Electronic copy available at: http://ssrn.com/abstract=1483380 NET Institute* www.NETinst.org Working Paper #09-23 September 2009 Testing Models of Consumer Search using Data on Web Browsing and Purchasing Behavior Babur De los Santos Indiana University Ali Hortacsu University of Chicago and NBER Matthijs R. Wildenbeest Indiana University * The Networks, Electronic Commerce, and Telecommunications (“NET”) Institute, http://www.NETinst.org, is a non-profit institution devoted to research on network industries, electronic commerce, telecommunications, the Internet, “virtual networks” comprised of computers that share the same technical standard or operating system, and on network issues in general.Electronic copy available at: http://ssrn.com/abstract=1483380Testing Models of Consumer Search using Data onWeb Browsing and Purchasing Behavior∗Babur De los Santos†Ali Horta¸csu‡Matthijs R. Wildenbeest§First version: April 2009This version: September 2009AbstractUsing a large data set on web browsing and purchasing behavior we test to what extentconsumers are searching in accordance to various classical search models. We find that thebenchmark model of sequential search with a known distributions of prices can be rejectedbased on the recall patterns we observe in the data. Moreover, we show that even if consumersare initially unaware of the price distribution and have to learn the price distribution, observedsearch behavior for given consumers over time is more consistent with non-sequential searchthan sequential search with learning. Our findings suggest non-sequential search provides a moreaccurate description of observed consumer search behavior. We then utilize the non-sequentialsearch model to estimate the price elasticities and markups of online book retailers.Keywords: consumer search, electronic commerce, consumer behaviorJEL Classification: D43, D83, L13∗We thank Ken Hendricks for his useful comments and suggestions. In addition we thank seminar participants atthe Industrial Organization Workshop of the NBER Summer Institute 2009 (Boston), the 2009 Far East and SouthAsia Meeting of the Econometric Society (Tokyo), and the SIEPR Conference on Internet Economics (Stanford). Wegratefully acknowledge financial support from the NET Institute (www.netinst.org).†Kelley School of Business, Indiana University, E-mail: [email protected].‡University of Chicago and NBER, E-mail: [email protected].§Kelley School of Business, Indiana University, E-mail: [email protected] IntroductionSince Stigler’s (1961) seminal paper, models of costly search have been at the heart of manyeconomic models trying to explain imperfectly competitive behavior in product and labor markets.The theoretical literature typically models consumer search in two ways: following Stigler’s originalmodel, a strand of literature assumes non-sequential search behavior, where consumers sample afixed number of stores, and choose to buy the lowest price alternative.1A much larger strand of theliterature, starting with McCall (1970) and Mortensen (1970), points out that consumers cannotcommit to a non-sequential search strategy in instances where the expected marginal benefit of anextra search exceeds the marginal cost. Thus, this literature argues that a sequential search modelprovides a better description of actual consumer search.2Unfortunately, beyond the a priori reasons put forth by the literature, there have been fewempirical studies of whether actual consumers follow sequential or non-sequential strategies. Thisis, no doubt, due to the difficulty of collecting data on individual search behavior. Therefore, most ofwhat we know about individual level search behavior is from laboratory experiments. The majorityof the experimental literature on search has focused on sequential search.3Schotter and Braunstein(1981) have reported that on average subjects tend to search in a fashion that is consistent withsequential search strategies, although subjects tend to search too little to be searching optimally.Kogut (1990) and Sonnemans (1998) find evidence that individuals are making decisions basedon the total return from searching instead of on the marginal return from another draw as theywould do if searching sequentially, resulting in too little search. Moreover, Kogut (1990) findsthat in about a third of the time individuals accepted old offers, which violates optimal policy.Zwick et al. (2003) also find large rates of recall among participants of an experiment in whicha randomly selected object with a known rank order has to be selected. Harrison and Morgan(1990) directly compare non-sequential and sequential strategies to so-called variable-sample-sizestrategies. The latter strategy is described in Morgan and Manning (1985) and is a generalizationof both non-sequential and sequential search since it allows individuals to choose both sample sizeand how many times to search. Harrison and Morgan (1990) report that experimental subjects1See also Burdett and Judd (1983) and Janssen and Moraga-Gonz´alez (2004).2Examples of sequential search models in the consumer search literature are Axell (1977), Reinganum (1979),Carlson and McAfee (1983), Rob (1985), and Stahl (1989).3See Camerer (1995, pp. 670-73) for a review of this literature.2indeed employ the least restrictive strategy if they are allowed to do so.Aside from experimental studies, Hong and Shum (2006) and Chen, Hong, and Shum (2007)are the only papers that we are aware of that have attempted to discriminate between sequentialand non-sequential search models using data from a real-world market. Hong and Shum (2006)collect data on textbook prices, and estimate structural parameters of search cost distributions(i.e. the demand parameters) that rationalize the prices set by competing firms. They find largersearch-cost magnitudes for the parametrically estimated sequential search model than for the non-parametrically estimated non-sequential search model. Similar data is used in Chen, Hong, andShum (2007) to conduct a nonparametric likelihood ratio test for choosing among the nonpara-metrically, moment-based non-sequential and parametrically estimated sequential search models.Although certain parameterizations of the sequential search model are found to be inferior, theyconclude that it is difficult to distinguish between the non-sequential search model and the log-normal parameterization of the sequential search model