We propose a new econometric study of the effects of managed care on preventive and curative utilization of health services - - a topic which occupies a central place in research and policy debates regarding the design and reform of public and private health care systems. The proposal recognizes that many existing quantitative and econometric studies that investigate the impact of managed care of medical utilization are flawed. First, most empirical studies are based on observational data and hence suffer from endogeneity bias caused by self- selection into managed care. Second, few of the studies are based on nationally representative samples. Third, those studies that are not subject to the two preceding criticisms are based on old data, so their conclusions may not generalize to the present. This proposal puts forward new econometric methodology for analyzing the role of managed care in health utilization. We propose to use the 1996 Medical Expenditure Panel Survey and focus on the sub-sample of non-elderly adults who have some form of private health insurance. We plan to examine two types of utilization variables. The first is a set of measures of preventive care services. The second set of utilization variables are more commonly used measures of visits to different types of providers. The effect of managed care is captured via two variables, which are more commonly used measures of visits to different types of providers. The effect of managed care is captured via two variables, enrollment in an HMO and enrollment in other types of Managed Care plans. Our econometric models will involve two sets of jointly dependent variables: an equation to explain enrollment in HMO or other managed care plan, and an equation to explain some measure of utilization. The outcome variables are discrete random variables that require use of nonlinear models. The occurrence of endogeneity bias in this context leads to significant econometric complications. These complications will be handled using modern computer-intensive methods.
