Personalized Price Discrimination Using Big Data

Person-specific pricing was rarely observed in the past because reservation prices were unobtainable. I investigate whether this still holds now that detailed individual behaviors are tracked. Individuals' expected demand functions are estimated by combining a classic economic model with machine learning techniques to address overfitting and high dimensionality. I find that tailoring prices based on web browsing histories increases profits by 14.55%, and results in some consumers paying nearly double the price others do for the same product. Using only demographics to personalize prices raises profits by only 0.30%, suggesting the percent profit gain from personalized pricing has increased 48-fold. This is a revised version. The original title of this paper was: 'First Degree Price Discrimination Using Big Data', working paper #58 from 2013.