USING NEURAL NETWORKS TO IDENTIFY COMPETITIVE MARKET STRUCTURES FROM AGGREGATE MARKET RESPONSE DATA

Elasticity estpdeltimates provide the brand manager with useful diagno stics for evaluating competitive market structure. However, an econome tric model must often be simplified due to the limited amount of data available to estimate the model's parameters, This results in a reduct ion in the structural insight one can gain from the model. Capitalizin g on the forecasting ability of neural networks, we introduce an innov ative method of extracting elasticity structure for a convenient consu mer retail product market. The resulting forecasting measures and elas ticity structures are then compared with those obtained from a differe ntial-effects multiplicative competitive interaction (MCI) aggregate m arket share model. We find that the neural network slightly outperform ed the differential-effects MCI model with regards to model fit, Our r esults also suggest that the neural network offered superior estimates of asymmetric cross-elasticities which resulted in superior forecasti ng ability of the holdout sample. (C) 1998 Elsevier Science Ltd. All r ights reserved.