Modeling of Crude Oil Prices with a Special Emphasis on Macroeconomic Factors

Abstrakt

This thesis aims to find out what drives price of crude oil. Since nowadays oil is rather a financial asset than industrial good its price evolves similarly to other financial assets prices. Traditional financial approach to modeling is based on assumptions about stochastic processes which allows for capturing partly random behavior of oil price changes. Besides estimating simple stochastic models (Geometric Brownian Motion (GBM) and Mean-Reversion) which are widely used for other commodities and returns modeling, influence of unique for oil factors is also assumed and tested. A unique feature of oil market is presence of precautionary demand reflecting expectations and concerns about future need for oil. Based on the common knowledge about global significance of oil, high market power of the producers, possible political instability and also observing soundness of macroeconomic background market participants form their expectations concerning future oil necessity. If the macroeconomic background is believed to be sound – optimistic mood in the market results into high precautionary demand. At this state market is very vulnerable to any announcements, and price response is immediate and sharp. In order to capture expectations impact on oil prices stochastic modeling is extended with factors describing macroeconomic conditions through the oil price volatility channel modeled within GARCH framework. The volatility models have better forecasting accuracy on the short horizon but produce only approximated long term expectation similarly to the simple ones. Simple stochastic models for oil prices demonstrate that drift estimations are very uncertain but are more reliable for the GBM model. The main finding here is that crude oil price process has a drift, but it changes once in a while – it may be assumed constant but for shorter than sample time periods. This conclusion may also hold for the mean-reverting property even though the latter is not supported by the findings. But the main attention should be paid to diffusion term, estimation of which is on average the same for all models and methods. It has strong serial dependence not consistent with theoretical properties of stochastic processes. It is reasonable to assume that serial dependence of stochastic term can be captured by heteroscedasticity modeling. GARCH models estimated reveal that oil price volatility positively responses to bond yield spread widening and depreciation of US dollar. Asymmetric property of the volatility is also documented meaning that oil prices become more volatile in occurrence of positive rather than negative shocks (the sign of the link is positive which is opposite to expected for asset returns. Typically asset prices are falling sharply in negative shocks presence, but in case of oil it may indirectly prove the precautionary demand hypothesis through the stronger impact of optimistic atmosphere rather than pessimistic). Convenience yield has a positive direct and very strong impact on oil price which is a fundamental theoretical assumption for commodity modeling consistent with theory of storage: oil price responses positively on inventories scarcity. US dollar to British pound exchange rate influences oil price also directly and this link is more stable than that in variance. Even though the obvious indicators of macroeconomic activity did not demonstrate strong influence on oil price but they apparently are not the best measure of expectations influencing the precautionary demand.