Consumer and firm behavior in advertising-supported Internet markets.

다국어 초록 (Multilingual Abstract)

This dissertation focuses on consumer and firm behavior in advertising-supported Internet markets. Chapter Two provides a description of market structure in this industry. Chapters Three and Four focus on methods of data analysis. The growth of the Internet has provided economists, marketers, and statisticians with a mountain of new data to analyze. One prevalent but relatively underused example of such data is clickstream data. This data format consists of each website visited by a panel of users and the order in which they arrive at these sites. Chapters Three and Four of this dissertation provide insight on how to interpret this data. Chapter Three aids with interpretation of the raw data and Chapter Four suggests that the econometric methods used to analyze grocery store purchases be used to analyze website choice.
Chapter Five provides a theoretical model of switching costs in advertising-supported online markets. The importance of switching costs and customer loyalty to Internet companies is an often-repeated mantra in business research. The land-grab mentality that pervaded Internet strategies in the late 1990's was often justified by switching costs. Once locked-in, the argument went, website visitors could be turned into profitable customers. In this chapter, I show that user switching costs can create market power, even though revenues come from advertisers.
Chapter Six brings the ideas of the previous three chapters together. It uses the findings of Chapters Three and Four to build a data set and develop an empirical framework. It uses this framework to measure true state dependence in the Internet portal market. From the theoretical perspective, the econometric phrase ‘true state dependence’ has the same implications as a switching cost. It finds large true state dependence, which the theoretical model in Chapter Five suggests leads to market power. Chapter Six also makes contributions to the econometrics literature on separately identifying true state dependence from spurious state dependence. It shows that random coefficients models overestimate true state dependence by underestimating heterogeneity. If it is necessary to do a random coefficients model, Chapter Six shows that models that allow for more heterogeneity are better.