Disaggregate Network Effects on Two-Sided Platforms

This dissertation empirically explores the mechanisms through which software variety affects hardware purchases in markets with a hardware-software structure. Software variety may provide value to consumers because (i) it allows them to find a product better matched to their preferences (heterogeneity), and (ii) it may satisfy their need for different products across consumption occasions (within-person demand for variety). Using household-level coffee purchases and Keurig machine adoption data, I first build and estimates a consumer demand model for coffee allowing both within-consumer demand for variety and cross-consumer heterogeneity. Repeated purchases and trip-level variations in product prices and availabilities identify the household preferences, which allows for more reliable estimates. With the preference estimates, I then calculate the value of consuming Keurig's coffee pods (K-Cups) for each household and week, which I call the option value of K-Cups in the paper. Finally, I directly link the option values to their Keurig machine adoption decisions via a dynamic discrete choice model. Estimation results indicate variety in K-Cups, particularly brand variety, increases the option values for households, and the higher option values, in turn, increases machine adoption. To understand the mechanisms through which K-Cup variety influence Keurig machine adoption, I simulate the counterfactual adoption rates if households (a) can only choose their favorite K-Cup brands or ground coffee, and (b) have a library of K-Cups to choose from but not their favorite brands. Simulation results show that adoption rate would be about 2/3 of the current level in the later case compared to 1/4 in the former scenario, and thus indicate demand for variety is the primary mechanism under which K-Cup variety is valuable to consumers. Moreover, I show without third-party brands, (i) Keurig adoptions would have been much lower at the end of 2013; and (ii) the median consumer welfare loss conditional on the adoption of Keurig machines is large compared to their spending on coffee with substantial heterogeneity. Furthermore, Keurig-owned K-Cups revenue would have been more than 40% without third-party brands because of lower adoption base.