Time Series Analysis: Using the Volatility Index as a Timing Indicator for the Commodity Futures Market

Abstract

Over the past decade, trading futures contracts has gained an increasing amount of popularity among investors. As trading activity of commodity futures proliferates, the knowledge that investors have regarding price movements relative to market volatility becomes essential. In times of economic distress, many macroeconomic factors including global supply and demand, political instability, and investors’ sentiment affect the price of commodities. However, one market concept that ties all of these macroeconomic factors together –- one that is both quantifiable and understandable to investors –- is volatility. By better understanding the volatility index and its impact on prices movements in different commodities, investors can make well-informed decisions during times of stability and instability.

Through utilization of time series regression, this thesis creates forecasting models to predict contract prices of WTI Crude Oil, Gold 100 Oz., and Soybean futures. After analyzing a 5-year stable market environment from 2014 to 2019, the forecasting model for each commodity is tested for its predictive strength during times of low volatility. A cross-validation method called forward chaining measures the robustness of these models by 20-fold iterations of 60-day testing intervals. Concluding Chapter 4, average RMSE values are calculated for each commodity, suggesting the predictive strength of each time series model.

Upon extracting meaningful statistics from this data set, this thesis then analyzes an unstable market environment as represented by the COVID-19 pandemic. By adding an additional testing iteration for the COVID-19 data, the forward chaining simulation measures how well the forecasting models formulated in Chapter 4 predict prices of crude oil, gold, and soybean during times of elevated volatility.

This thesis concludes by highlighting unique characteristics and trends in the individual commodities by analyzing the characteristics and forecasting strength of the models across the two dissimilar market environments. Unraveling future questions that address how investors can better predict commodity prices when facing society's next financial crisis, this thesis sets the stage for future academic research.