Essays in the economics of electronic commerce
- The widespread adoption of the internet has been one of the most significant events of the past ten years. While few would doubt, and many have predicted, the potential for the internet to vastly change many aspects of consumer markets, the data to test these predictions is only now becoming available. All three chapters of this dissertation were motivated, at least in part, by the need to turn casual speculation about the internet's impact on consumer markets into empirical fact. Chapter 1 investigates the effect of new competition from the online channel on the profitability of incumbent "brick and mortar" firms. The number of brick and mortar travel agents in the United States decreased by 38 percent between 1997 and 2003. In this chapter, I attempt to understand and quantify the extent to which the massive wave of internet adoption during this period was responsible for this observed pattern of exit. To do so, I combine data from four different sources to construct a panel data set of county level measures of demand for travel services associated with internet- adopting and non-adopting households. Importantly, these measures are disaggregated into demand for air travel services and non-air travel services, as the latter may be more difficult for households to purchase online. I then develop and estimate a model of brick and mortar agent profits. I find that household internet adoption primarily reduces profits associated with arranging air travel services, while leaving profits associated with other travel services relatively unchanged. Because internet adopters already accounted for 75 percent of air travel spending by 2003, counterfactual simulations suggest that the majority of travel agents that survived the current wave of internet adoption will remain profitable in the longer run, even as internet penetration rates approach 100 percent. Chapter 2 examines price dispersion in online retail markets. Previous empirical studies of price dispersion in the online retail channel have been constrained by a lack of data on actual sales. I employ a dataset of new and used compact discs listed on the online marketplace Half.com, where it is possible to infer when sales take place, to analyze the determinants of consumers' choices of sellers and sellers' pricing decisions. Estimates of a multinomial logit model of consumers' choices of sellers with whom to transact show substantial persistence in buyers' demand for sellers who do not charge the lowest price. Specifically, some proportion of consumers do not select the seller listing the lowest price for an item, even after adjusting for all available measures of seller quality. Through simulations, I demonstrate that these idiosyncracies often result in optimizing firms choosing not to be the lowest priced seller of a given item. When coupled with heterogeneity in the time horizons over which Half.com sellers intend to sell items, the simulations demonstrate that these pricing incentives are enough to rationalize observed levels of dispersion in sale prices. Finally, an analysis of the pricing decisions of new entrants offers evidence consistent with predictions generated by the simulations. Chapter 3 investigates the effect of broadband adoption on the activities in which individuals engage online. Estimating the causal relationship between broadband internet connection and individuals' online behavior is challenging, due to the fact that early broadband adopters have unobservable preferences over internet technology that influence their activities online prior to attaining a high-speed connection. I employ a unique identification strategy to estimate the direct causal effect of a broadband connection on the probability that individuals engage in thirteen common online activities (such as email, shopping, and banking). The strategy depends on individuals who identify themselves in a census survey as wanting to have a home broadband connection, but being constrained from doing so because the service is not available in their area. Because this group has the preferences associated with broadband adopters without the actual high-speed connection, I am able to decompose the positive correlation between broadband use and the probability of engaging in each online activity into the portion driven by the actual connection, and the portion driven by unobserved preferences. The results vary widely across activities, with the direct effect of broadband driving all of the positive correlation in some cases, none of it in others, and many cases in between. The importance of the direct effect relative to the effect of unobserved preferences increases with both the amount of information that must be transmitted over the internet, and the amount of time that must be spent online. In addition to a baseline probit specification, I also include propensity score matching and panel data regressions to verify my results.
|Type of resource
|electronic; electronic resource; remote
|1 online resource.
|2010, c2011; 2010
|Gulker, Matthew Morgan
|Stanford University, Department of Economics
|Bresnahan, Timothy F
|Wolak, Frank A
|Bresnahan, Timothy F
|Wolak, Frank A
|Statement of responsibility
|Submitted to the Department of Economics.
|Thesis (Ph.D.)--Stanford University, 2011.
- © 2011 by Matthew Morgan Gulker
Also listed in
Loading usage metrics...