Abstract: | The problem of predicting mortality among rural elderly (65 years and older) living independently is examined using 20-year panel data derived from a random multistage cluster sample. Fifteen independent variables, including social networks, age, sex, and health status, were hypothesized on theoretical and empirical grounds to predict mortality. Face-to-face structured interviews were conducted with the same respondents in 1966, 1974, and 1986/87. Logistic regression establishes that a model comprising age, sex, participation in formal organizations, relative and children association, and general health status is a powerful predictor of mortality. The authors conclude that the more heterogeneous formal, or secondary, social networks which may enhance self esteem are most functional for the elderly. |