Search

A method for estimating year of birth using only Social Security number is described. The method relies on estimating the year of issue of the Social Security number, using either precise information from a Social Security Administration table (for numbers issued since 1950) or extrapolation (for numbers issued before 1951). Age at issue was estimated using data from individuals in an occupationally defined population for whom both birth date and Social Security number were known. Year of birth is then year of issue minus age at issue. A highly statistically significant correlation of 0.91 for known vs. estimated year of birth was observed. Predictions correct within +/- 5 years were achieved for 78 per cent of the entire study population. For individuals born after approximately 1934, predictions within +/- 5 years of the correct year were achieved 95 per cent of the time, and within +/- 2 years 77 per cent of the time. The method may be useful in epidemiologic research in situations in which date of birth is missing. Some applications are discussed.