Clickstream data tracks the website visit history for each online user, thereby enabling a detailed analysis of user choice, are much larger than databases typically used to examine consumer behavior, and provides much more information about Internet users, and thus more complicated analyses can be attempted than with traditional data. Household-specific regressions mean doing a separate conditional logit regression on the portal choice of each household in the data. This method allows for more flexible substitution patterns than the panel methods typically used to study offline behavior choice. The method is computationally inexpensive and relatively easy to understand. The paper identifies the main drivers of Internet portal choice and thus provides a better understanding of online behavior. The success of previous searches is a particularly important driver of website choice. The ability to provide deep searches with many results is less important. Habit formation is shown to be relatively uncorrelated with brand preference. Several other conclusions are also reported. (13 refs.)
展开▼