Estimation and Testing of a Dynamic Sales Response Model with Data Aggregated over Time: Some Results for the Autoregressive Current Effects Model

Abstract

Using data aggregated over time, the author investigates the issues surrounding the estimation and testing of a sales response function where the dynamic response elements are confined to the random component of the model. This specification of a response function is elegant and helps circumvent some problems associated with models containing lagged sales terms. The estimation of such a model is theoretically straightforward. The study shows, however, that despite theoretical superiority, some estimators provide bad empirical values for dynamic response parameters; furthermore, the fact that the estimated parameter efficiency often is biased compromises the power of parameter significance tests. The author details the conditions under which the econometric results of such a model are interpretable and powerful, as well as the instances in which they should be evaluated carefully.