Maura Coughlin - Cornell University

Insurance Choice with Non-Monetary Plan Attributes: Limited Consideration in Medicare Part D

Abstract:  This paper evaluates insurance choices in a empirical setting where non-monetary plan attributes influence plan choice. Insurance choice is typically modeled using the expected utility (EU) framework, which does not immediately allow for non-cost plan attributes to impact choices. Reduced form approximations lose the structural focus on risk preferences and can imply difficult to interpret willingness to pay for attributes. In many health insurance markets a model without plan attributes cannot rationalize a substantial portion of observed choices. This model addresses such prevalent choice inconsistencies by relaxing the assumption that individuals evaluate all available plans at the time of enrollment. Limited consideration naturally provides a role for non-cost attributes to impact choice probabilities while preserving the structure of expected utility fundamentally underpinning the demand for insurance. Using a large administrative data set, I apply the alternative specific consideration model of Manzini and Mariotti (2014) to choices over Medicare Part D prescription drugs plans. I allow for the plan's consideration consideration probability to depend on the its attributes. This approach has a number of advantages over standard methods. In contrast to previous studies of Medicare Part D that have found little to no role of risk aversion, I find a mean coefficient of risk aversion that is substantial and comparable to estimates in other insurance markets. Alternative models estimated on the same sample return estimates of mean risk aversion below half of my baseline model. Consideration is largely driven by the brand, premium, deductible, and gap coverage of a plan. For example, plans with the highest premium and deductible in the market are considered 10% and 18% as equivalent plans with the lowest premium and deductible.