Does partnership predict mortality? Evidence from a twin fixed effects study design

This study examines the association between partnership status and mortality, addressing methodological concerns such as selection effects and unobserved heterogeneity. Utilizing Norwegian administrative data on birth cohorts from 1955 to 1975, we analyze mortality outcomes using Kaplan-Meier survival analyses and Cox proportional hazards regression models. The dataset includes over 1.2 million individuals and distinguishes between partnered and non-partnered individuals. While results for the general population show that non-partnered individuals face significantly higher mortality risks, with hazard ratios of 1.59 for men and 1.47 for women, twin fixed effects models reveal no significant relationship between partnership status and mortality. This finding suggests that much of the observed association may be due to shared genetic and environmental factors rather than a direct causal effect of partnership status. By leveraging the twin fixed effects research design, this study provides a robust test of the partnership-mortality hypothesis, highlighting the importance of controlling for unobserved heterogeneity in observational research. Our results underscore the complexity of the partnership-health nexus and call for caution in interpreting simple associations as evidence of causality.