The effects of bandwidth changes on model validity: an empirical study

Abstract

This simple study uses a variant of the Hsiao et al . (2006) test to analyse the impact of bandwidth changes on model validity using actual labour force data on workers in the Philippines. The test framework is used to verify the validity of several wage specifications. Results indicate that the tests are quite sensitive to chosen bandwidths for discrete regressors.

Acknowledgements

Financial and administrative support from the Japan International Cooperation Agency (JICA) and Japan International Cooperation Center (JICE) are gratefully acknowledged. I would also like to thank my supervisor, Prof. Kimio Morimune and Prof. Yatchew for allowing me to use and modify his Splus scripts for conducting the actual tests. Of course, the remaining errors are mine.

Notes

¹ Note that there are potential endogeneity problems caused by the inclusion of sectoral variables. These can be corrected using two-step parametric and semiparametric sample selection models. These issues will be dealt with in another paper.