AbstractCommodities are a broad class of economic consumption goods that differ in several key aspects from other asset classes. Commodity markets exhibit complex behaviours, such as seasonal trends and mean-reversion, and are characterised by high volatility. Rigorous empirical models that forecast commodity market prices and behaviour are thus crucial financial tools to evaluate capital investment decisions under commodity price uncertainty.
This research presents a comprehensive framework for estimating, analysing and applying multi-factor commodity pricing models (CPMs) and uses these CPMs to value and manage uncertainty in capital investment projects through real options (RO) analysis. A procedural framework for developing and selecting appropriate CPMs is developed and applied to energy, metal, and agricultural commodities. The number of latent factors required to describe these commodity classes are discussed, including unique methods of optimizing the computational efficiency of model estimation and commodity data fit. The sequential processing Kalman filter algorithm is empirically shown to be at least 6.5 times faster than traditional Kalman filtering, substantially reducing CPM estimation time. A unique approach allowing for flexibility in the error of observations increases the interpretability and performance of these models.
Three new open-source packages for the R statistical programming language are also developed and presented in this research: FKF.SP, NFCP and LSMRealOptions. These rigorously tested packages provide efficient and complete frameworks for Kalman filtering, commodity market modelling and capital asset valuation through RO analysis, respectively. The open-source publication of these packages promotes the application of these techniques in future research.
Multi-factor crude oil and European Union carbon allowance CPMs presented in this research are applied to value capital investment projects using RO models, with the multi-factor modelling of carbon allowances unique to this dissertation. Multi-factor CPMs are shown to substantially reduce bias and uncertainty and increase the robustness of calculated project value compared to a one-factor geometric Brownian motion model. These multi-factor models are further shown not to increase the dimensionality of RO models, providing substantial support for their application in capital asset valuation and RO analysis.
In a major empirical case study, the financial viability of retrofitting a western European coal-fired power plant with carbon capture technology under carbon allowance price uncertainty is evaluated through a novel two-stage RO model. The model considers the value of suspending or abandoning a capture unit at fixed costs. The option to abandon is shown to substantially reduce the irreversibility of investment, greatly promoting carbon capture. Immediate investment into the technology is concluded to be the optimal strategy in the EU’s energy sector due to high prevailing carbon market prices resulting from direct market intervention in 2018. This is the first case study within the literature to promote the commercial adoption of this emission abatement technology.
Future streams of research proposed include extensions to the commodity modelling framework by incorporating stochastic trigonometric factors or relaxing the assumption of Gaussian dynamics in the spot price process. Extensions to the RO case studies and applications of the modelling techniques presented in this research are also discussed.
|Date of Award
|15 Jun 2022
|Geoffrey Harris (Supervisor), Adrian Gepp (Supervisor), Simone Kelly (Supervisor), Colette Southam (Supervisor) & Bruce Vanstone (Supervisor)