ABSTRACT
Regression models include an additive error component that allows the regression model to reconcile any observation, however deviant, to the assumed model structure. The assumption of an additive error implies that all brands are considered to some extent, and as a result the model doesn't matter as much as it should. The solution to the additive error assumption is to consider employing structural and non-compensatory models of behavior, where hard constraints are present and some behavior is not explainable. The models require thought as to how errors enter into the model specification, providing sharper predictions about what matters to consumers. Assuming that consumers are rational in that they prefer lower prices to higher prices this means that pricing effects go in one direction only. When investigating the demand and price of goods in a competitive market, the presence of market prices that are not too extreme implies that the price elasticity is likely in the elastic range.
