Download FSU Oil Production and GDP Decline – Granger Causality and

FSU Oil Production
and GDP Decline
– Granger Causality and the
Multicycle Hubbert Curve
Marek Kolodziej
ASPO Workshop, Lisbon, Portugal
May 19, 2005
Presentation of the paper by
Douglas Reynolds and Marek Kolodziej
Reasons for breakup



Economic inefficiencies of central
planning
Reagan Administration policies
1987 oil production peak
Granger Causality


Correlation is not causation
Did the oil peak Granger-cause GDP
decline and the breakup?
GDP-to-oil hypothesis




Economic chaos decreased investment
Lower GDP: destroyed internal demand
Hypothesis rejected - we have 95%
confidence that GDP does not Grangercause oil production
Tested for 1-, 2-, and 3-year lags
Oil-to-GDP hypothesis



Oil Granger-causes GDP with 95%
confidence
This is 1970-2003 data, so oil Grangercaused GDP even prior to the peak
Tested for 1-, 2-, and 3-year lags
Granger Causality
GDP=f(lagged GDP)
GDP=f(lagged GDP, lagged oil)
5
4
0
0
-4
-5
-8
-10
-12
-16
-15
87
88
89
90
91
FSU GDP Growth
92
93
94
95
96
97
Forecas t FSU GDP Growth
98
87
88
89
90
91
FSU GDP Growth
92
93
94
95
96
97
Forecas t FSU GDP Growth
98
Coal-to-GDP,
Natural gas-to-GDP




GDP Granger-caused a coal production
decline (98% confidence) but not vice versa
No Granger-causality between NG and GDP
Therefore, Oil causes GDP but coal and NG
do not
This suggests oil affects the economy, but not
vice versa
Multi-cycle Hubbert curve





First cycle: 19th century to 1996, with a
1987 peak
Discovery peaked in the 1960s
1996 – privatization of Russian oil
Better property rights
Production at old fields resumed and
fueled recent growth
Multi-cycle Hubbert curve

The basic Hubbert model is
URR
CQP 
 a (t t 0 )
1 e

Take derivative to get
dCQP a *URR * e  a (t t0 )
QP 

dt
[1  e a (t t0 ) ]2
Multi-cycle Hubbert curve

Solve QP for
e  a ( t t0 ) 
 a(t  t0 )
to get
URR
1
CQP
 URR

 a (t  t 0 )  ln 
 1
 CQP 
a * URR * e  a (t t0 )
QP 
[1  e  a (t t0 ) ] 2
 URR

a * URR * 
 1
 CQP 

2
 URR 
1  CQP  1


URR 2
a*
 a * URR
a
CQP
QP 

a
*
CQP

* CQP 2
2
URR
URR
CQP 2
Multi-cycle Hubbert curve


Relabel a as b1 and
Model becomes

a
URR
as b2
QP  b1 * CQP  b2 * CQP

2
Dummy variable for 1996 institutional
change – Privatization
QP  b1 * CQP  b2 * CQP 2  b3 * DUM 96  b4 * DUM 96 * CQP
Actual Production
Estimated Production
2052
2046
2040
2034
2028
2022
2016
2010
2004
1998
1992
1986
1980
1974
1968
1962
1956
1950
1944
1938
1932
1926
1920
FSU Time Hubbert Curve
14
12
10
8
mb/d
6
4
2
0
FSU Cumulative Production
Hubbert curve
Figure 2. Former Soviet Union Oil Production as Function of
Cumulative Production--Forecast
Rate of Production (millions of
barrels per year)
6000
Actual
Production
95% Confidence
Interval
2009
5000
4000
1985
3000
2000
1996
1000
Forecast
Peak in
Production
0
0
50000
100000
150000
200000
Cumulative Production (millions of barrels)
250000
Multi-cycle Hubbert curve

Jean Laherrère (2002)
Conclusion


1987 oil production peak was a
contributing factor to Soviet collapse
Soviets were investing a lot in the
energy sector, but scarcity caused
production to decline
Conclusion




First peak in 1987 at 12.05 mb/d
Secondary peak expected in 2009-2010
at around 12.5 mb/d
FSU URR = 255 Gb
(Russia, Azerbaijan, Turkmenistan,
Kazakhstan, Uzbekistan)
Jean Laherrère: 250 Gb