Abstract
The purpose of this paper is to examine the hypothesis that marital and poverty status interact in their effects on mortality risks beyond their main effects. This study examines the epidemiological bases for applying an additive rather than a multiplicative specification when testing for interaction between two discrete risk factors. We specifically predict that risks associated with being nonmarried and with being poor .interact to produce mortality risks that are greater than each risk acting independently. The analysis is based on men and women who were ages 25–74 during the 1971–1975 National Health and Nutrition Examination Survey I (NHANES I) and who were traced successfully in the NHANES I Epidemiologic Follow-Up Study in 1982–1984. Overall, being both poor and nonmarried places nonelderly (ages 25–64) men, but not women, at risk of mortality greater than that expected from the main effects. This study shows that for all-cause mortality, marital and poverty status interact for men but less so for women; these findings exist when interaction is assessed with either a multiplicative or an additive standard. This difference is most pronounced for poor, widowed men and (to a lesser degree) poor, divorced men. For violent/accidental deaths among men, the interaction effects are large on the basis of an additive model. Weak main and interaction effects were detected for the elderly (age 65 +).
