Online Program

Effective Sample Size--Calibrated Multiple Comparison Methods for Long Memory US Stock Volatilities (308542)

Keywords: Autoregressive fractionally integrated moving-average, Degrees of freedom, Equivalent sample size, Time series

Volatilities in stock prices often show long-range dependence, representing significant autocorrelations even in large time lags. Multiple comparison methods can be used to identify different mean volatilities. However, the classical multiple comparison methods, including Fisher’s least significant differences test, Tukey’s honestly significant differences test, and Student-Newman-Keuls test, produce erroneously sensitive comparison results for long memory time series because these methods are developed for independent data. To accurately achieve the target significance level for long memory data, we propose using effective sample size (ESS) methods to calibrate these three popular multiple comparison tests. After using change point analysis to detect a sudden rise in mean stock volatilities of thirty prominent companies in January 2018, we analyze means before and after the changepoint using our ESS calibrated multiple comparison tests. With recent empirical evidence showing that low-volatility companies outperform high-volatility companies, our methods help accurately identify which companies are low or high volatility.