Riding Market Waves: Spectral Analysis and Signal Processing for market timing

In this research, I investigate the use of the Fourier transform for market timing.This tool is borrowed from Spectral Analysis and it allows to decompose a time-seriesas a linear combination of waves with different speeds. I apply it to market data todescribe market’s trends in terms of speed. The objective is to investigate whethermarket regimes indicated by the Fourier transform allow to improve the markettiming ability of time-series momentum strategies (TSMOM). Through the Fouriertransform, I extract meaningful regimes which denote the speed of the market. Ishow that TSMOM strategies with different speeds are related to these regimes:intuitively slow TSMOM rules benefit from regimes characterized by slow marketmovements, whereas fast rules benefit from fast market fluctuations. However, the sofound regimes do not carry predictive ability for the future performance of TSMOMstrategies and hence cannot be used for improving their timing ability. Volatilityon the other hand proves suitable to predict the performance of TSMOMrule withdifferent speeds. This result is translated into a dynamic trend rule which achievessuperior performance. In a second strand of research, I investigate whether theFourier transform is a suitable tool for predicting market direction: I assess itslimitations and conclude that it is not the case.