Americans had accumulated more than 1 trillion dollars in student loan debt as of 31 December, 2013. While the press conveyed widespread concern over this number, it might be that efficient credit markets are allowing more individuals to invest in education, with large rewards in terms of future earnings. After all, young college graduates earned 62.5% more on average than high school graduates in 2013 (Taylor et al. 2014). However, researchers have started paying more attention to the fact that the huge average ‘college wage premium’ masks large differences in post-college earnings. In particular, an extensive body of research, reviewed in Leuven and Oosterbeek (2011), has established that a sizeable fraction of workers hold occupations that do not require as much schooling as they have acquired. Think, for example, of college-educated secretaries or school teachers with a PhD.
A robust finding across geographical contexts is that overeducated workers earn more than less educated workers in the same occupation, but a lot less than similarly educated workers in occupations that do require their level of schooling.
The typical wage regression on American data finds that one year of surplus schooling, above the level required for one’s job, only yields an increase of 4.3% in earnings, which is about half of the usual ‘returns to schooling’ estimates for non-overeducated workers.1 Such numbers might lead some to reconsider whether more schooling is a good investment, at least financially, given increasing tuition costs.
Ultimately, the ability of overeducated college graduates to repay their pile of loans will hinge on how their careers progress. Is overeducation a temporary phenomenon, largely restricted to the first couple of years after graduation? Do the earnings of initially overeducated workers catch up over time with those of their non-overeducated peers? Or, instead, do these workers tend to remain in low-paying occupations durably?
Overeducation over time
In Clark et al. (2014), we consider the effects of being overeducated on future employment and wages. Our main data set, the NLSY79, is a survey that follows a representative cohort of Americans who were between 14 and 22 years old when first interviewed in 1979. We look at what happens to them over the first decade after they enter the labour market.
To classify a worker as overeducated in her current job, we must determine a level of ‘required schooling’ for her occupation. We use a larger data set, the Current Population Survey, to identify the most common level of schooling in each occupation among workers of the same cohort as the NLSY 79 respondents.
We obtain that 17.9% of workers in our NLSY79 pooled sample have more schooling than the typical worker in their occupation.2 Among college graduates only, this proportion goes up to 37.4%, and it is even higher for those workers with some post-secondary schooling who didn’t graduate from a 4-year college, at 65.7%.
Figure 1 plots the fraction of the NLSY79 workers with at least some college who are overeducated, against the number of years since labour market entry. Overeducation does become less frequent later in the career but the bulk of it persists after 10 years. Looking at raw annual transitions, an overeducated worker has a 66% chance of remaining overeducated one year later and only a 25% chance of switching to a job that matches her schooling. In other words, overeducation should not be viewed as a transitory alternative to unemployment some workers experience while waiting for their ideal job.
Figure 1. Overeducation incidence (NLSY79 respondents with some college or more), from Clark et al. (2014)
Heterogeneity among the overeducated
Not all workers are equally likely to remain overeducated. The probability of finding a job for which the worker is not overeducated is one third as low among overeducated blacks than among overeducated whites, after controlling for observed and unobserved worker characteristics.3 Workers with higher cognitive ability (as measured by the AFQT test), are also less likely to be overeducated and more likely to exit that state. Our statistical model also attributes an important role to unobservable individual attributes, which are not measured in the NLSY79. More research is needed to determine which they are and whether they can be impacted by public policies, but college major and college quality are examples of likely candidates that we are currently investigating.
Unemployment versus underemployment
- Not only is it hard for many workers to transition out of employment for which they are overeducated, but they are also likely to face wage penalties even after they do.
For example, a worker who lucked out after being overeducated for the last four years earns on average 14.6% less than her colleague who wasn’t previously overeducated. These penalties do appear to get smaller after a few years but a larger sample would be necessary to determine whether they truly disappear in the long-run. Overall, the persistence and wage penalties that characterise overeducated employment may help explain why up to 21% of outstanding student loan balances were considered delinquent in 2011 by economists at the NY Fed.
Further, our results suggest an interesting parallel between unemployment and underemployment (i.e. overeducation). Which of these two options hurts a worker’s career prospects the most? A number of studies have established that past unemployment lowers a worker’s subsequent earnings. This could happen if unemployed workers see their unused skills decline or because failing to secure a job sooner affects how prospective employers evaluate their potential.
- Our wage regression estimates suggest that past underemployment spells generate ‘scarring effects’ similar to those associated with past unemployment spells.
Many unemployed workers may have to choose between accepting a job for which they are overeducated or waiting for one that matches their qualifications. A well-defined, quantitative evaluation of the impact of this choice on a workers’ career, would ideally require to specify and estimate an empirical model of schooling and employment decisions that is able to reproduce the reduced-form evidence described above.
A more general policy implication of our results is that underemployment indicators might gainfully complement the usual unemployment indicators. For example, employment agencies are often evaluated on their ability to get workers out of unemployment quickly. This could generate more overeducation as workers accept jobs for which they are overqualified for fear of being disqualified from unemployment benefits.
Clark B, C Joubert and A Maurel (2014), “The career prospects of overeducated Americans”, NBER Working Paper No. 20167 .
Leuven, E and H Oosterbeek (2011), “Overeducation and mismatch in the labor market”, in E Hanushek, S Machin and L Woessmann (eds), Handbook of the Economics of Education, Vol. 4, Elsevier.
Schmieder J , T von Wachter, and S Bender (2013), “The causal effect of unemployment duration on wages: Evidence from unemployment insurance extensions”, NBER Working Paper No. 19772 .
Taylor P, K Parker, R Morin, R Fry, E Patten and A Brown (2014), “The rising cost of not going to college”, Technical report, Pew Research Center.
1 Leuven and Oosterbeek (2011)
2 This is on the conservative side, relative to the studies reviewed in Leuven and Oosterbeek (2011). In
fact, to avoid overestimating the incidence and persistence of overeducation, if the two most frequent
levels of schooling are within 15 percentage points of each other, we treat both as possible required
3 We do so using a mixed proportional hazard model of the duration of the first overeducated employment spell.
4 A recent example is Schmieder et al. (2013)