| During 1995-2001,over a dozen emerging market economies (EMEs) experienced severe financial crises. Arguably, these recent crises were more frequent and more painful than in the past, so more and more government officials and economists begin to pay attention to crises problems. They begin to discuss that whether something has changed in the economic environment and what are the reasons for these crises somehow different from earlier financial crises. As a result, international organizations and also private sector institutions have begun to develop Early Warning System (EWS) models with the aims of anticipating whether and when a financial crisis happens. The IMF has taken a lead in putting significant effort into developing EWS models for for EMEs, and they have got a lot of valuable results. The theory of currency crisis is being developed much, the study on banking crisis is also quite abundant. Economists are trying to give more credible explanation of why and how the financial crises happen. Simultaneously, the domestic and foreign economists are seeking the effective quantitative analysis method for financial EWS models, such as KLR Crisis Signals Model, discrete-choice (probit/ Logit) models, artificial neural network (ANN) and Markov-switching models etc., and these models are property in quantitative analysis methods. However, there are not absolutely credible EWS model, so it needs to go further in the study on the reasons of financial crises and EWS models. This paper aims to construct an appropriate financial crisis EWS model for China by reviewing the domestic and foreign literatures. China is an emerging market country, but it is unlike other countries which have high financial liberalization. China has much different environment, such as capital account are not fully open, interest rates are not determined by market, stock market has not developed well and the statistical data of financial market has not been fully constructed etc.. Therefore, when constructing financial EWS in China, it is a difficult work to choose the indicators predicting crisis. Otherwise, many of domestic literatures focus on qualitative analysis, and the main method is Crisis Signals model. This paper tries to use discrete-choice model to construct a EWS model predicting financial crisis. The results can be helpful in studying the relations between the independent variables and dependent variable. Also we apply ARIMA model to forecast the indicators in short term, and improve the predicting ability out-of-sample. The EWS model in this paper, like almost all EWS models on EMEs in the literature, focuses primarily on currency crises. Our EWS model employs a commonly used exchange market pressure (EMP) variable for defining a currency crisis. All things in our country considered, we will use macroeconomic indicators (the growth of GDP, inflation, the growth of export, current account/GDP, money/reserves, public debt/GDP, FDI/GDP and U.S. interest rate) and financial sector indicators (banking reserves/banking assets, the growth of domestic credit, the claims on private sector/GDP, banks'foreign liabilities/GDP and lending rate/deposit rate) to predict the financial crisis, monitoring the functioning of our financial system from macro and micro aspects. This paper uses Logit model as the financial crisis EWS model. One feature of Logit model is the non-linearity of the effect of the independent variables on the left-side index, describing the non-linearity relations between the factors in the theory of currency crisis. To realize the out-of-sample empirical test, this paper apply ARIMA model to study the time series structure of the indicators, and on the basis of this study build the forecast model for every indicator. Then estimate Logit model using the forecasting and original data of the indicators, and get the result of good-of-fit effects. The steps of forecasting ARIMA model are: (1) seasonal adjustment for time series of the indicators, and expand the seasonal adjustment factors in 2004; (2) unit root test for the time series seasonally adjusted, determining the integrated order and choose the optimal ARIMA model; (3) estimating the coefficients of ARIMA model; (4) forecasting using the optimal ARIMA model; (5) to multiply seasonal adjustment factor expanded by indicators forecasted. This paper takes the growth of GDP and inflation (the change of CPI) as examples to test ARIMA model, found a well forecast effect. Then changing the sample to 1995~2003 and estimating Logit model, we can use the forecasting monthly data in 2004 of the indicators to forecast the index predicting financial crisis. We can draw several conclusions from the results of the empirical study in this paper:...
|
PDF Full Text Request