eThesis - School of Business electronic theses

In this thesis I analyze the competitive storage model and how it has been used to model the price behavior of renewable commodities. This is done by first looking at the stylized facts of commodity prices through a time-series analysis in order to confirm the theoretical assumptions of the modeling approach. I then explicitly show the steps needed to set up the theoretical model and look at some extensions to the baseline model. The framework of the competitive storage model I analyze is pioneered by the work of Gustafson (1951) and Muth (1961), but the version I focus on is a more recent version of the model that follows the work by Deaton and Laroque (1992, 1995, and 1996).

I critically asses the empirical evidence regarding the estimation of the model by looking at four key papers, by Deaton and Laroque (1992, 1995, and 1996) and Wright et al. (2011), on the subject and comment on the methods used to bind the model with the data, mainly the maximum likelihood method. The empirical evidence is assed to see how the model performs in estimating commodity prices.

The final objective of this thesis is to look at an extension of the commodity storage model where it is used to estimate the price behavior of oil. This part differs from the conventional view where oil is modeled as a non-renewable commodity. This part of the thesis focuses on the work by and Dvir and Rogoff (2011).

The study is mostly conducted as a literature review, but has also a time-series analysis to calculate some of the common characteristics of commodity prices. This time-series analysis is done on a World Bank time-series for fourteen commodities for the period 1900-1987 and goes over key characteristics such as autocorrelation, mean and coefficient of variation.

The main results of the study can be summarized as follows: the key characteristics of historical prices for the fourteen commodities shown in this thesis are inline with the existing stylized facts. The competitive storage model can be setup to produce price data that corresponds to the historical price data and it can be expanded to take into account different assumptions regarding things such as the cost of storage and the interest rate. Further the evidence reviewed in this thesis shows that oil can be modeled using similar methods as conventional renewable commodities by extending the competitive storage model.