A Dynamic Model of Consumer Replacement Cycles in the PC Processor Industry

Product replacement plays a critical role in many high-tech durable goods markets. Since these markets frequently undergo both rapid improvements in quality and falling prices, the consumer’s replacement decision is most often due to product obsolescence, as opposed to wear and tear. Despite its importance, this issue has received insufficient attention in the literature, especially from an empirical perspective. Managers in these industries know that consumers follow replacement cycles, but little is known about how and why these replacement cycles change over time. And recognizing that consumers do follow such cycles, how should managers alter their own strategies to take advantage of this? To address these issues, I develop and estimate a dynamic consumer demand model that explicitly accounts for the replacement decision when consumers are uncertain about both future product price and quality. The replacement decision is inherently dynamic because the consumer that forgoes a purchase today can potentially buy a better product tomorrow for less money. Using a unique data set from the PC processor industry, I show how to combine aggregate data on sales and product ownership to infer replacement behavior. The results reveal substantial variation in replacement behavior over time. I find that a myopic model of replacement underestimates price elasticities by approximately 30 to 40 percent and overestimates the frequency of replacement by 50 percent. I demonstrate that this heterogeneity in consumer replacement behavior provides an opportunity for managers to tailor their product introduction and pricing strategies to target the particular segment of consumers that is most likely to replace in the near future. Among other results, a policy simulation shows that the marginal effect of innovation on the replacement cycle length has decreased over time, implying that firms should attempt to differentiate their products along non-speed dimensions. ∗This paper has benefited greatly from discussions with Ron Goettler, Robert Miller, Holger Sieg, Vishal Singh, and Kannan Srinivasan. Special thanks are also due to Michael Peress, Minjae Song, Fallaw Sowell, and Baohong Sun, and seminar participants at the 2005 International I.O. Conference who saw an earlier version of this paper. Financial support from Margaret Kyle and the William Larimer Mellon fund was greatly appreciated. All remaining errors are mine. Correspondence: Brett R. Gordon, Tepper School of Business, Carnegie Mellon University, GSIA Room 317A, 5000 Forbes Avenue, Pittsburgh, PA 15213, Phone: 412-268-8740, Fax: 412-268-7357, Email: [email protected].