Description: Presented at Living to 100 symposium, January 2011. This paper presents a method for using death records to infer exposure on non-extinguished cohorts, thereby allowing the development of a credible table for high ages.

Description: Odd patterns in mortality curves can be caused by bad data. The author hypothesizes that the negatively sloped mortality rates at the highest ages for U.S. males are caused by overstating age at death, and the low mortality rates for Japanese females by failure to report deaths. The paper upholds the first hypothesis but soundly defeats the second. The paper also demonstrates the impact of mishandling incurred but unreported deaths.