Does Buying High Mean Buying Often

Many durable goods, particularly in the high technology sector of the economy, experience rapid quality improvement. As a result, replacement is typically due to obsolescence rather than breakdown, and replacement cycles are generally short. When these goods are vertically differentiated, consumers must make two important, inter-related choices upon making a purchase: what quality level to buy, and how long to wait between purchases. In this paper, we empirically determine how these two decisions are related for personal computers using data on over 6,000 American households. We find that quality choice is persistent across purchases, and that quality choice and time between purchases are (strongly) positively correlated. Next, we explore two reasons why this correlation exists – heterogeneous replacement costs and heterogeneous marginal values of quality. Regarding the former, the data show that high-end buyers are more likely to upgrade without making a new purchase (e.g., buy more memory); however, this behavior cannot fully explain our findings. Regarding the latter, using proxies for marginal values of quality, we show that they are (approximately) normally distributed in our data. We also find that they: are unusually low for households who buy low and infrequently; are unusually high for households who buy high and frequently; and are similarly near the mean for households who buy low and frequently and households who buy high and infrequently. Then, using a simple theoretical model, we show that these empirical patterns can explain our result. Further, we find the last two groups (low/frequent and high/infrequent) to be strikingly similar along several other dimensions. Our results extend our understanding of durable goods demand for these types of products, and they show how market definitions and determinations of market segments crucially depend on consumers’ joint decisions of quality levels and time between purchases – not just one or the other. 1 Cornell University, Department of Applied Economics and Management. I would like to thank Daniel Simon, Fernando Vega-Redondo, Ted Welser, participants at the International Industrial Organization Conference and the Econometric Society North American Summer Meetings, and especially Rob Fairlie and Shane Greenstein for helpful discussions on this topic. All errors are mine.