Estimating Network Effects in a Dynamic Environment

This paper proposes methods for identifying indirect network effects with dynamically optimizing consumers purchasing a durable hardware good and associated software. We apply this model to data drawn from the DVD player and titles markets. We observe model-level prices, sales and characteristics of DVD players and sales and availability of DVDs at the level of the month for 10 years. We augment these aggregate data with household survey data on player holdings. In our model, forward looking consumers buy possibly multiple DVD players over time and benefit from the evolution of the titles market. We address issues of clustering, spurious correlation and endogeneity.