Versioning and Quality Distortion in Software? Evidence from E-Commerce Panel Data

We present a framework for measuring software quality using pricing and demand data, and empirical estimates that quantify the extent of quality degradation associated with software ver- sioning. Using a 7-month, 108-product panel of software sales from Amazon.com, we document the extent to which quality varies across di¤erent software versions, estimating quality degradation that ranges from as little as 8% to as much as 56% below that of the corresponding ?agship ver- sion. Consistent with prescriptions from the theory of vertical di¤erentiation, we also ?nd that an increase in the total number of versions is associated with an increase in the di¤erence in quality between the highest and lowest quality versions, and a decrease in the quality di¤erence between "neighboring" versions. We compare our estimates with those derived from two sets of subjective measures of quality, based on CNET editorial ratings and Amazon.com user reviews, and discuss competing interpretations of the signi?cant di¤erences that emerge from this comparison. As the ?rst empirical study of software versioning that is based on both subjective and econometrically estimated measures of quality, this paper provides a framework for testing a wide variety of results in IS that are based on related models of vertical di¤erentiation, and its ?ndings have important implications for studies that treat web-based user ratings as cardinal data.