A signalling model of school grades: centralized versus decentralized examinations

In this paper we examine the signalling value for skills of different examination systems in relation to errors that may affect grades obtained by students. Firm use school grades as a signal of the effective skills of workers, taking into account that evaluation are effected by stochastic shocks. We show that more precise evaluation systems, being associated to a higher reactivity of wages to school grades, induce an higher level of student effort. However, the effect is heterogeneous: low ability students tend to react less compared to high ability students. Moreover, from our analysis, it emerges that individuals endowed with low abilities may prefer less accurate evaluation systems. Nevertheless, when productivity increases the convenience of these systems reduces and the number of individuals preferring them shrinks. Our analysis highlights an important trade-off between centralized and decentralized evaluation systems. Frequent evaluations, typical of decentralized systems, reduce the impact on grades of errors that influence student performance and in this way diminish signal noise, on the other hand, different teachers generally adopt different performance assessment standards, and this tends to produce noisier evaluations. Conversely, centralized systems use common evaluation standards, but their frequency is limited by relevant administration costs and then produce evaluations that are more affected by errors influencing student performance. In the final part of the paper we investigate the relationship between the optimal class size and evaluation systems. We show that under decentralized evaluation systems the class size also affects the signal noise, since larger classes may reduce the frequency of evaluations undertaken by teachers.