A Study Of Flexible Margin System In Stock Index Futures Based On VaR Methods

After three years’ stimulation, the stock index future market is still at thestage of stimulation, due to the risk of such financial derivatives. If there is noeffective risk prevention system, we can’t bear the risk it brings. Marginsystem is one of the most important measures which can ensure that the risksbe controlled at an acceptable range. Now, there are two types of marginsystem which are the static margin system and the dynamic margin system.Because the former one can’t adjust margin level according to risk change,sometimes the margin level is high, but sometimes it’s superfluous. So theworld major exchanges have adapted the dynamic margin system, forexample, the SPAN system in the CME and the PRiME system in the HKE.But they are tailored for the exchange which has diverse financial product andmay be too complicated for china. So it’s necessary to develop a simple butreliable risk predicting system for china future market.So, the main work of this paper is to develop such a system for Chinausing VaR method. We analyze the statistical feature of the trading data. Theresults show they have significant volatility clustering effect and fat-tailedcharacter. Then we test several VaR model using the simulation trading dataand get the following conclusions. Firstly, the EWMA model is simple, but ifusing this model to deal with data which has fat-tailed character it mayunderestimate the risk. Secondly the GARCH-t model can resolve thefat-tailed character, but it’s not reliable. Thirdly the traditional Monte Carlosimulation method assumes that the price change follows the GeometricBrownian Motion, so it also cannot deal with data which has fat-tailedcharacter. Furthermore, we usually assume the mean and variance is aconstant in Geometric Brownian Motion, which can not reflect the volatilityclustering features of the return sequence. Finally we propose twoimprovements to the Monte Carlo model, on one hand we assume pricechanges follows jump-diffusion process in order to capture the fat-tail featherof the yields sequence. On the other hand, we use the EWMA model to update the parameters to embody the volatility clustering effect.We use HS300stock index futures stimulation transaction data, and theTaiwan stock index futures transaction data to make empirical analysis andfind that the improved Monte Carlo model is very suitable risk measurementmodels for a flexible margin system.