国际金融研究
國際金融研究
국제금융연구
Studies of International Finance
2013年
2期
4~15
,共null页
外汇风险 货币危机 早期预警系统 动态Logit模型 极值分位数
外彙風險 貨幣危機 早期預警繫統 動態Logit模型 極值分位數
외회풍험 화폐위궤 조기예경계통 동태Logit모형 겁치분위수
Foreign Exchange Risk; Currency Crisis; Early Warning System; Dynamic Logit Model; Extreme Quantile
本文采用极值分位数的估计方法和动态Logit预警模型来识别和预测中国的外汇风险。其中用极值分位数的估计方法替代传统的偏离均值若干倍标准差的方法来识别中国的外汇风险.并以此为基础,选取代表宏观经济、金融体系和国外冲击三方面的24个指标构建中国外汇风险动态预警模型。所采用的动态Logit预警模型不仅考虑了基本面对外汇风险的影响,而且加入了动态性因素,即考虑了外汇风险的自回归特性。从实证结果来看.极值分位数的估计方法较客观地识别了外汇风险,动态Logit预警模型无论在样本内还是在样本外的预测能力都比静态Logit模型有了较大的提高,我国的汇率改革没有对外汇风险产生显著影响。
本文採用極值分位數的估計方法和動態Logit預警模型來識彆和預測中國的外彙風險。其中用極值分位數的估計方法替代傳統的偏離均值若榦倍標準差的方法來識彆中國的外彙風險.併以此為基礎,選取代錶宏觀經濟、金融體繫和國外遲擊三方麵的24箇指標構建中國外彙風險動態預警模型。所採用的動態Logit預警模型不僅攷慮瞭基本麵對外彙風險的影響,而且加入瞭動態性因素,即攷慮瞭外彙風險的自迴歸特性。從實證結果來看.極值分位數的估計方法較客觀地識彆瞭外彙風險,動態Logit預警模型無論在樣本內還是在樣本外的預測能力都比靜態Logit模型有瞭較大的提高,我國的彙率改革沒有對外彙風險產生顯著影響。
본문채용겁치분위수적고계방법화동태Logit예경모형래식별화예측중국적외회풍험。기중용겁치분위수적고계방법체대전통적편리균치약간배표준차적방법래식별중국적외회풍험.병이차위기출,선취대표굉관경제、금융체계화국외충격삼방면적24개지표구건중국외회풍험동태예경모형。소채용적동태Logit예경모형불부고필료기본면대외회풍험적영향,이차가입료동태성인소,즉고필료외회풍험적자회귀특성。종실증결과래간.겁치분위수적고계방법교객관지식별료외회풍험,동태Logit예경모형무론재양본내환시재양본외적예측능력도비정태Logit모형유료교대적제고,아국적회솔개혁몰유대외회풍험산생현저영향。
The paper uses the method of estimating extreme quantile and the dynamic Logit early warning model to identify and predict the foreign exchange risk in china, which was identified by utilizing the method of estimating extreme quantile instead of the traditional method of several variance deviation from mean value. On the basis of the foreign exchange risk identified, the dynamic early warning model of foreign exchange risk in china was made by choosing 24 indicators which represent macro-economy, financial system, and foreign shocks. The dynamic Logit model not only took into account the effect of fundamental factors, but was added with dynamic factor, namely autoregressive character of foreign exchange risk. The empirical result showed that the method of estimating extreme quantile identified the foreign exchange risk objectively, and the forecasting ability of dynamic Logit model in sample improved greatly than that of static Logit model, as well as outside of sample.