作者
郭柏昇^* (中央銀行經濟研究處)

Author
Po-Sheng Kuo^* (Department of Economic Research, Central Bank of the Republic of China (Taiwan))

中文摘要
本研究探討本國銀行風險承擔決策是否受同儕銀行所影響，估計銀行風險承擔之同儕效果。為捕捉同儕銀行之間的相依關係，本研究利用 Fama and French (2015) 五因子資本資產定價模型 (capital asset pricing model, CAPM) 與 Diebold and Yilmaz (2014) 網路拓樸分析，建構同儕相依矩陣，並以動態空間自我迴歸 (dynamic spatial autoregressive, DSAR) 模型進行實證分析。樣本銀行包含本國 6 家泛公股銀行及 15 家民營銀行，合計 21 家上市銀行，樣本期間則為2009 年至 2024 年。實證結果顯示，全樣本期間內銀行風險承擔變化存在正向的同儕效果，表示銀行進行風險承擔決策時會觀察其他銀行的行為，並做出與同儕銀行相同的決策；惟動態效果為負向，表示銀行風險承擔的增幅並不會隨時間而持續上升，其可能為來自金融監理機關或股東的壓力所致。此外，新冠肺炎 (COVID-19) 疫情後之樣本期間同儕效果轉變為負向，顯示面對未來不確定性高的經濟金融環境，銀行傾向於採取與同儕銀行不同的風險承擔決策。

Abstract
This research investigates whether peer institutions influence the risk-taking decisions of domestic banks in Taiwan via estimating the peer effect in bank risk-taking. To capture interdependencies among peer banks, it constructs a peer-dependence matrix using the Fama and French (2015) five-factor capital asset pricing model (CAPM) combined with the network topology analysis of Diebold and Yilmaz (2014) and conducts empirical analysis employing a dynamic spatial autoregressive (DSAR) model. The sample comprises 21 publicly-listed banks, including six government-controlled banks and 15 privately-owned banks, covering the period from 2009 to 2024. The results indicate the presence of a positive peer effect in bank risk-taking throughout the sample period, suggesting that banks tend to observe the behavior of their peers and align their risk-taking decisions accordingly. However, the dynamic effect is negative, implying that increases in bank risk-taking do not persist over time, potentially due to pressure from financial regulators or shareholders. Like in the period following the outbreak of the COVID-19 pandemic, when the economic and financial environment becomes increasingly uncertain, the peer effect turns negative, indicating that banks are more likely to adopt risk-taking strategies that diverge from those of their peers.

中文關鍵字
銀行風險承擔、同儕效果、動態空間計量、系統風險

Keywords
Bank Risk-Taking, Peer Effect, Dynamic Spatial Models, Systemic Risk