Numerous studies have shown that appropriate drug use can reduce hospitalization and other more expensive therapies. Some commentators argue that these findings imply that drug coverage pays for itself. As appealing as this conclusion may be, there is no credible research literature to support it. OBJECTIVES: This study is one of a series to assess the extent to which extending drug coverage to all US Medicare beneficiaries would result in cost savings elsewhere in the Medicare budget. Using data from the Medicare Current Beneficiary Survey (1996-2000), we assess each link in the causal chain between drug coverage and other health care spending for a sample of beneficiaries diagnosed with COPD. We selected COPD for 3 reasons: it is a prevalent problem affecting about 16% of the population over 65, acute exacerbations of the disease are sensitive to medication therapy, and there are reports that under-medication may be associated with lack of prescription coverage. We compared expenditures for COPD patients with and without drug coverage controlling for selection effects and other confounders using two METHODS: 1) a general linear regression model with a gamma distribution and a log link function, and 2) by weighting sample observations based on propensity scores. The propensity scores were derived from a logit regression of prescription coverage status on a rich set of sociodemographic variables and a medical expenditure risk adjuster specifically designed for use with Medicare claims. Our results indicate that for every 10% increase in the generosity of drug coverage, drug spending rises between 3% and 4%, while inpatiet hospital and physician spending each decline by 3% to 4%. The generalizability of these results to older people with other chronic conditions is discussed.