| Founded by Alibaba Group in 2003, Taobao Marketplace has become the dominant online retailer in China. In Chapter 1, I utilize a unique dataset of iPhone5 posted-price transactions from TaobaoMarketplace to study expected consumerwelfare gain due to increased quality provision. Departing from product specific reputation measures constructed by counting cumulative past sales, I introduce a more accurate measure by counting the amount of positive and negative feedback given to products directly. I find that consumers respond positively to good reputation: utility from purchasing an iPhone5 increases when feedback ratings increase, and thus products that have better reputation enjoy greater market share. Reputation has a significant effect on consumer willingness to pay. For a one standard deviation increase in the iPhone5 specific positive feedback ratio, the consumer willingness to pay is 3 percent of the price of an iPhone5. I also find compelling empirical evidence in support of the positive relationship between good reputation and consumer welfare. Counterfactual analysis suggests that if every seller of the iPhone5 in Taobao Marketplace provides satisfactory service, consumer welfare would increase by 11.1 USD, about 2 percent of the price of an iPhone5.;Chapter 2 quantifies the welfare effects of reputation in this online marketplace. On the demand side, I employ a nested logit model and estimate consumer willingness to pay for reputation. On the supply side, reputation is modeled as an endogenous variable and cost parameters are estimated. Good reputation is found costly to maintain. My results suggest that it costs an average seller 123 USD per day to maintain his positive feedback ratio at 99.1 rather than 98.6 percent, a half standard deviation difference. Two counterfactual analyses are conducted to study the welfare effects of reputation.;Chapter 3 uses a laboratory experiment to study the effects of reputation systems. This is achieved by varying the information that short-run players receive concerning the past actions of long-run players. In particular, I compare the equilibrium behavior and payoffs of these two types of players under two different reputation monitoring systems. The first system mimics a classic reputation framework that allows short-run players to observe the long-run player's past choices in chronological order. Under the second reputation system, only the aggregate frequencies of the long-run player's past choices are observable to short-run players. Results show that efficient outcomes are reached more frequently under the second system than the first. These experimental results have strong implications for the design of feedback systems in online marketplaces. |
