Demand Estimation and Forecasting Using Neuroeconomic Models of Consumer Choice

A foundational problem in marketing and economics involves accurately predicting purchase decisions at both individual and aggregate levels. Building on recent advances in neuroeconomic models of decision making, we investigate the possibility of improving upon the prediction accuracy of popular existing approaches based on the multinomial logit model (MNL). Specifically, using a neuroeconomic model that incorporates response times in addition to choice data, we compare the out-of-sample prediction accuracy of both approaches using a series of consumer choice experiments. We show that our neuroeconomic model robustly outperformed the standard MNL approach in providing accurate forecasts on diverse measures including revenue, market share, and market cannibalization. Finally, we develop a generalizable framework to assess the relative strengths and weaknesses of our neuroeconomic approach compared to current modeling techniques.