An evaluation of scale-specific predictive systems

Predictions of excess market returns with market proxies seem to have hump-shaped behavior, as found by Bandi, Perron, Tamoni, and Tebaldi (2019). Firstly, Bandi et al. (2019) found that traditional predictive systems are tightly restricted and a novel predictive framework is developed. Time series are decomposed into scale specific components, each allowing for processes operating at specific frequencies. With the components, new predictive regressions are applied to decimated components (filtered observations). Their work is replicated and extended. Bandi et al. (2019) use observations exclusively from the U.S. In this paper, the framework is applied to different markets, namely the Asian-Pacific and European markets. Lastly, the framework is expanded, allowing for more freedom in the explanatory and predicted variables. This is found to have a positive impact on predictive power, up to an R2 of 99.6%. Although, it is also found that the scale-specific predictive systems are often not able to generate reliable results.