The over-education wage penalty among PhD holders: a European perspective

ABSTRACTUsing a unique pan-European dataset, we rely on two alternative measures of over-education and control stepwise for four groups of covariates in order to interpret the over-education wage penalty in light of theoretical models. Firstly, it appears that a significant fraction (i.e. between 1/5 and 1/3) of PhD holders in Europe are genuinely over-educated. Secondly, these genuinely over-educated PhD holders are found to face a substantial wage penalty (ranging from 15 to almost 30%) with respect to their well-matched counterparts. Finally, unconditional quantile regressions highlight that the over-education wage penalty among PhD holders increases greatly along the wage distribution.KEYWORDS: Phd graduatesover-educationover-skillingjob satisfactionwagesEuropeJEL CODES: J21J24 Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 In the European Union, the statistics point in the same direction: the number of newly enrolled doctoral students aged between 24 and 35 increased by almost 27% between 2013 and 2018 (from around 71,000 to almost 90,000), while the number of doctoral students rose from around 735,000 to 779,000 between 2013 and 2019 (European Commission Citation2020; Eurostat, Citation2023). Furthermore, in 2019, the number of new doctorate holders was around 121,000 in the EU-28 (Eurostat, Citation2023).2 This said, it should be noted that a significant number of people embark on a thesis for reasons other than obtaining a job requiring a PhD. Intrinsic motivation and intellectual development are also important drivers (Hnatkova et al. Citation2022). In addition, studies show that many PhD graduates, despite holding jobs for which a PhD is not essential (and for which they are therefore likely to be over-educated), can nevertheless leverage their degree to improve their career prospects. More specifically, as Boman et al. (Citation2021) point out, in many jobs, a doctorate, even if not required, is desired or valued, so that the person with a doctorate has a more interesting and rewarding job, which also makes it easier to access more responsibility, promotion or other benefits (pecuniary or otherwise).3 The term ‘voluntary’ should be interpreted with caution as it may obviously be a constrained choice.4 The study by Ermini, Papi, and Scaturro (Citation2017), based on four cohorts of Italian doctoral graduates (relating to the years 2004, 2006, 2008 and 2010), also finds that jobs held by doctoral graduates in academia and the research sector are more often associated with a successful match. The analysis by Boman et al. (Citation2021), which is based on a career tracking survey of doctoral graduates between 2006 and 2020 in nine European universities, concludes that almost half of doctoral graduates are employed in jobs that do not require a doctorate, but also that overeducation is most prevalent outside universities and research institutions.5 By relying on the WA method, over-education is constructed with replies from questions concerning the usefulness of the PhD. In this regards, it has been underlined how the exact phrasing varies substantially across studies: some interviews refer to recruiting standards (Duncan and Hoffman Citation1981), while others to the education needed to perform the job (Hartog and Oosterbeek Citation1988). Evidence also shows that the same person responds differently to similar questions, and it is not clear whether and to what extent these variations in framing and phrasing cause differences in the measured levels of required education (Green, Myerson, and Ostaszewski Citation1999). For these reasons, we calculated the over-education variable in two alternative ways, namely on the basis of the usefulness of the PhD title to do and to get the current job position respectively.6 The over-education variable computed on the basis of the educational level to get the job is used as a robustness test. Estimates based on this alternative measure are very similar to those based on the educational level required to do the job. Therefore, the latter are only reported for the stepwise OLS analysis (without interaction effects). The results of other specifications, which corroborate our findings, are available on request.7 To calculate the gross hourly wage, we first converted the variable Q50_X, i.e. the ‘gross monthly earnings’ of workers in the various countries analysed, into a common currency, namely the euro. To do this, we use the average exchange rate between national currencies and the euro in 2014. Next, we calculated the number of hours worked per month by multiplying the ‘hours worked per week’ (i.e. variable Q10_2) by 4.43 (as it is generally accepted that a month comprises an average of 4.33 weeks). Finally, by dividing the gross monthly wage (expressed in euros) by the hours worked per month, we obtain the gross hourly wage.8 A worker is considered as under-educated if her/his level of education is lower than that required for her/his job.9 Given the distribution of these variables, we have chosen to include 2 dummies for ICT skills (i.e. for moderate and advanced levels) and only 1 for numeracy and literacy skills respectively (i.e. for the advanced level).10 This fourth group of variables refers to the career aspirations of employees before they started working for their current employer. In practice, it turns out that for many of these employees, these aspirations are not, or are no longer, being met. This is particularly the case if we focus on the desire of certain employees to have a job that corresponds to their qualifications. Indeed, when we calculate the correlation between this variable and the overeducation variable (our variable of interest), we find that it is very weak (equal to -0.04) and therefore that there is no systematic relationship between these two variables.11 The European Skills and Jobs Survey relies on a quota approach. Quota sampling can achieve representativeness by using quotas and weights which align the sample with the population on key variables. This method ensures that the sample is representative for the key control variables and makes it likely for other variables that correlate with them. Consequently, we have used weights in our analysis. Specifically, we relied on variable 291 (i.e. ‘Weight_Country_with_education’). For more details, see Ipsos MORI (Citation2014).12 Specifically, we excluded from our analysis individuals who answered ‘no’ to the following question from the ESJS: ‘Did you do any paid work in the last 7 days, even if it was for one hour?’.13 To examine whether over-education among PhD holders is a temporary phenomenon (i.e. whether it occurs more often just after the doctorate is obtained), we calculated the incidence of over-education according to the year in which doctoral graduates obtained their degree. We considered different thresholds. In all cases, the incidence remains very stable, close to its mean value. Our estimates (available on request) therefore suggest that over-education among PhD holders is fairly persistent and stable across cohorts of PhD holders.14 We also examined the incidence of over-education, as well as the distribution of PhD graduates, by sector, occupation and field of study. Descriptive statistics presented in Appendix Table A-3 show that 46% of doctorate holders work in the private sector and that their probability of being over-educated there is 87%, i.e. 14% points higher than in the public sector (including other organisations, such as not-for-profit trusts, charities and non-governmental organisations). Furthermore, we note that almost all doctorate holders are either managers (11%), professionals (56%), technicians and associate professionals (15%) or clerical support workers (13%). In these occupations, the incidence of over-education varies from 70% (among professionals) to 96% (among clerical support workers). As for the breakdown of doctorate holders by field of study, we find that 22% of them hold degrees in Economics, business, law and finance, 15% respectively in Natural sciences and Engineering sciences, 14% in Humanities, languages and arts, and 11% in Teacher training and education science. Other fields of study (e.g. Computer sciences or Medecine and health-related sciences) gather less than 10% of PhD holders in our sample. Finally, as regards the incidence of over-education among doctoral graduates by field of study, it fluctuates between 49% for graduates in Medecine and health-related sciences and 88% for graduates in Humanities, languages and arts.15 The severity of over-education has also been investigated. Therefore, we classified over-educated PhD holders in three categories according to the degree required to do their jobs. We have qualified over-education as ‘mild’ when the level of education required for the job corresponds to ISCED 5 (i.e. tertiary education – first level), ‘severe’ when it corresponds to ISCED 3 or 4 (i.e. upper secondary education or pre-vocational post-secondary education), and ‘very severe’ when it corresponds to ISCED 1 or 2 (i.e. primary or lower secondary education at most). Unfortunately, when the level required to do a job corresponds to a tertiary diploma – first level (ISCED 5), our data do not allow us to have more details and in particular to identify whether it is a bachelor’s or a master’s degree that is needed. Clearly, having this additional information would have been a plus, as the consequences of over-education are probably most acute when only a bachelor's degree is required. Be that as it may, our results (available on request) show that in 80 to 86% of cases, depending on whether we consider our sample as a whole, the private or public sector, over-education can be described as mild, insofar as the PhD graduates actually do a job for which a tertiary degree is sufficient. Moreover, they suggest that the problem of over-education is somewhat less acute in the public sector than in the private sector. Indeed, while in the private sector over-education is severe or very severe for over 16% of over-educated PhDs, this is the case for less than 12% of over-educated PhDs in the public sector.16 The complete set of regression results, not reported here due to space constraints, is available on request.17 See Section 3.1 and Appendix Table A-2 for a detailed description of these characteristics.18 As a sensitivity test, we re-estimated our benchmark equation, including all covariates, separately for PhD holders working in the private sector (i.e. private companies and partnerships), the public sector (i.e. national, regional and local public organisations) and other organisations (i.e. not-for-profit trusts, charities, non-governmental organisations and other organisations) respectively. Our estimates, available on request, show that the wage penalty associated with over-education is greatest in the private sector (-23.2%), lower in the public sector (-16.2%) and not significant in other organisations. The lower penalty in the public sector is probably due to the fact that, in this this sector, wages are generally based on fairly precise job classifications, where seniority-based pay often continues to play an important role, with little or no scope for individual wage negotiation. Overall, this results in a relatively compressed wage distribution, where low-skilled workers tend to earn more than their private sector counterparts, while the reverse is true for high-skilled workers, and therefore also for PhD holders (Bargain and Melly Citation2008; Lucifora and Meurs Citation2006). In the private sector, on the other hand, salary dispersion is typically greater and highly qualified positions (which in some cases require a doctorate) are generally much better paid than in the public sector (Giordano et al. Citation2011). As the academic sector represents only a small proportion of total public employment, our results are not directly comparable with those of previous studies (e.g. Bender and Heywood Citation2009; Gaeta, Lavadera, and Pastore Citation2022) comparing wage penalties of over-educated PhD holders inside and outside academia.19 We also re-estimated our benchmark equation according to the severity of the over-education problem (e.g. qualified as mild, severe or very severe, see footnote 16 for the description of these categories). All sectors combined, our results (available on request) show that the wage penalty increases as the over-education problem worsens. The penalty is estimated at −24.9% and −34.5% in severe and very severe cases respectively. In mild cases, the penalty is not statistically significant. This non-significant outcome is likely to be at least partly driven by the fact that we are unable to distinguish PhD graduates in jobs requiring a bachelor’s degree from those in jobs requiring a master’s degree. Interestingly, though, if we turn to the estimates for the private sector, we find that the penalty is statistically significant and equal to −16.4% in mild cases and reaches −51.9% in more severe cases. In the public sector, the over-education wage penalty is estimated at −9.7 and −20.1% in mild and more severe cases respectively. However, none of these coefficients is statistically significant (again, probably because we can't distinguish between jobs requiring a bachelor’s or a master’s degree, but also because of a small sample issue, particularly for the more severe cases of over-education).20 The ORU (Over-, Required, and Under-education) approach, suggested by Duncan and Hoffman (Citation1981), enables to estimate the wage impact of over-education conditional on the level of education required to perform a job. The standard result from this literature is that over-educated workers earn a wage premium over their colleagues who do the same job but are adequately educated to do it (i.e. well-matched). In order to examine whether a similar result is obtained with our data, we re-estimated the over-education wage differential for doctoral graduates using an ORU approach. Our results, available on request, show that wages rise on average by 7.8% when the level of required education for a job increases by one year. Moreover, they indicate that a one-year increase in over-education among PhD holders improves wages by 3.9%. These estimates, as expected, suggest that over-educated PhD holders earn a wage premium over their well-matched colleagues. They are also in line with previous evidence showing that over-educated PhD holders can nevertheless take advantage of their degree to improve their career prospects (see e.g. Boman et al. (Citation2021)).21 See footnote 4.22 Detailed regression results, not reported here due to space constraints, are available on request.23 We thank an anonymous referee for making these valuable comments.