The Computation and Evaluation of VaR methods of Commercial Banks in Taiwan
| Autor: | Li-Fang Kao, 高儷芳 |
|---|---|
| Rok vydání: | 2006 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 94 Since the Ministry of Finance in Taiwan has enforced the law「New Basel Capital Accord」at the end of 2006, the markets of all major commercial banks face great challenges. Increased competition and growing pressures for revenue generation have led financial institutions to search for more effective ways to control losses. The importance on risk management to build up VaR models has been emphasized as one of the key issues. This study utilizes return of stock prices of the major commercial banks to measure Value-at-Risk (VaR). Considering various of operation types and conditions of banks, we select five banks from five different types of banks based on Shieh (2001). In order to enhance the practical applications of this research, we choose the topics of the influence of card-debt problems. We evaluated the accuracy and prediction of VaRs performances estimated by the different models before and during the problems. Furthermore, seeking to find out more stable VaR methods in two different periods influenced by this significant market events. In general, the simulation algorithms are mainly utilized in the historical simulation method, monte carlo simulation method and the variance-covariance method. Since the characteristics of most of financial asset returns hold leptokurtosis and heavy-tailed, we apply mixture of normal distributions method instead of the traditional single component normal distribution assumption. Advance model of the GARCH(1,1) is used to capture the variations of volatility. Finally, we use Forward testing、Binomial test、Kupiec test、Christoffersen test and RMSE to examine the accuracy of the VaRs calculated by the models mentioned above. We find that there is no universal best method to compute VaR value for different types of banks during two different periods or in different conditions (including window sizes and significant levels). To control the risk and make decisions, we suggest that banks could choose proper Value method according its operation environment, allocation of capital assets and strategy of risk management. In the period of card-debt problems, the exception rate of Chinatrust bank and Cosmos bank is higher under 1% significant level. And it exhibits the capability of capturing the tails is worse and unstable because of the increase of the volatility in this period. Besides, in this period, the RMSE of Chinatrust bank is obviously higher, it reveals that in case of causing loss, the situation of loss is more serious. When significant level is higher, mixture of normal distributions method underestimate VaRs more seriously. We also find out when mixing proportion of the left (low) return distribution is small, it would overestimate VaRs and cause the bias of estimation. The GARCH model can successfully capture the trend of volatility but hardly capture the VaRs in the shape change of the volatility. As to issues of assigning window sizes and significance level, we find the two situations: small window size with low significant level and big window size with high significant level, would make the performance worse, especially when volatility is heavily fluctuated. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|