Artifical intelligence and competition: prediction of consumer demand

the purpose of the article is to test the hypothesis that artificial intelligence and machine learning methods allow firms to gain a competitive advantage by reducing uncertainty in predicting consumer demand. Methods: The article compares two approaches to demand estimation: an econometric approach based on a micro-founded and interpretable BLP model and a statistical approach based on machine learning methods. In particular, the classical model is compared with linear models with L1 and L2 regularization, random forests, gradient boosting, and a neural network. Findings: Comparison of models applied to the US car market showed a significant advantage of machine learning models relative to the BLP model. The prediction accuracy of such models outweighed the accuracy of the econometric model by more than ten times. Conclusions: The results show that the use of machine learning methods in the problem of predicting consumer demand can indeed radically increase the accuracy of prediction relative to standard models. However, in exchange for such accuracy, one has to sacrifice the economic interpretation of the model parameters.

market, economic competition, consumer demand estimation, consumer demand forecasting, machine learning methods