ABSTRACT
This chapter examines the nature of aggregation bias associated with error type and suggests the appropriate course to follow when using the multinomial-logit model in assessing policy changes. Sampling randomly from the empirical distribution and forming the sample expectation, as an approximation to the population expectation, is the most flexible approach for aggregate forecasting from individual-choice models. Individual choice models prove a useful framework for obtaining empirical estimates of the monetary-equivalences of attributes. Assuming that one of the explanatory variables is expressed in monetary units then the value of an attribute can be determined. Market segmentation as a procedure aimed at identifying homogeneous groupings of the population has possibilities for reducing the errors of aggregation. Missing important markets may be the consequence of the researcher's method for seeking significant subgroups. F. Reid has reported empirical results which suggest that because F.S. Koppelman's findings are based on a very specific empirical issue, the magnitudes of bias should not be generalised.
