匯率制度與銀行危機關連性之實證分析

Abstract

自1980年代以來，許多國家都相繼發生銀行危機，不但重創了該國的金融部門以及經濟發展，並為該國帶來難以估計的社會成本。綜觀這些發生危機的國家，其中不乏有金融體系穩健且實行浮動匯率制度的工業化國家；也包含了因其國內金融環境還無法適應浮動匯率，不得不採行釘住美元之固定匯率制度的發展中國家。 本文的研究重點在於探討不同匯率制度的選擇對於銀行危機發生之間的關連性。故此，本文實證研究以1980年至2004年間的季資料做為研究期間，選取29個樣本國家，利用信號法以及固定效應之二元Panel Logit Model相結合，建立一個引入匯率制度變數的銀行危機預警模型，來探討匯率制度對於銀行危機發生機率的影響。考慮到匯率制度變數在銀行危機預警模型中具有內生化的問題，本文參考DomaDomaç and Martinez-Peria (2003)的作法，嘗試對匯率 制度解釋指標採用工具變數法，來解決解釋變數與應變數產生聯立性偏誤(simultaneity bias)的問題。 從本文綜合指標模型的實證結果發現，考慮匯率制度與銀行危機內生性，匯率制度工具變數大致上都顯著且一致呈現負相關，亦即較為穩定的匯率制度會降低銀行危機發生的機率。此外，實證結果也發現，一般預警模型會比基本預警模型的解釋能力稍弱，其原因可能為解釋變數間相關性高的干擾，模糊了原解釋指標對銀行危機發生機率的影響程度。

A lot of countries suffered from banking crises since 1980. In these countries, banking crises not only hurt financial sector and economic development, but also incur inestimable cost to society. Moreover, these countries which occured comprise not only industrialized countries adopting a float exchange rate but also developing countries adopting a fixed exchange rate. This paper investigates mainly the relationship between exchange rate regimes and banking crises. Using a seasonal data set including 29 sample countries for the period 1980-2004 and combining the signal approach with the binary panel logit model, we integrate the explanatory variable of exchange rate regimes to construct an early-warning model of banking crises to examine whether the choice of exchange rate regimes affects the likelihood of banking crises. To consider that the exchange rate regime could be endogenous, this paper refers to Domaç and Martinez-Peria (2003) and utilizes an instrumental variables regression to resovle the problem of simultaneity bias. According to the empirical findings, we find the explanatory variable of exchange rate regimes is mostly negative and significant. That is to say, the exchange rate stability may reduces the probability of banking crises. In addition, we also find that the basic model is comparatively superior to the general model, the reason is maybe that the higher correlation among explanatory variables blurs the impact of the explanatory variable on the likelihood of banking crises.