During the past two decades, the theoretical literature on semi-and non-parametric methods for estimating and testing micro-economic models has grown tremendously. See, for example, the surveys in Powell (1994) or Horowitz (1993). Applications of these methods, however, are relatively rare and have had mixed success. To bridge the gap between the theory of semi-and non-parametric methods and applied micro-econometrics, the department of Econometrics at Tilburg University, with support of CentER, the Journal of Applied Econometrics, and the Netherlands Royal Academy of Arts and Sciences (KNAW), organized a conference on the application of semiparametric methods for micro-data in October 1997.
The seven papers in this issue are a selection of those presented at the conference. The papers apply recently developed semi-and non-parametric methods to a variety of micro-econometric questions. They illustrate the state of the art in the area and show both the beneยฎts and the diculties of semi-and non-parametric methods. In particular, they address many issues that may inยฏuence the choice between these methods and conventional parametric methods. We hope these papers will help applied econometricians to better understand the relative merits of the various methods for applied research. We also hope that the papers will be useful to theoretical econometricians by providing information on the practical problems that need to be overcome to make newly developed methods truly useful for applied work.
Blundell, Duncan and Pendakur use semi-parametric methods to estimate Engel curves and the impact of demographics on consumer demand in the UK. Starting with a log cost function that is additively separable into a function of prices and demographics and a function of prices and utilities, they derive a system of semiparametric budget share equations for various groups of commodities. This model can be seen as an extended partially linear model (EPLM)
It is estimated using a recently developed approach to pooling in non-parametric regression. An augmented regression technique is adapted to account for endogeneity of total expenditure. Blundell et al. test the semiparametric EPLM against a fully non-parametric alternative and do not reject the EPLM. Tests of the EPLM against parametric alternatives reject the null of the Piglog speciยฎcation (linear in log total expenditures) for alcohol but not for food. The tests do