Reply to Trends in Education-Specific Life Expectancy, Data Quality, and Shifting Education Distributions: A Note on Recent Research

Replication studies are commendable: no single study should be taken at face value. Nevertheless, replications should be approached with an equal amount of healthy skepticism. I turned to the National Health Interview Survey (NHIS) first when writing the original paper (Sasson 2016a) and found it too limited for studying trends in life expectancy and higher-order life table functions by education—particularly for U.S. minority groups. Unfortunately, the NHIS suffers from two notable sources of error that seem to outweigh the benefits of more accurate education reporting: (1) systematic error, in that its population is significantly healthier than the U.S. population, irrespective of education; and (2) random error, whereby considerable sampling variability prohibits any meaningful inference about trends in lifespan variation. Based on new results presented in this reply, I maintain that the National Vital Statistics System (NVSS), in spite of its known limitations, remains the single most comprehensive data source...