I am currently working on the following topics:

Comparative Profitability of Product Disclosure Statements with J. Li (revision requested)

In insurance industry, product disclosure statements (PDSs) consist of descriptions of uncertain contingencies by the insurance plans (e.g.,"hospital coverage", "dental coverage", etc.) and are often very different. In this paper, we model PDSs as information partitions of the state space, which can influence how a consumer perceives the structure of her choice problem and hence her deductible choices. We study a model of an insurance company that aims to promote profit by designing the framing of its PDS. We compare the company's profits under two PDSs, one of which is coarser than the other. Our main results show that under simple conditions, the PDS consisting of finer partitions of the more expensive states is more profitable.

State Aggregation with α-MEU

This paper studies how ambiguity affects identification of the state aggregation model -- a model with an agent that combines several states into an event in order to simplify decision-making. We show that the state aggregation is partially identified under Maxmin Expected Utility (MEU) and it is uniquely identified if α-MEU model is used. In addition, we offer testable restrictions of the model.

Semi-Parametric Estimation of Ballot Stuffing

This paper introduces a structural model of electoral choice and ballot stuffing that allows for derivation of the joint distribution of turnout and voter share from the unobservable joint distribution of the costs of voting and preferences over candidates. We assume availability of a "clean" subset of data that provides us with the following results: if political preferences are heterogenous across the country, we are able to identify the density of ballot stuffing at any polling station of interest. If political preferences are rather homogenous, we are able to identify the bounds on average ballot stuffing. We also provide corresponding semi-parametric estimators, which are based on kernel density estimation. In addition, this paper offers an empirical illustration of the model estimation using the 2011 Russia parliamentary election data.