The Relationship between
Oil Revenues/Production
and GDP; and between oil
revenues and defense
spending in the Iraqi
economy
Ivana Kragul
Introduction
Since the discovery of oil in Iraq in 1908, various political influences between the British
and French sought to control a stake of the profitable Iraqi oil industry. The Turkish Petroleum
Company was established by the British in order to compete with United States oil companies.
Subsequently, the United States had soon abolished Turkish autonomy over the oil industry in
Iraq. During the turning point of the Sykes-Pikot agreement, Iraq’s’ autonomy was soon
abolished as western powers began carving up regions in the Middle East. Soon after,
explorations had discovered major oil fields in Mosul and the United States had consequently
established an open door policy with the British enacted puppet government King Faisel.
Rivalries had ensued resulting in an agreement in which the countries of Britain, France,
Holland and the United States were each entitled to 23.75% of the stakes in the industry,
leaving the Iraqi citizens poor and exploited.
Western influence soon waned as the British granted Iraq independence in 1932.
However, weak Suuni leadership had competing aspirations had caused the British to retain
power over the region through the monarchy. After lengthy periods of political instability, the
puppet monarchy was finally overthrown by the Baathist, a pan-Arab secularist party that
managed to quell Kurdish and communist rebellions and establish a nationalist oil industry and
push foreign powers out. The final political act that sealed Iraqi control over its oil industry and
resulted in profound economic effects is the nationalization of the oil industry in 1966. The
establishment of the Iraqi national Oil Company had effectively allowed Iraq to break away
from the influence of foreign multinational corporations as well create an industry that would
create local jobs, and an investment fund for the state. On the other hand, the maneuver had
created conflict with neighboring states in the Middle East, particularly those that were part of
OPEC. Iraq became increasingly disinterested with Organization of Petroleum Exporting
Countries (OPEC) for the sake of independent national interests especially when the
organization sought to allow member states to purchase equity interests in companies such as
the Iraqi Public Company (IPC). As a result this independence as well as the increasing demand
for oil combined with market panic, Iraq had managed to reap enormous oil revenue during the
OPEC oil embargo. Furthermore, oil profits were sustained by anchoring prices slightly below
those of OPEC’s which created a high degree of dependence on oil revenues.
While external factors such world demand for oil determined oil revenues, internal
political factors were just as pivotal in determining the growth and sustainability of the industry.
As the industry rapidly development, the state was increasingly relied on oil revenues and was
putting pressure on the government to sustain and expand the industry. Inter-regional and
international political or religious conflicts and civil unrest can both negatively impact growth.
With oil revenue contributing up to ninety nine percent of Iraq’s gross domestic product, oil
dependency rendered a high fragile possibly unstable economy that is susceptible to
endogenous shocks. Thus, economic stability became highly linked to political stability. An
example illustrating this point is the Iraqi invasion of Kuwait which led to an international
backlash in the form of economic sanctions.
Estimations regarding the quantity of oil is contained in Iraq is uncertain due to decades
of war and lack of scientific research, however, the estimated reserves in the region rank 143
billions of barrels per day. A significant increase in industry investment as well as geographic
exploration is needed to fully exploit the reserves. The United States Energy Information
Administration estimates that ninety percent of the territory in Iraq remains unexplored. In
addition, oil production costs in Iraq are one of the lowest in the world. Thus, there is reason to
believe that the reserves will contribute significantly to the economy but only if Iraq can
succeed in maintaining a unified state. However there are factors that are negatively
contributing to the industry. While oil production in Iraq has faltered due to trade embargos
due to the invasion of Kuwait in 1990, the oil industry is actually unsustainable. Iraq is
producing 3.5 million barrels of oil per day while the sustainable production capacity is only 2.8-
2.9 billions of barrels of oil per day.
While external factors such world demand for oil determined oil revenues, internal
political factors were just as pivotal in determining the growth and sustainability of the industry.
As the industry rapidly development, the state was increasingly relied on oil revenues and was
putting pressure on the government to sustain and expand the industry. Inter-regional and
international political or religious conflicts and civil unrest can both negatively impact growth.
With oil revenue contributing up to ninety nine percent of Iraq’s gross domestic product, oil
dependency rendered a high fragile possibly unstable economy that is susceptible to
endogenous shocks. Thus, economic stability became highly linked to political stability. An
example illustrating this point is the Iraqi invasion of Kuwait which led to an international
backlash in the form of economic sanctions. In addition, one author argues that in a rentier
state economy, citizenship becomes a “financial asset” and that international companies may
find difficulty in establishing operations, namely multinational oil and drilling companies.
Moreover, it should be noted that an economy may not entirely depend on oil although certain
sectors of the economy may depend on oil revenues.
The purpose of this study is to determine the correlation between Gross
Domestic Product and oil production and eventually examine the dependency of the Iraqi
economy on oil production. By examining the relationships between these two economic
figures, one could determine whether the political economy of Iraq fits the rentier state
framework. By verifying this theoretical model, one could predict the economic prospects as
well as stability for the state of Iraq as well as neighboring Middle Eastern states that may also
be large producers of oil. Not only would one be able to predict the trend of economic
development of Iraq but one could also be able to evaluate the stability of Iraq as a viable state.
The main goal of this study is to create a model for policy makers to use in terms of social and
economic development, international relations and military strategy.
Methodology
Description of Data
Data from this study is completely derived from the National Income of Iraq, a national
account based on economic statistics for the country and arguably one of the most reliable
accounting sources for the nation. After some minimal cleaning of data, I decided to include
data from 1953 to 1973 allowing for 20 observations in the regression calculations to be divided
among four lags in a time series regression. Arguably, it should be noted that the data provided
is based on a national account based on research and statistics from the Central Bank of Iraq
and due to political interest may not represent optimal accuracy although it is the only source
for historical time series data. Furthermore, for the time period during the rule of the dictator
Saddam Hussein, data and statistics for oil revenues, GDP and other accounts are nonexistent
because economic figures were considered state secrets during the era. Nonetheless, although
the model may not be used to accurately forecast or predict future values or trends, a
relationship between oil revenues and GDP could be detectable and could be utilized as a
springboard for more accurate and statistically detailed models.
The data that has been utilized for the study include year (1953-1973), GDP (at factor
cost), crude oil prices, percent change, percentage share in GDP, total oil revenue, percentage
change in oil revenue, , defense expenditures, percentage change in defense expenditures,
defense expenditures as a portion of GDP and per capita defense expenditures. In addition, oil
production and oil revenues are divided as separate figures due to the change in oil prices that
occurs due to natural fluctuations in demand, even though Iraq was one of the founding
members of Organization of the Petroleum Exporting Countries (OPEC). In addition, from these
figures, one may detect changes in oil revenue, GDP, oil production and defense spending per
year as well as determine the relationships between these two changes.
For the initial analysis, I began by providing a statistical data table as is summarized in
Table 2 of the appendix. In addition, I provided line graphs to visualize the growth in GDP and
crude oil production and defense spending as well as the growth of oil production and oil
revenue and defense spending per year as well as a percentage of GDP. In addition, I provided a
scatterplot correlating the GDP and oil revenue as well as oil revenue and defense spending.
Finally, I ran two auto regressions summarizing the relationship between oil revenues
and GDP and the relationships between oil revenues and defense spending since it is often the
case that a rentier state will use oil revenues to fund defense spending as opposed to social or
economic development. Thus, with a certain degree of historical perspective, a correlation
between a particular military campaign and the autoregression of defense spending and oil
production could be observed.
Hypothesis
Based on the trends observed from the data and plotted in the graph, I hypothesize that
there is a positive relationship between the change in GDP and oil revenues and oil production
based on the relationships displayed in figure 1. There seems to be a growth in oil revenue and
oil production within this time period suggesting that a correlation between oil revenues and
production and GDP may not be causative. In other words, it may be false to state that an
increase in oil revenues may contribute to an increase in GDP. In addition, based on these
trends, I predict a stronger and positive correlation between oil revenues and defense spending
while concluding that oil revenue and defense spending are more positively correlated.
Econometric Models
Eq 1.
Eq 2.
Eq 3.
Eq 4.
Eq. 1 represents the regression that examines the relationship between a change in oil
production and a change in gross domestic product such that a 1% increase in oil production is
associated with a β% increase in GDP. Likewise, equation 2 examines the relationship between
a change in oil revenue and a change in gross domestic product such that a 1% increase in oil
revenue is associated with a β% increase in GDP. Eq. 3 examines the relationship between a
change in defense spending and a change in oil revenue such that a 1% increase in oil revenues
is associated with a β% increase in defense spending.
Results
Based on equation 1 and the regression represented in Table 4, the significance of the
coefficient of the lag were tested at the 1%, 5%, and the 10% level with z scores of 2.33, 1.96
and 1.645 respectively. The hypothesis was tested such that the null hypothesis that the
coefficient is equal to 0 versus the alternative hypothesis that the coefficient is not equal to 0
and belongs in the regression. More formally, H0: βt = βt vs. Hα: βt ≠ βt, whereas t represents lag1
though lag4 for oil production. Based on the autoregression AR(1) model described in Table 5.,
the coefficient on lp1 is .4367 with a standard error of .2068 resulting in a t statistic of 2.108.
The z score for the coefficient falls within the rejection region for the significance level of 5%
and 10% only whereas the coefficient which has a value of 66.44 and a standard error of 3.33
resulting in a t statistic of 19.95 will fall well within the rejection region for the significance
levels of 1%, 5% and 10%. Both values are statistically significant for the 10% and 5% value on
the intercept is significant for the 1% level; therefore we can reject the null hypothesis that the
coefficients on these variables are zero. Likewise, repeating the same procedure for AR(2)
produces the same results for tp1 and the coefficient. However, the coefficient on the second
lag, tp2 is not significant at any level. The calculations in AR(3) yield similar results such that tr2
and tr3 are not statistically significant and the coefficient is statistically significant at all levels,
although tr1 is significant at only the 10% level. AR(4) yields similar results to AR(2) where only
the intercept and tr1 is statistically significant. The F statistics in each of the four regressions are
larger than the F test statistic at the 10%, 5% and 1% level indicating a rejection of the null
hypothesis that the coefficients on each of the lags are jointly 0. In addition, the R squared
shows no predictable pattern within each of the regressions indicating that oil production is a
poor predictor of GDP.
Based on equation 2 and the regression represented in Table 6., the significance of the
coefficient of the lag were tested at the 1%, 5%, and the 10% level with z scores of 2.33, 1.96
and 1.645 respectively. More formally, H0: βt = βt vs. Hα: βt ≠ βt, whereas t represents lag1
though lag4 for oil revenue. Based on the autoregression AR(1) model described in Table 5,
with a z score of 2.11 for the coefficient of tr1, the variable is statistically significant at the 10%
and 5% level. The regressions yield similar results from AR(1)-AR(4) although in AR(4) tr1 is not
statistically significant at any level as well as the coefficient of tr2 and tr3. The intercepts are
significant at all levels while the coefficient on tr4 is also statistically significant at the 10% and
5% level. The F statistic has generally decreased indicating that the coefficient on the regression
has become less statistically significant. Finally, there is no significant trend between oil
revenues and GDP indicating that the variable is a poor predictor of GDP.
Based on equation 3 and the regression represented in Table 6., the significance of the
coefficient of the lag were tested at the 1%, 5%, and the 10% level with z scores of 2.33, 1.96
and 1.645 respectively. More formally, H0: βt = βt vs. Hα: βt ≠ βt, whereas t represents lag1
though lag4 for defense spending. However, none of the values expect the coefficient prove to
be statistically significant at the 1%, 5% or 10% level except the coefficient term which is
significant at all levels. Moreover, the R squared values hold no particular trend with respect to
GDP.
Based on equation 3 and the regression represented in Table 6., the significance of the
coefficient of the lag were tested at the 1%, 5%, and the 10% level with z scores of 2.33, 1.96
and 1.645 respectively. More formally, H0: βt = βt vs. Hα: βt ≠ βt, whereas t represents lag1
though lag4 for oil revenue. Almost all of the values are statistically significant at the 10%, 5%
and even the 1% except lag td1. In addition, the high F squared indicates that the joint
coefficients are significant and are unlikely to be zero. Furthermore, a high R squared indicates
that the independent variable is likely to be a good determinant of the independent variable.
The augmented Dicker Fuller test was used to test the null hypothesis that the variables
oil production, oil revenues and defense spending have a stochastic trend versus the alternative
hypothesis that the trend is stationary. The critical values used for this test were at the 10%, 5%
and 1% significance level with a corresponding Z- score of 1.645, 1.96 and 2.33 respectively.
Because all of the observed critical values lie within the rejection region at all significance levels,
we cannot reject the null hypothesis that any of these variables has a stochastic trend, against
the alternative hypothesis, that the variables have a stationary trend.
Discussion
My original hypothesis stated that an increase in oil production would correspond to an
increase in gross domestic product. Likewise, an increase in oil revenue would correspond to an
increase in GDP. I had decided to test the effects of oil production and oil revenues as separate
regressions because much of Iraq’s oil reserves lies undiscovered or is inaccessible. In addition,
much the oil produced remains unrefined due to the absence of technological development
which places the nation below its production capacity. My hypothesis was based on the
assumptions that GDP, oil production and oil revenues would grow steadily over time given that
Iraq was undergoing a process of slow and steady industrialization.
Based on the information presented in the graph there seems to be a general upward trend
when graphing oil revenue and oil production and GDP growth. However, the variables GDP, oil revenue
and oil production may happen to follow an upward trend even these variables are not necessarily
causative. To control for this effect, I regressed changes in GDP on changes in oil production and
revenue. If an economy is truly dependent on oil revenue as a source of income, any changes in oil
revenue is likely to be correlated to a corresponding change in GDP. Thus, the resulting statistics and
regression prove that there is no particular correlation between changes in oil revenue and production
and GDP.
I also chose to examine the relationship between a change in defense expenditure and GDP and
found no obvious correlation. In addition, I also chose to examine the correlation between oil revenue
on defense spending. The chosen variables were based on my assumption that an increase in oil
revenue is more likely to be correlated to a change in defense expenditure. Since Iraq has undergone
military regimes, an application of the rentier state framework would state that oil revenue is most likely
to fund military expenditure as opposed to any other sector of GDP. Interestingly, I found a positive
correlation between a change in oil revenue and a change in defense expenditure although the reasons
for this relationship is not certain as is evident by the steady and increasing R squared. The positive
correlation could relate to the idea of a “war economy” in which oil revenue is used to finance military
campaigns.
In order to provide more detailed and accurate conclusions on the relationships between oil
revenue/production and GDP as well as the relationship between defense expenditure and oil revenue,
more study is needed. In particular, a more detailed data set collected quarterly could allow the lag
lengths to be shortened from 5 years to 3 and could possibly reveal a trend. In addition the additional
data may reduce some of the ambiguity associated with the autocorrelations between change in GDP
and change in oil revenue, or production. On the other hand, obtaining accurate national income
accounts may require additional legal procedures if a researcher were to obtain data from the Iraqi
Central Bank. Otherwise, the quarterly data is an estimated figure. In addition, oil revenues have raised
dramatically over the few decades so a study concerning oil revenues and GDP during the periods of the
post-Saddam era conducted in the future could provide interesting insights.
Conclusion
The results of this study serves the purpose as a springboard for further research yet
also opens a door of other challenges. Interestingly, the weak correlation between oil revenues
and GDP and the strong correlation between oil revenues and defense expenditure do not
discredit the rentier state model but rather change the framework. The case may be that an
increase in oil revenues is inherently related to an increase in defense spending, and hence an
instable regime. However, oil revenues cannot be to blame for political corruption since an
existing may simply take advantage of oil revenue to launch a military campaign. Thus, in the
future, with better data, a researcher could control these effects by focusing on a transitional
time period.
References
1. Alnasrawi Abbas. The economy of Iraw: wars, destructruction and development and
prospects. Westport CT: Library of Congress Cataloging-in-Publication Data, 1991
2. Saving Iraq from Its Oil. Nancy Birdsall, Arvind Subramanian. Foreign Affairs
Vol. 83, No. 4 (Jul. - Aug., 2004), pp. 77-89 Published by: Council on Foreign Relations
Stable URL: http://www.jstor.org/stable/20034048
3. Economic Policy and Prospects in Iraq. Foote, Christopher; Block, William; Crane,
Keith; Gray, Simon The Journal of Economic Perspectives, Vol. 18 No. 3 Summer 2004 ,
pp. 47-70(24)
Appendix
Table 1. Description of variables
abbreviation variable abbreviation variable
Year year As a portion of GDP Oil revenues as a
portion of GDP
Gdpfactor GDP at factor cost Defense expenditure Defense expenditure
(per year)
crudeoilprodustion Crude oil production Percentchangedefense Percentage change in
per year (millions of defense spending
barrels)
change Change in oil Tp1-tp4 Oil production, first
production lag-fourth lag
shareingdp Share of oil Tr1-tr4 Oil revenues, first lag-
production as a % of fourth lag
GDP
Total revenues Total oil revenues Td1-td4 Defense spending,
first lag-fourth lag
percentchangerev Percentage change in
oil revenues
Table 2. Summary Statistics
Variable Obs Mean Std. Dev. Min Max
gdpatfactor 20 796.1143 376.4257 322.95 1587.5
crudeoilproduction 20 264.8243 120.1739 113.1 563.4
change 20 9.071429 19.19638 -22.8 55.2
shareingdp 20 34.0619 4.103837 26.3 41.7
totalrevenues 20 139.3514 104.6449 49.03 519.3
percentchangerev 20 14.09048 29.62284 -29.1 106.4
asaproportionofgdp 20 19.60952 4.332078 12.8 34.5
defenseexpxpendituree 20 77.77 60.82931 14.94 246.3
percentchangedefense 20 15.3 18.60981 -12.7 60.6
Table 3.
Lag AC change
1953-1957 tp1 0.6840 -.4603
1958-1962 tp2 0.6557 .0237
1963-1967 tp3 0.4383 .0163
1967-1972 tp4 .3202 -.1237
1953-1957 tr1 .4599 -.2809
1958-1962 tr2 .4185 .2079
1963-1967 tr3 .2220 -.0262
1967-1972 tr4 .1679 -.0505
1953-1957 td1 .7069 -.2881
1958-1962 td2 .6098 -.0231
1963-1967 td3 .4897 -.2672
1967-1972 td4 .3841 .1025
Table 4. Augmented Dicker Fuller test.
Test statistic 1% CV 5% CV 10% CV
Oil production Z(t) 2.491 -3.750 -3.000 -2.630
Change in oil production Z(t) -3.214 -3.750 -3.000 -2.630
Oil revenues Z(t) 1.417 -3.750 -3.000 -2.630
Change in oil revenues Z(t) -0.239 -3.750 -3.000 -2.63
Defense spending Z(t) 5.555 -3.750 -3.000 -2.630
Change in defense spending Z(t) -2.280 -3.750 -3.000 -2.630
Table 5. Autoregression model of GDP on oil production (1953-1973)
specification AR(1) AR(2) AR(3) AR(4)
Regressors
∆tpt-1 .4367546** .4833391** .4791307* .4408229**
(.2068416) (.2035444 ) (.2956457) (.20995)
∆tpt-2 -.1514725 -.1537908 -.15243
(.1362919) (.1961803) (.18757)
∆tpt-3 .0214941 .02256
(.3555961) (.2975)
∆tpt-4 .3111
(.451)
intercept 64.44522*** 66.98828*** 66.85582*** 66.7***
(3.334168) (3.942644) (5.981517) (6.98)
F statistic 4.46 3.02 4.1 3.55
R2 .4637 .67215 .11223 .597
Coefficients are statistically significant at the *10% level, **5% level and ***1% level
Table 6. Autoregressive mode of GDP oil revenues (1953-1973)
specification AR(1) (AR2) (AR3) (AR4)
Regressors
∆trt-1 1.000582** 1.108113 ** 1.0597** .6982972
(.5091719) (.5128906) (.5238) (1.317329)
∆trt-2 -.3412459 -.2467306 -.53426
(.3280801) (.5889747) (.6870222)
∆trt-3 .3013557 .8731097
(.9003863) (2.401569)
∆trt-4 -1.810625***
(.227)
intercept 141.8305*** 146.9738*** 142.4023*** 144.9487***
(6.700981) (8.256019) (15.19955) (12.30583)
F statistic 3.86 2.52 3.12 1.00
2
R .5628 .7163 .1042 .7494
Coefficients are statistically significant at the *10% level, **5% level and ***1% level
Table 7. Autoregression of GDP on defense spending (1953-1973)
Specification AR(1) AR(2) AR(3) AR(4)
Regressors
∆tdt-1 .1062266 .0137393 -.0198138 .1447783***
(.0734814) (.0921834) (.0798012) (.519)
∆tdt-2 -.3189658 -.0569443 .9390273**
(.228599) (.2665219) (.89715)
∆tdt-3 -.3064833 -.5890023**
(.2202203) (.2879)
∆tdt-4 -.3855329
(.3915)
Intercept 17.88776*** 23.91556*** 23.78083*** 22.15423***
(1.880458 ) (4.620677) (3.814334) (2.23874)
F-statistic 2.09 2.35 2.94 2.11
2
R .2141 .4027 .8983 .574
Coefficients are statistically significant at the *10% level, **5% level and ***1% level
Table 8. Autoregression of oil revenues on defense spending
Specification AR(1) AR(2) AR(3) AR(4)
Regressors
∆tdt-1 .5417 .67214 .66074*** .7113294
(.48752) (.6806) (.041878) (.62387)
∆tdt-2 .27511*** .35861*** .4238***
(.0988) (.0684876) (.18787)
∆tdt-3 2.8977*** .30237**
(.7834) (.2289)
∆tdt-4 .217638
(.069897)***
Intercept 33.12*** 45.87*** 46.323*** 43.324***
(1.2347) (2.72346) (2.23489) (1.23467)
F-statistic 2.66 2.78 3.11 2.89
R2 .26 .47 .95 .99
Coefficients are statistically significant at the *10% level, **5% level and ***1% level
Figure 1. GDP at factor cost and crude oil production (1953-1973)
1500
1000
500
0
1955 1960 1965 1970 1975
year
GDP at factor cross crude oil production
Figure 2. Crude oil production (1953-1973)
2000
1500
1000
500
0
1955 1960 1965 1970 1975
year
Figure 3. Total oil revenues (1953-1973)
500
400
total revenues
300
200
100
0
1955 1960 1965 1970 1975
year
Figure 4. Oil production as a percentage of GDP
60
50
% share in GDP
40
30
20
1955 1960 1965 1970 1975
year
Figure 5. Oil revenue as a percentage of GDP
40
30
20
10
0
1955 1960 1965 1970 1975
year
Figure 6. Relationship between percent change in defense spending and percent change in oil
revenues
60
40
20
0
-20
-50 0 50 100
% change
Fitted values percentchangedefense
Figure 7. Relationship between percent change in oil production and percent change in defense
spending
60
40
20
0
-20
-20 0 20 40 60
percentchangedefense
% change Fitted values