|
Abstract:
|
This paper examines estimation of and inference about parameters of health care cost distributions. Using data from the Medical Expenditure Panel Survey (MEPS) and the Nationwide Inpatient Sample (NIS), the paper considers a set of parametric models for these distributions. The paper focuses on implications of "heavy tail" properties for estimation of and inference about parameters (e.g., (conditional-)mean costs) that play central roles in economic evaluation, risk adjustment, etc. Should health care cost distributions manifest such "Pareto-type" properties, estimates obtained from standard methods of estimation that rely on moment-based estimators (e.g., GLM) or on estimators that enforce finite moment assumptions (e.g., gamma distributions) may provide misleading guidance when such estimates are used to inform evaluations. Goodness of fit tests are deployed to assess tail fit.
|
