VaR 추정기법들의 성과비교 및 평가: 한국금융시장을 대상으로

Abstract

Evaluation and comparison on VaR Estimation Methods: Focusing on Korean Financial Market

Don Hwa Lee

Department of Economics, Graduate School, Pukyong National University

Abstract

The crisis started in 2007 to manage the global financial crisis increased the importance. As a result, many financial institutions became concerned about the Value-at-Risk. Value-at-Risk is conceptual simple and ease of use. Therefore, International financial institutions are commonly using method of risk management. Value-at-Risk summarizes maximum possible amount of loss for a specified period in normal market conditions, Estimation methods are parametric method under normal distribution and non-parametric method under not normal distribution. The main purpose of this study is to explain basic concept of Value-at-Risk and discusses various estimation methods(Delta-Normal Method, Historical Simulation Method, MonteCarlo Simulation Method, G and H Method). In addition to parametric and non-parametric methods are compared. Also, An alternative to existing methods of Value-at-Risk, I proposes a new method that is based on Tukey's G-and-H distribution. Delta-Normal Method is assumed Normal distribution in probability distribution of returns, Parametic methods are called. Delta-Normal Method, when estimating VaR, volatility is an important factor. Methods for measuring volatility, historical standard deviation method, EWMA method, GARCH(1,1) was used. Non-parametric methods were used Historical simulation and Monte Carlo simulation methods. G-and-H Method, distributional assumption is especially useful if asymmetries and fatter tails have to be considered and fast simulation from the model is of importance. To compare these methods, Value-at-Risk is calculated in confidence level to 95%, 99% using data of close-to-close return of KOSPI 200 and 3-year Government bonds. And sample period is January 2, 2007 to December 31, 2009. And method of back-testing was conducted Failure Ratio(FR) and Z-score test.

Keywords : Value-at-Risk, Delta-Normal Value-at-Risk, Montecarlo simulation, Historical simulation, G and H distribution, skewness-kurtosis, risk measurement, distribution fitting.