Health and Residential Mobility in Later Life: A New Analytical Technique to Address an Old Problem

For some time researchers have known that the relationship between health and the residential mobility of the elderly is not straight forward and changes with age. Attempts to examine this relationship in multi-variate models using cross-sectional data have resulted in contradictory or ambiguous findings. One solution has been to create separate models for different age groups. However, the onset of poor health differs considerably by individual, particularly for the "young-old". Multi-variate proportional hazards models using longitudinal data offer a new approach to address this problem. As an example, data from the Ontario Longitudinal Study of Aging have been analyzed using proportional hazards models as compared with logistic regressions. The logistic regressions yield typically ambiguous results. The proportional hazards models indicate a reversal with time in the relationship between one of the two mid-life health measures and residential mobility, and the results for both measures are consistent with the theoretical literature.