While the Internet has apparently become a major source of information about many items, including real estate, travel, computer equipment, books, records, and automobiles, little is known about how it is used, and about its impact on search behavior. The objective of this study is to shed light on this question by studying the use of the Internet as an information source for one good, automobiles. In order to accomplish our objective, we need to construct a model of the choice of information sources. We liken the time spent with each alternative source to the choice of factors of production. In our case, different sources are combined to produce current information, which, when combined with prior information, produces gains to information, defined as the difference in monetary value between the best choice given complete information and a random choice. Assuming that consumers choose the optimal amount of time to allocate to each source, given their time costs, we derive first-order conditions that lead to expressions for the share of time spent with each source and total search time. These expressions become the basis for our empirical analysis. Consistent with past literature and our data, we define four major categories of sources of information on automobiles. These are interpersonal sources such as friends and relatives, non-advocate or neutral sources of print information such as Consumer Reports, dealer/manufacturer sources, and the Internet. We use data on information source use obtained from recent new car buyers in one market in 1999, after the Internet became established, and compare this with data obtained from a survey of recent car buyers obtained in 1989 in the same market, before the Internet was introduced. The two data sets allow us to create baseline estimates of what buyers with a particular set of characteristics would have done in the absence of the Internet, and then to determine how the presence of the Internet is associated with deviations from this baseline. Our descriptive results indicate that about 40 percent of our sample used the Internet as a source in their purchase decision, which is very consistent with other evidence on Internet use in 1999, the time of our survey. The share on Internet use in total search is significantly related to younger age, higher education, higher wage, dissatisfaction with the dealer of the previous car, and no decision on which manufacturer to buy from prior to the search. A testable prediction of our model is that the Internet should draw proportionally from each other source. Since dealer/manufacturer sources get the biggest share of time in the absence of the Internet, this implies that the Internet has the greatest absolute impact on time with these sources. Tests of the proportional draw assumption indicate that this hypothesis cannot be rejected, and that the assumption provides a good approximation for our data. Our analysis of total search indicates that total search increases with factors that determine gains to search, such as price level of the chosen model, and decreases with prior information and efficiency at search. Specifically, total search decreases with a prior decision on a manufacturer or dealer to buy from, satisfaction with the previous dealer, and age. Search also decreases with wages, a measure of time costs. Putting all of the pieces of our model together allows us to estimate the effects of the Internet and other consumer characteristics on search, and on gains to search. We simulate overall behavior, and that of Internet users and certain other groups, both with and without the Internet. Our simulations indicate that the Internet both reduces total search time, and leads to improved purchases. These effects are largest for younger, more educated consumers, particularly those that were dissatisfied with their previous dealer. Conversely, the Internet has very little value for older consumers, and younger consumers with high school or less education. Our simulations also indicate that the Internet leads to a considerable reduction in time spent with dealer/manufacturer sources, most of which is time spent at the dealership requiring the presence of a sales person. This suggests that, by making it easier for consumers to be better informed when they arrive at the dealer and hence cutting down on the demand for sales person time, the Internet can lead to a substantial improvement in the efficiency of dealer operations. Thus the Internet is likely to benefit dealers as well as consumers.
