Measuring The Wti Crude Oil Spot Market Risk Based On Extreme Value Theory

Oil as one of the most important energy in a country’s economic development, it not only affects country’s security policy, but also affects the oil related enterprise’s development. However, the consumption of crude oil has exceeded Japan since2003,and become the world’s second and Asia first. With the increase of oil consumption, the ratio of the oil import dependency has more then50%in2009, which means more than half oil consumption depends on international market. Once the international crude oil price change will affects the fluctuation of oil price in our country, and bring certain risk to economy and financial field. Therefore, accurately measure the risk of international crude oil price will be more and more important.Nowadays, people do more and more focus on extreme risk, because the extreme event will lead inventors to bankruptcy and even cause economic collapse and social unrest. For such risk, the extreme value theory(EVT) has been very mature, it has the biggest advantage is that it doesn’t need to make any distribution assumption and it’s only care about tail information. But it has some deficiencies, such as it requires the sample obeys independent identical distribution and within high confidence level the measured risk is too conservative and so on. So in this paper we will try to use the improved tail index method to measure the extreme risk, because this method can relax restriction of the independent identical distribution. Although this two method can measure out extreme risk, the risk value are static, they can’t capture the characteristic of fluctuation time-varying. So we will adopt APARCH-EVT model to measure the dynamic extreme risk.This paper selects West Texas light crude oil(WTI) spot price returning as empirical objects, firstly, we found the time series has fat tail and its distribution is biased and the volatility is clustering and has "leverage effect" from simple statistical analysis; secondly, through extreme value theory, the tail index improved method and APARCH-EVT dynamic modeling methods calculated corresponding risk value, we used change point theory to select threshold value, because this method can avoid the impact of subjectivity, we also apply biased t distribution to capture the biased distribution; lastly, we used three back-testing ways to test effectiveness of the models.Conclusions show that within the low confidence level(95%,97.5%) the improved tail index method is more accuracy than traditional extreme value theory, it indicated that the improved tail index method weakened independent identical distribution and increased the accuracy of risk value, but within high confident level(99.9%) this two ways were too conservative, so will affect the enthusiasm of the investors and financial institutions; within the same confident level the APARCH-EVT dynamic model is the most accuracy, because it takes into account of the time-series fat tail, biased, volatility clustering and the "leverage effect". We hope supply some references to the investors or financial institutions through this paper.