以極端差額模型評定多個台灣銀行產業之優質者

Abstract

我們評量台灣 31 家銀行的績效,收集了多項指標的數據,每個銀行輪流作為被評量的 主角( Bank o ),量測它與其他銀行之間的相對綜合績效值及虛擬差額值。透過一線性 規劃之極端差額 EGM 模型,評量時分別找出優於 Bank o 之最大的虛擬差額值,線性規 劃模型目標式為在不同銀行間所找出優於主角 Bank o 之優質者(稱此銀行為 Bank α )的 虛擬差額(虛擬投入扣除虛擬產出)使其最小化。則所找到之參考對象銀行 Bank α 將為主 角 Bank o 之改善目標。此外,進行改善時分別找出單一的最優者( Bank α )為基準進行 改善之動作。若改善能力受限制,則可剔除 Bank α ,以次優者為目標重新評量。以此 類推,至其改善能力能達成為標的。本篇研究將利用 EGM 模型進行分析,並針對 EGM 模型與差額型模型 GBM 進行更詳細的比較與說明。

We need to assess a set of bank with a bundle of inputs and outputs items. Each Bank o evaluates his performance against the peer banks. We establish a min-max linear programming model to identify all the banks that superior than Bank o . The objective function of the model is minimizing the largest virtual gap (i.e. the virtual inputs minus virtual outputs) among those superior banks. If a bank has been identified has the largest virtual gap, it will be the improvement target of Bank o , name as Bank α . Our min-max model identified multiple Banks that are the target points for gradual improvement step by step. Bank o may not have the capability to improve to as Bank α . We could select the next superior Bank be the Bank α , repeat the process until a proper target is identified. The improvement process consists of determining a sequence of targets so that they are easily achievable from the previous ones and considered as a detailed path from the initial situation to the final target.