Endogenous Living Arrangement and Health of Older Adults: Evidence from Instrumental Variable Estimation

Abstract

Background: Health status of older adults living independently is of great concern. The literature has reported conflicting evidence on the association between independent living arrangements (ILAs) and several health outcomes. This study aims to address the endogenous relationship between ILAs and health of older adults by means of the instrumental variable (IV) approach and confirm essential mechanisms relating ILAs to the health of older adults. Methods: Data were obtained from a cross-sectional survey of older adults in Xiamen, China, in 2013 (N=14292). Using number of daughters and Governmental Employee Health Benefits as IV, we applied mixed effect linear models. General health and health outcomes of six dimensions were measured by the Multidimensional Functional Health Scale for Chinese Elderly. Results: Overall, 14,239 participants had complete data, of whom 40.45% lived independently. The results indicated that conventional single-equation models significantly underestimated the detrimental effects of ILAs on activities of daily living (ADL) and general health. After accounting for endogeneity, older adults in ILAs were worse in ADL by 3.445 points (95% CI: -8.207 – 0.279) and worse in general health by 2.759 points (95% CI: -5.343 – -0.740), compared with older adults of coresidence. The older adults who were unmarried and living independently had the worst general heath. The results also verified the mechanism of psychological well-being in determining the health outcomes of ILAs. Conclusions: We found that the IV estimator could effectually correct the endogeneity bias within ILAs and health of older adults, which provides further evidence that older adults in ILAs are more vulnerable in multiple dimensions of health. We demonstrated a hypothesis that Chinese older adults compensate for their health decline by selecting coresidence. The understanding of this reciprocal association and the social processes that generate the vulnerability of ILA can be conducive to enhance estimation in empirical studies and improve the long-term care supply.