Iceland認爲,所得增加、所得分配不均和家戶結構的改變,都是影響貧窮趨勢之主因。在台灣,由於家戶所得持續增加,預計將能改善貧窮水準;然而,此時家戶所得分配卻趨向不均,可能削弱所得成長的效果。另外,台灣的家戶結構也與過去不同,這些變化勢必影響整體貧窮趨勢。討論貧窮趨勢演變之前,得先定義何謂「貧窮」。由於官訂貧窮線無法真正代表貧窮水準,本文使用Citro and Michael的FCSU法來描繪貧窮線。接著,透過Danziger and Gottschalk's的方法,控制其他因素,模擬1990年至2004年(依時間劃分爲1990-1994年、1995-1999年與2000-2004年三個階段)之間,個別因素對貧窮趨勢的效果。結果發現,所得成長降低貧窮水準的效果僅發生在1990-1994年與1995-1999年兩階段;2000-2004年,由於家戶所得減少,故效果爲增加1%貧窮率,即「雨露均霑」效果隨時間遞減。同時,所得增加及家戶結構改變對貧窮趨勢影響之效果,亦隨時間增加而遞減。相對的,由於台灣所得分配朝不均發展,將造成貧窮率上升,特別是1995-1999年和2000-2004年兩階段,增加貧窮率將近4%。
Iceland holds that income growth, economic inequality, and family structure are the main factors to explain the poverty trends. In Taiwan, income growth has occurred steadily, and this might have had a negative effect on poverty. However, we found that economic inequality increased simultaneously, which could mitigate the impact of income growth. At the same time, changes in family structure also contributed to the alteration of poverty rates. What were the relative associations among trends of poverty in 1990-2004 and income growth, economic inequality, and changes in family structure? As the official poverty rates do not reflect important changes in such trends, it is necessary to first define ”poverty”. The FCSU method, devised by Citro and Michael, is then used in drawing the poverty line. To estimate the effect of income growth, economic inequality, and changes in family structure on poverty, we used Danziger and Gottschalk's method of exploiting direct standardization and simulations to calculate adjusted rates under different assumptions. Our results show the relationships between income growth and trends in poverty rates over the three periods (1990-1994, 1995-1999, and 2000-2004). Income growth decreased poverty during the 1990-1999 period. Between 2000 and 2004, income growth was not related to the decline in poverty rates (+1%), this was due to the dropping of family incomes. The effects of income growth and changes in family structure are becoming ever weaker. The outcome of economic inequality was an amazing contrast to other cases. In simulations, it was shown to have a significant effect (+4%) on the poverty rates during the periods of 1995-1999 and 2000-2004.