Vance Ginn
Chapter 1 of Dissertation
Texas Tech University
Sam Houston State University
Fall 2011
Vance.Ginn@SHSU.EDU
Introduction
 Gasoline prices impact consumers similar to a tax.
 Edelstein & Kilian (2009): SUVs and complements
 Diesel prices affect decisions made by many firms.
 Brown & Theis (2009): increase costs
 Monetary policymakers may react to fuel prices.
 Pindyck (1999): reversion to mean
 Bernanke (2010, 2011): transitory, but closely watching
 Friedman (1968): information about economic events
 My goal is to provide good models to forecast fuel prices
during different periods of volatility in an era of rising
petroleum prices.
2
Literature Review
 Anderson, Kellogg, and Sallee (2011) use the Michigan Survey
of Consumers: what do consumers believe the price of
gasoline will be in the future?
 Forecast by consumers is similar to a random walk
 Ginn and Gilbert (2009): prices of crude oil futures and gas
 There is a 2% increase in the average weekly price of gasoline for
every 10% increase in average weekly oil price futures.
 There were periods that the model did not perform very well.

This lack of efficiency indicates that there may be better measures to
predict gas prices.
 Crude oil is the main component (42 gallons in barrel)
 Chouinard and Perloff (2002): gas prices (19 gallons)
 Brown and Thies (2009): diesel prices (10 gallons)
3
Gasoline and Diesel Price Components
Retail Gasoline Price Averages Each Year
$1.34
$1.85
$2.57
$3.25
$2.99
Retail Diesel Price Averages Each Year
$1.37
$1.81
$2.71
$3.81
$3.25
Crude Oil
43%
45%
48%
56%
69%
32%
49%
53%
58%
68%
23%
35%
18%
27%
20%
63%
Federal &
State Taxes
Distribution
& Marketing
13%
16%
12%
12%
13%
2002
17%
2004
9%
17%
2006
13%
14%
7%
9%
12%
10%
12%
12%
11%
6%
8%
9%
2008
2010
2002
15%
2004
20%
17%
Refining
Costs &
Profits
9%
2006
2008
2010
4
Literature Review
 Futures prices:
 Fama and French (1987): efficient market hypothesis (EMH).
 Chinn and Coibion (2010): gasoline price futures and heating oil
futures are good predictors of their future spot prices.
 Oil price futures appear to not be as efficient:
 Alquist and Kilian (2010): not a reliable predictor in 2000s.
 Buyuksahin and Harris (2011): speculators distort price futures?

Not likely, due to investors following oil market fundamentals.
 Wu and McCallum (2005) show that light trading exists in longer
term contracts than short-term ones, reducing the ability for
prices to be valued correctly.
 So the question remains, what variable(s) will provide good
forecasts for future gas and diesel prices?
5
Data
 I use monthly data from the EIA for the sample period January
1983 to March 2010 for the following variables:
 Motor gasoline regular grade retail price (GP) (including all





taxes)
On-highway diesel fuel price (DP) (including all taxes)
New York Harbor No. 2 heating oil future contract 1 (DPFut)
Imported crude oil price (OP)
Crude oil price futures (OilFut) that are traded on the New York
Mercantile Exchange (NYMEX)
New York Harbor regular gasoline future contract 1 (GPFut),
which is available from January 1985 to March 2010.

Split into two types:
 Reformulated regular gasoline: January 1985 to December 2006
 Reformulated gasoline blendstock for oxygenate blending (RBOB)
includes a percentage of ethanol that was added to gasoline in 2005:
January 2007 to March 2010.
6
An Era of Rising Petro Prices
Gas oline Prices
Gas Price Futures
$5
$5
$5
$5
$4
$4
$4
$4
$3
$3
$3
$3
$2
$2
$2
$2
$1
$1
$1
$1
$0
$0
$0
2000
2002
2004
2006
2008
2010
$0
2000
2002
Dies el Prices
$5
$4
$4
$3
$3
$2
$2
$1
$1
2002
2004
2006
2006
2008
2010
Dies el Price Futures
$5
2000
2004
2008
$5
$5
$4
$4
$3
$3
$2
$2
$1
$1
$0
2010
$0
2000
Imported Crude Oil Prices
2002
2004
2006
2008
2010
Crude Oil Price Futures
$150
$150
$150
$150
$125
$125
$125
$125
$100
$100
$100
$100
$75
$75
$75
$75
$50
$50
$50
$50
$25
$25
$25
$25
$0
$0
2000
2002
2004
2006
2008
2010
$0
$0
2000
2002
2004
2006
2008
2010
Table 1: Statistics for Variables for Estimation Period from 1983:1-2002:12
Summary Stats
Correlation Stats
Mean
Std. Dev.
GP
$1.13
0.17
1
$0.61
0.15
0.86
1
$1.14
0.16
0.94
0.82
1
$0.60
0.15
0.75
0.88
0.83
1
Imported Crude Oil
Prices (OP)
$19.99
5.49
0.76
0.91
0.78
0.94
1
Crude Oil Price Futures
(OilFut)
$21.83
5.51
0.80
0.94
0.83
0.96
0.98
Retail Prices of
Gasoline (GP)
Gas Price Futures
(GPFut)
Retail Prices of Diesel
(DP)
Diesel Price Futures
(DPFut)
GPFut
DP
DPFut
OP
OilFut
1
Table 2: Statistics for Variables for Forecast Period from 2003:1-2010:3
Summary Stats
Mean
Std. Dev.
GP
GPFut
Correlation Stats
DP
DPFut
OP
Retail Prices of
Gasoline (GP)
$2.39
0.63
1
Gas Price Futures
(GPFut)
$1.70
0.60
0.98
1
Retail Prices of Diesel
(DP)
$2.52
0.76
0.96
0.94
1
Diesel Price Futures
(DPFut)
$1.73
0.69
0.96
0.96
0.99
1
Imported Crude Oil
Prices (OP)
$56.42
23.48
0.96
0.97
0.96
0.98
1
Crude Oil Price Futures
(OilFut)
$61.88
24.27
0.96
0.97
0.97
0.99
0.995
OilFut
1
Table 3: Augmented Dickey-Fuller Unit Root Tests
Null Hypotheses: GP, GPFut, DP, DPFut, OP, OilFut has a unit root
Exogenous: Constant, Lag Length: Automatic selection based on AIC, max lag is 14)
Log Levels:
Log(GP)
Log(GPFut)
Log(DP)
Log(DPFut)
Log(OP)
Log(OilFut)
PP Test Statistics:
-2.26
-2.95
-2.13
-2.91
-2.90
-2.51
DLOG:
ΔGP
ΔGPFut
ΔDP
ΔDPFut
ΔOP
ΔOilFut
PP Test Statistics:
-10.26
-13.05
-10.52
-11.73
-8.38
-11.04
Test Critical Values:
1% level
-3.46
-3.46
-3.46
-3.46
-3.46
-3.46
5% level
-2.87
-2.87
-2.87
-2.87
-2.87
-2.87
10% level
-2.57
-2.57
-2.57
-2.57
-2.57
-2.57
Table 4: Phillips-Perron Unit Root Tests
Null Hypotheses: GP, GPFut, DP, DPFut, OP, OilFut has a unit root
Exogenous: Constant, Bandwidth: (Newey-West automation) using Bartlett kernel
Log Levels:
Log(GP)
Log(GPFut)
Log(DP)
Log(DPFut)
Log(OP)
Log(OilFut)
ADF Test Statistics:
-2.26
-3.03
-2.33
-3.15
-3.09
-3.38
DLOG:
ΔGP
ΔGPFut
ΔDP
ΔDPFut
ΔOP
ΔOilFut
ADF Test Statistics:
-4.52
-12.02
-10.33
-11.80
-8.07
-10.24
Test Critical Values:
1% level
-3.46
-3.46
-3.46
-3.46
-3.46
-3.46
5% level
-2.87
-2.87
-2.87
-2.87
-2.87
-2.87
10% level
-2.57
-2.57
-2.57
-2.57
-2.57
-2.57
Note: The sample period is 1983:1-2002:12, except for gasoline price futures (GPFut) which is from 1985:1-2002:12.
Table 5: Engle-Granger Cointegration Tests
Dependent: GP
Tau-stat
Prob.*
z-stat
Prob.*
Gas Price Futures (GPFut)
-3.44
0.0409
-23.61
0.0228
Imported Crude Oil Prices (OP)
-2.43
0.3157
-11.57
0.2714
Crude Oil Price Futures (OilFut)
-3.03
0.1079
-18.12
0.0756
Tau-stat
Prob.*
z-stat
Prob.*
Diesel Price Futures (DPFut)
-2.52
0.2739
-11.65
0.2678
Imported Crude Oil Prices (OP)
-2.11
0.4706
-8.20
0.4765
Crude Oil Price Futures (OilFut)
-2.52
0.2750
-11.88
0.2570
Dependent: DP
*MacKinnon (1996) p-values.
Notes: The data are in logs and the sample period is 1983:1-2002:12, except for gasoline price futures (GPFut) which
is from 1985:1-2002:12. The null hypothesis is that the series are not cointegrated. The automatic lag specification is
10
based on the Schwarz Bayesian Criterion (SBC).
Table 7: Forecast Model Representations
AR
ARIMA
AROilFutS
AROPS
DPFut
GPFut
MA
OilFut
OilFutS
OP
11
Table 10: Gas Out-of-Sample Rolling Forecast RMSEs
Forecast Period
2003:1-2004:12
2005:6-2007:5
2008:4-2010:3
h=1
h=3
h=9
h=12
h=1
h=3
h=9
h=12
h=1
h=3
h=9
h=12
0.07
0.12
0.13
0.17
0.18
0.30
0.35
0.29
0.18
0.33
0.51
0.31
OilFut RMSE
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
ARCH(OilFut)
1.02
0.99
0.95
1.02
1.03
1.02
1.00
1.06
1.07
1.09
1.06
1.16
ARCH(OilFutS) 1.02
Random Walk 1.21
1.00
0.80
0.89
0.98
0.95
0.88
1.02
1.09
1.08
1.05
1.25
1.37
1.47
1.85
1.10
1.28
1.24
1.66
1.35
1.83
1.81
2.72
OilFut(AIC=1)
Relative RMSEs
ARCH(RW)
1.21
1.35
1.37
1.75
1.11
1.26
1.18
1.64
1.38
1.83
1.73
2.73
AR(3)S
1.20
1.41
1.30
1.65
0.99
1.14
1.18
1.50
1.35
1.73
1.72
2.65
ARIMA(1,1,2)
1.19
1.37
2.44
2.13
1.09
1.43
2.19
2.21
1.32
1.64
1.64
2.40
ARIMAS
1.18
1.30
1.42
1.56
0.97
1.11
1.15
1.45
1.33
1.71
1.47
2.62
GPFut
0.68
0.56
0.98
0.29
0.62
0.55
0.46
0.65
0.70
0.66
0.67
0.55
ARCH(AROFS) 0.96
MA(1)
1.16
0.99
0.96
0.80
0.93
0.94
0.81
1.06
1.03
1.09
0.98
1.33
1.26
1.50
1.84
0.99
1.23
1.30
1.59
1.33
1.84
1.67
2.74
MAS
1.18
1.33
1.36
1.58
0.98
1.13
1.21
1.53
1.30
1.73
1.60
2.67
ARCH(OP)
1.01
0.99
0.91
1.09
0.95
0.95
0.89
0.98
1.00
0.98
0.87
1.09
ARCH(OPS)
0.99
0.97
0.76
0.92
0.92
0.88
0.78
0.94
1.02
0.96
0.89
1.11
AROPS
0.91
0.95
0.80
0.84
0.79
0.84
0.78
0.82
0.87
0.83
0.71
0.89
Table 11: Diesel Out-of-Sample Rolling Forecast RMSEs
Forecast Period
2003:1-2004:12
2005:6-2007:5
2008:4-2010:3
h=1
h=3
h=9
h=12
h=1
h=3
h=9
h=12
h=1
h=3
h=9
h=12
0.05
0.08
0.10
0.10
0.15
0.19
0.23
0.19
0.10
0.16
0.21
0.21
OilFut RMSE
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
OilFutS
1.03
1.06
1.21
1.18
0.99
1.04
0.99
1.05
1.17
1.22
1.06
1.25
Random Walk
1.63
1.84
3.02
2.53
1.23
1.46
1.57
2.19
1.97
2.79
2.71
4.50
AR(2)
1.55
1.77
3.01
2.51
1.18
1.45
1.57
2.07
2.09
2.95
2.44
4.25
ARS
1.66
1.94
3.30
2.65
1.18
1.51
1.44
2.12
2.17
2.94
2.81
4.48
MA(1)
1.56
1.79
3.04
2.52
1.18
1.44
1.57
2.10
2.07
2.89
2.53
4.33
MAS
1.65
1.90
3.30
2.64
1.17
1.47
1.42
2.10
2.20
2.96
2.77
4.44
ARIMA(1,1,2)
1.61
1.95
3.28
2.83
1.27
1.80
3.09
3.24
2.08
2.64
3.33
3.47
ARIMAS
1.65
1.89
3.33
2.63
1.17
1.49
1.43
2.14
2.15
2.95
2.80
4.49
ARCH(DPFut)
0.88
0.59
0.63
0.94
0.86
0.84
0.82
0.88
1.00
1.01
0.77
1.23
AROilFutS
1.00
1.03
1.24
1.17
0.96
1.03
0.98
1.03
1.23
1.28
1.13
1.22
ARCH(AROPS)
1.08
1.17
1.46
1.39
0.89
0.94
0.77
0.92
1.17
1.19
1.68
1.12
OP(3)
1.11
1.20
1.30
1.28
0.96
0.95
0.80
0.91
0.96
0.98
1.65
0.98
ARCH(OPS)
1.12
1.23
1.67
1.42
0.94
0.96
0.77
0.95
1.77
1.93
1.19
1.17
OilFut(AIC=3)
Relative RMSEs
Rolling Out-of-Sample Forecasts: 2003:1-2004:12
DP Using ARCH(DPFut)
GP Forecast Using GPFut
1-Step Ahead Rolling Forecast
1-Step Ahead Rolling Forecast
$2.2
$2.2
$2.0
$2.0
$1.8
$1.8
$1.6
$1.6
$1.4
$1.4
$1.2
$1.2
$1.0
1/02
$1.0
4/02
7/02
10/02
1/03
4/03
7/03
10/03
1/04
4/04
7/04
10/04
$2.4
$2.4
$2.0
$2.0
$1.6
$1.6
$1.2
$1.2
$0.8
1/02
$0.8
4/02
7/02
10/02
1/03
12-Step Ahead Rolling Forecast
4/03
7/03
10/03
1/04
4/04
7/04
10/04
12-Step Ahead Rolling Forecast
$2.4
$2.4
$2.4
$2.4
$2.0
$2.0
$2.0
$2.0
$1.6
$1.6
$1.6
$1.6
$1.2
$1.2
$1.2
$1.2
$0.8
$0.8
1/02
$0.8
1/02
4/02
7/02
10/02
1/03
actual
4/03
7/03
rolling forecast
10/03
1/04
95% bound
4/04
7/04
10/04
$0.8
4/02
7/02
10/02
1/03
actual
4/03
7/03
rolling forecast
10/03
1/04
95% bound
4/04
7/04
10/04
14
Rolling Out-of-Sample Forecasts: 2005:6-2007:5
GP Forecast Using GPFut
DP Using ARCH(DPFut)
1-Step Ahead Rolling Forecast
1-Step Ahead Rolling Forecast
$3.6
$3.6
$3.6
$3.6
$3.2
$3.2
$3.2
$3.2
$2.8
$2.8
$2.8
$2.8
$2.4
$2.4
$2.4
$2.4
$2.0
$2.0
$2.0
$2.0
$1.6
1/05
$1.6
$1.6
1/05
4/05
7/05
10/05
1/06
4/06
7/06
10/06
1/07
4/07
$1.6
4/05
7/05
10/05
12-Step Ahead Rolling Forecast
1/06
4/06
7/06
10/06
1/07
4/07
12-Step Ahead Rolling Forecast
$4.0
$4.0
$3.6
$3.6
$3.5
$3.5
$3.2
$3.2
$3.0
$3.0
$2.8
$2.8
$2.5
$2.5
$2.4
$2.4
$2.0
$2.0
$2.0
$2.0
$1.5
1/05
$1.5
$1.6
1/05
4/05
7/05
10/05
actual
1/06
4/06
rolling forecast
7/06
95% bound
10/06
1/07
4/07
$1.6
4/05
7/05
10/05
1/06
actual
4/06
rolling forecast
7/06
10/06
95% bound
1/07
4/07
15
Rolling Out-of-Sample Forecasts: 2008:4-2010:3
GP Forecast Using GPFut
DP Using OP
1-Step Ahead Rolling Forecast
1-Step Ahead Rolling Forecast
$5
$5
$6
$6
$4
$4
$5
$5
$3
$3
$4
$4
$2
$2
$3
$3
$1
$2
1/08
$1
1/08
4/08
7/08
10/08
1/09
4/09
7/09
10/09
1/10
7/08
10/08
1/09
4/09
7/09
10/09
1/10
12-Step Ahead Rolling Forecast
12-Step Ahead Rolling Forecast
$5
$5
$4
$4
$3
$5
$5
$4
$4
$3
$3
$3
$2
$1
1/08
$2
4/08
$2
$1
4/08
7/08
10/08
1/09
actual
rolling forecast
4/09
7/09
95% bound
10/09
1/10
$2
1/08
$2
4/08
7/08
10/08
actual
1/09
4/09
rolling forecast
7/09
95% bound
10/09
1/10
16
Conclusions
 No matter whether gas prices are stable or not, their
futures price does better at forecasting them than others.
 Fama(1970): Efficient Market Hypothesis
 Similarly, diesel price futures perform well at predicting
diesel prices during periods of stable increases and
shocks, but spot oil prices helped predict prices during
the latter period.
 Therefore, I construct good models to forecast fuel prices
during recent periods of rising petroleum prices that
achieves a further understanding of their future prices.
17
Future Research
 What do futures prices tell us about business cycles?
 What are the impacts on the term structure of interest
rates from gas and diesel price shocks?
 What implications are there for monetary policy?
 Do asymmetric responses exist between fuel price
futures and the price at the pump?
18