本文旨在透過Translog 成本函數及潛在類別隨機邊界法，進行信用部群組區分與經營績效之評估。除藉由共同邊界（Metafrontier）概念進行信用部效率指標計算外，同時亦探討信用部成本結構狀況。研究對象為237 家農會信用部，期間涵蓋2001 至2009 年，九年總計2,133 筆縱橫（Panel）資料。主要實證結果顯示：採用潛在類別隨機邊界法，可將全體信用部區分為兩類群組；其中，群組一特徵表現與農業具較高相關性，而兩群組平均效率水準相當接近，且以群組一變動程度較高。再者，不論何種技術型態信用部，皆能發揮總規模經濟及範疇經濟效果，且隨著時間增加，無效率項會逐漸遞減，顯示具有技術進步現象。經由本文客觀且一致性的分組方式與效率之衡量，後續實務應用上，可作為相關單位輔導、衡量與評鑑信用部績效等任務之參考。另外，兩群組信用部尚有潛在成本節省空間，此時應更加重視商品組合多樣性，以充份發揮成本節省優勢。

This paper aims to measure cost efficiencies and to differentiate groups among the existing credit department of farmers’ associations (CDFAs) in Taiwan, via translog cost function and latent class stochastic frontier. Metafrontier is also used in this research to study cost efficiencies and structure of these CDFAs. The empirical analysis is based on 2,133 panels from 237 CDFAs over the period 2001-2009. This study reveals: these CDFAs can be divided into two groups according to the latent class stochastic frontier model. Although these two groups share comparable mean TGR (Technology Gap Ratio) efficiency, Group 1 shows higher relevancy to the agricultural financial system. Besides, no matter what technology either group uses, both are proven scale and scope economies, and also see technical progress. Finally, the outcomes of this study – the measurements on grouping and efficiencies – are applicable in the measurements and evaluation of the CDFAs in the future. Furthermore, CDFAs can reduce its average cost through further diversifying their line of financial services