Measuring product differentiation by observing consumer search

Abstract In this article, we calculate various distance metrics to measure product differentiation by observing consumer search behavior directly. We compare these measures by analyzing a dataset of 1,642 digital cameras with 85 product characteristics and show that differences in the product characteristics are correlated with some of the measures of distance to a surprisingly high degree. Furthermore, we demonstrate that products categorized as having more substitutes according to our metrics appear to face stronger competition, resulting in lower prices. We also find significant price effects from distant competitors, challenging traditional assumptions about limited competition. The Jaccard index and a newly proposed weighted frequency-based metric outperform the alternative measures of product differentiation and effectively capture substitutability in multi-attribute spaces when it is impossible to calculate cross-price elasticities.