Stock Return Predictability in Frequency Domain

Using the long-term wavelet component of monthly S&P500 excess returns as supervision information, we employ a partial least squares method to extract the common predictive information of 14 prevalent macroeconomic variables, and construct a new macroeconomic index aligned for predicting stock market returns. The macroeconomic index significantly forecasts aggregate stock returns in- and out-of-sample, and shows considerable economic value in asset allocation. Furthermore, the new predictor is negatively correlated with the future financial uncertainty depicting persistent market risk