有別於以往許多老人居住安排的文獻常將65歲以上的人當作一個群體，或有的橫斷面研究使用不同樣本，抑或即使以縱斷面研究亦使用不同樣本，進行老化研究，本文應用我國衛生署與密西根大學合作之「臺灣地區中老年身心社會生活狀況長期追蹤調查」的縱斷面資料，利用相同樣本縱斷面研究分析老年人在初老、中老、老老階段的居住安排，回答「隨著年齡的增長，老人與子女共住的比例是呈U型或直線下降的關係？」、「老化對居住安排有何影響？」的問題。研究結果顯示隨著年齡的增長，老人與子女共住的比例呈直線下降。本文發現一些因素在某階段有顯著影響，而在其他階段未發現有顯著影響，此即與老化有關；除此，對三個階段的居住安排均發生顯著影響的因素，其影響程度亦隨著年齡的增長而有不同，均為預測老人各階段居住安排決定的良好指標。 Most previous research on elderly living arrangements examined all people aged 65 and over as a whole. Some cross-sectional studies on aging employed different samples, as did some longitudinal studies. This study examines the living arrangements of young-old, old-old, and oldest-old in Taiwan using a longitudinal data source, produced by the Department of Health (Taiwan) and the University of Michigan. A panel study was used to answer the following questions: Is the proportion of coresidence with increased age U-shaped or decreased linearly? What influence does aging have on living arrangements? Analytical results show that the proportion of coresidence decreased linearly among the same samples for the young-old, old-old, and oldest-old. Empirical results show that some factors were significant at some stages, while not significant at other stages, that is, they were affected by aging. Additionally, some factors related to living arrangements for all three age groups, and their effects differed with increased age. All of them are good indicators in predicting the determinants of elderly living arrangements for the three groups.