Non- and semi-parametric stochastic frontiers : a penalised spline approach

Hajargasht, Gholamreza. (2005). Non- and semi-parametric stochastic frontiers : a penalised spline approach PhD Thesis, School of Economics, The University of Queensland.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
THE19533.pdf Full text application/pdf 7.36MB 2
Author Hajargasht, Gholamreza.
Thesis Title Non- and semi-parametric stochastic frontiers : a penalised spline approach
School, Centre or Institute School of Economics
Institution The University of Queensland
Publication date 2005
Thesis type PhD Thesis
Total pages 172
Language eng
Subjects 1401 Economic Theory
Formatted abstract
      Two Methodologies have been mainly used for technical efficiency measurement: nonparametric (deterministic) data envelopment analysis (DEA); and parametric stochastic frontier analysis (SFA). This study uses recent advances in nonparametric regression estimation to relax the restrictive parametric assumptions on stochastic frontier models.

      A number of nonparametric estimation techniques are available from the popular kernel smoothing to methods that have been rarely used. Most of the previous studies on non- and semiparametric stochastic frontiers have used kernel smoothing. However generalization of these approaches to more complex variants of stochastic frontier models is not usually straightforward or practical. This study employs a recently emerging nonparametric technique known as penalized splines (or P-splines) to provide a robust nonparametric approach to efficiency measurement. The P-spline approach involves replacing the nonparametric component of the regression function with a spline function that is linear in the parameters. Constraints (penalties) are imposed on the parameters to prevent over-fitting. Estimation then can be carried out in a least squares, mixed model or Bayesian framework.

      This dissertation shows how P-splines can be used to estimate nonparametric versions of the main stochastic frontier models appearing in the literature: cross-sectional models; fixed and random effects panel data models with time-invariant or time-varying inefficiency effects; and models that incorporate environmental variables. The performance of different P-spline estimators is assessed against the performance of several parametric alternatives in a series of simulation experiments and applications to real data. The performance of the nonparametric estimators is found to be comparable with the parametric estimators when the parametric models make correct assumptions concerning underlying functional forms and outperform parametric estimators when the true technology is of unknown nonlinear form.
Keyword Production functions (Economic theory)
Data envelopment analysis.
Stochastic analysis.

Document type: Thesis
Collection: UQ Theses (RHD) - UQ staff and students only
Citation counts: Google Scholar Search Google Scholar
Created: Wed, 20 Oct 2010, 19:21:04 EST by Muhammad Noman Ali on behalf of The University of Queensland Library