The effect of HMOs on overall hospital expenses: is anything left after correcting for simultaneity and selectivity?

Policymakers are interested in the effect health maintenance organizations (HMOs) have had not only on the hospital expenditures of their enrollees, but also on the expenditures of non-HMO consumers. Previous studies of the "HMO effect" have focused on the comparison between hospital expenditures of HMO enrollees and those of non-HMO groups within the same market area. To the extent that the expenditures of non-HMO groups are affected by the presence of HMOs, this comparison will not give an accurate measure of the change in expenditures due to HMOs for either group. Using SMSAs without any HMOs as the comparison group can provide an accurate measure of the HMO effect on overall hospital expenses, if any nonrandom selection process of HMOs into SMSAs is controlled. In this article, the effect of prepaid group practices (PGPs) on overall hospital expenses is estimated using a simultaneous-equation model and all 283 standard metropolitan statistical areas (SMSAs) in 1980, controlling for any nonrandom selection process. The results indicate that while a significant selectivity bias exists that must be corrected, there is no significant PGP effect on hospital expenses. Controlling for other factors, hospital expenses per capita, per admission, and per day do not change as the PGP market share increases. While hospital expenses per capita for PGP enrollees may be falling, those of non-PGP groups in those SMSAs must be increasing by offsetting amounts.