To study education as a complex production process in a noisy and heterogeneous setting, this paper suggests to using a stochastic frontier model estimated by a local maximum likelihood approach (LMLSF). The LMLSF smoothly combines the virtues of the non-parametric Data Envelopment Analysis model and the semi-parametric Stochastic Frontier model. Additionally, by the LMLSF approach one can deduce the effectiveness of resources by examining the impact of inputs on the frontier. Indeed, while efficiency estimations (i.e., doing the things right) received considerable attention in the literature, the analysis of effectiveness (i.e., doing the right things) is less explored. The approach is illustrated on a sample of Dutch primary education pupils. We examine the effectiveness of instruction time, experience of the teacher, and student care (both social worker and psychologist) on educational attainments of native and non-native students.
Keywords: Stochastic Frontier Analysis; Data Envelopment Analysis; Local Maximum Likelihood; Education; Student care
JEL-classi_cation: C14, C25, I21