引用
老年人居住安排的動態:Markov模型的設計與估計
Markov Model and the Dynamics of Elderly Living Arrangement in Taiwan
作者:陳寬政(Kuanjeng CHEN)、林子瑜(Tzuyu LIN)、 張雅君(Ya-Chun CHANG) | 首次發表於 2020-06-02 | 第 48 期 December 2011
DOI:https://dx.doi.org/10.6786/TJS.201112_(48).0005
研究紀要(Research Notes)
論文資訊 | Article information
摘要 Abstract
本文使用1986-2005年間,行政院主計處與內政部共十次老人生活狀況調查資料,就歷次老人居住狀態分布的關連建立馬可夫模型,估計前後期居住狀態間的移轉參數。老年人除了與子女同居之外,僅與配偶同居、獨居、居住於養護機構等皆是可能的選擇,到目前為止各項調查卻顯示與子女同居仍為臺灣老年人的主要居住型態,雖然歷年來與子女同居的比例顯有下降,代之而起的卻不是老年獨居的生活型態,而是老年人僅與配偶同居的生活型態。文獻指出,日本與臺灣老年人的居住型態相當穩定,與已婚子女同居是最常見且穩定的型態,其次為僅與配偶同居與獨居;獨居人口有相當比例從其他幾種居住型態移轉而來,主要是從僅與配偶同居移轉而來,而且後期居住型態受前期居住型態的影響很大。本文目的不在於這些移轉機率的調查與分析,而在於將橫斷面的老年人居住安排參數化,也就是建立模型說明各類居住狀態的比例分布及其變化,指出模型估計需要考量的幾個重要條件,提出參數估計及參數設限的方法與結論。
我們的估計程序顯示,在適當週邊條件的控制下,如果參數估計值落在適當的範圍0≦mij≦1內,馬可夫模型中不同等式對應參數累加為1的設限並非必要,自然累加為1;但是由於初步結果指出某些特定參數需要調整數值,則不同等式對應參數累加為1的條件就成為參數調整時的重要考量。調整參數當然就偏離最小平方法的結果,但是我們所設立的計算程序不但取得收斂結果,而且其差方和並不顯著不同於最小平方法所產生的差方和。參數估計的結果顯示,「與子女同居」表現為最穩定的居住狀態,其次為「僅與配偶同居」,「獨居」與「其他」居住狀態則穩定度較差;由於「其他」居住狀態包含高比例的養護機構老年人口,一般而言健康狀況較差,產生偏高的死亡率。與子女同居不但是最 穩定的居住狀態,也是移出所有其他居住狀態者的主要流向,仍然是老年人的首要歸宿,晚年生活的可靠依據。

關鍵詞:老人居住安排、馬可夫模型、移轉機率、參數設限
While the elderly may choose to live with children, with spouse only, live alone, or with others, it appears that the elderly in Taiwan predominantly live with the children and the spouse, reflecting the traditional family values. Based on longitudinal surveys, past researches in Taiwan and the East Asian countries indicate that living with children constitutes the most stable living arrangement at old age, with some transfers in between the different states in living arrangement. This paper parameterizes a series of the observed cross- sectional distributions of the elderly living in different states, estimates a Markov model to reveal a stable structure of flows underneath the observed distributions.
We employed parameter restrictions 0≦mij≦1 and Σi mij=1 to confine the parameters in estimating the transition matrix. Living with children is found the most favored and stable way of life at old age, followed by living with spouse only. Living alone is considered transitional in between living with spouse only and the other states. Living with others, mostly composed of elderly living in the institutions, is also transitional probably resulting from degradation in health. The high mortality rate implied by living with others compared to the other states suggested that elderly living arrangement is related to health, with death being the final state.

Keywords: elderly living arrangement, markov model, transition matrix, parameter restriction