Appendix A. Statistical Appendix Methodology for Reweighting the Medical Expenditure Panel Survey

The first reweighting stage uses the Current Population Survey (CPS) Northeast sample to estimate a statistical model of the probability of living in New York as a function of socio-demographic variables common to the CPS and the Medical Expenditure Panel Survey (MEPS). The variables include age, race and ethnicity, health insurance status (public, private, or uninsured), gender, education, income, labor force status, marital status, household composition, urban residence, and self-reported health.1 (Tables A.A.1 and A.A.2 report the parameters of the probit models for adults and children, respectively.) The parameters from this model are then combined with the corresponding variables from the MEPS sample to generate a predicted probability of living in New York for each MEPS respondent. These predicted probabilities are then combined with the actual probability of New York residence observed in the CPS Northeast sample to adjust the MEPS survey person-weight using a formula (described below) developed in the statistical literature.2 The revised person-weights have the effect of making the MEPS Northeast sample “look like ” a sample of New York residents. Statistical Reweighting Formula The following formula describes how the person-weights for the MEPS Northeast regiona