台灣基層農業金融營運空間模型之研究

Spatial Model on the Basic Agricultural Finance Operation in Taiwan

基層農會信用部的盈餘，地區性差異很大，都會區或鄰近都會區的農會，其盈餘均比遠離都會區或純農業區的農會為高，但過去的研究，並未考慮基層金融之間的鄰近效果，因此，本文將透過空間模型來探討基層農業金融間的空間相關性及其鄰近效果，採用台灣245個鄉（鎮、市）農會信用部為樣本，以存放比率、逾放比率、信用部總收益為自變數，盈餘為依變數。實證結果發現盈餘具有高度的空間相關性，顯示與鄰近鄉鎮地區有相似的屬性，且有聚集的現象；盈餘的空間相關區域指標分佈圖中，農會本身盈餘高，其鄰近農會的盈餘亦高者多集中於北部，農會本身盈餘低，其鄰近農會的盈餘亦低者多集中於中南部。在模型設定方面，加入空間因素後解釋力提高，參數檢定結果，各年度的逾放比率和盈餘均呈現負向的極顯著關係，總收益和盈餘均呈現正向的極顯著關係，此與實務的認知一致。但傳統迴歸模型會高估自變數對依變數的影響，如逾放比率對盈餘的影響，比空間落遲模型高14.86％，比空間誤差模型高5.52％。總收益對盈餘的影響，則比空間落遲模型高9.53％，比空間誤差模型高4.57％。 For the earnings of the credit departments of the basic farmers' association (FA) in Taiwan, it has a great difference between urban areas and rural areas. Earnings of the FA in metropolitan area or adjacent metropolitan area are higher than that of the areas away from the metropolitan or pure agricultural areas. However, for the researches on the past, it did not consider the adjacent effect among the basic financial units. Therefore, this study will examine the spatial correlations and its adjacent effects with the spatial model. With 245 FA as a sample, independent variables are the ratio of deposit to loan, non-performing loan ratio (NPL ratio), and total revenue of the FA. Dependent variable is the earnings. From the results of the empirical study, it finds that earnings with high degree of spatial correlation. It shows that has a similarity among adjacent township areas and with phenomenon of agglomeration. From scatter diagram of the LISA on the earnings, the FA fall on the HH quadrant concentrated in northern area and fall on the LL quadrant concentrated in middle and southern areas. As to the model, after adding spatial factor to the model, it can increase the coefficient of determination. From parameter of test results, it shows that a very significant negative relationship between NPL ratio and earnings, and a highly significant positive relationship between total revenue and earnings. It is consist with the practices. However, for the traditional regression model will overestimate the impact of independent variable on the dependent variable, such as the impact of NPL ratio on the earnings, higher than spatial lag model 14.86%, higher than spatial error model 5.52%. For the impact of total revenue on the earnings, traditional regression model is higher than spatial lag model 9.53%, higher than spatial error model 4.57%.

For the earnings of the credit departments of the basic farmers' association (FA) in Taiwan, it has a great difference between urban areas and rural areas. Earnings of the FA in metropolitan area or adjacent metropolitan area are higher than that of the areas away from the metropolitan or pure agricultural areas. However, for the researches on the past, it did not consider the adjacent effect among the basic financial units. Therefore, this study will examine the spatial correlations and its adjacent effects with the spatial model. With 245 FA as a sample, independent variables are the ratio of deposit to loan, non-performing loan ratio (NPL ratio), and total revenue of the FA. Dependent variable is the earnings. From the results of the empirical study, it finds that earnings with high degree of spatial correlation. It shows that has a similarity among adjacent township areas and with phenomenon of agglomeration. From scatter diagram of the LISA on the earnings, the FA fall on the HH quadrant concentrated in northern area and fall on the LL quadrant concentrated in middle and southern areas. As to the model, after adding spatial factor to the model, it can increase the coefficient of determination. From parameter of test results, it shows that a very significant negative relationship between NPL ratio and earnings, and a highly significant positive relationship between total revenue and earnings. It is consist with the practices. However, for the traditional regression model will overestimate the impact of independent variable on the dependent variable, such as the impact of NPL ratio on the earnings, higher than spatial lag model 14.86%, higher than spatial error model 5.52%. For the impact of total revenue on the earnings, traditional regression model is higher than spatial lag model 9.53%, higher than spatial error model 4.57%.