The Status of Health in Demand Estimation
Beyond Excellent, Good, Fair, and Poor
This study addresses two issues. (1) What can one gain by using more comprehensive measures of health status in demand estimation than a common single item measure? Would you rate your health as excellent, good, fair, or poor? The authors find that by using multidimensional and less-coarse health status measures they achieve an increase in precision approximately equivalent to a 10 percent increase in sample size. (2) What is the consequence of employing postdiction (i.e., predicting utilization from health status measured after the fact) rather than prediction? Using a simple, but plausible, model, the authors show that such measures cause the estimates to be inconsistent; the direction of the inconsistency generally cannot be signed a priori. Empirically the direction is generally away from zero.