An Econometric Model of the Medicare System: Reply

Abstract

Louise Russell's comment focuses on an important feature of the Medicare program. The designers of Medicare understood that, by covering a broad range of services, they would reduce the bias in favor of hospital inpatient care that is inherent in many insurance programs. Moreover, they hoped that insuring treatment in extended care facilities 1 would shorten the duration of hospital stays enough to make the total cost of hospital plus extended care services less than the cost of hospital services alone if extended care facilities were not covered. As a by-product of estimating a general model of the Medicare program, my previous paper showed that, contrary to the hopes of the designers of the Medicare program, the use of extended care facilities actually raised the total cost of institutional care. Louise Russell, using somewhat different data, has come to the opposite conclusion. More specifically, we both consider a regression equation relating the Medicare reimbursement for hospital and extended care services per hospital episode to measures of hospital cost per day, extended care admissions per hospital episode, and several other variables. The equation is estimated with a cross section of state observations. I found that extended care admissions per hospital episode raises total reimbursements; the estimated elasticity is 0.12 with a standard error of 0.06. Russell found very small and statistically insignificant elasticities: -0.002 with a standard error of 0.02 for 1967 and -0.032 with a standard error of 0.04 for 1968. The basic reason for our contrasting conclusions is that, because of our different ways of measuring hospital costs, I estimate a total effect of extended care facilities and Russell estimates a partial effect. More specifically, I measure the general level of hospital costs in the state by the average cost per patient day in short-term general hospitals. This is an average cost for all patients, not just Medicare patients. Russell uses the average reimbursement per patient day of