Author(s): Joshua Nicholls, Jacob Hutchings
Mentor(s): Joe Price, Sven Wilson
Institution BYU
We use a new dataset that contains the lifespan of about 15 million couples drawn from the US census from 1880-1940 to examine long-run patterns in the spousal correlation of lifespan. We use a sample of 26 million opposite-gender sibling pairs as a comparison group for our couples. First, we provide novel descriptive methods for examining the spousal correlation in lifespan. One method is the 10-year co-mortality rate, where we consider the fraction of couples that both die within an interval of time and compare this to randomly assigned marriages. Our second novel method is examining the density of the difference in death year within a spouse pair compared to randomly assigned marriages. Both of these methods suggest results consistent with existing literature: that spouses die in close proximity to each other at surprisingly high rates. We find that this is especially true for young couples. We show that the correlation in lifespan (measured by the regression coefficient of husband lifespan on wife lifespan) between couples nearly doubled from 0.05 to over 0.10 between the 1860 and 1920 birth cohorts. We estimate these same correlations for key subgroups of our sample, including immigrants and non-white racial groups, and find evidence of heterogeneity in the estimates. Specifically, the correlations are stronger for non-whites (~0.85) and weaker for immigrants (~0.4). An instrumental variables model using the lifespan of the spouses' same-gender siblings to account for endogeneity reveals that the true effect may be closer to .2 and has also increased markedly over time. Considering the stochastic nature of lifespan, these estimates are very noteworthy.