The Impact of Data Limitations on the Statistical Reliability of Mortality Improvement Rate Estimates

December 2024

Author

Society of Actuaries Research Institute
Mortality Improvement Reliability Committee

Description

Estimating historical mortality trends—which this report refers to as “mortality improvement rates”—is statistically challenging. The frequently cited rule-of-thumb that 1,082 deaths are needed to achieve a credible mortality rate estimate is not applicable to improvement rate estimation. This rule-of-thumb assumes that a single rate, rather than the ratio of two rates, is being estimated. To produce a statistically meaningful estimate of a mortality improvement rate, a larger pool of data is required. This report examines how the number of persons (“lives”) covered in a mortality study, and the length of the study period (in years), affects the statistical reliability of improvement rate estimates. In addition, the report examines practical strategies for addressing data limitations, such as estimating improvement rates by age group instead of by single age, as well as smoothing results across single ages.

The results presented in this report were produced using an Excel/VBA simulation tool. The tool is available for download, along with the report.

Materials

The Impact of Data Limitations on MI Estimates

Simulation Tool for Assessing the Reliability of MI Estimates

Questions or Comments?

Give us your feedback! Take a short survey on this report. Take Survey

If you have comments or questions, please send an email to Research-ML@soa.org.