Commodity price co-movement: Comparing models and correlation measures
Details
- Authors: Luca Kozian
- Title: Commodity price co-movement: Comparing models and correlation measures
- Supervisor: Dr. Jörg Osterrieder, Dr. Marcos Machado
- Degree: Master of science
- University: University of Twente
- Year: 2023
- Status: Master thesis
Summary
Commodity price co-movement: Comparing models and correlation measures
Abstract
This thesis investigates how well different types of methodologies and co-movement measures can explain cross-commodity price co-movement using macroeconomic variables. Thereby, VAR, VARX, multiple regressions and Random Forest regressions are applied as models, and Pearson correlations and Gerber statistics as co-movement measures. They are implemented on a dataset of price co-movement of 20 major commodities and various macroeconomic factors from mid-2003 to early 2023. The findings show that VAR and VARX models outperformed Random Forests and multiple regressions, reaching R^2 values as high as 89%. Random Forest regression, however, only showed slightly better performance metrics than multiple regressions. Moreover, did the use of Gerber statistics over Pearson correlations enhance model performance for VAR and VARX models, but for Random Forests and multiple regressions this is less clear. By examining and comparing different methodologies, this thesis contributes to existing literature on commodity price co-movement and lays a groundwork for further assessments of the performance of various methods in modelling this phenomenon.