Stochastic Nonparametric Envelopment of Panel Data: Frontier Estimation with Fixed and Random Effects Approaches

Stochastic nonparametric envelopment of data (StoNED) combines the virtues of data envelopment analysis (DEA) and stochastic frontier analysis (SFA) into a unified framework of frontier estimation. StoNED melds the nonparametric piece-wise linear DEA-type frontier with stochastic SFA-type inefficiency and noise terms. We show that the StoNED model can be estimated in the panel data setting in a fully nonparametric fashion. Both fixed and random effects approaches are adapted to the StoNED framework. To disentangle changes in technology and efficiency, a dynamic semiparametric variant of the StoNED model is developed. An application to the wholesale and retail industry illustrates the approach. Key Words: data envelopment analysis (DEA), nonparametric least squares, nonparametric regression, productive efficiency analysis, stochastic frontier analysis (SFA