Crude oil returns predictability in the frequency domain

Abstract

In this dissertation we follow Faria and Verona (2018, 2020a,b, 2021) investigation and apply a wavelet-based method (MODWT MRA) that allows to decomposing a time series into different frequency components. Therefore, one can unveil hidden information in time domain useful to forecasting. Taking into account the Conlon et al. (2021) paper on the time domain, we expand out-of-sample crude oil returns predictability literature, and find strong statistical and economic gains from using frequency domain information. Results are also robust to different settings. The best result achieved is a 𝑅𝑂𝑆 2 of 2,35% in the high-frequency component of predictor Chicago Board Options Exchange volatility index (VIX), outperforming out-of-sample R-Squareds obtained in literature on time domain based on end-of-month crude oil returns. In fact, for investors and policymakers interested in oil market developments, the short-term dynamics of Treasury bill rate (TBL), Change in Treasury bill rate (CTBL), and VIX are promising predictors to look at. We also conclude that evidence of predictability is stronger during NBER-dated recessions.

Date of Award	12 Jul 2023
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Gonçalo Faria (Supervisor) & Fabio Verona (Co-Supervisor)