Commodities vs. S&P 500: Causal Interaction, Temporal Analysis and Predictive Modelling Using Econometric Approach, Machine Learning, and Deep Learning

Abstract

Due to the presence of inherent complex behaviour with nonlinear dynamics, irregular temporal behaviour, high volatility, both in commodities and stock prices, this research aims to quantify, understand, model, and predict such irregular fluctuations of crude oil, gold and silver prices in comparison to S&P 500 index. This is done by employing powerful data modelling techniques followed by econometric approach using Granger causality, impulse response, forecast error variance decomposition and instantaneous phase synchrony prior to predictive modelling. S&P 500 index and commodities exhibiting non-stationary behaviour during the period 1 April 2000-27 July 2019 are considered for the entire analysis. The best model has been chosen comparing the error rate on the results of six powerful algorithms, KNN, DT, SVM, GBM, EF and ANN. The numerical analysis on the modelling and prediction of irregular fluctuations in commodities and stock indexes would practically support delineating the nexus between these two.