Title: A robust evaluation of macro-financial predictive content for realized volatility
Authors: Robinson Kruse-Becher - FernUniversität in Hagen (Germany) [presenting]
Abstract: Predictive regressions for realized volatility are studied. Dating back to 1989, there is an ongoing debate whether financial and macroeconomic series have predictive power for financial volatility. We make use of the econometric recently provided framework to study the role of several potential predictors for realized volatility in a robust way. In contrast to standard approaches, the applied methodology accounts for long memory, multiple structural breaks (in levels and persistence) and spurious regressions. As most predictors are highly persistent, and realized volatility has typical long memory features, it is important to account for these phenomena when testing for predictive power. Standard predictive regressions method fail in this context and might even provide spurious evidence in favor of predictability. We employ an updated monthly data set used previously which covers a long time span and several important predictors. Among these are credit spreads, term spreads, price-dividend and price-earnings ratios from 1885 to 2016. Our findings indicate important differences in the outcomes when properly accounting for changes in persistence and long memory. We further study the role of time-variation in the predictive content by recursive estimation and test for spurious long memory in a number of robustness checks.