Individual reporting heterogeneity in self-assessed health

Abstract: Self-assessed health (SAH) questions have often been shown to correlate well with objective health measures and a large literature has focused on treating SAH responses as proxies of health and econometrically modelled them in various contexts. However, concerns that SAH questions are susceptible to individual reporting heterogeneity biases suggest caution in estimation and interpreting of findings. We build on past methodology to allow explicit testing of the presence and extent of reporting bias in SAH questions. To do this, we develop a structural model that yields a set of latent factors that can be interpreted as physical and cognitive heath indicators. The general premise is that if SAH is a true measure of health, once controlling for true health and its socioeconomic/demographic gradient, the remaining heterogeneity should be minimal. Simulations test the power of our framework to retrieve true parameters of the DGP under various contexts and assumptions. Comparisons with standard ordinal models are drawn and reporting bias is formally derived for the predicted values to allow calculations of its direction and magnitude.