RESEARCH

I am an economist trained in empirical industrial organization.  I study how data and AI technologies generate useful information in imperfect markets, and the mechanisms through which this information creates and distributes economic value.  

 

I focus on applications in insurance and digital platforms as well as the design of pricing, contracting, and matching mechanisms. I specialize in structural modeling, field experiments, and independent collaboration with firms.

exp_eff_handheld_mi_per_mi_0mph_2_1.png

In many economic settings -- e.g. consumer lending, criminal justice, labor contracts -- the intuition of moral hazard has given rise to a doctrine of deterring bad behaviors by increasing punishment. In auto insurance, firms combat risky driving almost exclusively by raising accident punishment. However, not only does this undermine risk-sharing, this paper shows that it fails at deterring risky driving due to drivers' inattention to risk. We do so with novel sensor data as well as observational and experimental methods. On the other hand, these data make risky driving behavior contractable, so we develop and estimate a structural model to simulate the "first-best" contract, which features full insurance and a direct and hence salient price on risky driving behavior (Holmström 1979). Our results illustrate the important interaction between bounded rationality and market failure, and how behavioral data allow insurers to expand risk management strategies from selective risk-sharing to risk mitigation.

(first author | w/ Thomas Yu)

Paper & Slides
gg_welfare_bar_new.png

Firms in many markets directly elicit large amounts of data from consumers. The data is used to mitigate information problems, gain competitive advantage, and extract rent from consumers. We develop an equilibrium framework to trade-off these countervailing

forces, and use it to study the creation of detailed driving data in U.S. auto insurance through a voluntary monitoring program. Surprisingly, the incentive effect from the data collection process, not the data itself, is the main source of social value creation. Despite large switching inertia, the product market is still competitive enough so that forcing the firm to make the data public hurts short-term consumer welfare by discouraging data creation.

 

In macro/finance, our model is typically referred to as "rational inattention:" in a market with asymmetric information, firms endogenously acquire information subject to a cost (switching inertia and consumers' unwillingness to be monitored).

(first author | w/ Shosh Vasserman)

Screenshot 2020-04-23 13.46.34.png

We collaborate with a large e-commerce platform and conduct a year-long experiment among new sellers on the platform. The treatment group receives access to a free and customized entrepreneur training program, in which an AI algorithm dynamically assigns training materials to sellers based on their operations data. With a 24% take-up rate, new sellers that are eligible for training see 1.7% higher revenue on average (6.6% ATT), largely driven by higher traffic and enhanced marketing activities. To investigate consumer-side benefit while boosting statistical power, we construct a panel dataset of consumer-seller pairs across consideration sets that are observed in search sessions. Using exhaustive controls and fixed effects, we show that training also improves new-seller conversion, which is not driven by selection through heightened entry barrier but due to a direct effect on quality. We then estimate an empirical model to capture unobserved consumer preference heterogeneity and compute welfare. Although only 0.25% of products in consumers' consideration sets are from treated new sellers, removing the program would reduce consumer surplus by 0.07%.

(w/ Zhengyun Sun)

ctf_growth_hhi.png

What kind of firms collect what types of data? Can such data facilitate growth? If so, by influencing what strategies? We analyze the adoption and the effect of analytics tools by firms on an e-commerce platform. We then conduct a high-stake experiment among non-adopting stores and found evidence of information friction despite low take-up.

(w/ Zhengyun Sun)