Problems of Mortality Measurement at Advanced Ages

Leonid A. Gavrilov, University of Chicago

Mortality measurement at advanced ages suffers from several problems: (1) small numbers of survivors to advanced age, which requires mixing different birth cohorts with different mortality;(2) extremely high risks of death at old ages, which make the standard assumptions of hazard rate estimates to be invalid; (3) age misreporting by old persons. The Social Security Administration Death Master File (DMF) was used to alleviate the two of the above mentioned problems and to obtain more precise monthly estimates of hazard rates after ages 85-90 years. Study of several single-year extinct birth cohorts showed that mortality grows steadily without deceleration from 80 to 102-105 years of age. Then statistical noise rapidly increases and mortality tends to decelerate. The study shows that mortality deceleration effect at advanced ages is not particularly strong when data for more homogeneous single-year birth cohorts are analyzed.