兩岸銀行投入效率之比較研究：DEA與SFA之應用

Abstract

近年來兩岸經濟活動益趨繁盛，隨著兩岸金融監理備忘錄(Memorandum of Understanding, MOU)的簽訂，兩岸銀行業密切合作可期。本文旨為探討兩岸銀行業之投入要素效率，分析兩岸銀行個別要素優劣勢，以便作為型塑合作策略的參考。本研究採用資料包絡分析法 (DEA)與隨機邊界法(SFA)，同時進行兩者之比較；本文延伸Zhou et al. (2012)之橫斷面資料模型，並使用Battese and Coelli (1992)之縱橫面模型進行實證估計。本研究資料採用兩岸共35家銀行之年資料，自2006年至2011年，總計有210筆資料，同時以2000年為基期進行平減。本研究參酌Hu and Wang (2006)，計算出兩岸銀行業的總要素效率值。共計有三項投入(員工數、總客戶存款、固定資產)，兩項產出(總放款、總營利性資產)，及四項環境變數(營運年限、國內儲蓄率、貪腐指數、經濟自由指數)。實證研究的結果包含銀行的資深員工對於工作領域較為熟悉，因此對於投入要素效率值產生正向影響；成立較久之銀行在制定策略上較為保守。兩岸銀行之效率值均在簽訂MOU之後隨之提高，而大陸銀行在客戶存款與固定資產的投入效率值平均高於台灣，台灣則在員工的品質與效率上展現優勢。本文針對DEA與SFA兩者方法個別產生之效率值進行比較，結果顯示DEA產生之效率值接近效率前緣，數值普遍高於SFA；兩者效率值之先後排序呈現相當低的相關性。

This study aims to analyze disaggregate input efficiency of banks in Taiwan and mainland China by employing a stochastic frontier approach (SFA) and makes a comparison to the data envelopment analysis method (DEA). We extend the cross-section SFA model of approach of Zhou et al. (2012) and to a panel data model which can use the SFA model of Battese and Coelli (1992). The disaggregated input efficiency scores for 35 banks across the Taiwan-strait during from 2006 to 2011 are estimated, up to a total six-year amount of 210 observations. There are three inputs (number of employees, total customer deposits, and fixed assets), two output (gross loans and total earning assets) and four environment variables (the operation year, gross domestic saving rate, corruption perception index, and index of economic freedom) which are adjusted by GDP deflator adopting 2000 year as the base period. Our major empirical findings are as follows: The senior employees may have more sense of familiarity toward the work field, hence making positive influence on input efficiency scores. A senior bank in drawing up its market strategy is more conservative. The efficiency values of the banks across the Taiwan-strait are going higher with time after signing the MOU and ECFA. The input efficiency of customer deposits and fixed assets of the banks in mainland China is averagely higher than those in Taiwan, although Taiwan owns strength in employees’ quality. This study also gives the comparison of the DEA and SFA approaches. The scores of DEA are not only more close to the efficiency frontier but also are higher than the ones using the SFA method and exists a weak correlation of the SFA and DEA paired data.