台灣銀行信用卡外溢效果網絡分析

Abstract

本研究採用向量自我回歸模型和廣義預測誤差變異數分解方法，探索銀行間信用卡外溢效果。我們利用2015年1月至2023年6月台灣的銀行信用卡資料，涵蓋30家銀行的數據。通過收集到的外溢指標，將資料分為全樣本及分類樣本來製作網絡圖，探討銀行彼此之間的外溢效果及傳導結構，並分析銀行之間的關係。 研究中利用PR(Percentile Rank)百分等級來篩選銀行之間發卡成長率及簽帳金額成長率的鏈結指數。PR是一種統計工具，用於評估某一特定數據點在整個數據集中相對於其他數據點的位置。例如，當我們設定PR70%時，這表示我們只關注那些數據值高於70%的情況，這些數據點在整個數據集中屬於前30%的範圍，通常代表具有較高影響力或重要性的數據點。 在取得上述結果後，我們進一步設定條件，將高於PR70%的外溢效果納入網絡圖，並比較前後的結果差異。最後，使用簽帳金額成長率來做外溢效果的分析，以觀察其銀行之間在發生外生衝擊時的變化。我們發現在(1)發卡成長率估計出較高鏈結指數的節點會群聚在一起，其中民營銀行之間位置較緊密、(2)當發卡成長率鏈結指數設定高於PR70%之上時，以華南銀與中信銀為網絡圖的中心、(3)每卡簽帳金額成長率所繪製的網絡圖中，較高鏈結指數的節點並無群聚在一起，而網絡圖以中信銀為中心。

This study employs the Vector Autoregressive (VAR) model and the Generalized Forecast Error Variance Decomposition (GVD) method to explore the spillover effects on credit card issuances among banks. We utilize credit card data from 30 banks in Taiwan,spanning from January 2015 to June 2023. By collecting spillover indicators and using a network gravity layout, we divided the data into full sample and classified sample to investigate the spillover effects and transmission structures among banks, and analyze the relationships between these banks. After obtaining the aforementioned results, we further set the condition to include spillover effects higher than PR70 into the network graph and compared the differences before and after. Finally, we analyzed the spillover effects using the growth rate of transaction amount per card to observe the changes when exogenous shocks occur between banks. We found that: (1) Nodes with higher linkage indices estimated by card issuance growth rate tend to cluster together, with closer positions among private banks; (2) When the linkage index of card issuance growth rate is set above PR70, HNCB and CTBC Bank become the centers of the network graph; (3) In the network graph estimated and drawn by the growth rate of transaction amount per card, nodes with higher linkage indices do not cluster together, and CTBC Bank is the center of the network graph.