Forecasting International Oil Price Using an ECM Model

Wu Haiying and Shengchu Pan

Central University of Finance & Economics, Beijing China

Presenting at the 19th INFORUM World Conference South Africa, August 2011

Introduction

• The rapid increment of China’s oil demand is thought to be one of the main reasons for the rising of international oil price in recent years.

• Our model intends to empirically answer if and how China’s oil demand influences NYMEX crude oil future price (WTI), and forecasting of the WTI price is also made.

Main Contents

• The model’s characteristics • WTI equation • WTI forecasts in 2011 • Conclusions

The Model’s Characteristics

• Exams the significance of China’s oil demand on global oil price based on world oil import and export markets.

• Explanatory variables: – Growth changes in oil exports (supply side) – Growth changes in major countries’ oil imports (demand

side) – Other factors ( US dollar index and US crude oil stocks)

• Dynamic modeling techniques and error correction model (ECM) are adopted. The software used is PcGive10.0.

• Quarterly model based on data from 1995 to 2010.

Variables of WTI Equation Description Units Source

WTI WTI crude oil future price $/barrel Wind Database

UsIm US crude oil & petroleum product imports Million barrels CEIC Database

ChnIm China crude oil & petroleum product imports Million barrels Wind Database

OPEC OPEC crude oil production capacity Million barrels Wind Database

UsInv US crude oil stocks Million barrels CEIC Database

USD Nominal major currencies dollar index March 1973 = 100 www.federalreserve.gov

Note: WTI and USD are quarterly average data.

Figure 1: Oil Imports1 - US & China (% of World Exports of Oil2) 1US & China oil imports include imports of crude oil and petroleum products; 2World exports of crude oil and refined products. Source: Calculations based on Wind database and OPEC Annual Statistical Bulletin.

0.0

5.0

10.0

15.0

20.0

25.0

1995 1997 1999 2001 2003 2005 2007 2009

US

China

19.1

9.1

Figure 2: Change in Oil Imports – World1, US and China (Million Barrels)

1World here refers to world exports of crude oil and refined products. Source: Calculations based on Wind database and OPEC Annual Statistical Bulletin.

-1000

-500

0

500

1000

1500

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

World US China

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

0.0

1.0

2.0

3.0

4.0

5.0

6.0

1995

-1

1996

-1

1997

-1

1998

-1

1999

-1

2000

-1

2001

-1

2002

-1

2003

-1

2004

-1

2005

-1

2006

-1

2007

-1

2008

-1

2009

-1

2010

-1

LWTI LChnIm

6.3 6.4 6.5 6.6 6.7 6.8 6.9 7.0 7.1 7.2 7.3

0.0

1.0

2.0

3.0

4.0

5.0

6.0

1995

-1

1996

-1

1997

-1

1998

-1

1999

-1

2000

-1

2001

-1

2002

-1

2003

-1

2004

-1

2005

-1

2006

-1

2007

-1

2008

-1

2009

-1

2010

-1

LWTI LUsIm

Figure 3: US, China Oil Imports VS WTI Future Price

Left axis - LWTI Right axis - LChnIm Left axis - LWTI Right axis - LUsIm Note: Lx means logarithm of x. Source: Calculations based on Wind and CEIC database.

Figure 4: OPEC Capacity and US Crude Oil Stocks VS WTI Future Price

Left axis - LWTI Right axis - LOPEC Left axis - LWTI Right axis - LUsInv

Note: Lx means logarithm of x. Source: Calculations based on Wind and CEIC database.

7.7 7.7 7.8 7.8 7.9 7.9 8.0 8.0 8.1 8.1

0.0

1.0

2.0

3.0

4.0

5.0

6.0

1995

-1

1996

-1

1997

-1

1998

-1

1999

-1

2000

-1

2001

-1

2002

-1

2003

-1

2004

-1

2005

-1

2006

-1

2007

-1

2008

-1

2009

-1

2010

-1

LWTI LOPEC

7.7 7.7 7.8 7.8 7.9 7.9 8.0 8.0 8.1 8.1 8.2

0.0

1.0

2.0

3.0

4.0

5.0

6.0

1995

-1

1996

-1

1997

-1

1998

-1

1999

-1

2000

-1

2001

-1

2002

-1

2003

-1

2004

-1

2005

-1

2006

-1

2007

-1

2008

-1

2009

-1

2010

-1

LWTI LUsInv

Figure 5: US Dollar Index VS WTI Future Price

Note: Lx means logarithm of x. Source: Calculations based on Wind and CEIC database.

4.0

4.1

4.2

4.3

4.4

4.5

4.6

4.7

4.8

0.0

1.0

2.0

3.0

4.0

5.0

6.0 19

95-1

19

95-4

19

96-3

19

97-2

19

98-1

19

98-4

19

99-3

20

00-2

20

01-1

20

01-4

20

02-3

20

03-2

20

04-1

20

04-4

20

05-3

20

06-2

20

07-1

20

07-4

20

08-3

20

09-2

20

10-1

20

10-4

LWTI LUSD

Left axis - LWTI Right axis - LUSD

Unit-root Tests LWTI LUsIm LChnIm LOPEC LUsInv LUSD

-3.26 -0.92 -3.73* -2.89 -1.98 -2.68

DLWTI DLUsIm DLChnIm DLOPEC DLUsInv DLUSD

-6.25** -9.88** -7.97** -7.50** -7.65** -5.67**

Note: 1Data are quarterly from 1995-2010; 2LX represents logarithm of X, DLX means the first-order difference of LX; 3LX ADF tests use constant, trend and seasonal dummies, while DLX tests just include constant. ** for 1% significance level, * for 5% significance level.

Econometric Techniques Adopted

• ECM can be generated by rewriting autoregressive distributed-lags model (AD). For a simple example, the AD model in (1) is equivalent to the ECM model given by (2).

(1)

(2) Where : Δ is the first-order difference; is lagged equilibrium error (ECM variable); is the adjustment factor which is expected to be negative and

greater than -1.

ttttt zyzy εββββ ++++= −− 131210

ttt zKyz εβββ +−−+∆+=∆ −11210t ))(1(y2

311 1 β

ββ−+

=K

11 )( −− tzKy)1( 2 −β

Estimation Process

– First, we set up an ADL model with LWTI as dependent variable: • LWTIt = f(LUsImt, LChnImt, LOPECt, LUsInvt, LUSDt, Seasonal)+εt， (3)

• Adequately long lags were used to ensure that εt is a white noise process.

– Then, adopting the “general to specific” dynamic approach, we gradually reduced the model (3) and derived its ECM equation as the format expressed in model (2).

– The final estimated equation is as follows:

• DLWTI = 4.979 + 0.1724*DLWTI(-1) + 1.218*DLOPEC - 5.776*DLUsInv - 0.948*DLUsIm + 0.2018*DLChnIm + 0.2325*Seasonal + 0.3139*Seasonal_1 - 0.4405*WTI_ECM(-1) (4) Where, DLWTI = LWTI – LWTI(-1) and so on;

• At the same time, we got the ECM term: WTI_ECM = LWTI - 0.51*LChnIm - 1.97*LUsIm + 2.1*LUSD+1.88*LOPEC

• From the ECM term, In the long run , we have LWTI = 0.51*LChnIm + 1.97*LUsIm -1.88*LOPEC - 2.1*LUSD

Estimation Results Coefficient Std. Error t-value t-prob

Constant 4.98 0.8022 6.21 0.000 DLWTI (-1) 0.17 0.0889 1.94 0.058 DLOPEC 1.22 0.6347 1.92 0.060 DLUsInv -5.78 0.9386 -6.15 0.000 DLUsIm -0.95 0.3456 -2.74 0.008 DLChnIm 0.20 0.0878 2.30 0.026 Seasonal 0.23 0.0390 5.96 0.000 Seasonal_1 0.31 0.0493 6.37 0.000 WTI_ECM (-1) -0.44 0.0695 -6.34 0.000

R2 = 0.702153 DW = 2.08 AR 1-4 test: F(4,49) = 0.46567 [0.7606] ARCH 1-4 test: F(4,45) = 0.32070 [0.8626] Normality test: Chi^2(2) = 0.65040 [0.7224] hetero test: F(14,38) = 1.9549 [0.0509]

China’s Influence on WTI Price

• China’s oil imports do have effects on global oil price both in the long run and short run. – In the long run , its elasticity to WTI is 0.51. – In the short run, change in China’s oil imports also have a

positive influence on WTI price.

• Although China affects oil price, its strength is the weakest among other three long determinants, which are US dollar index, US oil imports and OPEC crude oil production capacity. The elasticity of oil price with respect to them is -2.1, 1.97 and -1.88 respectively.

1995 2000 2005 2010

-0.50

-0.25

0.00

0.25DLWTI Fitted

1995 2000 2005 2010

-2

-1

0

1

2 r:DLWTI (scaled)

Figure 6: Fitness of DLWTI and its Residual Distribution

2000 2005 2010-0.250.000.250.50 DLWTI_1 × +/-2SE

2000 2005 20102.55.07.5

10.0 Constant × +/-2SE

2000 2005 2010

0.02.55.0 DLOPEC × +/-2SE

2000 2005 2010

-10.0-7.5-5.0-2.5 DLUsInv × +/-2SE

2000 2005 2010

-2

0

2 DLUsIm × +/-2SE

2000 2005 2010

0.00

0.25

0.50 DLChnIm × +/-2SE

2000 2005 20100.10.20.30.40.5 Seasonal × +/-2SE

2000 2005 2010

0.25

0.50

0.75 Seasonal_1 × +/-2SE

2000 2005 2010

-0.75-0.50-0.25 WTI_ECM_1 × +/-2SE

Figure 7: Recursive Analysis of Coefficients in DLWTI Equation (95% confidence interval)

Figure 8: Fitness of LWTI and its Residual Distribution (R2=0.9789)

1995 2000 2005 2010

3.0

3.5

4.0

4.5

5.0LWTI Fitted

1995 2000 2005 2010

-2

-1

0

1

2 r:LWTI (scaled)

Note: LWTI equation is simply derived from DLWTI equation (4) with LWTI = DLWTI + LWTI (-1).

Assumptions for Explanatory Variables

• OPEC: According to Short-term Energy Outlook, August 2011 from Energy Information Association(EIA). EIA expects OPEC crude oil production capacity will decline from 33.76 in 2010 to 32.86 million barrels per day in 2011, in large part due to the supply disruption in Libya.

• USD：2011 Q1 and Q2 data are actual observations. For the next two quarters in 2011, we assume US dollar will keep a slight appreciation with 70 and 71, largely due to Europe debt crisis.

• UsInv: Based on quarterly growth rates from EIA short-term energy outlook, August 2011.

• UsIm & Chnim: From their own autoregressive models.

Exogenous Variables (Million barrels)

US oil imports China oil imports OPEC crude oil production capacity

US crude oil stocks

US Dollar index March 1973=100

2011 Q1 1024 548 3051 3240 72.1

2011 Q2 1103 536 2959 3282 69.8

2011 Q3 1099 549 2991 3272 70.0

2011 Q4 1061 554 2991 3302 71.0

Note : Data in red color is forecasted based on the assumption listed.

Oil Price Forecasts

1 Dynamic (ex ante) forecasts. 2 The standard error based on error variance only.

Actual price Point forecasts1 Confidence intervals2

2011 Q1 94.60 87.4 72.5 - 105.4

2011 Q2 102.34 84.9 70.2 - 102.6

2011 Q3 NA 85.1 70.4 - 102.9

2011 Q4 NA 80.1 66.2 - 96.8

Figure 9: Oil price forecasts ($/Barrel)

0

20

40

60

80

100

120

140

2005-1 2006-1 2007-1 2008-1 2009-1 2010-1 2011-1

WTI_Actual WTI_Fore

Upper Limits Lower Limits

Conclusions

• The rapid growth of China’s oil imports has already influenced the global oil price, while the extent is much lower than that of US and OPEC factors.

• WTI oil future price will hold around 84 dollar per barrel through 2011.

Thank you!