This paper studies the economic impact of direct transactions of consumer data in the context of auto-insurance monitoring programs, in which insurers incentivize consumers to have their driving behavior monitored for a short period of time. We acquire proprietary datasets from a major U.S. auto insurer that offers a monitoring program. The data is matched with price menus of the firm’s main competitors. We first develop a structural model of consumers’ monitoring opt-in choice in relations to their insurance demand and the cost to insure them. Key parameters are estimated using rich data variation in insurance claims, prices, contract space, and monitoring status. We then conduct counterfactual simulations using a dynamic pricing model that endogenizes the firm’s information set. We find three main results. (i) Data collection changes consumer behavior. Drivers become 30% safer when monitored, which boosts total surplus and alters the informativeness of the data. (ii) Safer drivers are more likely to opt in. But monitoring take-up is low due to both consumers’ innate preference against being monitored and attractive outside options from other insurers. Nonetheless, introducing monitoring raises both consumer welfare and total surplus. (iii) Proprietary data facilitate higher markups but protect the firm’s ex-ante incentives to produce the data. A counterfactual equilibrium in which the firm must share monitoring data with competitors harms both profit and consumer welfare. This is because the firm offers smaller upfront incentives for monitoring opt-in, so that fewer drivers are monitored in equilibrium.
This paper studies the impact of uncertainty on market competition. Michigan requires auto insurers to cover all expenses related to injuries from auto accidents. However, prices are unregulated as long as competition exists. Starting in the early 2000s, long-term medical costs have ballooned, leading to rapidly increasing injury coverage premiums. Detroit, for example, had an average auto insurance rate more than five times the national average in 2017. However, insurer profit also increased during this period, largely fueled by rising markups on non-injury and smaller coverages. Using publicly available data on insurance quotes and firm-level cost, I show that uncertainty played a key role in mitigating market competition. Specifically, I propose a model of insurance pricing that incorporates firm learning and risk aversion in a market with changing fundamentals. Comparing similar neighborhoods with different realizations of catastrophic claims over time, the model can explain why larger unexpected loss developments lead to higher markups ex-post, despite already exacerbated adverse selection.
This paper provides evidence on online firms’ decision to acquire information and on the impact of data access on firm performance and strategy. We empirically examine firms operating on a large Chinese e-commerce platform that have access to a data analytics product. The product provides detailed and real-time information on traffic breakdown, customer characteristics, and competitor strategies. These data are largely proprietary to the platform, and firms need to pay lump-sum fees for access. Focusing on several consumer electronics and peripherals markets, we find four main results: (i) scale economy is important. 9% of sellers pay for data but generate over 60% of sales. (ii) We construct a matching estimator and find that sellers that acquire data, on average, experience 15% higher sales growth. This is largely driven by persistent traffic growth and higher conversion rates. Importantly, most of these effects are driven by large gains among small sellers, despite lower take-up rates. (iii) Data access influences firm strategy. In particular, small sellers that acquire data significantly increase ads purchase and the adjustment frequency of product titles. (iv) Data facilitate persistent learning, especially among entrants. New sellers are more likely to pay for data, conditional on firm size. Non-paying sellers that have purchased data in the past also outperform those that have never purchased data.