Better educated workers may earn higher wages for two possible reasons: because they acquire more skills or because more education signals their intellect and enthusiasm. Economists believe that signalling plays a role and worry that this may cause overeducation. Georg Graetz writes that signalling may lead students to invest less effort into acquiring skills than they would if employers had perfect information about their abilities.
A long-standing question in labour economics is why better educated workers earn higher wages. Is it because they acquire more skills (the human capital view) or because education is simply a way of proving that they are smart and hard-working (the signalling view)? Common sense as well as empirical evidence suggest that the truth lies somewhere in between. Both explanations have some merit, with the precise importance of each in accounting for the wage return to schooling being a matter of ongoing debate. But there is a widespread view among economists that the more important signalling is, the more we need to worry about people getting “too much” education. This is because under the signalling mechanism, the private gains from education are not only due to improved productive skills, but also due to being able to differentiate oneself when competing for jobs. In a new working paper, I add nuance to the discussion, challenging the view that people may get too much education due to signalling. I argue that a world in which signalling is important may be one in which students put in less effort into their education than they would if employers had perfect information.
Existing signalling models highlight the difficulty that employers face in assessing an applicant’s productivity. In the simplest textbook model, workers’ productivity is fully determined by their innate (that is, pre-schooling) abilities, while schooling has no productive effects whatsoever. However, if schooling requires considerable effort, but less so for higher-ability individuals, then obtaining an education will serve as a credible signal for productivity in the labour market. And assuming workers are compensated according to their productivity, education will carry a wage premium. In this case, education is socially wasteful in the sense that it is costly but does not enhance the productive capacity of the workforce. Its gains are purely private, not social.
For greater realism, consider three extensions to this simple story. First, suppose that schooling does raise students’ productivity. This means that education is no longer purely wasteful, but students will still tend to obtain too much of it, relative to an efficient, full-information benchmark. This is because they gain from it not only because of its productivity-enhancing role, but also because it signals their innate ability.
Second, in fact education does not reveal productivity perfectly. While in principle there is a wealth of information that employers may consider – detailed transcripts, reference letters, extracurricular activities – this is costly to process, and in many cases may not be relevant for the job. In practice employers pay attention to only a limited set of criteria, such as degree class, at least at the stage of initial screening.
Third, while employers may struggle to assess the productivity of graduates newly entering the labour market, workers’ productivity becomes less of a secret as they spend time in the market. This process is known as “employer learning” and has been extensively documented.
I incorporate these features into a formal model in which students make educational choices, taking into account the (monetary and other) costs of studying and the gains in future earnings. Unlike prior literature, I explicitly distinguish between time spent in education and the amount of productive skills obtained. This distinction matters because initially, the labour market rewards these two aspects differently. While length of education – whether someone has attended college, for instance – is one of the most salient aspects of any CV, how many and which skills someone has acquired is much harder to ascertain and will only be revealed gradually. Importantly, allowing for skill acquisition to be imperfectly observed helps account for the patterns of employer learning that we see in the data – indeed, employers must be uncertain about something for there to be subsequent learning.
Imperfectly observed skill acquisition has important consequences for the incentives that students face when making educational choices. Students may focus on achievements that are visible to employers, as opposed to working hard to acquire productive skills. For instance, getting a college degree may be more important than the content studied; and students may prioritise good grades and thus prefer “easy” courses if employers pay attention to degree class but not to the actual skills acquired. We may then end up with a workforce that is less skilled than it would be if acquired skills were immediately visible to employers.
To gauge the relevance of such scenarios for the real world, I estimate and quantify my model using data from the US National Longitudinal Survey of Youth (NLSY). I demonstrate that my model is fully consistent with prior studies of employer learning that used these same data. I then quantify the extent to which skill acquisition during school is initially imperfectly observed by employers. I find that individuals entering the labour market right after high school face particularly severe information frictions – employers know essentially nothing about the productive skills they acquired in school. College graduates seem to face less severe information frictions, consistent with earlier work. I estimate that information frictions cause substantial losses to GDP and that in a perfect-information scenario, the fraction of a cohort attending college would be lower. Information frictions cause students to spend too much time in education, even as students learn too little.
What can policy do to address such inefficiencies? If there was a way to reduce informational barriers (for instance, taking advantage of recent advances in artificial intelligence), would it be desirable to do so? Note that “perfect information” in the model means that employers not only see exactly what skills an applicant has acquired, but also all other traits of an applicant that are relevant for productivity, such as their pre-schooling abilities and their personality. If we invented a test that would perfectly reveal skill acquisition only, then my model predicts that students would over-invest in order to signal those other traits. Thus, any efforts to lessen information frictions would need to have a broader focus than just skill acquisition.
Besides addressing informational frictions directly, there are of course other ways of incentivising students to study hard, such as exams and other mandatory assignments. While educators have known the value of exams for millennia, when seen through the lens of economic theory, exams have not appeared in a positive light. Human capital theory has no role for them, as students would acquire the optimal level of skills regardless. And in conventional signalling models, exams have a clear purpose but are socially wasteful. In contrast, my model suggests a key role for exams in inducing students to acquire skills that are productive in the labour market. Educators should keep this in mind when selecting examinable material.
- This blog post is based on the CEP Discussion Paper 1919 Imperfect Signals by Georg Graetz.
- The post represents the views of its authors, not the position of LSE Business Review or the London School of Economics.
- Featured image provided by Shutterstock
- When you leave a comment, you’re agreeing to our Comment Policy.