Measuring the Biases in Self-Reported Disability Status: Evidence from Aggregate Data December 2007 Naoko Akashi-Ronquest * – California State University, Fullerton Paul Carrillo † – George Washington University Bruce Dembling ‡ – University of Virginia Steven Stern § – University of Virginia Abstract There is an extensive amount of literature that seeks to explain both a) the Disability Insurance's (DI) application and award process and b) the relationship between DI generosity and labor force participation. In addition, it has been widely documented that self-reported health status may be subject to endogeneity problems and measurement error. While endogeneity issues may overstate the effect of disability status, attenuation bias may understate it. In this paper, we employ county level aggregate data to analyze the determinants of variation in Social Security Disability rates and use instrumental variables to control for these possible biases. We find two surprising results. First, we provide evidence that, as the proportion of disabled people in a county increases, the proportion of SSDI beneficiaries rises more than proportionally. This finding suggests that there may be synergies for applying for SSDI when the disabled population is larger. Second, we show that measurement error is the dominating source of the bias and that the main source of measurement error is sampling error. * 800 N. State College Blvd, Fullerton, CA 92834 (nakashi@fullerton.edu) † Old Main, 226, 1922 F St. NW, Washington, DC 20052 (pcarrill@gwu.edu) ‡ Public Health Sciences, University of Virginia, Charlottesville, VA 22903 (bpd6n@virginia.edu) § Department of Economics, Dynamics Building, University of Virginia, Charlottesville, VA 22903 (sns5r@virginia.edu)