A general framework for the time series cross section estimation

This paper presents the design of a program to handle the specific estimation problems asso ciated with time series-cross section data. In order to minimize the costs of dealing with this kind of data, the program design relies on the forming of the appropriate moment matrices by summing over one or the other dimension, weighting or not, as desired. It is shown how this method may be used to estimate several different models, such as OLS with individual constant terms, GLS, Balestra-Nerlove, and the variable coefficients model of Mundlak. In addition, the method is extended to a nonlinear model, the binary probit model with variance