Estimating Demand Elasticities Using Nonlinear Pricing, October 2013 (Revision Requested) [PDF]
(Previous version: Estimating Health Expenditure Elasticities Using Nonlinear Pricing, November 2009)
Step-by-step directions and code for implementing the estimation method [Code]
Stata ado-file [locpolyslope]
Nonlinear pricing is prevalent in industries such as health care, public utilities, and telecommunications. However, nonlinear pricing in these sectors causes bias when estimating elasticities for welfare analysis or policy changes. This paper develops a local elasticity estimation method that takes advantage of nonlinear price schedules to isolate consumers' expenditure choices from selection and simultaneity biases. This method improves over previous approaches because it uses commonly-available observational data and requires a single general monotonicity assumption. Using detailed claims-level health insurance data with two nonlinearities, this paper measures declining elasticities from -0.26 to -0.09 along initial expenditure ranges, in contrast to previous work which aggregates over a skewed spending distribution. These estimates are then used to calculate moral hazard deadweight loss. This method enables estimation on many policies with nonlinear pricing which previous tools could not address.
Moral Hazard, Adverse Selection and Health Expenditures: A Semiparametric Analysis (with Patrick Bajari, Han Hong, and Ahmed Khwaja), September 2013 (2nd Revision requested) [PDF]
Theoretical models predict asymmetric information in health insurance markets may generate inefficient outcomes due to adverse selection and moral hazard. However, previous empirical research has found it difficult to disentangle adverse selection from moral hazard in health care. We empirically study this question by a using unique data set with confidential information from a large self insured employer to estimate a structural model of the demand for health insurance and medical care. We propose a two-step semi-parametric estimation strategy that builds on the work on identification and estimation of auction models. We find significant evidence of moral hazard and adverse selection.
Why Do Firms Use Insurance to Fund Worker Health Benefits (with Sarah Holland), October 2013 [PDF]
When a firm offers health benefits to workers, it exposes the firm to the risk of making payments when workers get sick. A firm can either pay health expenses out of its general assets, keeping the risk inside the firm, or it can purchase insurance, shifting the risk outside the firm. We analyze the firm's decision to manage this risk. Using data on the insurance decisions of publicly-traded firms, we find that smaller firms, firms with more investment opportunities, and firms that face a convex tax schedule are more likely to hedge the risk of health benefit payments. These financial characteristics explain more of the hedging decision relative to commonly-cited state insurance mandates. We also show that hedging health risk mitigates investment-cash flow sensitivities.
Outsourcing and Ownership: Theory and Evidence from California General Care Hospitals (with Patrick Warren), October 2013 [PDF]
For-profit hospitals in California contract out to a much greater extent than either public hospitals or private nonprofit hospitals. To explain why, we build a model in which the outsourcing decision is a trade-off between net revenues and "quality", any factor of interest to the hospital manager over-and-above its impact on net-revenues. Since nonprofit firms must consume profits indirectly through perquisites, they trade off differently from for-profit firms. This difference is exaggerated when "quality" is particularly important or the firm is hit with a favorable fixed-cost shock. We test these predictions in a panel of California hospitals, finding evidence for each. These results suggest that a model of public or nonprofit make-or-buy decisions should be more than a simple relabeling of a model derived in the for-profit context.
Nonprofit vs. For-profit Competition: A Study of Hospice Care (with David Bradford), March 2013
Many health care sectors have persistent coexistence of nonprofit and for-profit. What is the government buying with tax exemptions to only the nonprofit firms in such a market? This paper develops a model to explain the coexistence of both nonprofit and for-profit health providers within the same market. This paper addresses the impact of these government tax exemptions by modeling empirical differences in patient mix by ownership status as product positioning. The product position of a health facility is measured by the mix of diagnoses served by the facility. In particular, we find that coexistence of both nonprofit and for-profit allows a larger range of patients, of varying profitability, to be served in markets that are heavily influenced by an inflexible Medicare price schedule. Nonprofit facilities take on the costs of soliciting donations, which allows the choice of a less-profitable patient mix. The model is then tested on data covering over 90 percent of U.S. hospice facilities, and finds significant evidence of product positioning.
Estimating Elasticity and Drug Adherence in Medicare Part D (with Gautam Gowrisankaran and Robert Town), September 2013
The introduction of Medicare Part D in 2006 has had a clear impact on its beneficiaries. Early evidence indicates that the program has lowered Medicare beneficiaries' out-of-pocket costs while increasing prescription drug consumption. Nonetheless, the program has its critics who point to the possibility that enrollees may reduce consumption of medicines during the doughnut hole with adverse consequences for health. This paper estimates drug adherence response of Medicare Part D enrollees to changes in pharmaceutical out-of-pocket cost sharing controlling for differential enrollee selection and the non-linear structure of the Part D benefit design. We use unique data from proprietary pharmacy claims data from Express Scripts, a pharmacy benefits manager whose portfolio of business includes Medicare Part D plans. Express Scripts manages Medicare Part D benefits for approximately 30 different employer-sponsored Medicare Part D plans with 100,000 enrollees. In our empirical strategy, we estimate the impact of changes in the marginal out-of-pocket expenditure on drug consumption (i.e., enrollees' out-of-pocket drug adherence and consumption elasticity). To control for selection into part D plans, we propose a regression discontinuity approach that exploits the changes in marginal incentives around the entry and exit to the doughnut hole to obtain consistent estimates of the elasticities.
The Effect of Urban Resident Basic Medical Insurance on Household Consumption (with Gordon Liu and Fei Xu), December 2012
Changing from an export-driven growth mode to a more stable domestic consumption orientation is a priority for many emerging economies, notably China. After exports were seriously disrupted by the world financial crisis in 2008, China's policy makers are increasingly interested in boosting domestic consumption to promote sustainable economic growth. This paper examines the role of introducing health insurance to economic growth. Protecting urban residents from unexpected medical expenditures could reduce income uncertainty of urban households and consequently increase their current consumption. Using a new, detailed longitudinal household survey, we analyze the effects of the introduction of Urban Resident Basic Medical Insurance (URBMI) on household consumption in urban China. Our difference-in-differences estimates show that participating families increase annual non-medical consumption by about 13%. The largest component of the consumption increase is for education expenses, which is an additional benefit of introducing insurance to developing countries seeking to improve their human capital advantage. URBMI increases non-medical consumption the most for low-income families, by over 20%.