Understanding Consumer Search

Consumers are often imperfectly informed about prices or quality of products offered in the market, and they must engage in costly search activities to obtain the information relevant for their purchasing decisions. Consumer search theory provides a framework to study market outcomes in such settings. An essential ingredient of any model of consumer search is the assumed search protocol, i.e. the exact form in which consumer search takes place. Consumers are assumed to either search sequentially or simultaneously. A sequentially searching consumer samples one price after the other. After each sample she decides whether to buy and/or whether to obtain another price quote. A simultaneously searching consumer first decides how many prices quotes to obtain, and then decides whether to buy at lowest observed price or not. This project aims to test the implications of different search protocols. In particular, it will implement a test that is robust to the possibility that consumers learn not only about prices actually charged in the market, but about the underlying distribution from which prices are sampled. We show that the simultaneous search model can be tested against the sequential search model by testing for the joint statistical significance of the highest and lowest observed price of a consumer in explaining the number of sampled prices. To implement the test, we will follow two approaches. First, we use web browsing data on a sample of individuals in multiple retail sectors. Second, we will complement the real-world data by conducting a field experiment. We will then use discrete choice methods to test the implications on the obtained data. This project aims at providing a robust understanding of the nature of consumer search. While we will apply the proposed testing framework to a specific market, the framework will be useful for other applications as well. Knowledge of the appropriate search protocol is important for the policy implications drawn from the analysis of search models, as the search protocol affects price elasticities and the efficiency of market outcomes. A key advantage of our proposed testing framework is that it does not assume that consumers are fully aware of the distribution of prices that generates the prices quoted.

In the project "Understanding Consumer Search" we combined theory and empirics to enhance our understanding of markets. A particular issue that attracted our attention during this period is the use of dynamic pricing tools by firms and the response by consumers who actively search for prices to such a policy: rather than search prices across firms consumers may search for prices across time. A new technique was developed to estimate demand using high-frequency data that results from using algorithmic price recommendations to set prices. An application to the hotel industry also provides one of the first estimates in the literature of the proportion consumers who actively search for prices over time when firms use dynamic pricing. We find that around one third of consumers consider waiting to purchase later following a price hike. This tool is a practical solution for firms in the hospitality industry who are interested in the impact of price changes to their demand. Following the best practices, we have provided an open-access version of our code that can be easily implemented. Further, the role of personalized pricing as a barrier to competition was examined. We identify the impact of price targeting based on willingness-to-pay information on market outcomes, depending on whether the information is public or privately held by firms. It is shown that regardless of whether the information is private or public, it harms consumers and may also reduce social welfare.