Abstract
Health is typically imperfectly measured. How important is this imperfect ob- servability to evaluate the costs of bad health? We estimate a dynamic, structural life-cycle model of savings and labor supply with health risk under two assump- tions on the observability of health. The first one, which is prevalent in much of the literature, is that health is perfectly observable. The second one is that, while health is not observable, a battery of noisy measures of health is available to the researcher. We find that ignoring measurement error in health leads to substan- tially underestimating both the persistence of health and the time costs of being unhealthy. Ultimately, measurement error has an effect on the estimated lifetime costs of bad health—as measured by labor earnings, hours worked, consumption, and assets—leading to underestimate these by as much as 300%. A key message of our paper is that estimating the lifetime costs of bad health using structural economic models requires researchers to worry about measurement error in health.
