以停駐者──漂移者模型評量銀行經營績效之持續性

A Study of Performance Persisence in Taiwan Banks-By Mover-stayer Model

本研究利用「漂移者──停駐者」模型探討臺灣上市銀行之績效持續性，資料取自臺灣經濟新報的財務資料庫（TEJ Finance DB），樣本期間自2003年至2013年，選取股價報酬率及11類財務指標的季資料進行分析。就報酬率指標看，各組別銀行下一季停留在同一組別的短期持續性大致相同；在績效漂移行為方面，報酬率最高組別銀行下一季跌至最差組別的機率最大、報酬率中等組別銀行下一季跌至績效最差組別的機率較高、報酬率最差組別銀行下一季在各組間漂移機率大致相同；就長期持續性而言，各組別銀行均無法長期停駐在同一組別。再就他財務績效指標討論，在短期績效持續性方面，績效最差組別銀行的短期績效持續性最強，代表績效最差的銀行較難改變現況，呈現弱者恆弱的現象。在績效漂移方面，各財務指標顯現一致的漂移行為：（1）前一季績效落於最佳組別的銀行，停留在原組別的機率最高。（2）績效中等組別銀行大多維持前一季績效，一旦發生績效漂移，績效下滑至最差組別的可能性較高。（3）績效最差組別銀行，下一季表現仍以維持現況為主。在長期績效持續性方面，績效最差組別銀行的持續性最強、其次為績效最佳組別，績效中等組別銀行則無長期績效持續性，且多數指標所估計的銀行長期績效持續性機率值偏低。根據結果，本研究認為屬於金融控股公司體系下的銀行較獨立經營的銀行有更佳的持續性，原因在於金控體系下的銀行，能藉由業務多元化、人員專業化的服務經營模式為銀行帶來綜效，因此有較佳的績效表現及持續性。

This paper uses the mover-stayer model to study the performance persistence of banks listing in Taiwan Stock Exchange (TWSE). We use quarterly data by different financial indices from 2003 to 2013 to estimate the transition probabilities of movers and stayer's ratio in Taiwan banks industry. Our study finds that the short-term performance persistenace are almost the same among different groups when we use quarterly returns. In the drift behaviors, the probabilities of the best group changing to the worst group are higher, the probabilities of the middle group changing to the worst group are higher, and the probabilities of the worst group changing to other group are similar. In the long-term, there exists no bank that shows lon-term performance. As for the other indices, we conclude that: (1) The probabilities of the best performance group staying in the same group are highest in sequent periods. The probabilities of the middle group deteriorating their performances are higher than those of improving performances. (2) In the performance behaviors of movers, the best performance group shows P_(11) ＞ P_(12) ＞ P_(13) patterns, the worse performance group shows P_(33) ＞ P_(32) ＞ P_(31) patterns and the middle performance group shows P_(22) ＞ P_(23) ＞ P_(21) patterns. (3) In long term performance persistence, we find that there exists significant performance persistence in the worse performance group. And the evidences about performance persistence in the middle performance group are absent.