Overeducation and skill-biased technical change

The literature is vast. See Autor, Katz, and Kearney (2008) for an overview of the evidence.

Acemoglu (1999), Albrecht and Vroman (2002) and Dolado, Jansen, and Jimeno (2009).

A worker is defined as overeducated when she is employed in an occupation whose required qualifications are lower than her own educational level.

See also Wolff (2000, table 2.2) for additional evidence.

See also Chevalier (2003).

See Bowles, Gintis, and Osborne (2001) for a review.

See Goos and Manning (2007), Goos, Manning, and Salomons (2009) and Autor, Katz, and Kearney (2008) for evidence on job polarization in the UK, Europe, and the US.

Slonimczyk (2013) also invokes the idea of polarization to suggest an empirical connection from SBTC to overeducation.

Formally, overeducation arises in situations in which educated workers accept job offers from both educated and non educated firms (pooling equilibrium), as opposed to situations where educated workers do not take low qualification jobs and wait for better opportunities (separating equilibrium). Dolado, Jansen, and Jimeno (2009) extend that model with on the job search where overeducation is a more transitory situation.

Chassamboulli (2011) presents a similar model although does not analyze the issue of SBTC and focuses instead on the cyclical behaviour of skill mismatches.

In the literature there are directed search models with heterogeneous agents and wage posting (our model has wage bargaining instead), but they do not deal with the issue of overeducation. In Shi (2002) wage offers are made conditional on workers’ types, while in Peters (2010), Eeckhout and Kircher (2010), and Lang, Manove, and Dickens (2005) wage offers cannot be conditioned on them.

Chevalier and Lindley (2009), Chevalier (2003), Pryor and Schaffer (1997) and papers reviewed in Leuven and Oosterbeek (2011).

A similar assumption is made in Mortensen and Pissarides (1999).

We will consider only equilibria in which the educated and skilled will always target vacancies of the educated type, so we omit this choice to save notation.

Note that, as (8) must hold, from (3) and (6) workers always accept the job offers they receive.

Skilled educated workers must decide to participate in the educated segment. Otherwise non-skilled workers would be the only participants in it, which would be inconsistent with the restriction on p_j∣i’s.

Specifically, z_s∣e=μ(s, e)/(μ(s, e)+μ(ns, e)) when π_ns∣e=1, and z_s∣e=1 otherwise. Instead, for the non-educated segment, the analogous mapping has no discontinuity as both skills are hired there.

In their steady-state model, like in ours, equilibrium unemployment is neutral with respect to the common trend in market productivity, hence, after de-trending, a permanent skill-biased shock to technology can be interpreted as a mean-preserving spread change. Cuadras-Morató and Mateos-Planas (2006) perform the same type of exercise. In Acemoglu (1999) the cost of vacancies is endogenous and increases when, in the course of SBTC, the productivity of skilled workers increases, thus also rendering non-skilled workers less productive relative to aggregate variables. This is also true in Moore and Ranjan (2005) where a SBTC brings about shifts in sectoral demands.

For instance, Katz and Murphy (1992) or, more recently, Autor, Katz, and Kearney (2008), as well as Krusell et al. (2000) study the role of technology for the wage structure focusing on changes in the quality of labor inputs relative to skill-neutral changes. Thus, abstracting from changes in absolute output arising from factor-neutral technological change and from changes in the scale of the economy, non-skilled productivity falls and skilled productivity rises relative to aggregate output and, hence, relative to other aggregate variables. See, for example, Greenwood and Yorukoglu (1997) for a macroeconomic analysis containing this feature.

This is the version of SBTC in Albrecht and Vroman (2002) and Shi (2002).

First, as one of two possible equilibria, but eventually as a unique one.

This calibration follows Cuadras-Morató and Mateos-Planas (2006) closely. Further details and data sources can be found there.

Other specifications leave the results largely unaffected. This is the case, for example, when the changes in productivity are weighted by demographic size.

The quantitative implications for inequality and unemployment are obviously less satisfactory. The significance of overeducation for those variables is preserved though.

However, once the equilibrium has switched to the overeducation regime – perhaps because of SBTC – this change in the quality of educated workers would further increase, as a direct composition effect, the proportion of overeducated individuals. Evaluating its actual contribution to observed changes in overeducation would require the quantitative analysis of a more elaborated model, possibly involving a richer distribution of skill levels.

We thank the editor Arpad Abraham for his suggestions on this.