This study quantifies the moral hazard effect of health insurance on medical expenditure by estimating a dynamic model of within-year medical care consumption that allows for insurance selection, endogenous health transitions, and individual uncertainty about medical care prices in an environment where insurance has nonlinear cost-sharing features. The results suggest that moral hazard accounts for 53.1%, on average, of total annual medical expenditure when insured. This estimate is significantly different, and generally larger, than that produced by an alternative model that is representative of the annual medical care decision-making models commonly found in the literature. CRONIN Allowing an individual's medical care optimization problem to evolve endogenously over the course of a health insurance year has implications for the analysis of moral hazard. For example, if medical care consumption affects future health transitions, then a generous insurance plan, relative to a less generous plan, may lead an individual to consume more medical care early in the year, experience health improvement, and spend less late in the year. Additionally, if an individual's medical care decisions are determined in part by his forecast of future health and medical expenditure, then health insurance can affect spending patterns through its impact on the forecast horizon in addition to its impact on out-of-pocket prices. The cumulative effect that these mechanisms have on moral hazard is ex ante ambiguous, which motivates this empirical investigation.The principal innovation of this research is the development and estimation of a WYDM model, which is motivated by theoretical (Grossman, 1972;Keeler et al., 1977) and empirical (Gilleskie, 1998;Blau and Gilleskie, 2008;Khwaja, 2010) models of health production and medical care demand. An individual's optimization problem consists of an annual health insurance decision, followed by a sequence of monthly medical care decisions made over the course of a health insurance year. I model monthly medical care decisions to allow the unique benefits and costs associated with the timing of unexpected illness and dynamic insurance cost-sharing features to impact behavior within the model. Within each month, an individual responds to an endogenous, stochastic health event by consuming medical care. The anticipated primary benefit of medical care consumption is improved future health. The anticipated primary cost is financial (i.e., a decrease in the current consumption of nonmedical goods). When health insurance has dynamic cost-sharing features (i.e., a deductible and maximum out-of-pocket expenditure [MOX] level), an additional benefit of medical care consumption is lower future out-of-pocket prices. 4 Importantly, the model assumes that individuals are forward looking, meaning they consider the impact that their actions have on the probability of future outcomes (e.g., health and prices) when making decisions. Evidence of such forward-looking health behavior can be found in Arcidiacono et a...