QUESTION: What is the key premise of the study?
ANDREW CHING: Most of the economics and marketing research that looks at consumer preferences takes those consumer choices and reverse-engineers them to uncover what product characteristics buyers prefer and to what extent they prefer them. This includes inferring product quality based on things like market share. But such inferences by researchers can be misleading, given that consumers do not always have complete information about products when they choose them. For instance, many consumers prefer national brands to a store brand because they think their quality is better. But some store brands’ quality could be just as good.
In this study, we propose using observed inter-purchase time (instead of market share) as another way to measure product quality. We argue that this method is more objective than using market share or consumer buying choice because the inter-purchase time primarily depends on product durability, and it is free from the caveat about consumers’ biased perceptions of quality due to their incomplete information.
You state that your paper is, to the best of your knowledge, the first to examine inter-purchase periods as a way of determining a product’s durability and popularity. How did you arrive at that method?
I arrived at it when I read some blogs by new parents. They talked about how some diapers did not hold very well, and that they needed to be changed more often, etc. So I thought: Wait a second, the length of replacement cycle – in other words, the inter-purchase time – can be used to infer product quality, and we can get the data from consumer purchase scanner data. (Many economists and marketing researchers are using this type of data.)
Why did you choose diapers for a study of product durability, when diapers are generally used once and then thrown away?
As you may imagine, this idea can be applied to many different products, like cars. Japanese car brands are well-known to have higher durability; they can keep running for 20 years or so. But other brands need to be replaced much more often. So why not study cars or other durable goods, like TVs? The problem is that the replacement cycle of products like these is still long (at least a few years). For any given individual using that kind of product, we wouldn’t be able to record too many observations.
On the other hand, diapers need to be replaced much more often. So even with just two years of data, we can observe a lot of repeated purchases per individual. Such a rich data set on inter-purchase time allows us to control for individual differences, which can be another threat in empirical research when we try to infer product quality more precisely.
In the study, how did you account for consumption rates, which probably vary among households using a product such as diapers?
Yes, varying consumption rates must be considered, as these can offer other explanations about why people use diapers differently. Roughly speaking, our model uses data based on durability and quantity purchased per household over a wide data window (say, two years), and assumes that the household uses up all the quantity purchased in this two-year window. That’s how we get a proxy for a household-specific consumption rate.