Wavelet Decomposition in Financial Forecasting: Analyzing Term Spread Predictability on Equity Risk Premiums

Abstract

This thesis investigates the predictive power of the term spread of interest rates on the equity risk premium across eight countries: the United States, United Kingdom, Germany, France, Norway, Belgium, Netherlands, and Ireland. By applying the Maximum Overlap Discrete Wavelet Transform we decompose the term spread into three frequency components capturing oscillations at multiple frequencies. Our findings suggest that the low-frequency component is a strong and robust out-of-sample predictor for both the U.S. and the U.K., particularly for longer forecasting horizons. In addition, the business-cycle frequency component is identified as a reliable predictor for Norway when forecasting between three months and one year ahead. For Germany, France, Belgium, Netherlands, and Ireland, findings indicate that none of the frequency components serve as reliable predictors.