Spatial Analysis of Divorce Rate in Taiwan

離婚率上升是近年來越來越受到重視的一個社會現象，是值得探討其究的，以俾利政府在相關政策之決策、施行或推動配套措施時有所參據。而本文是第一篇從空間分析的角度，探究經社背景變數及擴散效應對於離婚率變化的影響，並分別比較金融危機前後二個時期，探討不同時期自變數的變化，以及可能產生結構性差異。研究是先從實證的角度，以2004 年與2013年之離婚率並選取22 縣市之368鄉(鎮、市、區)影響離婚率變動的因素之變數資料為基礎，探究其離婚率分別在非空間模型及空間模型下之實證及分析結果。 另本研究方法係透過空間計量經濟學理論結合空間Geoda 軟體，首次嘗試將離婚率在空間上的自我相關納入空間模型中，先對資料做全域型空間自我相關(Moran’s I)測試，與地域型 LISA 分布圖，除探討空間自我相關特性對離婚率之影響，與了解空間自我迴歸是否能解釋與鄰近區域呈現空間自我相關。最後比較傳統迴歸模型與空間落遲模型與誤差模型作分析比較，找出最合適的空間模式。The rising divorce rate, a social phenomenon which gets more and more attention in recent years, is worth exploring and could become the reference for our government when making and executing relative policies, or promoting supporting measurements. This article is the first one to explore the variables of Economic and Social background, and the impact of spillover effects on the changes in divorce rate from the perspective of spatial analysis, and compared the two periods before and after the financial crisis for studying the changes of independent variables at different times and the possible structural difference. The research started with an empirical perspective, based on the factors which influenced the divorce rate in 368 districts out of 22 cities in 2004 to 2013, and explored the divorce rate under non-spatial models and spatial models. Moreover, the research is the first attempted to apply the spatial autocorrelation model on divorce rate, through the method of combining Spatial Economics theory with GeoDa. First, to complete the test of full domain autocorrelation (Moran’s I) and the geographical distribution plot of LISA, then study the effect of spatial autocorrelation on divorce rate, and understand the ability of explaining and presenting the spatial autocorrelation in neighboring regions for space self-regression. Finally, to identify the most appropriate spatial patterns through a comparative analysis of traditional regression models, Spatial Lag Model, and Spatial Error Model