Improving procurement through regression analysis: A case study of predicting argentine jet fuel prices
| Date | 01 March 2009 |
| Pages | 1-16 |
| Published date | 01 March 2009 |
| DOI | https://doi.org/10.1108/JOPP-09-01-2009-B001 |
| Author | Juan A. Salaverry,Edward D. White III |
| Subject Matter | Public policy & environmental management,Politics,Public adminstration & management,Government,Economics,Public Finance/economics,Texation/public revenue |
JOURNAL OF PUBLIC PROCUREMENT, VOLUME 9, ISSUE 1, 1-16 2009
IMPROVING PROCUREMENT THROUGH REGRESSION ANALYSIS:
A CASE STUDY OF PREDICTING ARGENTINE JET FUEL PRICES
Juan A. Salaverry and Edward D. White III*
ABSTRACT. Of all oil products consumed by the Argentine Air Force (AAF), jet
fuel is the resource with highest demand and at the end of the day the most
expensive support item procured by the AAF. Accurate predictions of
Argentine jet fuel prices are necessary to improve AAF financial and logistics
planning. Multiple regression analysis is one such tool that can aid in
accurately forecasting the amount required when procuring this valuable
commodity. Using this methodology, we develop and illustrate a highly
predictive model that has an adjusted R2 of 0.98 and an average percentage
absolute error of 4%.
INTRODUCTION
Oil distillates are considered important elements to accomplish
the missions of the Argentine Air Force (AAF). Of all oil products
consumed by the AAF, jet fuel is the resource with highest demand
and at the end of the day the most expensive support item procured
by the Argentine Air Force. The AAF consumes more than 12 million
gallons each year and spends almost 35% of its total material budget
in the acquisition of this resource (Argentine Air Force Command of
Material, 2006).
Crude oil is the main element in the production of jet fuel. During
recent years, crude oil price instability has brought additional
--------------------------------
* Juan A. Salaverry, MA, is Lt Colonel, Argentine Air Force and currently
works as Planning Division Chief within the Planning Department of the Air
Material Command located in the Condor Building in Buenos Aires,
Argentina. Edward D. White III, Ph.D., is an Associate Professor, Department
of Mathematics and Statistics, Air Force Institute of Technology. His
teaching and research interests are in design of experiments, linear and
nonlinear regression, and statistical consulting.
Copyright © 2009 by PrAcademics Press
2 SALAVERRY & WHITE III
problems to budget and logistics planning. Inaccurate fuel price
forecasts can cause major problems in the AAF budget. Predicting
too high of a price results in the AAF receiving more funds than
required for this commodity. This in turn results in fewer funds
available to meet other priorities. In contrast, low jet fuel predictions
mean that the received funds are not sufficient to pay for the cost of
fuel, prompting the AAF to either request a supplemental
appropriation or transfer funds from another account which produces
other significant negative effects over the organization.
Accurate oil predictions are therefore an important strategy for
the AAF to protect taxpayer contributions. Individual efforts have
been attempted in the AAF to solve this issue such as the use of
simple regression models, but the results have never been universally
accepted in the organization. Not only is there a lack of
understanding of the variables that affect the problem, but also there
are difficulties in finding the appropriate tools to address this issue.
This article attempts to rectify this by demonstrating the utility of
using multiple regression techniques to improve the prediction of jet
fuel prices for the AAF and thereby minimizing errors when procuring
this valuable commodity.
BACKGROUND
Accurate oil price predictions have not been easy to achieve since
the oil embargo occurred in 1973. Especially during the last decade,
jet fuel, a light oil distillate obtained by a chemical process called
hydrocraking, has displayed extremely volatile prices, led by the
erratic behavior of crude oil prices, the main component in jet fuel
production. Figure 1 illustrates the erratic behavior of crude oil prices
(WTI) and jet fuel prices (JetKero 54) from 1994 to 2006.
Several methods have been used to predict jet fuel prices with
varied results over the years. Artificial networks (Kasprzak, 1995),
multiple regression models (United States Department of Energy,
2002) and econometric forecasting (Coloma, 1998; Mercuri, 2001)
have proved to be effective to forecast oil distillates prices like gas,
fuel oil and jet fuel prices. Unfortunately, these models have been
developed to forecast the variable of interest in the particular
environment of the market, primarily the United States, but not in the
Argentina market. Because of this, the direct application of these
To continue reading
Request your trial