Underestimation of Disabled Life Expectancy 335 Demography, Volume 41-Number 2, May 2004: 335–361 335 A A NEW METHOD FOR CORRECTING UNDER- ESTIMATION OF DISABLED LIFE EXPECTANCY AND AN APPLICATION TO THE CHINESE OLDEST-OLD* ZENG YI, GU DANAN, AND KENNETH C. LAND This article demonstrates that disabled life expectancies that are based on conventional multistate life-table methods are significantly underestimated because of the assumption of no changes in functional status between age x and death. We present a new method to correct the bias and apply it to data from a longitudinal survey of about 9,000 oldest-old Chinese aged 80–105 col- lected in 1998 and 2000. In our application, the age trajectories of disability (activities of daily living—ADL), status-specific death rates, and the probabilities of transitions between ADL states of the oldest-old were investigated for the first time in a developing country. In this article, we report estimates of bias-corrected disabled and active life expectancies of the Chinese oldest-old and dem- onstrate patterns of large differences associated with initial status, gender, and advances in ages. Using combined information on ADL disabilities and length of having been bedridden before dying, we analyze gender and age patterns of the extent of morbidity before dying among the oldest-old and their implications for debates on the hypothesis of compression of morbidity. t the population level, demographers have used the Sullivan method and multistate life-table models to summarize information about functional capacity in activities of daily living (ADL) obtained in surveys in the form of estimates of active and disabled life expectancy. These estimates provide easily understandable indicators of the functional capacity and caregiving needs of the elderly population (Crimmins, Hayward, and Saito 1994, 1996; Crimmins, Saito, and Ingegneri 1997; Robine, Mathers, and Bucquet 1993; Rogers, Rogers, and Belanger 1990; Rogers, Rogers, and Branch 1989). The Sullivan method (Crimmins, Saito, and Ingegneri 1989; Robine et al. 1986) is a useful approach for providing cross-sectional period-specific summary measures of disability status when lon- gitudinal data on transitions between functional statuses and differences in mortality among people with different functional statuses are not available. When such longitudinal data are available, the multistate life-table method, which accounts for dynamic changes in func- tional statuses and differences in mortality by functional statuses, is preferable (Crimmins et al. 1994, 1996; Crimmins et al. 1997; Laditka and Hayward 2003; Land, Guralnik, and Blazer 1994; Manton and Land 2000a; Rogers et al. 1990; Rogers et al. 1989). A recent *Zeng Yi, Research Professor, Center for Demographic Studies and the Department of Sociology, Duke University; Professor of China Center for Economic Research, Peking University; and Distinguished Research Scholar, Max Planck Institute for Demographic Research. Gu Danan, Research Associate, Sanford Institute for Public Policy, Duke University. Kenneth C. Land, John Franklin Crowell Professor of Sociology and Director of the Center for Demographic Studies, Duke University. Address correspondence to Zeng Yi, Center for Demo- graphic Studies, Duke University, Campus Drive, Box 90408, Durham, NC 27708; E-mail: zengyi@duke.edu. This article is based on data that were derived from the Chinese Longitudinal Healthy Longevity Survey, which was supported by Grant P01 AG 08761 from the National Institute on Aging awarded to Duke University and Chinese matching input toward personnel costs and some local expenses. The Max Planck Institute for Demo- graphic Research provided support for international training. We greatly appreciate the invaluable comments provided by the three anonymous referees and the editor. An earlier version of this article was presented at a seminar of the International Union for the Scientific Study of Population, entitled “Increasing Longevity: Causes, Consequences and Prospects,” Rockefeller University, New York, October 20–22, 2003; we are grateful to E.M. Crimmins and J.M. Robine for their useful comments. Downloaded from http://read.dukeupress.edu/demography/article-pdf/41/2/335/884086/335yi.pdf by guest on 25 May 2022