Understanding and estimating consumer demand in durable goods has been a challenge to many marketing scholars and economists since they do not have access to good data to study them. In studying consumer preference and brand competition, researchers typically have used aggregate level sales data for durable goods which are often less than ideal to study such a topic..

Jun B. Kim, Paulo Albuquerque and Bart J. Bronnenberg have developed a new model to study consumer demand using “consumer search data”, different kinds of data set that are becoming available in many online stores. In addition, they provide a framework to study well documented market frictions such as consumer search cost.

“Marketing scholars and economists have long recognized that consumers don’t generally search or consider the universal choice set due to reasons such as search cost” they said.

“Hence there has been a chasm between theory and reality that not taking into account the limited nature of choice sets leads to incorrect estimation off the demand.”

As a demonstration of the new model, consumer search data in the form of view-rank data was drawn from May 2007 when more than 91 camcorders were on sale at Amazon.com (after removing outliers). The model showed about 40 per cent of consumers searched for fewer than five of these products, and the average set contained only 14 products. This finding had important implications for competition.

“This means that online competition between many products in our analysis was effectively zero because many of them are not jointly searched by consumers. In fact, the large majority of possible product pairs, about 70 per cent, is viewed together by less than five per cent of the population. This implies limits on substitution, which in turn causes many cross-price elasticities to be numerically zero,” the authors said.

On the market friction side,, they found “virtually all” consumers benefited from seller-sponsored product recommendation features offered by Amazon.com, via reduced search cost, although this, too, had a double edge.

“Such tools benefited the better-selling items and tended to concentrate the online market for camcorders into demand for popular items. We note that this ‘polarization’ effect may be larger once we simulate over multiple time periods, and further research is needed,” they said.

The authors tested their model at the market level data, and found it to be applicable in such situation. They also noted that it could be based on a variety of product search data, not just the view-rank data, and in other search contexts besides shopping for durable goods.