Bond return predictability in the frequency domain

Abstract

In this dissertation we analyse the out-of-sample forecast of the long-term government bond yield using frequency domain techniques. We follow Faria and Verona (2018, 2019, 2020a, b, 2021) and use a wavelet-based approach to break down a time series into different component frequencies: high-frequency, business-cycle frequency and low-frequency. In this way it is possible to reveal hidden information that time domain techniques cover in their prediction. Almadi et al. (2014) predictors were used to predict out-of-sample bond returns with frequency domain techniques in two different sample periods. The results show that the model has a significant shortcoming in terms of bond return forecasting accuracy, especially when compared to other successful models. The efficacy of the predictors decreases when applying frequency domain approaches like wavelets. Our research highlights the model's limitations to unfavourable market situations and provides insight into how the model performs during times of financial and macroeconomic instability.

Date of Award	16 Jul 2024
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Gonçalo Faria (Supervisor) & Fábio Verona (Co-Supervisor)