level required to perform a job successfully. By contrast, our strategy relies on defining a sharper indicator of individuals’ human capital than the conventional measure (attained schooling), which can be accurately compared with the educational requirements of jobs, hence providing more reliable figures of educational mismatch than observed mismatch.

The structure of this article is as follows. In the second section, after a brief review of the literature, we examine the two pillars supporting our methodological proposal: (i) educational and skill mismatches have significant and distinct effects on wages and (ii) educational and skill mismatches might compensate for each other in terms of wages. To do this, the third section presents the data used in this study and offers an exploratory analysis. The fourth section then analyses the substitution between skill and educational mismatches in terms of wages, giving rise to the definition of effective schooling and effective educational mismatch. In the fifth section, we use these concepts to assess the level of effective over- and undereducation in a set of countries. The article closes, in the sixth section, with some conclusions.

2. Literature review

The idea that individuals who have received the same education are heterogeneous in terms of skills has led to three lines of research. The first uses panel data techniques to control for unobserved heterogeneity (Bauer 2002; Frenette 2004; Lindley and McIntosh 2009; Mavromaras et al. 2013), finding that the omission of unobserved abilities overstates the pay penalty associated with overeducation. The second line of research differentiates between educational and skill mismatches, showing that overeducation weakly correlates with skill underutilization and that the wage effects of schooling mismatch are unchanged when controlling for skill mismatch (Allen and Van der Velden 2001; Di Pietro and Urwin 2006; Green and McIntosh 2007). The third strand of this literature takes account of skill heterogeneity to distinguish between genuine and formal overeducation and estimate the monetary returns to overeducation depending on workers’ skill levels. This suggests that, when skills are taken into consideration, overeducated workers do not constitute a homogeneous group (Chevalier 2003; Green and Zhu 2010; Pecoraro 2014; Mateos-Romeros and Salinas-Jiménez 2017). All in all, the results reported by these contributions clearly reject the hypothesis of skill homogeneity within educational levels.

There is broad consensus surrounding the idea that any level (or area) of education other than that required to get or do a job creates educational mismatch. This educational mismatch can be measured through job analysis, workers’ self-assessment or the “realized matches” method, using objective, subjective or statistical methods, respectively (Leuven and Oosterbeek 2011). By contrast, both the concept and the measurement of skill mismatch appear to be less specific. As stated by McGuinness, Pouliakas and Redmon (2018), the term “skill mismatch” is used in the literature to refer not only to over- and underskilling related to an individual’s attributes, but also to skill gaps and skill shortages at the firm level. Objective measures of workers’ skills are rarely available in data sets and authors approach over-/underskilling using workers’ subjective responses to questions about the extent to which their knowledge is above, below or unrelated to their job requirements. On the other hand, skill gaps and skill shortages are usually measured using employers’ information about the difficulties that they have finding workers or applicants with the right competencies to carry out jobs or fill vacancies adequately.

Some papers have used PIAAC data to define new measures of skill mismatches. Allen, Levels and van der Velden (2013) compare individual standardized scores for academic skills with the standardized average use of skills at work. They define a scale according to which workers are well-matched if the difference between skill levels and skill use is no more than 1.5 points above or below zero; otherwise, skills would be over- or underutilized. Pellizzari and Fichen (2017) focus on workers who describe themselves as being well-matched, in the sense that, according to their skills, they do not require any additional training to perform their tasks but cannot manage more demanding duties. They define minimum and maximum levels of assessed skills by occupation for well-matched workers, and individuals whose skills are below or above the interval are classified as underskilled and overskilled, respectively.² Flisi et al. (2017) analyse occupational mismatch based on both educational and skill variables. They compute a wide set of both objective and subjective measures indicating whether individual educational/skill levels differ from what is required to perform the job. The main conclusion of these studies, as explicitly stated by Flisi et al. (2017), is that, rather than counterparts, educational mismatch and skill mismatch are two distinct phenomena.

The above-mentioned papers contribute to a better understanding of skill mismatches. However, the extent to which skill use at work is representative of the knowledge required by the labour market is a matter of concern. As Pellizzari and Fichen (2013) acknowledge, skill use at work strongly depends on the effort made by the individual. By contrast, educational level appears to be a more general indicator, although the use of schooling to measure human capital has the disadvantage of ignoring the heterogeneity of skills within educational levels. To sum up, a focus on either skill use at work or attained schooling to measure labour market mismatches results in unclear indicators. This article addresses this limitation by looking not only at the years of attained schooling but also at workers’ real skill levels. By defining “effective schooling”, we propose an indicator that is sharper than attained education and broader than skill use at work. Since effective schooling measures the years of attained education adjusted by the skills actually acquired by the individual, we can draw a fair comparison with the education required by the job, thus providing more reliable educational mismatch figures than the data documented in previous papers.

3. Wage effects of educational and skill mismatches

4. An analysis of the substitution between skill and educational mismatches: Effective schooling and effective educational mismatch

The wage equation (5) can be used to compare how the market evaluates the mismatch in schooling and the mismatch in skills in monetary terms. Once the equation has been estimated, it is possible to calculate the rate of substitution between educational mismatch and skill mismatch to keep the individual’s wage unchanged. In other words, we can calculate how many additional years of schooling are needed to counterbalance a one-unit skill mismatch and vice versa – how many units of skill surplus (deficit) are needed to compensate for one year of undereducation (overeducation).

Accepting the hypothesis that there is a direct correspondence between productivity and wages, a unit of skill mismatch can, based on the estimation of equation (5), be compensated for in terms of wages by r years of schooling, where r is the rate of substitution:

$r\text{=}\frac{{\beta }_5}{{\beta }_6}$ (10)

This ratio can be used to derive the mismatch of skills necessary to compensate for the mismatch in schooling and ensure that the worker is adequately qualified (that is, has no educational or skill mismatch) in terms of wages. Specifically, r units of skills over (under) the expected level can be compensated for by one year of undereducation (overeducation).

Following this reasoning, it is also possible to distinguish between attained schooling declared in the PIAAC survey (S) and effective schooling as the years of schooling that account for potential skill mismatches expressed in equivalent terms (S_e):

$S_e\text{=}S\text{+}r.\mathit{\text{skmm}}$ (11)

Hence, effective educational mismatch – as opposed to observed educational mismatch – can be defined as:

$\begin{array}{l}\mathit{\text{Observed}} \mathit{\text{educational}} \mathit{\text{mismatch}}=S–S^{*}\\ \mathit{\text{Effective}} \mathit{\text{educational}} \mathit{\text{mismatch}}=S_e–S^{*}\end{array}$ (12)

where S* indicates the years of attained schooling required by the job. To sum up, an individual who claims to be overeducated is only effectively overeducated if:

• (i)  their observed competencies are at least equal to the expected skills corresponding to their education and experience; or

• (ii)  the years of declared overeducation more than compensate for their skill deficit.

In turn, an individual who claims to be undereducated might be effectively overeducated if they have a skill surplus that more than compensates for their lack of schooling.

Table 4 provides the results of the estimation of equation (5) using PIAAC data and table 5 shows the rate of substitution between educational and skill mismatch, calculated from the values provided by this estimation.^{8, 9}

As has been widely documented in the literature, the returns to attained schooling are positive but their magnitude depends on the extent to which the years of attained schooling meet the requirements of the job, and the penalty associated with the years of educational mismatch is lower than the above returns. Moreover, it is the competencies actually acquired by workers throughout their formal education, rather than just the quantity of schooling, that play a significant role in wage determination: individuals with higher-than-expected skills receive a wage premium with respect to their less capable classmates.¹⁰ Lastly, as table 5 indicates, an individual might compensate for a one-unit mismatch of standardized skills with around 1.6 years of additional schooling. In other words, on average, if a student does not acquire the expected skills within the course year, it takes 1.6 years of additional education to reach the skills of another student who has acquired them within the year. In this case, the extra schooling of the less able student should not be seen as overeducation, but as compensation for their lower ability.

5. Observed educational mismatch and effective educational mismatch: Results by country

In this section, we present the results of estimating effective educational mismatch by country, using our data set. The figures reported in table 6 illustrate the idea that an individual who has the educational level required by a job will not necessarily have acquired the expected skill level from that education.¹¹ Hence, it is important, for some countries especially, to make a distinction between observed and effective educational mismatch. Thus, the percentage of adequately educated workers whose skills are lower than expected ranges from less than one third in Japan to more than three quarters in Ireland.

In view of the above, we can expect one year of attained schooling to lead to different years of effective schooling depending on the competencies actually acquired by individuals. Table 7 provides information on the difference between these two variables as well as the average attained and effective schooling,¹² by country and for the workforce as a whole, while table 8 reports the same information, focusing on workers who claim to be over- or undereducated. Note that effective schooling is slightly lower than attained schooling for the sample as a whole and varies greatly by country. Thus, effective schooling falls far below attained schooling in Ireland – and, to a lesser extent, in Spain and the United Kingdom – while the opposite is the case for Japan and Finland. These results are in line with the findings reported by Calero, Murillo Huertas and Raymond (2021) who, using the same data set, conclude that the efficiency with which the education system transforms schooling into skills varies notably by country. In particular, they single out Japan and Finland as being among the most efficient, with Ireland, the United Kingdom and Spain at the opposite end of the scale.

Related to the above, we should note that the average years of educational mismatch observed in the countries in our sample differ substantially when considering compensation between skills and schooling (table 7 and figure 1). Especially striking is the case of Ireland, declaring on average 0.77 years of overeducation for their effectively undereducated individuals, with Italy showing a similar pattern. Finland and Japan, by contrast, underestimate their corresponding average years of overeducation, as effective educational mismatch increases by about half a year, taking into account workers’ competencies.

Figure 1

Observed educational mismatch vs effective educational mismatch.

Source: Own calculations based on PIAAC data.

Lastly, table 8 presents the values for observed and effective educational mismatch focusing particularly on mismatched individuals. We find that Ireland, followed by the Mediterranean countries, has the highest value for the difference between attained and effective schooling for overeducated workers. This indicates that undereducation is exacerbated in these countries – as well as in the Republic of Korea – when effective schooling is considered. By contrast, and in line with the results reported in tables 6 and 7, Japan and Finland have higher effective than observed over- and undereducation figures, suggesting that overeducated (undereducated) workers in the two countries have, on average, a higher-than-expected skill level given their education and experience.

The values reported in tables 7 and 8 and in figure 1 are in line with the findings reported by Flisi et al. (2017). They put Ireland, Italy and Spain in the category of countries with high levels of overeducation whose workers do not, however, possess extra skills. This is consistent with the sizeable difference between observed and effective overeducation that we find for these countries. Flisi et al. (2017) also find that workers in Finland have the right educational level required by their jobs but possess a high level of skills, which is consistent with our figures for the effective educational mismatch being higher than the observed educational mismatch.

An analysis of the reasons for the disparate educational mismatch figures by country is beyond the scope of this article and is available in other studies (see table 9). Nonetheless, we would like to point out that, as stated by Flisi et al. (2017), a plausible explanation would lie in the idiosyncrasy of national educational systems. Thus, schooling in Ireland, Italy and Spain provides very little tracking and a high level of general education, meaning that students are poorly matched to labour market opportunities. By contrast, Scandinavian educational systems are more stratified and offer solid vocational education tracks, resulting in less overeducation.

6. Conclusions

The issue of educational mismatch has been studied at length in the literature. Typically, researchers identify mismatched workers by comparing individuals’ formal education with the level required to do or to get a job. However, the empirical evidence shows that the years of attained schooling do not accurately reflect individuals’ actual skills. New trends in the literature on labour mismatches suggest that the wage effects of educational mismatch might be associated with workers having different skills to those that are expected given their level of education. Our article has built upon these trends, aiming to shed light on the question of whether individuals might compensate for a deficit (surplus) in skills with an excess (shortage) of education. The objective was to analyse whether at least some of the observed educational mismatch is down to apparent rather than actual overeducation (undereducation) considering that mismatched workers may have different skills to those expected.

The PIAAC’s Survey of Adult Skills provides data on both individuals’ attained years of education and their observed skills. We have used these data to implement an objective method to compare formal education with the expected skills that individuals should have acquired through schooling. The skill mismatch is calculated by comparing individuals’ observed skills with those expected given their human capital. We then explicitly measure the compensation between educational and skill mismatches in terms of wages in the labour market and define a new indicator of educational mismatch taking into account this substitution between educational and skill mismatches. We apply this methodology to study the level of effective – or real – educational mismatch in a sample of countries included in the PIAAC database.

We have found that, overall, observed competencies are frequently lower than expected, especially in certain countries where overeducation is, thus, more apparent than real. Therefore, overeducation figures tend to decrease when the compensation between surplus schooling and deficient skills is considered, with some countries declaring overeducation for effectively undereducated individuals. By contrast, other countries have populations whose average skills surpass the competencies that can be expected based on human capital. As the inputs considered in our study to approximate expected skills are the years of formal education and the years of labour market experience, a skill shortage can be attributed to either ineffective knowledge transmission or a production process of goods and services that does not properly exploit employee potential.

Although the topic of educational and skill mismatches has been extensively analysed in the literature, some aspects still remain underdeveloped and warrant further research. This is the case of the analysis of the substitution between education and skills. As McGuinness, Pouliakas and Redmon (2018) emphasize, the incidence and effects of labour mismatch strongly depend on how it is measured (e.g. in terms of schooling or skill mismatches). How accurately labour mismatch is defined is even more important, since the concept of educational mismatch could play a different role if it is compensated for, to some extent, by skill mismatches.

Being able to measure educational mismatch reliably is also significant from an economic policy perspective. If overeducation truly exists, it suggests an inefficient allocation of resources, where overeducated workers fail to take full advantage of the capabilities provided by their formal education. Thus, from both an individual and a societal standpoint, these resources are misallocated. In this regard, it is essential to take skill heterogeneity among workers with similar schooling into account when designing effective policies. Policymakers should acknowledge that overeducation does not necessarily imply skill underutilization and tailor interventions accordingly.

Our findings suggest that it is crucial to enhance the effectiveness with which educational systems transform schooling into skills, especially in countries such as Ireland or Italy, where effective educational mismatch is capable of turning the observed overeducation figures around. Policies focusing on enhancing the quality of education and training programmes should be a priority to ensure that graduates have the skills and knowledge needed to perform their jobs. Promoting programmes that meet the evolving needs of the labour market and provide individuals with key skills can help reduce overeducation by bringing education into line with job requirements. Furthermore, addressing the informal sector and promoting formal employment opportunities can help reduce overeducation by providing workers with jobs that match their skills and educational levels. Closing the gap between actual and expected competencies, together with a better match between individuals’ education and changing labour market needs, would maximize both individual and social returns to education. As stated by Brunello and Wruuck (2021), this challenge is a responsibility not only for individuals and schools but also for firms and governments.

Acknowledgements

We thank Séamus McGuinness and María del Mar Salinas for their valuable comments. Inés P. Murillo Huertas acknowledges financial support under project AEI/10.13039/501100011033.

Competing interests

The authors declare that they have no competing interests.

Appendix 1

Appendix 2. Results using literacy skills to approximate skill mismatch

Figure A1

Attained schooling (S) vs effective schooling (S_e)

Source: Own calculations based on PIAAC data.