Major online platforms such as Amazon and eBay have invested significantly in search technologies to direct consumer searches to relevant products. These technologies lead to targeted search, implying consumers are visiting more relevant sellers first. For example, consumers may directly enter their desirable attributes into search queries, and the platform will retrieve relevant sellers accordingly. The platform may also let consumers refine the search outcomes by various criteria. This study characterizes the role of targeted search, and examines how targeted search affects market equilibrium and platform design.
I model targeted search in a differentiated market with many firms where consumers search sequentially for the best product match. Within this setup, I endogenize the search design by allowing the platform to choose the precision of targeted search and the revenue model contract. One of the central results of the analysis is how targeted search affects equilibrium prices. I find its impact on price is not monotonic. When targeting is not too precise, targeted search lowers the equilibrium price. It makes sellers more similar and intensifies price competition, despite the fact that all consumers face sellers with better fit. However, once the targeting becomes sufficiently precise, the equilibrium price increases, because highly targeted search discourages active consumer search and gives sellers monopoly power. Furthermore, I consider two major platform revenue models, commission and promoted slots, with consumer search. The platform, by providing targeted search with precision up to the aforementioned limit, can extract more consumer surplus through higher commission rates, because targeted search improves consumer surplus by lowering search cost, increasing fit, and lowering price. With targeted search up to the limit, the platform can also extract more surplus from sellers by offering promoted slots, because sellers can use promoted slots to better target consumers. However, once targeted search becomes too precise, the market will face a price hike, hurting the platform revenue in both models. Therefore, I find that in both revenue models, the platform may want to limit the precision of targeted search even if improving it is costless, with or without consumer entry. Using a unique dataset from Taobao, I find suggestive evidences that are consistent with the model predictions.
I model targeted search in a differentiated market with many firms where consumers search sequentially for the best product match. Within this setup, I endogenize the search design by allowing the platform to choose the precision of targeted search and the revenue model contract. One of the central results of the analysis is how targeted search affects equilibrium prices. I find its impact on price is not monotonic. When targeting is not too precise, targeted search lowers the equilibrium price. It makes sellers more similar and intensifies price competition, despite the fact that all consumers face sellers with better fit. However, once the targeting becomes sufficiently precise, the equilibrium price increases, because highly targeted search discourages active consumer search and gives sellers monopoly power. Furthermore, I consider two major platform revenue models, commission and promoted slots, with consumer search. The platform, by providing targeted search with precision up to the aforementioned limit, can extract more consumer surplus through higher commission rates, because targeted search improves consumer surplus by lowering search cost, increasing fit, and lowering price. With targeted search up to the limit, the platform can also extract more surplus from sellers by offering promoted slots, because sellers can use promoted slots to better target consumers. However, once targeted search becomes too precise, the market will face a price hike, hurting the platform revenue in both models. Therefore, I find that in both revenue models, the platform may want to limit the precision of targeted search even if improving it is costless, with or without consumer entry. Using a unique dataset from Taobao, I find suggestive evidences that are consistent with the model predictions.
|
|
The success of e-commerce has largely relied on reputation systems. This study examines the demand and supply implications of reputation systems in online marketplaces. One of the most pervasive features of many reputation systems is that platforms assign sellers into different levels based on continuous rating scores. I utilize this institutional feature to make causal inferences on how the number of ratings affects demand and how sellers respond. Using a unique dataset from Taobao, a regression discontinuity design shows that one extra level increases sales significantly and substantially even for well-established sellers. This level effect is robust to a wide range of model specifications and falsification test. More importantly, a dynamic pricing model predicts sellers respond by lowering their prices before the thresholds of levels, and increasing prices afterward. The evidence from the data supports such predictions.
|
Individuals today have almost ubiquitous access to online information and content and can choose to be constantly online through multiple devices and platforms. There is a growing trend in many countries including the U.S., China and Korea of compulsive Internet usage and individuals spending ever more time checking and consuming information. For example, in the U.S. about 16% of 18-25 year olds are deemed to be addicted to compulsive Internet checking and consumption. We propose a theory of information addiction and develop an economic decision making framework that incorporates a disutility for carrying uncertainty based on the psychological foundation of curiosity. New information arrives through a Poisson process and is available to agents. Two distinct types of uncertainty characterize an agent’s information consumption: i) whether or not there is a new piece of information and ii) conditional on the existence of new information whether or not it is valuable (i.e., its utility level). Consumers have costs of checking for new information and they dynamically choose an optimal frequency of checking to maximize their expected utility. We then explore the design of an optimal information structure by an information provider. The design the information structure can involve the choice of the advertising intensity (which increases cost of checking), or the decision to “push” information or in other words whether to offer a notification about new information arrival. Interestingly, while push notification lowers consumer uncertainty, it could also induce more frequent checking by inducing consumers to check immediately after new information arrival. The paper also investigates the role of several consumer self-regulation and commitment strategies in the face of their information addiction.
|
In this online field experiment, we investigate consumer choice between two identical firms whose listings differ only in price, in a setting where listings are sorted by price. Many empirical studies of online price dispersion show that sellers post different prices for homogeneous goods. However, seller heterogeneity is difficult to control for and posted prices may not reflect price dispersion in actual transactions. We contribute to this literature by selling identical simple goods (cell phone credits) at different prices from sellers that were identical except in name and with minimal ratings. The only way consumers could find us in this extremely thick market is to rank by price from lowest to highest. Out of 514 sales, 73 were of the higher priced item, for which we had non-negligible demand even when the price gap was 2.5%. We show that price sensitive consumers do not necessarily buy the lowest priced item, all else being equal.
|
|
