Why do leaders nationalize the oil industry? The politics
of resource expropriation
Paasha Mahdavi
Department of Political Science, University of California, Los Angeles, 4289 Bunche Hall, Box 951472, Los Angeles, CA 90095, USA
H I G H L I G H T S
 I model determinants of oil nationalizations for 65 producing countries 1945–2005.
 I offer a new measure of nationalization using the establishment of NOCs.
 Oil prices, political institutions, cross-country diffusion predict nationalization.
 Nationalization is also likely when revenue is perceived to be shared unfairly.
 Operator-led contract renegotiation can reduce likelihood of nationalization.
a r t i c l e i n f o
Article history:
Received 20 May 2014
Received in revised form
5 August 2014
Accepted 27 September 2014
Available online 1 November 2014
Keywords:
Nationalization
National oil companies
Resource nationalism
Bayesian statistics
a b s t r a c t
Why do leaders nationalize the oil industry? In line with a general utility-maximizing theory, I argue
that leaders nationalize to maximize state revenues while minimizing costs. The latter includes
international retaliation and domestic political constraints. Using a novel longitudinal dataset on the
establishment of national oil companies (NOCs), the empirical evidence presented in this paper lends
support to four primary findings. States are most likely to establish NOCs (1) in periods of high oil prices,
when the risks of expropriation are outweighed by the financial benefits; (2) in non-democratic systems,
where executive constraints are limited; (3) in “waves”, that is, after other countries have nationalized,
reflecting reduced likelihood of international retaliation; and, though with weaker empirical support,
(4) in political settings marked by resource nationalism. This last factor is proxied by OPEC membership
in large-N analysis and, in a two-case comparison, by the difference in retained profits between the host
and foreign governments. The theory and empirics presented here offer some clues for policy makers
and multinational companies alike as to when to expect leaders to opt for nationalization.
& 2014 Elsevier Ltd. All rights reserved.
1. Introduction
As of 2012, between 73 and 95 percent of global oil reserves are
controlled by national oil companies (NOCs).1
The majority of these
NOCs were established through nationalizations in the 1970s, though
several states opted for NOCs in the 1930s and 1990s (see Fig. 1).
Though these kinds of nationalization are rare events – occurring
only 45 times since the 1930s – the impacts of state expropriation are
game-changing in both international markets and the domestic
political environment. Scholars working on the political resource
curse – that is, the hypothesized relationship between oil and
authoritarianism – point to the wave of nationalizations in the
1970s as a turning point for autocrats in gaining control over
lucrative resource revenues (Aslaksen, 2010; Dunning, 2008; Haber
and Menaldo, 2011; Ross, 2012). I provide insight on the determinants
of these events; that is, I aim to answer the question, why do
political leaders nationalize the oil industry?
Using both statistical analysis of historical nationalizations and
a quantitative case comparison, I show that the decision to nationalize
is motivated by state revenue maximization, risk of international
retaliation, and resource nationalism. While researchers have
put forth a handful of theories on why leaders expropriate the oil
industry (see Victor, 2013 for a review), there exists no comprehensive
assessment of political and economic factors of oil nationalization
in the context of domestic perceptions and international risks.
Some provide strong theoretical frameworks for the economic
underpinnings of expropriation (Chang et al., 2010; Guriev et al.,
2011), while others expound on domestic political factors in the
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/enpol
Energy Policy
http://dx.doi.org/10.1016/j.enpol.2014.09.023
0301-4215/& 2014 Elsevier Ltd. All rights reserved.
E-mail address: paasha@ucla.edu
1
The variance in figures stems from how reserves are attributed to operating
firms. The 95% figure is drawn from Ernst and Young (2013), Global Oil and Gas
Reserves Study, which assumes that all reserves in a country with a nationalized
sector belong to the NOC. The lower bound of 73% is drawn from Victor et al.
(2012), who use the classification of reserves based on actual share of production
from a given field (also known as “working interest”).
Energy Policy 75 (2014) 228–243
formation of ownership structure in the oil sector (Luong and
Weinthal, 2010; Warshaw, 2012). But current scholarly work has
yet to incorporate a systematic discussion of the cost-benefit
structure of nationalizations which takes into account resource
nationalism and fears of foreign intervention.
This study offers three contributions to the extant literature. First,
an often omitted factor in the empirical analysis of oil nationalization
is the diffusion effect (Kobrin, 1985; Vernon, 1971). That is, the
cost of nationalization in a given country is substantially reduced the
more that other countries nationalize. Stephen Kobrin termed this
phenomenon “the domino effect” of nationalization. While the
theoretical implications of Kobrin's work have not been subjected
to statistical analysis except by Kobrin himself – who was able to
provide support for a “cumulative” or wave effect in the 1970–1984
period – other scholars have discussed the contagion effect of
nationalizations (Adelman, 1993; Warshaw, 2012). Here, I extend
this analysis to a broader time frame to confirm that diffusion is a
strong predictor of nationalization. Based on these findings, I
conclude that nationalization is substantially more likely to occur
after a first-mover has reduced the risk of international retaliation
and paved the way for further nationalizations – a phenomenon that
occurred both in the early 1970s and the early 1990s.
Second, I hypothesize that a state's perceptions of “unfairness” in
how profits are shared between host and operating countries
influences the likelihood of nationalization. When a leader perceives
that her share of oil profits is lower than the share taken home by
the foreign operating company's government, this prompts government
and public sentiments of resource nationalism and provides
motivation for nationalization to eliminate the profit-sharing gap.
These perceptions of unfairness are difficult to test cross-nationally
due to data availability, so I show this effect both by employing
a proxy variable in the longitudinal analysis and by analyzing a case
comparison of Iran and Saudi Arabia. Whereas the ratio of profits
shared between Iran and the UK was consistently in favor of the UK
in early years of production (1930–1950), the profit-sharing ratio
between Saudi Arabia and the US was nearly equal in the same
period. Not surprisingly, Iran nationalized the oil industry in 1951,
while Saudi Arabia waited until 1974 to nationalize and until 1980 to
fully expropriate its oil sector (and was the last OPEC member to
nationalize).2
This explanation, I argue, helps to understand cases
where current models get the prediction of nationalization wrong:
Iran nationalized during a time of low oil prices and during an
era with relatively (among non-democracies) high executive constraints,
both of which are factors predicting a low probability of
nationalization.
Third, I provide a methodological contribution to the existing
literature on resource nationalization. Because the decision to
nationalize is tested in the context of longitudinal data with a
discrete outcome – a leader either nationalizes or not in a given
country in a given year – researchers typically use either ordinary
least squares or maximum likelihood regression techniques including
unit fixed effects to account for country-specific potentially
omitted factors. As I discuss in greater detail in the pages that
follow, the application of conventional methods to these data is
problematic. As such, I operate within a Bayesian estimation framework
to mitigate these concerns. As this method requires specification
of prior distributions for the parameters to be estimated, I
combine expert interviews and previous scholarly findings to
estimate informative priors for the analysis.
The findings of this study speak to the complexity of a state's
decision to nationalize the oil sector. With many moving parts to
this decision, it is difficult to pin down any one explanation for
nationalization. My aim is to augment scholarly understanding of
such events by providing evidence for two additional factors –
resource nationalism and the diffusion effect – that help to improve
the predictive accuracy of arguments for why leaders nationalize. In
the sections that follow, I begin with a presentation of the puzzle in
theoretical context. I then formulate hypotheses and discuss the
methods and data I use to test them. The subsequent sections
include empirical results from a statistical analysis and a case
comparison of Iran and Saudi Arabia. I conclude with a discussion
of policy implications based on these findings.
2. Methods
2.1. Theoretical determinants of nationalization
A leader's decision to nationalize the oil industry is inherently
based on a delicate cost-benefit analysis.3
A leader must maximize
his expected utility from nationalization while considering the
potential benefits to state ownership and avoiding the potential
costs of expropriation.
The primary benefit to nationalization is a short- to mediumterm
increase in the state's take of revenues from the sale of oil
(Victor et al., 2012; Victor, 2013; Wolf, 2009). Other benefits
include direct oversight of operations and production decisions,
and control over lucrative state-owned enterprise management
positions to use as tools of patronage (Nolan and Thurber, 2010;
Golden and Mahdavi, 2015). By expropriating foreign assets, the
Year
CumulativeNationalizations
1905 1925 1945 1965 1985 2005
0
5
15
25
35
45
Year
Nationalizationsperyear
1905 1925 1945 1965 1985 2005
0
1
2
3
4
5
6
7
Fig. 1. Plotting cumulative and annual nationalizations for the period 1905–2005. Nationalization is measured as the establishment of a national oil company (NOC) in
a given year. Sample includes 61 oil-producing countries.
2
This excludes Gabon, which joined OPEC in 1975 and nationalized in 1979,
and left OPEC in 1995.
3
Though I use the term “leader” here referring to an individual political agent,
the concept applies equally to consensus-based decisions to nationalize such as
those by a parliament, junta, oligarchy, etc.
P. Mahdavi / Energy Policy 75 (2014) 228–243 229
state not only gains by controlling new hard assets (e.g. rigs,
pipelines, and drilling equipment), but more importantly increases
the share of profits collected by the treasury from the oil industry
(Marcel, 2006; Stevens, 2007). Rather than having to split profits
with a foreign company or government, the state can collect 100%
of profits from the sale of oil via the NOC and decide how best to
reinvest profits back to the company, though more often states
such as Brazil, Nigeria, and the UAE have chosen to rely on other
operators for production that is facilitated through profit-sharing
agreements (McPherson, 2010). In the realm of oil nationalizations,
this framework is adopted by Guriev et al. (2011) to show
that leaders nationalize the oil industry when petroleum prices are
high as this maximizes the short-term revenues from expropriation
and outweighs the potential costs of nationalization.
An additional benefit to nationalization is satisfying domestic
sentiments of “resource nationalism”. This refers to the public's
perceived “lost profits” from the perspective of the state vis-a-vis
private oil operators (Bremmer and Johnston, 2009; Tordo et al.,
2011; Vivoda, 2009). More specifically, a leader may feel “cheated”
by private operators of her fair share of oil profits if she sees
private companies benefiting more from oil production than they
are entitled to. In other words, if the state perceives its share of oil
profits to be low relative to the operators’ share, the state may see
the difference in profits as the opportunity cost of maintaining
a private ownership structure.
When private operators are foreign-owned in particular, xenophobic
feelings arise that foreigners are “stealing” a country's oil,
which leaders and the public feel is the sovereign right of an
independent nation. Referring to the period prior to expropriation,
scholars use nationalistic quotations from political leaders such as
“it's our oil”, “the oil belongs to the people”, and “driving out the
foreign devils” (Karl, 1997; Yergin, 1991). If there exists a noticeable
gap between what the state treasury collects from oil profits
and what foreign operators collect, this could influence the
decision to nationalize. Though this is itself a form of revenue
maximization, it is a combination of resource nationalism and
maximizing state revenues. As Vlado Vivoda has noted, “it is
natural that during a period of high prices the phenomenon of
resource nationalism comes to the surface, as it is a by-product of
high prices” (Vivoda, 2009, 518).
Research by Manzano and Monaldi (2009) similarly finds that
high oil prices induce pressure to renegotiate fiscal contracts.
Because of the lack of price contingencies in many existing contracts,
high oil prices translate to disproportionately higher operatorretained
revenues compared to what is allocated to the state. This
imbalance, the authors argue, can stoke grievances over revenuesharing
that result in contract renegotiation in the form of increasing
taxes and royalties to outright nationalization of assets.
Though it has not been tested empirically, this notion of resource
nationalism – specifically, public perceptions of unfairness in how
resource revenues are divided between foreign operators and the
host state – has been recognized by existing scholarly work in the
context of oil nationalizations. Building off of Manzano and Monaldi
(2009), Berrios et al. (2011) recognize resource nationalism as one of
the mechanisms behind their explanation for the political Left's
expropriation of oil and gas in Latin America. Solberg (1979) and
Smith (2007) cite resource nationalism as potential factors in the
nationalizations of Argentina and Iran, respectively. Singh (1989) and
Stevens (2008) note the cyclical patterns of resource nationalism, not
just in the oil and gas sector, but also in metals such as copper, iron
and steel. Similarly, Kretzschmar et al. (2010) identify resource
nationalism as the reason for limited foreign investment after a
nationalization due to strong elements of xenophobia and mistrust.
The costs of nationalizing the oil sector are more complex. The
most straightforward cost is the expected loss of efficiency when
switching from a privately run oil company to a state-run firm
(Hartley and Medlock, 2008), though this may be reduced in the
long run by improving NOC efficiency (Victor et al., 2012). A more
dangerous cost is the loss of oil exports due to international
retaliation, as foreign governments may enforce an oil embargo
on the nationalizing country. Such was the case after Mexico's
nationalization in 1938 and Iran's nationalization in 1951. Lesser
sanctions may be enacted by the international community following
nationalization, such as trade and financial sanctions, which
were considered by Spain and the EU after nationalization of
Repsol in Argentina in 2012.
Work in the 1980s by Stephen Kobrin highlighted the so-called
“diffusion effect” of nationalizations in other countries affecting
the probability of nationalization in a given country (Kobrin, 1984,
1985). Consider that in December 1936, the Bolivian state nationalized
Standard Oil's assets to form the national oil company
Yacimientos Petroliferos Fiscales Bolivianos (YPFB). Two years
later, Brazil's dictator Getulio Vargas proposed a new government
agency with “extensive powers over all sections of the oil
industry”, an agency which was formally established in 1938 and
early 1939 as the country's national oil company, the CNP (Philip,
1982, 230). These two events are not independent of one another:
indeed, Vargas' energy commission directly cited the establishment
of Bolivia's YPFB and Argentina's YPF (formed in 1911) as
examples to be followed (Cohn, 1968). One explanation for this
pattern is that the diffusion effect of nationalizations could simply
be a “copycat” effect, whereby countries nationalize the industry
to emulate the ownership structure of perceived “pioneer” countries,
similar to patterns of “copycat entrepeneurship” among
small to medium sized enterprises (Philip, 2002).
In minimizing the costs of nationalization, leaders must also
take into account the constraints of expropriation given domestic
politics and the institutional environment. Work by Luong and
Weinthal (2010) considers institutional constraints on expropriation:
countries with strong political institutions will find it harder
to expropriate without incurring large and possibly disastrous
political costs. On the other hand, countries with either weak
institutions or institutional settings that favor state control will
find it easier to nationalize.
Guriev et al. (2011) similarly characterize strong institutions as
impediments to expropriation, and find that executive constraints
– checks and balances on the executive branch of government –
increase the costs of nationalization. In non-democratic systems in
particular, weak or non-existent parliaments or other veto points
may allow a leader to nationalize the oil industry with little
political opposition, though this might have long-term costs in
the form of diminished foreign investment (Tsebelis, 2002; Henisz,
2004). Further, elites may have the ability to influence dictators
and monarchs to press for nationalizations in a way not possible in
democracies (Gandhi, 2008).
Executive constraints may also reduce the probability of nationalization
by decreasing the leader's share of rents resulting from
expropriation (Warshaw, 2012). Compared to a personalist dictatorship
or monarchical regime, a leader in a power-sharing position
must divide these rents among other high-ranking elites, as in
a single-party dominant autocracy such as the USSR or pre-1994
Mexico, or in the case of democracy, among other government
branches and bureaucracies.
Before turning to a discussion of testable hypotheses, it is
important to note exactly how rents from the sale of oil are allocated
to the government. Even in states with nationalized sectors, there
are a variety of rent allocation options. A government can collect
rents directly from its NOC or from IOCs making payments via the
NOC on royalties, license fees, acreage fees, dividends (from joint
ventures), income taxes, bonus payments, or what are sometimes
called “special profits taxes.” A government can also collect rents
indirectly either from its NOC or from IOCs through profit oil
P. Mahdavi / Energy Policy 75 (2014) 228–243230
(typically from production-sharing contracts), infrastructure projects,
or social/training funds (McPherson, 2010). Centralizing rent
allocation through a government-owned entity is often stated as a
justification for creating a NOC in the aims of narrowing the
technical knowledge gap between state and operator (on this point,
see Stevens, 2008). Similarly, Warshaw (2012) cites closing information
asymmetries as one factor in why governments nationalize the
oil sector.4
2.2. Hypotheses
Taken together, these costs and benefits suggest specific factors
at play in a leader's decision when and whether to nationalize the
oil industry. I offer four testable hypotheses based on observable
implications from the theoretical determinants discussed above. I
begin by re-framing the question of why do leaders nationalize the
oil industry into statements that capture the likelihood of oil
nationalization based on a given factor or set of factors.
First, oil nationalization is more likely when oil prices are high. If
leaders are revenue maximizers, then the timing of nationalization
should occur in or after moments when the financial return to
doing so is at its peak. When global oil prices are high, then oil
revenues are high (assuming constant or increasing production).
Further, nationalizing when the short-term gains are at high levels
outweighs the potential financial costs of nationalization – notably
inefficiency and retaliation.
Second, oil nationalization is more likely when others nationalize.
Vernon (1971) and Kobrin (1985) suggest that first-movers can
defray the probability of retaliation for followers. Here the notion
of first-movers is relaxed somewhat to refer to the sum of
nationalizers occurring in a previous period with respect to a
given country that has not yet nationalized. This probability
decreases with the number of countries that nationalize in this
prior period. For instance, if six countries nationalize in a given
year (t), this reduces the retaliation probability in the following
year (tþ1) more than if only two countries nationalize.5
Third, oil nationalization is more likely when there is perceived
unfairness in revenue-sharing between host government and operating
company. The resource nationalism hypothesis follows from
the notion that nationalization is a function of the perception of
fairness by the producing country with respect to the operating
company's take-home share of resource revenues. Thus nationalization
is more likely when there are perceptions of unfairness
with regards to how oil revenues are shared between host and
operator. Yet these perceptions are latent characteristics and by
nature unobservable.6
As such, an observable implication of this
hypothesis is that nationalization is more likely as the gap
between state and foreign/private revenue collection increases.
Absent data on how oil revenues are divided, a second observable
implication is that states joining OPEC are more likely to
nationalize. OPEC was founded in September 1960 on principles
of revenue fairness as manifested through price control – its
mission is “to unify petroleum policies among Member Countries
in order to secure fair and stable prices for petroleum
producers”7
– and xenophobic ideologies with respect to operating
companies backed by Western governments (Park et al.,
1976). Still, the determinants of joining OPEC are likely endogenous
to other determinants of nationalization, namely autocratic
government, a long history of oil production, and generally high
levels of oil exports. However, these factors should be considered
necessary but not sufficient determinants of OPEC membership
given the absence of major autocratic producers in the 1960s and
1970s such as the Soviet Union, Mexico, and Malaysia. In the
absence of data on revenue-sharing, OPEC membership is a
proxy, albeit one with measurement error, for countries with
resource nationalistic tendencies and revenue-maximization
ideals.
Fourth, oil nationalization is more likely when institutional
quality is weak. Luong and Weinthal (2010) and Guriev et al.
(2011) show that leaders with more executive constraints find it
harder to expropriate private assets for fear of domestic backlash.
With increasing numbers of institutional veto points a leader will
find it difficult to push nationalization through the requisite
legislative and judicial channels (Tsebelis, 2002; Henisz, 2000).
Alternatively, Warshaw (2012) contends that executive constraints
can limit a leader's consumption of rents by virtue of having to
share rents from expropriation with power-sharing elites. On the
other hand, leaders with few constraints can nationalize without
overcoming institutional roadblocks. Thus we should expect
nationalization to be more likely in authoritarian states than in
democratic states. However, the empirical tests conducted here do
not discern between which mechanisms drive the resulting
relationship between institutional quality and nationalization.
2.3. Data
The outcome of interest is the probability of nationalization in a
given country in a given year. I measure nationalization as a binary
variable according to whether or not a state establishes a majoritystate-owned
NOC.8
All years prior to NOC establishment are coded
zero; the year of nationalization is coded one. For all years after
nationalization, the country is removed from the data given that
nationalization as defined here cannot occur twice in the same
country, unless a state privatizes a previously nationalized oil
industry.9
To create the NOC dataset, I rely on primary and
secondary sources to code the timing of the establishment of a
NOC and, if applicable, the government's shareholder structure. To
this end, I use 25 petroleum laws, 80 United States Geological
Survey Minerals Yearbooks, and roughly 100 scholarly accounts of
individual countries' oil histories. A full list of references, as well as
4
This is often referenced to the broader Principal-Agent Theory whereby state
intervention is presumed to reduce the information asymmetries between governments
and market operators. For a review of principal-agency theory in the context
of energy markets, see Nikogosian and Veith (2012).
5
In Tables 7 and 8 in the Appendix, I test lag times longer than one-year
(tþk4tþ1), as well as test models with cumulative effects over the course of
multiple years. See footnote 11 for further discussion of these results.
6
In theory, survey methods could be applied to measure public perceptions of
revenue-sharing fairness in the years preceding nationalization, though it still
could be argued that elites are driving public opinion to believe in “perceived
unfairness” when in fact some other factor is at the core of the decision to
nationalize.
7
OPEC. (n.d.). “Brief History.” Accessed 4 August, 2014, from http://www.opec.
org/opec_web/en/about_us/24.htm.
8
A similar measure can be constructed based not on majority ownership by
the state, but rather state ownership of what is sometimes referred to as a “golden
share” whereby the state may not control the majority of shares but retains the
authority to outvote other shareholders on key company decisions. There are only
three differences between these coding rules: France (1995–2003), Italy (1998–
2014/present), and the UK (1983–1985). The dates refer to periods where the two
coding decisions differ – for example, in France, starting in 1995 the state held a
minority stake in Elf-Aquitaine (later Total) with a golden share until 2003, at
which point the company became fully privatized (though the state continues to
own a small proportion of shares). If coding under the majority-ownership rule,
these years would be marked zero; if coding under the golden share rule, these
years would be marked one.
9
Consider the example of Canada, which nationalized oil in 1975 upon the
establishment of Petro-Canada based on existing assets held by the private
companies Panarctic and Syncrude. In 1995, under the premiership of Brian
Mulroney, Petro-Canada was privatized, with the government holding only a 19%
share in the company (Grayson, 1981). Thus, Canada “exits” the data set after
nationalization in 1975 but “re-enters” in 1996 upon privatization. After 1996, since
it becomes logically possible for Canada to re-nationalize the oil industry, Canada
remains in the data set with the nationalization measure set to zero.
P. Mahdavi / Energy Policy 75 (2014) 228–243 231
coding decisions for each country, is available as online supplementary
material.
Predictors are chosen to reflect the theoretical implications
discussed above. The measure of oil price is a de-trended, residual
price of oil, used to proxy for oil price shocks to observe long-term
oil price cycles. A de-trended price is used for ease of interpretation
and to reduce year-to-year noise in price changes.10
To capture the diffusion effect, I use a count measure of nationalizations
occurring in each year and then lag each by one year. For
example, if in 1995 there were no nationalizations and in 1996
there were two nationalizations, the count measure in 1996 would
simply be zero and in 1997 would be two.11
I also include a variable
for the count of previous nationalizations within the same region as
a given country; for example, Iran's nationalization in 1951 would
only be counted in the diffusion variable for other countries in the
Middle East & North Africa for the year 1952.12
For the resource nationalism hypothesis, in the longitudinal
analysis I use an OPEC dummy indicator for whether or not a
country is a member of the cartel in a given year. In the two-country
case comparison, I employ a more refined measure of the ratio of
state oil revenues to foreign and/or private company oil revenues.
Lastly, to measure political constraints and institutional strength,
I employ the “Polity” index of non-democracies and democracies
(Marshall and Jaggers, 2006).
To control for geological factors in the decision to nationalize,
two proxies for the oil production cycle are included in the empirical
analysis. The first is a measure of a country's “oil history”, or more
specifically, a measure of how long a country has been producing
oil. This is measured simply as the number of years since first oil
production. The second measure is one that captures growth in the
production cycle. This is measured as the year-to-year growth in oil
production, calculated using the reported oil production figures
published cross-nationally by the United States Geological Survey.
Nolan and Thurber (2010) argue that both measures should be
positively correlated with increased probabilities of nationalization.
Countries must deal with higher risks early in the production cycle
and when production begins declining. Both reflect the inherent
risks of exploration and risks associated with aging oil fields.13
To control for broad economic and political factors not covered
in the theoretical discussion above, I add to the analysis a measure
of economic development and political stability. The first is
measured using the GDP per capita indicator, collected from the
work of Maddison (2007) on global incomes. The second is
measured using the Cheibub et al. (2010) regime age indicator,
which reflects the duration of the current governing regime.
The sample includes 62 oil-producing countries across the period
1945–2005. Though the nationalization measure is coded beginning
in 1905, the lack of data on covariates reduces the time frame of the
multivariate analysis. However, this is not an egregious loss of data
given there were so few nationalizations prior to 1945 (only six
occurred in the period 1905–1944). The selection of cases is
determined by the universe of oil producers among all 175 sovereign
states in the period with populations above 100,000 (Ross, 2012).
Defining “oil producer” as a state which at any point in time
produces more than 1000 tonnes of oil per year (or about 20 barrels
per day), 62 of these 175 states qualify as oil producers.14
To put the
figures into context, 1000 metric tonnes of production could supply
the Maldives with enough oil for one day.15
2.4. Empirical methods
Hypotheses are tested using longitudinal statistical analysis of
cross-sectional time-series data on oil nationalizations. The decision
to nationalize in a given country in a given year is treated as a
dichotomous variable which is a function of country-level and timespecific
covariates. The first two hypotheses are temporal in nature;
inference is made via within-country analysis over time. The second
two hypotheses are both temporal and spatial in nature; inference is
made via between-country and within-country analysis over time.
Though I am unable to make strong causal inferences with this
research design, the identification strategy for each hypothesis relies
on capturing within-country variation over time augmented with
techniques for statistical control.
It is unrealistic to control for all possible determinants of nationalization,
particularly country-specific factors. The typical solution
to this problem in political economy studies is to add country
dummy variables or, as they are better known, country fixed effects.
This paper takes a different approach to the MLE unit fixed effects
problem.16
Here, I employ Bayesian methods with Markov Chain
Monte Carlo estimation of a conventional logistic regression model.
To account for country-specific factors, instead of adding country
dummy variables I estimate country-specific intercepts in the form of
a random variable with a standard normal distribution. I estimate
a Bayesian hierarchical logistic regression model, though I also provide
results from conventional logistic regression, hierarchical logistic
regression, and the linear probability model with country fixed effects.
The full model specification and information on Bayesian priors used
for the analysis are presented in the Appendix.
10
Calculated first by Pindyck (1999) and adapted by Guriev et al. (2011), the
formula for creating this residual is ln pt ¼ αnln pt À 1 þβ1 þβ2nt þβ3nt2
þϵt where pt
is the price of oil at time t and pt À 1 is the lagged price of oil, for each year
t A½1945; 2005Š. The deviation from this price trend is the corresponding price
shock, so we can estimate the shock by computing yearly residuals, ϵt. As a
robustness check, the nominal oil price is used based on data from the British
Petroleum Statistical Review of Energy.
11
Results in Appendix Table 7 show that when using lags, there is little effect
of diffusion beyond two years after a given nationalization. When looking at
cumulative lags, presented in Appendix Table 8, the diffusion coefficient is positive
and significant up to four years after a given nationalization (or set of nationalizations).
The correlations decay over time, with the largest coefficient estimated with
a one-year lag.
12
Results from using this variable instead of the “global” diffusion variable are
presented in Table 6, models 1–3. When including both the global and regional
counts, I find a stronger correlation for regional diffusion, suggesting that the
diffusion effect is largely driven by geographically proximate nationalizations.
13
In Appendix Table 6, I also include controls for an interactive effect between
oil production levels and changes. Results from model 5 in particular suggest a
positive correlation: there is a decline in nationalization probability when production
is declining and overall levels of oil production are low. This scenario reflects
two possibilities: (1) declining production in a typically small producer, such as the
case in post-1980 Bahrain or post-1990 Austria, or (2) declining production in a
once-major producer whose production levels have declined significantly over
time, such as the case in post-1991 Indonesia or post-1995 Gabon (both were
former members of OPEC).
14
For comparison, the median production level among producers is 1.2 million
tonnes per year. Changing the threshold to any state producing more than
0 tonnes/year only adds nine cases for a total of 71; changing the threshold to
10,000 tonnes/year drops ten cases for a total of 52.
15
Based on EIA estimates of international oil consumption in 2012. The
Maldives consumed 7,311 barrels per day, or equivalently 997 metric tonnes of
oil per day.
16
Adding country fixed effects to longitudinal analysis with a dichotomous
variable is subject to inconsistent estimates due to poor convergence of maximum
likelihood methods (Weiss, 2005). Further, Greene has shown that even the
commonly held belief that probit regression is robust to unit fixed effects is
incorrect in finite samples. One solution to this problem is to apply linear ordinary
least squares models to these fundamentally non-linear data given that the OLS
estimator is unbiased and consistent in finite samples (Heckman and Jr., 1977). This
is the approach taken by Guriev et al. (2011) in analyzing the determinants of acts
of expropriation in the oil sector, and robustness checks using non-linear methods
show similar substantive results. Yet the linear probability model, as the OLS
estimator with a dichotomous outcome is known, suffers from improper bounding
on the 0-1 interval of probabilities and implies heteroskedasticity of the residuals
(Horrace and Oaxaca, 2006). Lastly, the Bayesian approach improves the interpretation
of model results as compared to the OLS and maximum likelihood
approaches (Gelman et al., 2013).
P. Mahdavi / Energy Policy 75 (2014) 228–243232
3. Results
3.1. Statistical findings
Results from empirical analysis lend strong support for the
revenue maximization, resource nationalism, and diffusion (international
retaliation) hypotheses, and weak to modest support for the
domestic constraints hypothesis. To aide in interpretation of model
output, I present visual results in the form of added-variable plots
and modeled probability plots. A full table of statistical results from
the Bayesian analysis is presented in the Appendix in Table 4, along
with Table 5 which shows results from non-Bayesian regressions.
Added-variable plots are shown in Fig. 2 for the four variables of
interest: oil price shock, regime index (Polity score), the OPEC
dummy variable, and the number of nationalizations in the previous
year (diffusion). Each plot shows the predictor of interest on the
x-axis with the model-predicted probability of nationalization (conditional
on the full set of controls) on the y-axis. The top left plot
shows the nationalization probability for a given country in a given
year that is predicted by the model corresponding to changes in the
oil price shock variable (measured in standard deviation units). With
an oil price increase from the mean price across time (0) to a price
3 standard deviations above the mean, the predicted probability of
nationalization increases from roughly 1% to 4%. The regime index
partial regression plot shows the negative correlation between Polity
scores and nationalization probability: as a country becomes more
democratic, or more specifically, transitions from “full autocracy” to
“full democracy” the predicted nationalization probability drops
from 3% to less than 1%. As the OPEC dummy is a discrete predictor,
the visualization of the “OPEC effect” is clear and easily interpreted:
joining OPEC increases the nationalization probability from 1% to
just less than 5%.
To better place the magnitude of these effects in context, I show
scatterplots in Fig. 3 of country-specific within-sample predicted
probabilities for two periods of interest. The first is the change in
predicted nationalization probability from 1959 to 1960, reflective of
the formation of OPEC in 1960. If there were no change in predicted
probabilities, then all countries would lie along the dotted line. This
is generally the case, except for three of the 26 countries – Iraq,
Saudi Arabia, and Venezuela – which are three of five the founding
members of OPEC.17
For these three states, the act of joining OPEC
had a noticeable predicted effect on the probability of nationalization.
For Iraq, the pre-OPEC nationalization probability is predicted
to be 3%, which jumps three-fold to just under 10% the year Iraq
joins OPEC.18
Saudi Arabia, on the other hand, despite nearly tripling
in probability to nationalize after joining OPEC from 4% to 11%, did
not nationalize until much later in 1974.
The second two-year period of interest is 1973–1974, shown in
the bottom plot of predicted probabilities. This year-to-year change
represents the increase in the price of oil following the 1973 Arab oil
Fig. 2. Added-variable plots for logistic regression coefficients with 95% credible intervals for selected variables. Based on posterior estimates from the Bayesian hierarchical
logistic model with informative priors.
17
There are only 26 cases plotted here instead of the full 61 due to the fact that
21 countries were not yet sovereign (independent) and 14 countries had already
nationalized and thus removed from the sample post-nationalization. The other
two founding members of OPEC had either already nationalized (Iran) or were not
yet technically independent (Kuwait, which gained sovereignty from the UK
in 1961).
18
Iraq indeed nationalized its oil sector soon thereafter, when in 1961 the
“Free Officers” led by Abd al-Karim Qasim passed Public Law 80 expropriating the
privately owned Iraq Petroleum Company and in 1964 established the Iraq National
Oil Company (Alnasrawi, 2002).
P. Mahdavi / Energy Policy 75 (2014) 228–243 233
embargo when prices increased by 4.5 standard deviations, the
highest one-year increase in the sample. With this price shift, the
modeled nationalization probability increased for all 29 countries in
the sample in that year, as indicated by the representation of all
countries being above the dotted line. Gabon's predicted nationalization
probability increases from roughly 9% in 1973 to 34% in 1974,
while Qatar's probability increases from 35% to 62% (Qatar was also
an OPEC member at the time). The highest predicted probabilities of
the sample are estimated for the year 1974, indicative of substantively
large coefficient estimates for the oil price shock and the
OPEC dummy.
To assess predictive accuracy, I look at within-sample modeled
outcomes compared to actual outcomes. This is accomplished by
comparing the model's predictions over time for each country to
its actual year of nationalization. Instead of plotting all 62
countries, I present six cases illustrative of the model's predictive
strengths and weaknesses in Fig. 4. For Canada, Malaysia, and
Nigeria, the predicted probability of nationalization is indeed
highest during the year of or before actual nationalization, which
is represented in the plots by the dotted vertical line. Note that for
all cases, predictions are absent during the years of a nationalized
industry; for Canada, we see predicted values after nationalization
due to the sector's privatization in 1995 (Grayson, 1981).19
These
cases represent instances of model predictions with relative
accuracy of predicted vs. actual nationalization – note, however,
that the magnitude of predictions (typically peaking at 20–30%) is
relatively low due to the rare-event nature of nationalizations as
discussed above.
For the United States, Sudan, and Equatorial Guinea, however,
the model's predictions are noticeably weak. Sudan, for instance,
is predicted to nationalize in the late 1970s when the model's
predictions are at their highest levels (around 10%). Yet Sudan did
Fig. 3. Year-to-year changes in predicted nationalization probability, based on posterior estimates from the Bayesian hierarchical logistic model including all control variables
identified in the text. The dotted line represents the 45-degree line. Countries on the line reflect no year-to-year changes; countries above (below) the line reflect increased
(decreased) predicted probability of nationalization. Size of points represents oil production (in millions of metric tons per year). Dark gray points are OPEC member
countries; black points are all others.
19
Privatization began as early as 1990, when the administration of PM Brian
Mulroney (1984–1993) of the Progressive Conservative Party began its reform of
the economic sector and announced privatization of Petro-Canada. In 1991, the
government offered the first stage of public shares (30% of the company was
privatized); and by 1995, the government had sold its majority shares and retained
19% of the company. Finally, in 2004, the government sold its remaining shares and
in 2009, Petro-Canada merged with the private firm Suncor.
P. Mahdavi / Energy Policy 75 (2014) 228–243234
not nationalize until much later, when in 1996 the government
established Sudapet on the 15% expropriated stake of the Greater
Nile Operating Company international consortium (Hansohm,
2007). Interestingly, Sudan's government pursued broad nationalizations
in 1970s, expropriating assets in the agricultural, manufacturing,
and financial sectors, yet did not nationalize the nascent
oil industry. The United States never nationalized though the
modeled probability of nationalization peaks at 10% in 1974.
Getting a sense of how the model predicts out-of-sample nationalizations
is difficult given how rare nationalizations have been
since 2005, the last year in the sample. However, there are two
such cases of oil and gas nationalization for which I test the
model's predictions: Bolivia (2006) and Uganda (2012). Based on
the covariate characteristics of each country (GDP, years producing,
oil price, etc.), the modeled probability for Bolivia is 4.80%,
and for Uganda it is 0.28%. For Bolivia, while 4.80% may seem low,
consider that the modeled probabilities for the previous two years
(2004 and 2005) are 0.21% and 0.20%, respectively. The model
clearly does not perform well for the Ugandan nationalization in
2012, though it should be noted that this new NOC plays almost no
role in production and is merely a holding company for overseeing
production-sharing agreements.20
Although these models include unit effects to capture the
idiosyncracies of individual countries in the sample, predictive
accuracy can still be reduced by potential omitted variables. One
Canada
Year
PredictedProbability(%)
1950 1960 1970 1980 1990 2000
0
5
10
15
20
25
PredictedProbability(%)
0
5
10
15
20
25
PredictedProbability(%)
0
5
10
15
20
25
PredictedProbability(%)
0
5
10
15
20
25
PredictedProbability(%)
0
5
10
15
20
25
PredictedProbability(%)
0
5
10
15
20
25
Malaysia
Year
1950 1960 1970 1980 1990 2000
Nigeria
Year
1950 1960 1970 1980 1990 2000
United States
Year
1950 1960 1970 1980 1990 2000
Sudan
Year
1950 1960 1970 1980 1990 2000
Equatorial Guinea
Year
1950 1960 1970 1980 1990 2000
Fig. 4. Predictive probabilities of nationalization for selected countries based on posterior estimates from the Bayesian hierarchical logit model with informative priors.
The dashed vertical line indicates actual year of oil nationalization.
20
As of 2012, the state has approved the establishment of the Uganda National
Oil Company to take a participatory role in production-sharing agreements with
Tullow Oil, Total, and CNOOC. See Fred Ojambo (2014), “Uganda Draws Up Plan for
National Oil Company to Steer Industry,” Bloomberg News Jun 6, 2013. Retrieved
May 13, 2014, from http://www.bloomberg.com/news/2013-06-06/uganda-draw
s-up-plan-for-national-oil-company-to-steer-industry.html.
P. Mahdavi / Energy Policy 75 (2014) 228–243 235
such factor is the economic ideology of political leaders, given that
left-leaning governments are more likely than right-leaning governments
to expropriate assets (Boix, 1997; Berrios et al., 2011). Another
potential omitted factor is a leader's time horizon: Warshaw (2012)
speculates that leaders with longer time horizons might be less
likely to nationalize given the long-run costs of expropriation via
reduced productive efficiency. However, both factors are difficult to
include in statistical analysis given the lack of reliable cross-national
data on government ideology (especially in non-democratic governments
and developing democracies) and on leader time horizons.
How do these results compare to those obtained using other
model specifications? Does conducting a Bayesian analysis provide
noticeably different point estimates? To answer these questions,
I analyze the data using numerous model specifications and tech-
niques,21
the results from three of which are plotted in Fig. 5.
In terms of direction, all models provide consistent coefficient
estimates: negative estimates for regime; positive estimates for the
oil price shock, OPEC dummy, and diffusion terms; and null
estimates for all others. In terms of certainty, the Bayesian estimates
are more precise, with smaller variance than the Bayesian model
with flat priors and the maximum-likelihood estimation of the
hierarchical logistic model.22
The OLS linear probability model with
and without unit fixed effects gives the same substantive results
(shown in Table 5 in the appendix). The key message from these
results, however, is that neither model specification nor prior
selection substantively alters the findings of the study.
Taken together, the empirical analyses presented here lend
support to hypotheses that the likelihood of nationalization increases
(1) with the price of oil, (2) when states join OPEC, (3) when
other states nationalize in the previous year, and (4) in nondemocratic
states.
3.2. Resource nationalism in Iran vs. Saudi Arabia
There are cases where the modeled nationalization probability is
near zero, yet nationalization occurs, or inversely the modeled
probability is relatively high (25%) when no nationalization occurs.
What accounts for this under- and over-prediction? What other
factors could be driving the decision to nationalize? In this section,
I explore one such determinant and apply a case comparison
method to examine its relevance.
In both Iran and Saudi Arabia oil production began before 1940
and initial operations were conducted by foreign-owned oil companies.
The production profiles shown in Fig. 6 illustrate the nearconvergence
of production levels by the late 1940s. Yet the leaders of
each country – King Abdulaziz ibn-Saud and Shah Mohammad Reza
Pahlavi – took separate paths in natural resource ownership. Iran's
government infamously nationalized the Anglo-Iranian Oil Company's
(AIOC) operations in 1951. Saudi Arabia, on the other hand, waited
until 1974 to nationalize the Arabian-American Oil Company (Aramco)
and did so only gradually, with Saudi Arabia taking full ownership by
1980. Why did Iran nationalize the industry in the early 1950s while
the Saudi leadership waited another quarter-century to follow suit?
I explore the differences in both oil-producing countries during
their early years of production to show initial evidence in support of
the resource nationalism and revenue maximization hypotheses. In
Iran, the state sought to maximize the government's share of
resource revenues in conjunction with popular demand for resource
sovereignty. In Saudi Arabia, the state did not nationalize because
revenues from the sale of oil were perceived to be at a “fair” level
vis-a-vis Aramco, as negotiated between King Abdulaziz ibn-Saud
and the U.S. government (which oversaw operations by Aramco).
3.2.1. Iran
The perception of an unfair split in oil revenues between
Britain and Iran began to be disputed politically with the passage
of a new concession, now referred to as the 1933 Agreement. The
discrepancy between the two countries' collected revenues is
plotted in Fig. 7. Immediately after 1933 there existed a small
but noticeable gap between host and foreign state; from 1940
onwards Iran's absolute oil revenues did increase but clearly did so
at a much slower rate than that of the UK.
Fig. 5. Logit coefficient estimates with 95% confidence/probability intervals for four model specifications. The top (circle) in each grouping is the Bayesian Hierarchial Logit
model with informative or “proper” priors, as discussed in the text. The second (triangle) is the Bayesian Hierarchial Logit model with non-informative or “flat” priors. The
third (square) is the Maximum Likelihood Hierarchial Logit model. The bottom is the Maximum Likelihood conventional Logit model.
21
Other models analyzed: pooled linear probability model (OLS), linear
probability model with country fixed effects, maximum likelihood (MLE) logistic
model with country fixed effects, MLE probit model with and without country fixed
effects, Bayesian mixture model clustering by country and region, and a survival
(Cox Proportional Hazards) model.
22
This is simply a result of using more precise priors for estimation – even the
quasi-informative priors have smaller prior variance than the typically used “flat”
priors with mean 0 and standard deviation 1000.
P. Mahdavi / Energy Policy 75 (2014) 228–243236
Fifteen years after the 1933 agreement, on October 22, 1947,
Iranian MPs passed an amendment to the 1933 oil concession law
requiring government officials to be present in AIOC policy
decisions when they pertained to Iran's oil royalties. The motivation
for this amendment originated in MPs’ perceptions that “the
rights of the nation were impaired” when it came to AIOC royalty
receipts (Noori, 1965) and the 1933 Agreement was invalid as it
“had been forced upon Iran” (Makki, 1950). Two years later, AIOC
and the government negotiated the 1949 Supplemental Agreement,
which provided for a modest increase in Iran's royalty
payments. But noticeably absent from the new terms was any
discussion of an even “50-50” split between the two countries.
Hossein Makki, a member of the parliamentary opposition and the
right-hand-man of future Prime Minister Mossadegh, charged
before parliament that the 1949 agreement was still unfair to
the Iranian people, especially when compared to the terms given
to other oil producers at the time:
The new agreement deprived Iran of her share in the reserves of
subsidiary and allied companies. ...While the Supplemental
Agreement gave Iran seven shillings a ton, the government of
Iraq was negotiating for a new agreement which would bring that
country eighteen shillings per ton. In Venezuela, the government
was receiving approximately thirty shillings per ton.23
With this speech, and several others by prominent opposition
members, members of parliament rejected the agreement. The
British, upon seeing the failure of the negotiated agreement to make
it through Iran's parliament, told then-premier Saed that Iran should
either take the 1949 agreement or leave it; but in no way was Britain
to agree to a 50-50 sharing agreement (Noori, 1965).
After a disputed parliamentary election in 1949, the Shah tried
to placate nationalistic demands with a new election in 1950. With
campaign promises of a new oil settlement, Mohammad Mossadegh
and his National Front Party secured only eight of the 143
seats in parliament. The new PM, Ali Razmara, was a conservative
ex-soldier who sought to ease tensions between Iran and
Britain with new negotiations between the government and AIOC.
Upon Britain's rejection to opening a new round of negotiations,
Razmara was forced to take the unpopular 1949 Supplemental
Agreement again before parliament where it was rejected in 1951
for the second time. In the face of growing public opposition to the
AIOC and Britain's involvement in Iran – accompanied by street
demonstrations throughout the Winter of 1950/1951 – Razmara
publicly rejected nationalization in March 1951, calling it “imperialistic
and unwise” (Lenczowski, 1949, 17). This was to be his last
speech, as the Premier was assassinated on March 7 by the
extremist xenophobic group Fedayan-e Islam. Thirteen days after
Razmara's death, both the Parliament and the unelected Senate
passed Mossadegh's nationalization bill on March 20, 1951 and
soon thereafter established the National Iranian Oil Company.24
The Iranian nationalization of AIOC was a complex and drawnout
endeavor. There were grievances based on fair treatment of
Iranian workers, demands for substituting foreign managers by
Iranian nationals, adequate transparency of AIOC's accounting
books, and the company's unwillingness to renegotiate terms of
payment to the state. Yet the general message of the Iranian
government's decision to nationalize was that the 1933 concession
and 1949 agreement were fiscally unfair to Iran: the amount of
revenues collected by the British was substantially higher than the
payments sent to the Iranian government. Failure to renegotiate a
payment plan or even consider a 50-50 profit-sharing agreement
ultimately led to outright nationalization of AIOC in late 1951. Had
the British agreed to share profits from oil production in an
equitable manner, Iran's parliament may not have approved
nationalization. Thus a strong element in the government's decision
to nationalize was the state's desire not only to increase its
collected oil revenues but also to be on a level playing field with
respect to profits collected by AIOC and the British.
3.2.2. Saudi Arabia
The first successful oil concession in the Kingdom was the 1933
drilling agreement between King ibn-Saud and the Standard Oil
Company of California (Socal). After the discovery in 1938 of what
Socal called a “veritable oil bonanza” the company joined forces
with Texaco in setting up Aramco.25
Aramco soon discovered that it
enjoyed a substantial market advantage over Western oil companies
because of the relative ease in extracting Saudi oil: compared to the
unit costs of producing American oil at $1.01 per barrel (or in
Venezuela, $0.50/barrel), oil could be produced in Saudi Arabia for
only $0.23 per barrel (Mikdashi, 1966, 94).
King ibn-Saud's desire to maximize oil revenue came not in
the form of nationalization, but instead pushing Aramco for higher
levels of production. So much so that Texaco's president later
remarked to the Federal Trade Commission, “In order to keep King
ibn-Saud satisfied with the operation of the concession, it is
important that production be increased substantially so that the
Year
Oilproduction(milliontonnesperyear)
1930 1935 1940 1945 1950
0
1
2
3
Iran
Saudi Arabia
Fig. 6. Oil production in the 1930–1950 period, Iran compared to Saudi Arabia.
Data collected from Mu'assasat al-Naqd al-Arabi al Saudi (1960), Noori (1965), and
Mikdashi (1966).
Year
MillionsofBritishSterling
1920 1930 1940 1950
0
2
4
6
8 Britain
Iran
Fig. 7. Oil revenues collected by the Iranian government versus the British
government, 1911–1950. Data collected from Noori (1965) and Mikdashi (1966).
23
(Makki, 1950, 33240).
24
Though the consequences of nationalization are outside the scope of this
vignette, the immediate response by the British was to send the Royal Navy to
Iranian waters to threaten an occupation of the oil city of Abadan in order to protect
British interests overseas. While the British never occupied Abadan, the navy was
used to enforce an embargo of Iranian oil exports. After two years of back-and-forth
negotiations and subsequent sanctions, Mossadegh was ousted as premier by the
CIA and MI6 in what was then termed “Operation Ajax.” The Shah was reinstated in
full, and in 1954 reversed the nationalization bill to establish a joint consortium of
the National Iranian Oil Company with American, British, and French oil companies.
For more on the AIOC nationalization, see Mahdavi (2012).
25
Standard Oil Company of California (1946), Autumn Bulletin, 33 (7): 1-2.
P. Mahdavi / Energy Policy 75 (2014) 228–243 237
King would receive greater royalties.”26
To expand production, the
Aramco consortium grew from a joint venture of two companies to a
group of four companies including Socal, Texaco, Jersey, and Socony.
Prior to renegotiating the existing concession, the Saudi government
pressed Aramco to give them a greater share of its income tax
payments that were being paid to the United States. In 1948,
Aramco's Chairman testified before the U.S. Congress that the
company acquiesced to the king's demands to divert the money that
was supposed to be paid to the IRS instead to the king's treasury:
[The Saudi government] wanted more. They asked as early as
1948, “Isn't there some way in which we can get a greater
take?” and a little later than that they said, “Isn't there some
way in which the income tax you pay to the United States can
be diverted to us in whole or in part?”27
On December 30, 1950, the company signed an agreement
whereby the Saudi government taxed the net operating revenue of
Aramco to the point of creating an equal 50-50 split of revenues. By
1952, Aramco revised the 1950 agreement and provided for a
complete split of the profits before any payments were sent to the U.
S. in the form of income taxes, effectively giving the Saudi government
a greater share in operating revenues and profits than Aramco itself.
As opposed to the Iranian case, the Saudi experience in the
1940s and early 1950s reflects successful negotiations to reduce
the revenue gap between host government and foreign operator.
Instead of risking outright nationalization, Aramco understood the
risks of not negotiating with their producer government, quite the
opposite from AIOC's refusal to negotiate to create a fair sense of
revenue-sharing with the Iranian state.
3.2.3. Comparing kings: Was revenue shared “unfairly” in Iran?
To shed light on the mechanism driving the resource nationalism
hypothesis, I use a measure of the gap between host state
and foreign state and/or company revenues from oil production.
This is accomplished with data collected by Mikdashi (1966) and
Noori (1965) on how revenues were divided between host state,
foreign state, and foreign operating company.
Three possible measures can be constructed for the period 1938–
1951, with the starting point being Saudi Arabia's first year of
commercial production and the end point being Iran's nationalization.
The first is the host state's share of total revenues from the sale
of oil. A share of 50-50 is a natural benchmark for “parity”, though
the higher this share is for the host country the more “fair” a contract
is deemed by the public and the leadership. The second is the host
state's collection of oil revenues on a per-unit basis. This metric does
not have a natural benchmark since fairness will only be determined
in its relation to the foreign state's collection of revenues. As such, an
additional comparative metric is used, namely the ratio of collected
oil revenues between foreign and host state. The benchmark for this
metric is 1; values above the benchmark are more fair from the
perspective of the host state. The foreign-to-host-state ratio of
revenues is plotted for each case across time in Fig. 8.
Comparing Iran to Saudi Arabia during this period using any one
of these metrics reveals the stark differences between the two
countries in how revenues were shared with their respective foreign
states and operating companies. The mean values for each metric
(averaged across time) along with one-tailed t-tests are presented in
Table 1. Not only is the take-home revenue per unit higher in Saudi
Arabia than in Iran on an absolute basis, but also on the two relative
metrics. Indeed, for the 13-year period in consideration, Saudi Arabia
received more revenues than its operating partner, the United States.
Whereas Iran only received 27% of oil revenues with Britain
retaining the remaining 73%, Saudi Arabia collected 62% of oil
revenues compared to the United States' 38%.
Looking at these metrics for Saudi Arabia, it is no surprise that the
state opted not to nationalize Aramco in the 1940s and 1950s. At least
according to the 50-50 profit-sharing agreement of 1950, the Saudi
government was satisfied with the outcome of negotiations to the
point of allowing Aramco to continue its operations unfettered from
the prospect of expropriation. The document itself notes that the
government, prior to 1950, “for a period of many months has been
seeking additional revenue from Aramco [and] has held views
different from those of Aramco on many long-standing interpretations
of Aramco's concession and other agreements.”28
But upon completing
negotiations, the government formally acknowledged that the
agreement “constitutes a complete satisfaction of all outstanding
claims and demands of the Government with respect both to the
past and to the future; the Government agrees that Aramco may
continue to conduct its operations in accordance with the Aramco
concessionary agreements in the same manner as in the past” (§3.1).
Unlike in Iran, the Saudi government's grievances over profit-sharing
were resolved by Aramco and the U.S. government, thereby reducing
the chances of outright nationalization by King ibn-Saud.
The Iranian government similarly renegotiated its existing
concessions with the signing of the 1949 agreement, as discussed
above. Though the agreement was ultimately rejected by Iran's
Table 1
Comparative metrics of oil revenue-sharing between host government and foreign
government and/or operating company, Iran vs. Saudi Arabia, 1938–1951. One-sided
t-statistics in parentheses. (1) The data period for the host-state share of revenues
is limited to 1943–1951 while the rest of the metrics are calculated using data from
1938 to 1951.
Metric Iran Saudi
Arabia
Difference
Host-state share of revenues1
(%) 27.07 61.88 À34:81nnn
(4.58)
Host-state take of revenues per ton of
production ($/ton)
0.41 3.16 À2:74nnn
(5.81)
Foreign-to-host-state ratio of collected oil
revenues
3.14 0.99 2:16nnn
(4.54)
Note: n
p o 0.1; nn
p o 0.05; nnn
p o 0.01.
Fig. 8. Ratio of collected oil revenues between foreign and host state, Iran versus
Saudi Arabia, 1938–1951. Higher values indicate more revenue for the foreign
operator and state relative to the host state.
26
Subcommittee on Monopoly of the Select Committee on Small Business, U.S.
Senate, The International Petroleum Cartel: Staff Report to the Federal Trade Commission
(Washington, DC: 1952, 124).
27
United States Congress, Emergency Oil Lift Program and Related Oil Problems,
Hearings (Washington, DC: 1957, 1429). Cited in Mikdashi (1966, 149).
28
The full text of the Agreement can be found in J.C. Hurewitz, (ed.), Diplomacy
in the Near and Middle East: A Documentary Record, vol. 2: 1914–1956 (Princeton: D.
Van Nostrand, 1956), pp. 314–21.
P. Mahdavi / Energy Policy 75 (2014) 228–243238
parliament, negotiations resulted in a more favorable foreign-tohost-state
ratio in 1949 that, as reflected in Fig. 8, was roughly the
same ratio as in neighboring Saudi Arabia. Yet given the historically
higher levels of revenue-sharing inequity between the British
and Iranian governments, the agreement was not enough to
resolve Iranian grievances: upon news of the British government's
unfriendly offer, the public turned noticeably towards outright
nationalization of AIOC. Left-wing and moderate newspapers such
as Rahbar, Zafar, and Razm carried editorials that the oil fields
belonged to Iran and that the Iranian government should end its
policy of “generosity” towards Britain and the AIOC. Even the most
right-wing newspaper of the country, Ettelat, charged the British
government with “refusal to recognize the seriousness of the
situation in Iran” (quoted in Noori, 1965, 177). Shortly thereafter,
popular support for Mossadegh's platform of AIOC nationalization
carried him to the premiership in March 1951.
Though it is difficult to ascertain a causal effect of how perceptions
of revenue-sharing increased the risk of nationalization, it is
clear that operating companies saw it as a determinant of nationalization.
At the 1951 shareholders meeting for Standard Oil (New
Jersey), then-president Eugene Holman outlined “a better basis than
that now provided against the danger of nationalization” (Kuhn, 1951,
711). Among the factors that reduced the risk of nationalization, he
stressed that operating companies needed to “recognize that a
foreign government which lets oil concessions may rightfully expect
that an adequate participation in the proceeds should accrue to the
government” (quoted in Kuhn, 1951, 711). Concession agreements
that did not provide for satisfactory revenue-sharing were at risk of
being forcibly renegotiated through nationalization, a process which
Yergin and Stanislaw characterize as one that “was meant to put an
end not only to the concessions themselves but to the humiliation
that went with them” (Yergin and Stanislaw, 1998). Even in 2014,
public grievances against foreign operators can be stoked by perceptions
of unfair revenue-sharing in oil and gas contracts. In Tanzania in
July 2014, for instance, a leaked gas contract between the government
and the Norwegian company Statoil has prompted local media
to declare that the Tanzanian government is getting “fooled” out of
almost $1 billion in lost profits because of how production-sharing is
structured over the long term.29
Though there have been no serious
calls to expropriate Statoil's concessions, it is likely that nationalization
will be a bargaining threat if the contract is not renegotiated.
These vignettes lend initial support to the resource nationalism
hypothesis proposed in this study. Going beyond a crude indicator of
resource nationalism captured by OPEC membership, as employed in
the cross-national analysis above, the data on host vs. operator
revenue-sharing shed light on one mechanism for why leaders
nationalize the oil industry that is reflective of public and state
sentiments of unfairness which necessitates state control of precious
resources. Specifically, I have argued here that AIOC's failure to resolve
these grievances led to its nationalization by the Iranian government,
while Standard Oil's understanding of this “danger of nationalization”
led to a renegotiated agreement with the Saudi government that
effectively delayed expropriation until 1974.
4. Conclusion and policy implications
The list of oil-exporting states has been consistently growing,
with new producers emerging as of 2014 along the Gulf of Guinea
(Ghana, Ivory Coast), East Africa (Uganda, Tanzania, Kenya),
Central America (Belize), Central Asia (Afghanistan), and South
America (Suriname, Guyana). Governments in these states have
yet to decide on ownership structures of their nascent petroleum
industries, with several in the midst of implementing new
petroleum regulation at the time of writing.30
For oil companies
operating in these states prone to expropriation, the findings in
this study provide predictive factors that signal the likelihood of
nationalization. The establishment of a NOC is most likely ...
(1) in periods of high oil prices, when the risks of expropriation
are outweighed by the financial benefits;
(2) in non-democratic systems, where executive constraints are
limited;
(3) in “waves”, that is, after other countries have nationalized,
reflecting reduced likelihood of international retaliation;
(4) and, though with less empirical support for this finding, in
political settings marked by resource nationalism when there
exists a considerable profit-sharing gap between host and
foreign governments.
Similarly for governments of operating firms which are in positions
of political negotiations with host countries, the findings here
imply that leverage will be with the host country in periods of high
oil prices and nationalizations in other states. This position is further
strengthened when revenues from the sale of oil are perceived to be
shared unfairly between operating firm and host government. To
reduce the likelihood of nationalization in these contexts, operating
firms and their governments are recommended to consider renegotiating
contracts to narrow the gap between host and operator. To
wit, this recommendation will come as no surprise to veteran firms
with experience operating with resource nationalistic leaders. But to
new operators and the ever-growing number of “independents”,
understanding the negotiation context and pursuing renegotiation
of contracts can be decisive in reducing outright nationalization.
Acknowledgements
The author is thankful to Michael Ross, Jeffrey Colgan, James
DeNardo, Barbara Geddes, Mark Handcock, Jeffrey Lewis, Robin
Mills, Felipe Nunes, Michael Thies, Antonio Ramos, Robert Weiss
and three undisclosed reviewers for their constructive comments
and the editor for facilitating the process.
Appendix A. Model specification
The outcome variable πi;j is the country-specific probability of
nationalization at time j; xi;j are a matrix of mean-centered and
Table 2
Questions used for prior elicitation for the intercept term α0. Question #3 in
particular is used for m0 and v0, mean and variance of α0.
(1) What is your guess for how many oil nationalizations will occur in the
coming decade?
(2) What is your guess for the maximum number of oil nationalizations in the
next decade?
(3) In a given country, what is your guess for the probability of oil
nationalization next year,
all other factors equal? What is the maximum probability? The minimum
probability?
(4) What information are you using to make these guesses? Experience, data,
risk models?
(5) Have you ever used or heard about the Kobrin expropriation dataset?
29
The Citizen Reporter, 2014, “$1 bn. loss: Who's fooling Tanzanians?”, The
Citizen. Accessed 1 Aug 2014 from http://www.thecitizen.co.tz/News/-1bn-loss–
Who-s-fooling-Tanzanians-/-/1840392/2382948/-/ya2a1oz/-/index.html.
30
Indeed, many have already established NOCs, such as Staatsolie in Suriname,
GPNC in Ghana, or the National Oil Corporation in Kenya.
P. Mahdavi / Energy Policy 75 (2014) 228–243 239
standardized predictors; α is a vector of the intercept and fixed
effects; βi is the country random effect; and ϵi;j is the over-dispersion
parameter. The inverse-Wishart prior for the random effects variance
τ is chosen for mathematical convenience (Box and Tiao, 1973).
The model is given by
yi;jjπi;j $ Bernoulliðπi;jÞ ð1Þ
logitðπi;jÞ ¼ x0
i;jαþz0
iβi þϵi;j ð2Þ
α $ Nðm; vÞ ð3Þ
βijτ2
$ Nð0; τ2
Þ ð4Þ
τ2
$ InvWishartðr; RÞ ð5Þ
ϵi;j $ Nð0; DÞ where D ¼ 1 ð6Þ
Prior information for m; v is a combination of three sources: an
elicited prior, estimates from previous work, and the quasiinformative
range method for parameters which have not yet
Table 4
Posterior estimates for fixed effects from Bayesian hierarchical logistic regression
with informative priors. Coefficient estimates are on the logistic scale.
Variable Mean SD 2.5% 97.5% P(4 0 j Y)
Intercept À4.477 0.235 À4.930 À4.059 0.000
Years producing 0.269 0.222 -0.185 0.692 0.890
Oil price shock 0.473 0.116 0.235 0.694 1.000
Regime age (years) À0.290 0.293 À0.891 0.275 0.162
Regime (Polity) À0.557 0.253 À1.062 À0.077 0.011
OPEC dummy 0.274 0.109 0.058 0.479 0.992
GDP/capita (log) 0.143 0.168 À0.174 0.478 0.802
Change in oil production 0.096 0.125 À0.188 0.290 0.821
Diffusion effect 0.285 0.140 À0.005 0.537 0.974
GDP growth 0.036 0.151 À0.251 0.331 0.611
Table 3
Regression estimates from Guriev et al. (2011, : 316; Table 3, Column 7) as Odds
Ratios with 95% CI. Sample size is 1718. ‘Executive constraints’ is one measure
included in the Polity index of regime type used in this study. These estimates –
which the authors transformed from the Logit coefficients to odds ratios – are
transformed back into the Logit scale for prior estimates.
Variable Odds ratio 2.5% 97.5%
Oil price 1.038 1.0125 1.0635
Executive constraints 0.994 0.9920 0.9956
Log GDP per capita 1.000 0.9902 1.0098
Table 5
Regression results from alternative model specifications. OLS refers to the linear probability model; OLS cfe refers to the linear probability model with country fixed effects.
Independent Variable Dependent variable:
NOC nationalization Guriev expropriation
OLS OLS, cfe logistic OLS OLS, cfe logistic
(1) (2) (3) (4) (5) (6)
Years producing 0.005 0.044 0.222 À0:0002n
À0:003nnn
À0:001
ð0:005Þ ð0:028Þ ð0:206Þ ð0:0001Þ ð0:0004Þ ð0:005Þ
Oil price shock 0:024nnn
0:026nnn
0:432nnn
0:073nnn
0:064nnn
1:721nnn
ð0:005Þ ð0:005Þ ð0:106Þ ð0:013Þ ð0:013Þ ð0:341Þ
Regime age À0:010 À0:006 À0:367 À0:001nnn
À0:001nn
À0:073nnn
(years) ð0:007Þ ð0:013Þ ð0:303Þ ð0:0002Þ ð0:0003Þ ð0:015Þ
Regime index À0:018nn
À0:026nn
À0:539nn
À0:001 0.001 À0:048
(Polity) ð0:007Þ ð0:013Þ ð0:255Þ ð0:001Þ ð0:002Þ ð0:046Þ
OPEC dummy 0:021nnn
0:031nnn
0:208n
0:065nnn
0.021 1:553nnn
ð0:005Þ ð0:006Þ ð0:113Þ ð0:010Þ ð0:027Þ ð0:283Þ
GDP/capita (log) 0:014n
0.010 0.323 0:015nnn
0:113nnn
0.192
ð0:008Þ ð0:023Þ ð0:256Þ ð0:005Þ ð0:013Þ ð0:136Þ
GDP growth 0.001 0:010nn
0.021 0.036 À0:037 0.876
(percent) ð0:005Þ ð0:005Þ ð0:132Þ ð0:054Þ ð0:054Þ ð1:311Þ
Change in oil production 0.007 0.007 0:124n
0.00001 0.00000 0.0004
ð0:005Þ ð0:004Þ ð0:071Þ ð0:00001Þ ð0:00001Þ ð0:0003Þ
Count of previous nationalizations 0:012nn
0:011nn
0:255nn
0:026nnn
0:020nnn
0:465nnn
ð0:005Þ ð0:005Þ ð0:125Þ ð0:003Þ ð0:003Þ ð0:069Þ
Constant 0:032nnn
À3:902nnn
À0:096nnn
À4:957nnn
ð0:005Þ ð0:221Þ ð0:036Þ ð1:075Þ
Observations 1260 1260 1260 2314 2314 2314
R2
0.065 0.256 0.096 0.188
Adjusted R2
0.059 0.218 0.093 0.166
Log likelihood À150:909 À247:881
Akaike Inf. Crit. 321.818 515.762
Note: n
p o 0.1; nn
p o 0.05; nnn
p o 0.01.
P. Mahdavi / Energy Policy 75 (2014) 228–243240
been estimated in the literature.31
The elicited prior is used for the
intercept (α0), and is drawn from interviews in October, 2012,
with oil experts working at petroleum consulting firms in Dubai,
UAE (set of questions listed in Table 2). The experts noted a 0.5%
percent chance of nationalization in a given country in a given
year, all other factors equal, with a minimum estimate of 0.1% and
maximum of 1%.32
Priors for some predictors – oil price, regime
index, and GDP – are collected from previous work by Guriev et al.
(2011), who employ a different dataset (adapted and updated from
Kobrin (1985)), replicated in Table 3 in the Appendix.33
Priors for
the remaining predictors – the OPEC indicator, diffusion effect,
GDP growth, years since first production, and offshore dummy –
are estimated using the range method.34
Table 6
Additional regression results, controlling for regional diffusion (models 1–3) and
changing oil production (models 4 and 5). Compare to main results from Model 1,
Table 5. All models are run using the linear probability model specification with
country fixed effects.
Independent Variable Dependent variable:
NOC Nationalization
(1) (2) (3) (4) (5)
Years producing 0.043 0.042 0.040 0.021 0.021
(0.028) (0.026) (0.026) (0.028) (0.028)
Oil price shock 0.026nnn
0.009n
0.009n
0.025nnn
0.026nnn
(0.005) (0.005) (0.005) (0.005) (0.005)
Regime age À0.006 À0.006 À0.007 À0.006 À0.005
(years) (0.013) (0.013) (0.013) (0.013) (0.013)
Regime index À0.027nn
À0.027nn
À0.028nn
À0.027nn
À0.025n
(polity) (0.013) (0.012) (0.012) (0.013) (0.013)
OPEC dummy 0.031nnn
0.021nnn
0.021nnn
0.030nnn
0.029nnn
(0.006) (0.006) (0.006) (0.006) (0.006)
GDP/capita (log) 0.011 0.011 0.013 0.0004 0.004
(0.023) (0.022) (0.022) (0.024) (0.023)
GDP growth 0.009n
0.006 0.006 0.008 0.006
(percent) (0.005) (0.005) (0.005) (0.005) (0.005)
Count of previous 0.009n
À0.004 0.009nn
0.008n
nationalizations (0.005) (0.005) (0.005) (0.005)
Count of previous 0.060nnn
0.061nnn
nationalizations (0.005) (0.005)
(within region)
Change in 0.007 0.006 0.006 0.005 0.004
oil production (0.004) (0.004) (0.004) (0.005) (0.004)
Oil production 0.030nnn
0.034nnn
levels (log) (0.010) (0.010)
Oil production 0.120nnn
change  levels (0.029)
Observations 1260 1260 1260 1260 1260
R2
0.253 0.331 0.331 0.258 0.269
Adjusted R2
0.215 0.297 0.297 0.221 0.231
Note: n
p o 0.1; nn
p o 0.05; nnn
p o 0.01.
Table 7
Additional regression results, controlling for differing lags in the diffusion effect.
For example, NOC2 is the count of nationalizations two years prior to the current
given year. Compare to main results from Model 1, Table 5. All models are run using
the linear probability model specification with country fixed effects.
Independent Variable Dependent variable:
NOC nationalization
(1) (2) (3) (4)
Years producing 0.043 0.038 0.040 0.039
(0.028) (0.028) (0.028) (0.028)
Oil price shock 0.025nnn
0.028nnn
0.026nnn
0.027nnn
(0.005) (0.005) (0.005) (0.005)
Regime age À0.005 À0.007 À0.007 À0.007
(years) (0.013) (0.013) (0.013) (0.013)
Regime index À0.026nn
À0.030nn
À0.028nn
À0.029nn
(polity) (0.013) (0.013) (0.013) (0.013)
OPEC dummy 0.031nnn
0.032nnn
0.031nnn
0.031nnn
(0.006) (0.006) (0.006) (0.006)
GDP/capita (log) 0.011 0.016 0.013 0.016
(0.023) (0.023) (0.023) (0.023)
GDP growth 0.009n
0.009n
0.009n
0.009n
(percent) (0.005) (0.005) (0.005) (0.005)
Change in 0.006 0.007 0.007 0.007
oil production (0.004) (0.005) (0.005) (0.005)
NOC2 0.011nn
(0.004)
NOC3 À0.003
(0.005)
NOC4 0.005
(0.005)
NOC5 À0.002
(0.005)
Observations 1260 1260 1260 1260
R2
0.254 0.251 0.251 0.251
Adjusted R2
0.217 0.213 0.214 0.213
Note: n
p o 0.1; nn
p o 0.05; nnn
p o 0.01.
31
The prior for the inverse-Wishart degrees of freedom parameter r is set at k
number of countries (k¼60). The prior for the hyperparameter R1Â1 is estimated
using a variation of the range method. In this case, R1Â1 is set to
ðlogitð:95ÞÀlogitð:05ÞÞ=9
À Á
¼ 0:654.
32
This is transformed to the Logit scale to a mean of À5:29 with a variance of
logitð:01ÞÀlogitð:001Þ=2
À Á2
¼ 1:34.
33
Guriev et al. (2011) also estimate a logistic model, the values for m and v are
the coefficient estimates and inflated standard errors (obtained from reported 95%
CIs, and downweighted to represent 1% as many observations as the data; this
specification is relaxed in sensitivity analysis).
34
This assumes a normal distribution with mean zero and variance
1
2ðlogitð:95ÞÀlogitð:05ÞÞ=ðxH ÀxLÞ
À Á2
.
P. Mahdavi / Energy Policy 75 (2014) 228–243 241
References
Adelman, M., 1993. The Economics of Petroleum Supply. MIT Press, Cambridge.
Alnasrawi, A., 2002. Iraq's Burdens: Oil, Sanctions, and Underdevelopment. Greenwood
Press, Westport, CT.
Aslaksen, S., 2010. Oil and democracy: more than a cross-country correlation?.
J. Peace Res. 47, 421–431.
Berrios, R., Marak, A., Morgenstern, S., 2011. Explaining hydrocarbon nationalization
in Latin America: economics and political ideology. Rev. Int. Polit. Econ. 18,
673–697.
Boix, C., 1997. Privatizing the public business sector in the eighties: economic
performance, partisan responses, and divided governments. Br. J. Polit. Sci..
Box, G.E.P., Tiao, G.C., 1973. Bayesian Inference in Statistical Analysis. AddisonWesley,
Reading, MA.
Bremmer, I., Johnston, R., 2009. The rise and fall of resource nationalism. Survival
51 (2), 149–158.
Chang, R., Hevia, C., Loayza, N., 2010. Privatization and Nationalization Cycles. NBER
Working Paper.
Cheibub, J.A., Gandhi, J., Vreeland, J.R., 2010. Democracy and dictatorship revisited.
Publ. Choice 143, 67–101.
Cohn, G., 1968. Petroleo e nacionalismo. Sao Paolo.
Dunning, T., 2008. Crude Democracy. Cambridge University Press, New York, NY.
Gandhi, J., 2008. Political Institutions under Dictatorship. Cambridge University
Press, New York, NY.
Gelman, A., Carlin, J.B., Stern, H.S., Dunson, D.B., Vehtari, A., Rubin, D.B., 2013. Bayesian
Data Analysis. Chapman & Hall/CRC Texts in Statistical Science, 3rd edition.
Golden, M.A., Mahdavi, P., 2015. The institutional components of political corruption.
In: Gandhi, J., Ruiz-Rufino, R. (Eds.), Routledge Handbook of Comparative
Political Institutions. Routledge, New York, NY.
Grayson, L., 1981. National Oil Companies. Wiley Press, New York, NY.
Guriev, S., Kolotilin, A., Sonin, K., 2011. Determinants of nationalization in the oil
sector: a theory and evidence from panel data. J. Law Econ. Organ. 27 (2), 301–323.
Haber, S., Menaldo, V., 2011. Do natural resources fuel authoritarianism? A
reappraisal of the resource curse. Am. Polit. Sci. Rev. 105 (1), 1–26.
Hansohm, D., 2007. Oil and foreign aid: curse or blessing? The case of Sudan. In:
Wohlmuth, K. (Ed.), Reconstructing Economic Governance after Conflict in
Resource-rich African Countries. Learning from Country Experiences. Institute
for World Economics and International Management, Berlin, pp. 117–150.
Hartley, P., Medlock, K.B., 2008. A model of the operation and development of a
national oil company. Energy Econ. 30 (5), 2459–2485.
Heckman, J., 1977. Linear probability models of the demand for attributes with an
empirical application to estimating the preferences of legislators. Rand J. Econ.
28, 142–189.
Henisz, W., 2000. The institutional environment for multinational investment.
J. Law Econ. Organ. 16 (2), 334–364.
Henisz, W.J., 2004. Political institutions and policy volatility. Econ. Polit. 16 (1), 1–27.
Horrace, W.C., Oaxaca, R.L., 2006. Results on the bias and inconsistency of ordinary
least squares for the linear probability model. Econ. Lett. 90 (3), 321–327.
Karl, T.L., 1997. The Paradox of Plenty: Oil Booms and Petro-States. University of
California Press, Berkeley, CA.
Kobrin, S.J., 1984. Expropriation as an attempt to control foreign firms in LDCs:
trends from 1960 to 1979. Int. Stud. Q. 28, 329–348.
Table 8
Additional regression results, controlling for differing lags in the diffusion effect, calculated as cumulative sums. For example, NOCc2 refers to the cumulative count of
nationalizations within the previous two years. Compare to main results from Model 1, Table 5. All models are run using the linear probability model specification with
country fixed effects.
Independent Variable Dependent variable:
NOC nationalization
(1) (2) (3) (4)
Years producing 0.045 0.043 0.044 0.043
(0.028) (0.028) (0.028) (0.028)
Oil price shock 0.024nnn
0.024nnn
0.024nnn
0.024nnn
(0.005) (0.005) (0.005) (0.005)
Regime age À0.005 À0.006 À0.006 À0.006
(years) (0.013) (0.013) (0.013) (0.013)
Regime index À0.025n
À0.026nn
À0.026nn
À0.027nn
(polity) (0.013) (0.013) (0.013) (0.013)
OPEC dummy 0.030nnn
0.030nnn
0.030nnn
0.031nnn
(0.006) (0.006) (0.006) (0.006)
GDP/capita (log) 0.010 0.011 0.010 0.011
(0.023) (0.023) (0.023) (0.024)
GDP growth 0.009n
0.009n
0.010n
0.010n
(percent) (0.005) (0.005) (0.005) (0.005)
Change in 0.007 0.007 0.007 0.007
oil production (0.004) (0.005) (0.005) (0.005)
NOCc2 0.007nnn
(0.003)
NOCc3 0.004n
(0.002)
NOCc4 0.003n
(0.002)
NOCc5 0.002
(0.001)
Observations 1260 1260 1260 1260
R2
0.255 0.253 0.253 0.252
Adjusted R2
0.218 0.215 0.215 0.215
Note: n
p o 0.1; nn
p o 0.05; nnn
p o 0.01.
P. Mahdavi / Energy Policy 75 (2014) 228–243242
Kobrin, S.J., 1985. Diffusion as an explanation of oil nationalization: or the domino
effect rides again. J. Confl. Resolut. 29, 3–33.
Kretzschmar, G., Kirchner, A., Sharifzyanova, L., 2010. Resource nationalism: limits
to foreign direct investment. Energy J. 31 (2), 27–52.
Kuhn, A.K., 1951. Nationalization of foreign-owned property in its impact on
international law. Am. J. Int. Law 45 (4), 709–712.
Lenczowski, G., 1949. Russia and the West in Iran. Cornell University Press, New York.
Luong, P.J., Weinthal, E., 2010. Oil Is Not a Curse: Ownership Structure and
Institutions in Soviet Successor States. Cambridge University Press, New York.
Maddison, A., 2007. Contours of the World Economy 1-2030 AD. Oxford University
Press, Oxford, UK.
Mahdavi, P., 2012. Oil, monarchy, revolution, and theocracy: a study on the National
Iranian Oil Company (NIOC). In: Thurber, M.C., Hults, D., Victor, D.G. (Eds.), Oil
and Governance. Cambridge University Press, Cambridge, UK.
Makki, H., 1950. Kitabe Siah (The Black Book). Elmi Press, Tehran.
Manzano, O., Monaldi, F., 2009. The political economy of oil production in latin
America. Economia 9, 59–98.
Marcel, V., 2006. Oil Titans: National Oil Companies in the Middle East. Brookings
Institution Press, Baltimore, MD.
Marshall, M.G., Jaggers, K., 2006. Political Regime Characteristics and Transitions
1800-2004. Center for International Conflict Management & Development;
http://www.systemicpeace.org/polity/polity4.htm.
McPherson, C., 2010. State participation in the natural resource sectors: evolution,
issues and outlook. In: Daniel, P., Keen, M., McPherson, C. (Eds.), The Taxation
of Petroleum and Minerals: Principles, Problems, and Practice. Routledge,
New York, pp. 263–288, chapter 9.
Mikdashi, Z., 1966. A Financial Analysis of Middle Eastern Oil Concessions: 1901–65.
Praeger Publishers, New York.
Mu'assasat al-Naqd al-Arabi al Saudi, S.A.M.A., 1960. Annual report. Jeddah: Research
and Statistics Dept., Saudi Arabian Monetary Agency, Kingdom of Saudi Arabia.
Nikogosian, V., Veith, T., 2012. The impact of ownership on price-setting in retailenergy
markets—the German case. Energy Policy 41 (0), 161–172, Modeling
Transport (Energy) Demand and Policies.
Nolan, P.A., Thurber, M.C., 2010. On the state's choice of oil company: risk management
and the frontier of the petroleum industry. PESD Working Paper, 99.
Noori, H.S.-H., 1965. A Study on the Nationalization of the Oil Industry in Iran (Ph.D.
thesis), Colorado State College.
Park, T.W., Abolfathi, F., Ward, M., 1976. Resource nationalism in the foreign policy
behavior of oil exporting countries, 1947–1974. Int. Interact. 2 (4), 247–262.
Philip, G., 1982. Oil and Politics in Latin America. Cambridge University Press, New
York, NY.
Philip, K., 2002. The quest for rural enterprise support strategies that work.
Small Enterp. Dev. 13 (1), 13–25.
Ross, M.L., 2012. The Oil Curse: How Petroleum Wealth Shapes the Development of
Nations. Princeton University Press, Princeton, NJ.
Singh, C., 1989. Multinationals the State and the Management of Economic
Nationalism. Praeger Publishers, New York, NY.
Smith, B., 2007. Hard Times in the Land of Plenty: Oil Politics in Iran and Indonesia.
Cornell University Press, Ithaca, NY.
Solberg, C.E., 1979. Oil and Nationalism in Argentina: A History. Stanford University
Press, Stanford, CA.
Stevens, P., 2007. Investing in Oil in the Middle East and North Africa: Institutions,
Incentives and the National Oil Companies. Technical Report, ESMAP, The
World Bank.
Stevens, P., 2008. Resource nationalism and the role of national oil companies in the
middle east: history and prospects. J. World Energy Law Bus. 1 (1).
Tordo, S., Tracy, B., Arfaa, N., 2011. National oil companies and value creation. World
Bank Working Paper 218, The World Bank.
Tsebelis, G., 2002. Veto Players: How Political Institutions Work. Princeton
University Press, Princeton, NJ.
Vernon, R., 1971. Sovereignty at Bay: The Spread of U.S. Enterprises. Basic Publishing,
New York.
Victor, D.G., 2013. National oil companies and the future of the oil industry. Annu.
Rev. Resour. Econ. 5, 445–462.
Victor, D.G., Hults, D., Thurber, M.C. (Eds.), 2012. Oil and Governance: State-owned
Enterprises and the World Energy Supply. Cambridge University Press,
Cambridge, UK.
Vivoda, V., 2009. Resource nationalism, bargaining and international oil companies:
challenges and change in the new millennium. New Polit. Econ. 14 (4), 517–534.
Warshaw, C., 2012. The political economy of expropriation and privatization of
national oil companies. In: Victor, D.G., Thurber, M.C., Hults, D. (Eds.), Oil and
Governance: State-owned Enterprises and the World Energy Supply.
Cambridge University Press, Cambridge.
Weiss, R.E., 2005. Modeling Longitudinal Data. Springer, New York, N.Y..
Wolf, C., 2009. Does ownership matter? The performance and efficiency of state oil
vs. private oil (1987–2006). Energy Policy 37 (7), 2642–2652.
Yergin, D., 1991. The Prize. Simon and Schuster, New York, NY.
Yergin, D., Stanislaw, J., 1998. The Commanding Heights: The Battle for the World
Economy. Simon and Schuster, New York, NY.
P. Mahdavi / Energy Policy 75 (2014) 228–243 243