High Diesel Fuel Prices

Author : Joseph W. Bajjalieh
Publisher :
ISBN 13 :
Total Pages : pages
Book Rating : 4.:/5 (776 download)

DOWNLOAD NOW!

Download or read book Forecasting Diesel Fuel Prices written by Joseph W. Bajjalieh and published by . This book was released on 2011 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: Midwest agriculture depends heavily on corn, soybean, and wheat production which requires considerable diesel fuel to meet producer output objectives. An ability to anticipate fuel price movements may allow producers, farm managers, and fuel distributors to better plan their transactions to reduce costs or to hedge against price change. Despite potential gains from understanding diesel fuel price movements, little research has been performed to generate and assess diesel price forecasts. This research focuses on developing and evaluating the effectiveness of futures-based, structural-based, and time-series models to forecast diesel prices. Several composite forecast techniques, such as averaging, least squared regressions, and Harvey, Leybourne, and Newbold encompassing procedures also are evaluated to determine if they can improve forecasting performance. Models are specified and evaluated during an in-sample period, March 1994-February 2002, for one-, two-, and three-month forecast horizons. Using the in-sample specifications that are re-estimated each month for an out-of-sample period, March 2002-December 2008, diesel fuel price forecasts are developed and assessed. For the out-of-sample period, models are re-estimated recursively, and by using 4-year, 8-year rolling-window regressions to allow for the effects of structural change on forecast performance. During the in-sample period, no individual model consistently outperforms its rivals. However, composite forecasting methods improve forecast precision, but they are not statistically superior to the individual forecasts. Further, composite methods are generally not able to anticipate price changes as well as the individual models, which are led by the ARIMA specification. During the out-of-sample period, model performance decreases relative to the earlier period as new information and events emerge. In particular, mean squared errors gradually increase throughout the out-of-sample period. Among the models, forecast errors are highly correlated and anticipated information gains from combining forecasts do not arise. As a reflection of changing market conditions, 4-year rolling-window estimates clearly dominate forecast performance for all models and forecast horizons. Despite coefficients that are rarely significant, relative inventory-based models consistently are most accurate, suggesting the importance of allowing for inventories particularly in periods of low stocks and high prices. Interestingly, most models are able to anticipate price changes relatively well, with crude oil futures-based and time series models correctly identifying the direction of price change 60-80 percent of the time. Clearly, forecasting in periods of structural change is challenging. The research here suggests forecasts based on a relatively short rolling-window framework provide the most precise diesel fuel price estimates which are able to identify the direction of price change with reasonable accuracy. Further research may find it useful to consider allowing for more flexible model specification during periods of structural change. However, particularly in a rolling-window context, this approach will require careful monitoring of forecast performance to reduce the likelihood that large anomalies don0́9t unwarrantedly drive specifications leading to reduced accuracy. Another avenue for research might focus on longer-term forecasting, but based on the findings here this challenge seems somewhat formidable. In contrast, it might be useful to further investigate the ability to forecast the direction of price change, and to assess its economic value to decision makers.