A New Test for ARCH Effects and its Finite-Sample Performance
提出一种基于加权平方样本自相关的ARCH效应检验，通过给低阶滞后更大权重提升有限样本表现，无需设定备择假设且可用数据驱动方法选择滞后阶数。
We propose a test for autoregressive conditional heteroscedasticity based on a weighted sum of the squared sample autocorrelations of squared residuals from a regression, typically with greater weight given to lower-order lags. The tests of Engle, Box and Pierce, and Ljung and Box are equivalent to the test with equal weighting. Our test does not require formulation of an alternative and permits choice of the lag number via data-driven methods. Simulation studies show that the new test performs reasonably well in finite samples especially with greater weight on lower-order lags. We apply the test in two empirical examples.