Sample Selection in the Health Insurance Experiment

This Note compares the sample of individuals enrolled in the experimental portion of the Health Insurance Experiment (HIE) with the sample that was not enrolled but completed a baseline interview. Failure to enroll could result from refusal, a move out of the area, not being asked to enroll, or other reasons. The results show that these two samples are more similar on the key variables of pre-experimental physician visits and self-rated health status than would be expected from simple random sampling. The favorable degree of matching reflects the use of the Finite Selection Model in the design of the experiment. The group that enrolled, however, did include more children than the group that did not enroll. Provided that appropriate adjustments are made for age, this analysis suggests that HIE samples will support inferences to the population from which they were drawn at least as well as traditional random sampling methods with a 100 percent enrollment rate.