In this paper we examine to what extend it is possible to explain and forecast oilprice volatility. This is done by evaluating the comparative performance of different volatility models using weekly returns of West Texas Intermediate (WTI)crude oil prices. This approach builds on a dataset holding the historical development of WTI oil prices in the period 1986-2020. As an alternative approach we introduce the regression models Lasso, Ridge and Elastic Net. These models are based on a much smaller dataset only containing WTI oil prices in the period 2015-2020 adding Google Trend data which are expected to influence the oil price fluctations. This is done ad an attempt to strengthen the forecast.
We introduce ARIMA, GARCH, APARCH and EGARCH as theoretical methods for forecasting the oil price volatility. The selection and estimation of the best suited model is explained and applied. This model is then tested in-sample and out-of-sample before using it to forecast the expected volatility from the 19th of June and 6 months ahead. The model performs acceptable in-sample and out-of-sample, but is not very sufficient forecasting 6 months ahead. The conclusion suggests that it is not sufficient using only the historical development of the oil prices as model input.
For the second part of the analysis regression models are introduced. As an attempt to get a better forecast Google Trend data is used holding search words directly related oil and some for alternative energy sources. The findings are that the Lasso regression has the most precise in- and out-of-sample forecast and reaches a more realistic 6 months ahead forecast. Oil prices a proved to be very volatile and difficult to forecast.
|Educations||MSc in Business Administration and Mathematical Business Economics, (Graduate Programme) Final Thesis|
|Number of pages||114|
|Supervisors||Niels Buus Lassen|