Site Selection Criteria for the Health Insurance Study

Sites for drawing Health Insurance Study (HIS) experimental samples should be chosen to generate estimates that (1) reflect national norms and (2) vary minimally. Selecting several sites rather than a dispersed national sample will reduce variance but may increase bias. Sites should be chosen purposively (rather than randomly) to avoid the biasing effects of "concomitant characteristics," unique local conditions that cannot be controlled by specifying certain demographic characteristics in the sample. Concomitant characteristics most significant to the HIS are the capacity utilization of local physicians and hospitals, and regional effects on the demand for health care. Variance can also be reduced by lowering experimental costs, which permits a larger sample for a given budget. Intersite cost differences in the HIS will ultimately depend on local hospital charges, local physicians' fees, and minimum coinsurance rates and deductibles in participants' preexisting health insurance policies.