关键词: 波动率, 结构变化, 厚尾数据, 非参数估计

Abstract: This paper constructs two robust test statistics to examine structural changes in volatility. The way of constructing the new tests are based on the absolute value instead of the squared volatility. By doing so, we can relieve moment conditions and make the new tests more robust to various kinds of non-normal heavy-tailed data. We also employ the nonparametric residuals to construct the long-run variance, which can help improve the testing powers under alternatives. Under certain assumptions we prove that the two statistics have asymptotically standard null distributions, and exhibit strong powers to find single or multiple breakpoints and smooth changes in volatility. Monte Carlo simulations show the better performance of the new tests relative to other popular tests when the data presents non-normal distribution. Finally, new empirical applications to detect structural changes in volatilities of U.S. dollar/Russian Ruble exchange rate, S&P 500 index and CSI 1000 index highlights the usefulness of our tests in real datasets.

Key words: Volatility, Structural Changes, Heavy-tailed Data, Nonparametric Estimation