Cold War Geopolitics and the Making of the Oil Curse

Abstract

Does oil hinder democracy? The prevailing wisdom holds that, since 1980, oil has hindered democracy by enabling oil wealth to flow to state-owned oil companies, breaking the fiscal contract with society, and endowing oil-rich regimes with means to invest in repression and accommodation. However, these arguments do not account for system-level factors that might affect the oil-democracy relationship. I argue that a structural break in the oil-democracy relationship occurred at the end of the Cold War when the United States and the Soviet Union reduced support for both oil-poor and oil-rich authoritarian regimes in the developing world. The rollback of support facilitated post–Cold War democratization of the resource-poor regimes, while oil-rich regimes were better positioned to stave off pressures to democratize. Based on a re-analysis of two prominent studies, I find the oil curse to be a post–Cold War phenomenon, with negative consequences for democracy of a magnitude roughly 80 percent larger than previously estimated. I further explore these dynamics via comparative case studies of Azerbaijan and Georgia. The evidence shows that the oil curse is a function of geopolitical dynamics, not just international market conditions.

oil, geopolitics, Cold War, resource curse, Georgia, Azerbaijan

Introduction

Geopolitical dynamics have not loomed large in the literature on the oil curse—the purported negative relationship between oil and democracy—despite oil being the world's most widely traded commodity and a strategic resource necessary for the projection of military power in the modern era. Recent contributions acknowledge certain systemic factors, such as the role of price movements and successive oil shocks in global markets, but are comparatively silent on the role major powers might play in either amplifying or mitigating the oil curse. This oversight is curious. Oil appeared to play a significant role in the strategic calculations of major powers during the twentieth century and especially the Cold War. Major powers backed oil-rich autocracies, such as the Gulf States, Angola, and the Republic of Congo, and subverted democratic processes in others, such as the overthrow of Mosaddegh in Iran. Thus, I investigate the role the Cold War may have played in shaping domestic politics in oil-exporting countries.

The literature identifies temporal breaks in the oil-democracy relationship but usually attributes them to price effects and changes in resource ownership (Ahmadov 2014; Andersen and Ross 2014; Lall 2016). While previous studies place the temporal break in the oil-democracy relationship as either in 1980 or the 2000s, I argue the end of the Cold War constitutes a more significant break. My argument revolves around the fiscal contract theory of democracy, in which both internal and external pressures to democratize can be obviated with access to resource rents. During the Cold War, geopolitical competition between the United States and the Soviet Union led to extensive programs of overt and covert political and economic support for both oil-rich and oil-poor authoritarian governments, providing those governments with unearned sources of revenue. With the end of the Cold War, these assistance programs were significantly diminished and in some cases reversed, replaced by more conditional or even avowedly democracy-promoting ones. Regimes with oil revenue were less affected by this system-level shock and thus comparatively better positioned to weather post–Cold War pressures for democratization. The democracy-suppressing effects of oil only emerged once the more general democracy-suppressing effects of Cold War geopolitical competition ended and post–Cold War democracy promotion began.

I support my argument with a mixed-methods approach. Basing my quantitative analysis on replications of Haber and Menaldo (2011) and Andersen and Ross (2014), I find that, while there is a significant negative relationship between oil wealth and democracy post-1980, the effect is much larger for the post–Cold War (1990-) era. Furthermore, the negative effect of oil on democracy post–Cold War is roughly 80 percent larger than that previously estimated by Andersen and Ross (2014). During the Cold War, oil wealth is either not associated with democracy or has a modest positive relationship. This finding suggests that geopolitical context—and not just international market conditions—plays a strong role in conditioning the oil-democracy relationship. I bolster these large-N findings with comparative case studies of Azerbaijan and Georgia, two former Soviet republics that have followed very different democratic trajectories since independence. Azerbaijan is oil-rich and highly autocratic. Georgia is not and has democratized.

These results have significant implications beyond the oil-democracy link. First, they point to mechanisms by which hegemonic shocks—periods of general upheaval in the international system brought on by the sudden rise or decline of major powers—affect prospects for democratization and explain why some countries are more resilient to these hegemonic shocks than others (Gunitsky 2014). Second, they illuminate one of the ways the end of the Cold War affected democratization outside of the immediate environs of the former Soviet Union and Eastern Europe (Boix 2011). An accumulating body of evidence shows that the end of the Cold War represents a structural break not only in the effectiveness of democracy promotion aid, but also the decline of insurgency relative to other forms of intrastate war (Dunning 2004; Bearce and Tirone 2010; Bermeo 2016; Kalyvas and Balcells 2010). My findings add to this literature by demonstrating the end of the Cold War is responsible for the emergence of the oil curse as well.

The remainder of this article proceeds as follows. I begin by reviewing the relevant literature on the oil-curse debate, placing particular emphasis on influential studies by Haber and Menaldo (2011) and Andersen and Ross (2014). I then provide a brief overview of relevant historical developments during the Cold War, which grounds the theoretical argument I develop in the following section. I then conduct the re-analysis, which includes discussion of the data, estimation strategy, and results. The next section presents the case studies of Azerbaijan and Georgia. I conclude by revisiting the conventional wisdom on the oil curse in light of my findings.

An Oil Curse or Blessing?

There is considerable evidence that, at least since the early 1980s, oil wealth has inhibited democracy, although disentangling the specific influence of oil is confounded by the fact that, in cross-country data, oil wealth is collinear with other influences that may inhibit democracy (Ross 2001; Aslaksen 2010; Ahmadov 2014; Andersen and Ross 2014). Yet, Haber and Menaldo (2011) strongly dispute the existence of an oil curse. Their estimates of the impact of oil income per capita on democracy employ fixed effects, leveraging within-country variation to assess a causal relationship. Using data covering more than two centuries (1800–2006), they find no evidence of an oil curse and some evidence of a modest oil blessing. Building on these findings, Menaldo (2016) argues that the cross-national correlations between oil wealth and authoritarianism are due to factors such as weak rule of law and authoritarian institutions in earlier periods that both cause authoritarian rule and exploration for and reliance on oil (Menaldo 2016). Reliance on oil and authoritarian institutions are comorbid but not causally related, with both resulting from an “institutions curse” at earlier stages of political development. In a similar vein, Brooks and Kurtz (2016) argue that levels of resource dependence and democracy are likely affected by earlier attempts at state-led industrial development that in some cases led to the formation of capitalist and middle classes that augured for democratization. They use instrumental variables techniques and recover evidence of a modest oil wealth blessing for the period 1970–2009, with more oil-rich countries being more democratic.¹

The importance of initial, prediscovery conditions to whether oil wealth hinders economic development has been explored in some detail, with most studies emphasizing the importance of domestic political and economic institutions. If these institutions are strong and the size of the mineral sector does not dwarf the rest of the economy, resource wealth provides additional capital for productive investment in industrial development and human capital. Under these circumstances, resource income is growth- and democracy-promoting, and the curse becomes a blessing (Brunnschweiler and Bulte 2008; Alexeev and Conrad 2009; Jones Luong and Weinthal 2010; Brooks and Kurtz 2016). However, Menaldo shows that weaker states—states with less government revenue and less investor-friendly institutions—are more likely to undertake energy exploration and discover larger amounts of oil, either through state-owned enterprises or in partnerships with private firms. Thus, most-likely cases for the emergence of resource curse dynamics are those most likely to look for oil in the first place (Menaldo 2016, chap. 5).

Andersen and Ross (2014) counter by arguing that the positive relationship between oil wealth and democracy is confined to the pre-1980 period. Since then, oil is negatively related to democracy. In this argument, they place greater emphasis on the international context in which oil is produced and traded. They argue a structural break in the oil-democracy relationship occurred during the early 1980s. This break reflected an important shift in both the volume of oil wealth and the way it is managed. Before the 1970s, oil rents were relatively small and accrued primarily to the private sector. Moreover, exploitation in developing and middle-income countries—the countries most associated with the oil curse—was generally conducted by large, Western multinationals with headquarters in major powers like the United Kingdom, the United States, and France.

Both realities—low prices and predominately Western ownership—began changing in the 1970s. The 1973 oil embargo more than doubled prices, and the Iranian Revolution of 1979 sent prices even higher.² As prices rose, nationalizations became more frequent; between 1970 and 1980, governments in developing countries asserted direct control over their oil reserves. Governments that did not nationalize renegotiated contracts with multinationals on more favorable terms (Ross 2012; Andersen and Ross 2014). By 2010, only four of the world's twenty largest oil companies (BP, ExxonMobil, Lukoil, and Royal Dutch Shell) were fully private corporations. Their total reserves amounted to just 6.4 percent of reserves held by the top twenty, whereas national oil companies held 92.3 percent.³ This transformation in ownership has meant the unprecedented high oil prices of the past forty years have been captured as resource rents by governments, a significant windfall, even if prices have waned in the past several years.

Recent quantitative work mostly substantiates the idea of a structural break occurring in the 1980s. Andersen and Ross (2014) use a variety of time-series estimators to demonstrate that post-1980 oil wealth has been associated with negative changes in democracy. These changes become more pronounced using longer lag structures. Andersen and Ross (2014) show a modest effect of total oil income (TOI) per capita on democracy; increasing TOI per capita by one standard deviation—roughly equivalent to the difference between Denmark and Gabon in 2009—is associated with a small (2.2 percent) diminution of democracy.^4,5 Using a different estimation strategy, Aslaksen (2010) finds a 10 percent increase in oil revenue as a percentage of gross domestic product (GDP) is associated with a 7.4 percent long-run decrease in a country's democracy score.⁶ In a thorough meta-analysis, Ahmadov (2014) finds a general negative correlation between oil wealth and democracy, but one that is partially mediated by a host of factors: (1) oil wealth appears to suppress democracy in the Middle East and North Africa region and sub-Saharan Africa, but to be democracy-promoting in Latin America; (2) oil wealth was democracy-promoting in the 1960s, but democracy-curtailing in the 2000s; and (3) the resource curse is most pronounced when prediscovery regime attributes are not modeled, suggesting that, in regressions that do not control for prior regime type, “the oil variable may be picking up the bad effect of the previous political institutions” (Ahmadov 2014, 1259). Lall (2016) investigates these dynamics in light of problems of missing data and finds, via multiple imputation, additional support for the post-1980 structural break.

Geopolitical considerations are largely absent from these studies despite the geostrategic significance attributed to oil resources during the Cold War by historians (Yergin 1991; Blum 2003; Little 2008; Stern 2016). These scholars highlight US and Soviet concerns about access to strategic resources such as oil, which caused them and their respective allies to subvert democracy in resource-rich countries. A brief historical review of the period makes clear that geopolitics must factor into a complete account of the oil curse.

The Geopolitics of Oil During the Cold War

Oil has been the dominant source of global energy since overtaking coal in the 1960s and, owing to its portability, is an especially crucial input for industrial economies and the projection of military force (BP 2015).⁷ Securing access to oil and stabilizing prices have been central to the foreign policies of major powers since at least the 1920s.⁸ Other than the United States and the Soviet Union/Russia, all of the mid-twentieth-century major powers (Germany, France, the United Kingdom, Italy, Japan, and China) and current permanent members of the United Nations (UN) Security Council are highly oil import-dependent.⁹ The ability of these major powers to fuel their economies and wage war continues to be and has been dependent on access to crude oil sourced from abroad since the early twentieth century.

Historically, major-power access to oil was secured via a close relationship between major power governments and private firms—British Petroleum (UK), French Petroleum Company/Total (France), and Standard Oil's various successors (Amoco, Chevron, ExxonMobil USA) —that operated in partnership with local governments (Price-Smith 2015). Because of information asymmetries between the major-power-backed oil companies and host-country governments (and collusive deals between the two), these partnerships often failed to produce tangible benefits for host-country nonelites. These circumstances allowed populist politicians like Mohammed Mosaddegh in Iran and Abd al-Karim Qasim in Iraq to exploit popular rancor and promote plans to either renegotiate with foreign oil companies or nationalize their assets completely.

As with any investment in or reliance on specific assets, these multinationals faced extraordinarily high costs to adjusting to host-country policy changes and thus had strong incentives to meddle in their domestic politics (Alt et al. 1996). These incentives were perhaps strongest during the Cold War, as both the United States and the Soviet Union developed extensive ties with governments in resource-exporting countries to secure access to the raw materials that would be necessary for any sustained war effort. Oil may have been the most obvious strategic resource, but Cold War–era powers also targeted bauxite and copper, among other industrial metals (Gendron, Ingulstad, and Storli 2013).

This meddling in resource-rich countries was overwhelmingly antidemocratic. From Allende in copper-exporting Chile to Lumumba in copper- and uranium-rich Zaire (now the Democratic Republic of the Congo [DRC]) and Mosaddegh in oil-rich Iran, the historical record is littered with US and Western involvement in or support of coups against elected leaders whose political platforms and policies threatened Western—primarily US and UK—economic interests in strategic natural resources (Blum 2003). In each case, democratic institutions were adversely affected. Despite rhetorical commitment to supporting democracy, during the Cold War the United States tolerated and often supported authoritarian governments (Pinochet in Chile, Mobutu in Zaire/DRC, and the Shah in Iran) that did not threaten the economic status quo vis-à-vis their country's natural resource wealth specifically and their core economic institutions and policies more generally (Kirkpatrick 1982).

US policymakers and their allies in Europe were obviously alarmed and concerned at the prospect of oil nationalization. Domestically, the US and European governments experienced significant domestic political pressure from major oil firms and separately feared that these resources would fall under communist influence or at least outside of Western control. These concerns were amplified by successive “peak oil” forecasts in the 1950s and 1970s that predicted a future of zero-sum competition over rapidly dwindling supplies. These predictions heightened US policymakers’ beliefs that the Middle East's energy reserves were vital to national security (Stern 2016). The Eisenhower and Carter doctrines reflected the tenor of the period, committing the United States to use military force, if necessary, to deter the spread of communism and protect its interests in the Persian Gulf from “outside force” (Carter 1980).

For its part, the Soviet Union supported communist revolutions and national self-determination movements in oil-rich Angola and the Republic of Congo. In both cases the toppled governments were nondemocratic.¹⁰ Perhaps owing to its exporter status, the Soviet Union's relations with energy exporters were generally less fraught. The Soviet Union largely supported moves to nationalize oil wealth in the non-Western world and happily free rode on the price manipulation arising out of the 1970s Arab oil embargoes and the Organization of Petroleum Exporting Countries' (OPEC) collusive behavior (Chubin 1980). More generally, Soviet intervention favored avowedly communist parties that subverted competitive elections in favor of one-party rule.¹¹

During the Cold War, covert and overt major-power intervention took various forms that defy easy operationalization and quantification (Kalyvas and Balcells 2010; Boix 2011). Some observational data, however, can be parsed. Data on Cold War–era covert interventions¹² by both the United States and the Soviet Union support the notion that both superpowers targeted and supported more authoritarian governments than democratic ones (Berger et al. 2013). Authoritarian regimes held power during 72 percent of country-years with active Cold War–era influence by the United States or the Soviet Union—years when either country covertly or overtly intervened to “either install a new leader or to provide support to an existing leader to help maintain the power of the regime” (Berger et al. 2013, 866). Both the United States and the Soviet Union focused significantly more effort on supporting authoritarian leaders than democratic ones.¹³

Given these examples, we might expect an amplification of the oil curse during the Cold War as major powers sought to limit the policy autonomy of oil-exporting countries by subverting democratic elections and institutions therein. This interpretation is consistent with the received wisdom regarding the role of Cold War geopolitics in resource-rich countries. In the next section, however, I present a revisionist interpretation that brings into question the conventional account of the role of geopolitics in the oil curse.

A Revisionist Interpretation

This section forwards a revisionist interpretation. In addition to meddling in resource-rich authoritarian countries, Cold War dynamics led the United States, the Soviet Union, and their allies to provide extensive overt and covert political and economic support for comparatively resource-poor authoritarian governments. This support provided those governments with both unearned sources of revenue and enhanced repressive capacity that helped them stave off democratization pressures. With the end of the Cold War, the United States and the Soviet Union significantly decreased the size of these assistance programs, even to the point of reversal, and replaced them with more conditional or even avowedly democracy-promoting programs. Regimes with oil revenue were less affected by this system-level shock and thus were comparatively better positioned to weather post–Cold War pressures to democratize, resulting in a negative association between oil wealth and democracy in the post–Cold War era.

While the preceding discussion highlights some instances of antidemocratic meddling by major powers in oil-exporting countries during the Cold War, these interventions were part of a general policy of subverting democracy via material support. The United States provided support for anticommunist authoritarian governments, and the Soviet Union supported communist authoritarian governments (Carothers 1991). Several resource-rich countries were significant fronts during the Cold War (Indonesia, Angola, Iran), but comparatively resource-poor countries were as well (Afghanistan, Ethiopia, Korea, Nicaragua, and Vietnam, among others). US and Soviet influence in the developing world was not confined to oil-rich countries and, on balance, was more targeted toward comparatively resource-poor regimes.

In accordance with the fiscal contract theory of the state, material support for resource-poor authoritarian regimes during the Cold War—both covert and overt—should have had predictable consequences for democratization in those countries. Fiscal contract theory holds that governments that are reliant on their citizens for revenue have general incentives to defer to those citizens’ preferences over policy and process (Bates and Lein 1985; Ross 2001). The existence of a plentiful, windfall natural resource like oil wealth exerts downward pressure on the need to tax society at large as governments are able to finance themselves through the monopolization or taxation of these industries. This in turn reduces citizens’ incentives to monitor and critique incumbent regimes (Paler 2013).¹⁴ Because these resources generate substantial revenue, oil-rich states tend not to tax their populations as much. As a result, their populations do not have this source of leverage with which to bargain for a greater say in governing. The fiscal contract linking government behavior to citizens’ policy preferences and representative institutions does not develop (Chaudhry 1989). Additionally, oil wealth allows states to invest in large systems of patronage and means of coercion that can further insulate them from pressures to democratize (Ross 2001, 2012). Finally, oil wealth provides countries with significant autonomy in their international affairs. Oil exporters have little trouble finding markets for their products and as such have reduced incentives to join international organizations that might constrain their behavior. A plentiful resource sector obviates the need to reform governance institutions in order to secure foreign investment and/or development assistance. The effect is to make oil exporters less susceptible to economic inducements to cooperate with or defer to the policy preferences of other states (Ross and Voeten 2016). This effect is distinct from but consistent with the logic of fiscal contract theory, which operates through economic leverage.

During the Cold War, the United States, the Soviet Union, and their respective allies provided overt and covert financial and military support for anticommunist and communist dictatorships and intervened to influence democratic elections (Gaddis 1986; Boix 2011; Bermeo 2016; Levin 2016).¹⁵ Much like oil income, these external sources of revenue and repressive capacity insulated these regimes from pressures to democratize and make policy concessions to domestic constituencies. Even US aid that was specifically tied to democratic reform during the Cold War was unlikely to be withdrawn in the absence of reform due to fears that such withdrawals would lead anticommunist authoritarian regimes to be replaced by communist ones (Dunning 2004).

With some caveats, this support and/or meddling was disproportionately targeted at oil-poor regimes. Looking at US development assistance during the Cold War—another form of unearned income that can sustain otherwise resource-strapped authoritarian governments—reveals a negative relationship between oil income per capita and US development assistance (see Table 1), with few truly resource-rich (oil income per capita > one thousand dollars) regimes receiving any substantial development assistance.¹⁶ Indeed, several wealthy oil exporting countries—most notably Saudi Arabia—contributed to at least a dozen US-backed anticommunist insurgencies or covert operations, including those in Afghanistan, El Salvador, and North Yemen.¹⁷

Table 1.

Oil income and US official development assistance per capita, 1946–1989

Oil income per capita	N	Mean	75th percentile
$0	2,591	$16.34	$3.54
$0.01–$999.99	1,965	$13.61	$0.35
≥$1,000	311	$1.82	$0.00

Source: Haber and Menaldo 2011, Miller 2017.

A similar pattern emerges regarding US and Soviet interventions. Regimes experiencing either US or Soviet intervention were, on average, roughly one-third as oil-wealthy ($280 vs. $780, t = 3.80) as those that did not.¹⁸ Available data on US development assistance and US and Soviet covert interventions suggest oil-rich regimes were less likely to be targets for these types of influence than resource-poor authoritarian regimes. These data do not constitute conclusive evidence, but they are consistent with the notion that comparatively resource-poor countries did benefit from substantial US and Soviet support during the Cold War.

As the Cold War wound down in the 1980s, but especially after the collapse of the Soviet Union in 1991, three changes occurred that affected material support for resource-poor authoritarian regimes. First, the total volume of US and Western aid to some regions—particularly Central America, Northern Africa, and Southern Asia—declined. This occurred in response to a changed geostrategic environment and, perhaps relatedly, because aid reductions were a policy objective of the newly elected Republican Congress of the United States. Second, aid was increasingly conditional, with donors freer to withdraw aid if democratic reforms were not forthcoming or reversed (Dunning 2004; Bearce and Tirone 2010; Bermeo 2016). Third, the Soviet Union's support for communist authoritarian regimes evaporated as the Soviet Union collapsed. These newly independent states undertook their own processes of democratization with considerable Western support. Specifically, an increasing share of US assistance commitments (up to 23 percent by 1994, per Tierney et al. [2011]) flowed to the mostly former communist countries of Eastern Europe by the mid-1990s. Support for the newly independent states of the former Soviet Union and the newly democratized former Warsaw Pact countries was part of a general pattern of post–Cold War democracy promotion (Gleditsch and Ward 2006).

Thus, the end of the Cold War resulted in a large retrenchment in external resources—both overt and covert—to all authoritarian governments previously supported by the United States and the Soviet Union. According to fiscal contract logic, this should have resulted in movements toward democracy in these countries as their governments became more reliant on domestic sources of revenue and coercive capacity. The effects of post–Cold War major-power retrenchment, however, would be most pronounced for those governments that did not have continuing access to oil revenues that could otherwise sustain authoritarian systems of cooptation and coercion. Thus, I expect that, after the Cold War, the effect of oil wealth on democracy would be more pronounced as non-oil-rich regimes began democratizing. This expectation would be consistent with the observation that oil wealth does not undermine democracy so much as it preserves authoritarian rule, while simultaneously offering an alternate explanation for the timing of the oil curse (Smith 2004; Morrison 2009; Bueno de Mesquita and Smith 2010).

H1:Conditional on the Cold War period, oil wealth is unrelated to democracy. Before and after the Cold War, oil wealth is negatively associated with democracy.

Data, Estimation, and Results

I begin the quantitative section of this study with a re-analysis of Andersen and Ross (2014), which itself builds on a re-analysis of Haber and Menaldo's (2011) highly influential study. My sample consists of 163 countries for the period 1800–2006. The core dependent variable is the change in the country's normalized (0–100) revised combined Polity score, which for this sample ranges from –95 to 80, with a mean of 0.35 and a standard deviation of 8.29.¹⁹ The Polity score is preferable to binary indicators of democracy/autocracy because it captures the effects of partial movements toward and away from democracy such as the imposition/removal of restrictions on particular political parties or checks on executive authority (Marshall and Jaggers 2002). Thus, it better captures both incremental democratization, such as in Mexico and South Korea in the 1980s, and authoritarian backsliding, such as in Venezuela and Russia in more recent years.

I use TOI as the main independent variable, measured in thousands of constant 2007 US dollars per capita.²⁰ Values vary widely based on population and volume of production. For example, Nigeria is endowed with more than double Norway's reserves, but median oil income per capita ($208, in 2000) was only a fraction of much less populous Norway ($9,096). The modal value is zero, with only one-third of observations registering any oil income at all. This variable is interacted with indicator variables for the post-1980,²¹ post-1990,²² and Cold War period to capture temporal breaks in the oil-democracy relationship.

Why not model overt and covert support—in the form of aid, arms transfers, military commitments, and diplomatic support—directly? Temporal breaks rather than proxies are used for overt and covert support for three reasons. First, the data on US and Soviet covert support (Berger et al. 2013) likely suffer from measurement error and were only coded for the Cold War period, even though the United States and Russia continued to influence domestic politics abroad after 1991. For example, the United States was involved in both covert and overt attempts to remove Saddam Hussein from power in Iraq following the First Gulf War, and Russia's current involvement in Ukraine represents an attempt to undermine a democratic transition there (Washington Post 2003). Moreover, because interventions were both covert and only coded as a dummy variable, there is little information available to parse cases where intervention was limited and where it was more encompassing. Second, while data on official development and official military assistance are available for the United States, the United Kingdom, and several of their more important Cold War allies (Tierney et al. 2011), these data are not available for the Soviet Union or its Eastern European allies and proxies. This would introduce significant and systematic bias into attempts to force the Cold War effect to operate through aid proxies. Even if such data were available, it is well-known that US and Soviet intervention and financial support often flowed through and from intermediaries, as in the Saudi example earlier. Thus, calculating the volume of support that could reasonably be attributed to the United States and the Soviet Union would at best provide highly unreliable estimates. Third, earlier research using temporal breaks to mark the Cold War has yielded important insights on the impacts of Cold War dynamics on democratization and civil conflict (Kalyvas and Balcells 2010; Boix 2011).

In this analysis, I define the Cold War era as the period 1946–1990. Although the Soviet Union did not collapse until 1991, 1990 witnessed three landmark events that signaled the end of militarized competition between the North Atlantic Treaty Organization (NATO) and the Warsaw Pact. First, the parties to the Two Plus Four talks signed the Treaty on the Final Settlement with Respect to Germany. Second, and, perhaps more importantly, on November 19, 1990 the then sixteen NATO and then six Warsaw Pact members signed the Treaty on Conventional Armed Forces in Europe, which established comprehensive limits on military equipment and personnel and oversaw the destruction of excess arms. Third, in 1990 the United States’ and Soviet Union's proxy war in Yemen ended, which signaled the temporary termination of US-Soviet competition for position and influence in the developing world.

All models include control variables common to both Haber and Menaldo (2011) and Andersen and Ross (2014) (1) per capita income (logged), (2) the incidence of civil war, and (3) regional and world trends in democratization. Following Haber and Menaldo (2011) and Andersen and Ross (2014), I use error correction models (ECMs) with country and year fixed-effects and Drischoll-Kraay standard errors as my core specification. ECMs are appropriate in situations where longer-term levels of independent variables affect the dependent variable, but there may also be shorter-term transitory effects that should be modeled, as well.²³ The ECM framework allows me to estimate both short- and longer-term effects and is suitable for cointegrated data. The coefficients on the uninteracted oil wealth terms represent the effect of oil wealth during the period not captured by the relevant indicator (post-1980, post-1990, or Cold War). To assess the impact of oil wealth over the various periods, one must calculate the conditional slope as an additive function of the coefficients on oil wealth and the interaction terms.

Model 1 replicates Haber and Menaldo (2011) core result, and Model 2 replicates that of Andersen and Ross (2014). In both instances, the results match those reported in the respective articles. Model 3 extends Andersen and Ross (2014) by including the indicator variable for the post-1980 period, resulting in a positive coefficient on the indicator but unchanged coefficient estimates for TOI_t–1 and TOI_t–1 × post-1980.²⁴ Models 4–6 investigate whether the oil curse (or blessing) is a Cold War/post–Cold War phenomenon. Model 4 includes two more terms, TOI_t–1 × post-1990 and post-1990. Post-1990 is simply an indicator for the period 1991–2006.

As in Model 2, in Model 4 TOI is positively correlated with changes in Polity for the pre-1980 period, but negatively correlated with changes in Polity in the post-1980 period. The coefficient on TOI_t–1 × post-1990 is negative, statistically significant, and substantively large. For both the period 1980–1990 and 1991–2006, oil income per capita is negatively associated with changes in Polity, but the effect is more than twice as large for the post-1990 (β post-1980 = –0.095, β post-1990 = –0.169, cumulative effect = –0.264) period. Moreover, the coefficient on Δ TOI × post-1990 is negative and significant, indicating that short-run increases in oil income per capita are associated with democratic retrenchment. Neither Δ TOI or Δ TOI × post-1980 are significant in this specification. Beginning in 1991, both levels of oil income and changes in oil income are negatively associated with democracy.

Models 5 and 6 provide direct tests of the Cold War hypothesis, providing additional evidence in favor of the revisionist interpretation: the oil curse emerges in the post–Cold War period. Models 5 and 6 include interactions between oil income per capita and the indicator for the Cold War era. Model 5 is estimated using the full sample (1800–2006) while Model 6 only includes the post–WWII era (1946–2006). The uninteracted coefficient on TOI_t–1 is negative and statistically significant, indicating a negative effect of oil wealth on democracy after the Cold War. The coefficients on TOI_t–1 × Cold War are positive, significant, and larger than the uninteracted coefficients in both specifications, indicating that, for the Cold War period, oil wealth is positively associated with democracy. Moreover, there is no evidence for an oil curse or blessing prior to 1946. Model 7 restricts the analysis to the pre-1946 period. Though coefficients on TOI_t–1 and Δ TOI are both large and positive, the confidence intervals are large as well. The negative relationship between oil and democracy in the non–Cold War period is thus driven by post–Cold War (1991–2006) observations.

The substantive effects (based on Model 6) are presented in Figure 1. During the Cold War period (light gray lines), TOI is modestly and positively associated with changes in the Polity score, indicating a small prodemocracy effect. However, after the Cold War (the black lines), the relationship is negative and much stronger. Comparing coefficients between the replicated findings of Andersen and Ross (2014) (Model 2) and Model 6, we see the post–Cold War effect is 80 percent larger than the post-1980 antidemocratic effect of oil estimated by Andersen and Ross (2014). Since the end of the Cold War, oil has been even more antidemocratic than we had previously thought.

Figure 1.

Marginal effects of total oil income per capita on changes in democracy, Cold War, and post–Cold War eraNotes: Polity scores have been normalized to 0–100. Dashed lines are 95 percent confidence intervals.

Despite the robustness of these findings to different estimators and lag structures,²⁵ some might worry the results are driven by (1) the coincidence of the end of the Cold War with the 1980s oil glut, (2) highly influential cases, particularly the Gulf Cooperation Council (GCC) countries,²⁶ or (3) are overly dependent on the inclusion of post-Soviet successor states. The mid-1980s to mid-1990s were marked by both significant democratization (the third wave) and comparatively low oil prices. The end of the Cold War may be coincidental with but independent of the oil glut, and the results may be due to the effect of the former rather than the latter. Second, it may be worrisome if the results are dependent on the inclusion of the Arabian Peninsula cases, which have been a central focus of US foreign policy—especially external security support, via the Eisenhower and Carter doctrines—since the 1950s (Tètreault 1991; Stern 2016). Third, the collapse of the Soviet Union created fifteen successor states, several of which had significant oil endowments (Azerbaijan, Kazakhstan, Russia, and Turkmenistan) and many of which returned to authoritarian rule following a brief period of Western-influenced democratization in the immediate post–Cold War period. In these cases, post–Cold War democratic retrenchment may be more attributable to the persistent influence of Soviet-era institutions (Pop-Eleches 2007).

To address these potential issues, Table 3 replicates the models in Table 2 with three subsets of the data excluded from the sample: Models 8–9 exclude the years 1985–1993, thereby excluding the first years of the oil glut and the years surrounding the end of the Cold War;²⁷ Models 10–11 exclude the countries of the GCC; and Models 12–13 exclude the fifteen successor states to the Soviet Union.²⁸ Point estimates for the relevant variables and interactions are substantively and statistically similar to the main findings. This allows me to conclude that the negative relationship between oil and democracy is a post–Cold War phenomenon.²⁹ The findings are an artifact neither of the immediate pre- and post-end of Cold War cases nor of the GCC or former Soviet Union cases.

Table 2.

Error correction models for the impact of total oil income (TOI) on Polity score (country and year fixed effects)

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity
	Sample:	Sample:	Sample:	Sample:	Sample:	Sample:	Sample:
Variables	full	full	full	full	full	post-1945	pre-1946
Polityt–1	–0.087***	–0.087***	–0.087***	–0.087***	–0.087***	–0.115***	–0.080***
	(0.008)	(0.007)	(0.007)	(0.007)	(0.007)	(0.011)	(0.014)
TOIt–1	0.055**	0.048*	0.048*	0.032	–0.185*	–0.208*	2.477
	(0.019)	(0.024)	(0.024)	(0.024)	(0.072)	(0.081)	(2.107)
TOIt–1 × post-1980		–0.147**	–0.147**	–0.095*
		(0.052)	(0.052)	(0.045)
TOIt–1 × post-1990				–0.169**
				(0.064)
TOIt–1 × Cold War					0.220***	0.264***
					(0.059)	(0.067)
Δ TOI	–0.020	–0.059*	–0.059*	–0.048*	–0.182*	–0.201*	3.852
	(0.021)	(0.027)	(0.027)	(0.023)	(0.084)	(0.085)	(3.490)
Δ TOI × post-1980		–0.048	–0.048	–0.008
		(0.036)	(0.036)	(0.031)
Δ TOI × post-1990				–0.125
				(0.081)
Δ TOI × Cold War (1946–1990)					0.160	0.189*
					(0.086)	(0.088)
Post-1980			5.703***	6.285***
			(0.471)	(0.467)
Post-1990				–0.527***
				(0.110)
Cold War (1946–1990)					3.030***	5.806***
					(0.236)	(1.087)
log per capita incomet–1	–0.279	–0.273	–0.273	–0.268	–0.271	–0.856*	0.061
	(0.319)	(0.317)	(0.317)	(0.317)	(0.317)	(0.345)	(1.394)
Civil War t–1	0.065	0.057	0.057	0.063	0.070	0.157	–0.774
	(0.448)	(0.448)	(0.448)	(0.448)	(0.448)	(0.473)	(1.289)
Regional democratic diffusion t–1	0.025***	0.025***	0.025***	0.024***	0.025***	0.038***	0.024
	(0.007)	(0.007)	(0.007)	(0.007)	(0.007)	(0.010)	(0.015)
World democratic diffusion t–1	0.058*	0.059*	0.059*	0.060*	0.004	0.358***	0.226
	(0.029)	(0.029)	(0.029)	(0.029)	(0.023)	(0.076)	(0.464)
Δ log per capita income	1.289	1.199	1.199	1.253	1.334	1.874	–1.499
	(1.734)	(1.736)	(1.736)	(1.736)	(1.729)	(2.197)	(3.162)
Δ Regional democratic diffusion	0.375***	0.375***	0.375***	0.375***	0.375***	0.408***	0.168***
	(0.070)	(0.070)	(0.070)	(0.070)	(0.070)	(0.077)	(0.027)
Δ World democratic diffusion	–0.277*	–0.277*	–0.277*	–0.278*	0.211**	0.420***	–0.137
	(0.109)	(0.109)	(0.109)	(0.109)	(0.071)	(0.096)	(0.597)
Constant	6.989**	6.959**	6.959**	6.945**	7.287**	0.000	0.000
	(2.299)	(2.292)	(2.292)	(2.290)	(2.324)	(0.000)	(0.000)
Observations	10,195	10,195	10,195	10,195	10,195	7,439	2,756
R-squared	0.098	0.098	0.098	0.098	0.098	0.118	0.088
Number of groups	163	163	163	163	163	163	47
Country FE	yes	yes	yes	yes	yes	yes	yes
Time FE	yes	yes	yes	yes	yes	yes	yes

Notes: (1) Drischoll-Kraay standard errors in parentheses. (2) Statistical significance:

***p < 0.001, **p < 0.01, *p < 0.05.

Table 3.

Probing robustness for the impact of total oil income (TOI) on democracy (country and year fixed effects)

	(8)	(9)	(10)	(11)	(12)	(13)
	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity	DV: Δ Polity
	Excluding	Excluding	Excluding	Excluding	Exc. former	Exc. former
Variables	1985–1993	1985–1993	GCC	GCC	Soviet Union	Soviet Union
Polityt–1	–0.081***	–0.081***	–0.087***	–0.087***	–0.086***	–0.086***
	(0.007)	(0.007)	(0.008)	(0.008)	(0.008)	(0.008)
TOIt–1	0.013	–0.235**	0.314**	–0.313**	0.032	–0.183*
	(0.023)	(0.078)	(0.107)	(0.119)	(0.025)	(0.076)
TOIt–1 × post-1980	–0.041		–0.508**		–0.097*
	(0.027)		(0.153)		(0.046)
TOIt–1 × post-1990	–0.218***		–0.129		–0.164*
	(0.059)		(0.148)		(0.068)
TOIt–1 × Cold War		0.250***		0.521***		0.216***
		(0.063)		(0.109)		(0.063)
Δ TOI	–0.046*	–0.138	–0.161	–0.160	–0.048*	–0.201*
	(0.021)	(0.113)	(0.109)	(0.158)	(0.023)	(0.084)
Δ TOI × post-1980	0.015		–0.344**		–0.010
	(0.019)		(0.109)		(0.032)
Δ TOI × post-1990	–0.128		0.274		–0.142
	(0.115)		(0.193)		(0.081)
Δ TOI × Cold War		0.100		0.099		0.178*
		(0.115)		(0.164)		(0.085)
Post-1980 dummy	2.627***		1.544***		6.273***
	(0.515)		(0.461)		(0.475)
Post-1990 dummy	0.926*		4.014***		-0.679***
	(0.449)		(0.181)		(0.092)
Cold War dummy		3.143***		3.076***		3.073***
		(0.239)		(0.251)		(0.245)
log per capita incomet–1	–0.295	–0.296	–0.355	–0.341	–0.283	–0.286
	(0.348)	(0.348)	(0.342)	(0.343)	(0.324)	(0.325)
Civil wart–1	0.191	0.194	0.047	0.059	0.002	0.010
	(0.422)	(0.422)	(0.459)	(0.459)	(0.464)	(0.465)
Regional democratic diffusion t–1	0.026***	0.026***	0.024**	0.024**	0.023**	0.023**
	(0.007)	(0.007)	(0.007)	(0.007)	(0.008)	(0.008)
World democratic diffusion t–1	0.045	–0.005	0.066*	0.007	0.063*	0.004
	(0.030)	(0.023)	(0.031)	(0.024)	(0.030)	(0.024)
Δ log per capita income	1.014	1.045	1.211	1.267	1.004	1.090
	(1.898)	(1.894)	(1.853)	(1.851)	(1.830)	(1.823)
Δ Regional democratic diffusion	0.257***	0.257***	0.376***	0.376***	0.378***	0.378***
	(0.026)	(0.026)	(0.072)	(0.072)	(0.072)	(0.072)
Δ World democratic diffusion	–0.110	0.327***	–0.290*	0.213**	–0.288*	0.222**
	(0.066)	(0.030)	(0.112)	(0.073)	(0.113)	(0.073)
Constant	6.859**	7.156**	7.551**	7.784**	6.964**	7.321**
	(2.455)	(2.490)	(2.473)	(2.507)	(2.330)	(2.364)
Observations	8,920	8,920	9,935	9,935	9,931	9,931
R-squared	0.081	0.081	0.099	0.099	0.096	0.096
Number of groups	163	163	157	157	149	149
Country FE	yes	yes	yes	yes	yes	yes
Time FE	yes	yes	yes	yes	yes	yes

Notes: (1) Drischoll-Kraay standard errors in parentheses. (2) Statistical significance:

***p < 0.001, **p < 0.01, *p < 0.05.

Investigating the Mechanisms: Azerbaijan and Georgia

The quantitative results strongly support the assertion that the oil curse is a post–Cold War phenomenon. These results, however, are aggregate and do not provide direct insight into the mechanism at work: that the end of the Cold War resulted in a massive scaling back of US and Soviet support for both oil-rich and oil-poor autocracies, leaving oil-rich regimes comparatively better resourced to stave off pressures to democratize. In this section, I present case studies of Azerbaijan and Georgia that provide additional support for the purported mechanism. Both are small former Soviet republics in the Caucasus that were governed by similar political and economic institutions from the 1920s until 1991, and they thus faced similar institutional preconditions in the immediate post-independence era, as well as similar levels of economic development and urbanization, two key factors in explaining democratization (Ross 2001, see Table 4). Both are characterized by dominant ethnic groups (85–92 percent of the population) and have complex, at times conflictual relations with neighboring Russia. Both also have restive ethnic minority enclaves that have fought for independence, resulting in civil conflicts of similar magnitudes in the early 1990s (Nagorno Karabakh in Azerbaijan, Abkhazia and South Ossetia in Georgia).³⁰

Table 4.

Comparing Georgia and Azerbaijan

Variable	Georgia	Azerbaijan
Population (1991)	4.8 million	7.3 million
GDP per capita (1991), 2015 USD	$1,315	$1,209
Pop. share, dominant ethnic group	86.6	91.6
% urban (1991)	54.8	53.4
Absorbed by USSR	1921	1920
Independence declared	April 9, 1991	October 18, 1991
Date of first post-ind. pres. election	May 26, 1991	June 7, 1992
Dominant religion	Orthodox Christian (83%)	Shia Muslim (97%)
Separatist conflict deaths, 1991–1995	2,311	4,880
Avg. total oil income, 1991–2006	$4.60	$370
Max. total oil income, 1991–2006	$9.45	$1,090

As the Soviet Union collapsed in the early 1990s, both countries declared independence following national referendums and both held democratic elections within a year. In terms of demographics, world region, extent of influence by a major power during the Cold War (total, as both were republics in the Soviet Union), conflict history, and the specific timing of initial democratization, the two countries constitute a logical basis for a most-similar systems paired analysis (Bennett and Elman 2007). The principal differences between the two are the majority religion—Islam in Azerbaijan, Orthodox Christianity in Georgia—and oil wealth. Azerbaijan is one of the world's wealthier oil states, with a mean oil income per capita of $370 for the period 1991–2006. In 2006, Azerbaijan's oil income per capita roughly equaled that of Russia. Oil exports provide the majority of government revenue (74 percent in 2011) and export earnings (95 percent). In Georgia, on the other hand, oil income is a small proportion of national income, amounting to less than ten dollars per capita even in the most lucrative year in the sample.

For these cases to be consistent with my argument, I expect the see the following dynamics:

1. Major-power (Soviet Union) support for authoritarian institutions and rulers during the Cold War;

2. A retrenchment of major-power (Soviet Union and later Russia) support for authoritarian regimes after the Cold War;

3. Attempts by Western governments to promote democracy therein, via democracy promotion assistance and/or public statements, after the Cold War;

4. In oil-poor Georgia, financial pressures leading to a collapse of authoritarian patronage networks and repressive capacity, resulting in popular pressures to democratize; and

5. In oil-rich Azerbaijan, a lack of financial pressures leading to a continuation of patronage networks and maintenance of repressive capacity based on oil rents.

The first two points barely need elaboration. Both Azerbaijan and Georgia were forcibly incorporated into the Soviet Union in the early 1920s and ruled from Moscow via incorporation and extension of patronage networks that partially predated Soviet rule (Bunce and Wolchik 2006a). Available budget data indicate both Azerbaijan and Georgia were net recipients of intra-USSR fiscal transfers, receiving more money from the Soviet Union than they contributed via direct budget transfers or distribution of underpriced goods such as oil (Orlowski 1992). As the Soviet Union collapsed, its commitment to supporting authoritarian institutions in Azerbaijan and Georgia obviously waned. Intra–Soviet Union fiscal transfers ceased, as did Soviet military commitments to support the governments of both republics. While Russian troops were involved in both Azerbaijan's and Georgia's early 1990s separatist conflicts, their involvement was primarily focused on ensuring de facto independence for Russian enclaves and did not seek to depose the central governments of either.

Western efforts to influence both countries in the post-Soviet era centered upon democracy promotion. Postcommunist countries were a “clear priority” for United States Agency for International Development's democracy assistance programming (Bunce and Wolchik 2006b, 12). In the 1990s, the United States and its allies, both individually and via the World Bank, contributed more than $100 million to government and civil society reform in Georgia and $60 million in Azerbaijan (Tierney et al 2011). Much of this assistance was targeted at developing civil society and nongovernmental organizations, though the effect of this assistance seems to have cut both ways. While it did contribute to better organized and trained prodemocracy activists and opposition parties, it also fueled perceptions that these prodemocracy movements were puppets of Western forces (Mendelson 2002; Stewart 2009).

Initial attempts at democratization in both countries were both stymied by coups. In oil-rich Azerbaijan, democratically elected president Abulfaz Elchibey was overthrown in a coup that brought Heydar Aliyev, father of current president Ilham Aliyev and a former Soviet official, to power. The Aliyev family has ruled since, maintaining control of the government through extensive patronage networks, electoral fraud, and opposition intimidation. Since 1995, Azerbaijan has been considered a consolidated autocracy, with a Polity score of –6 from 1995 to 1997 and a score of –7 from 1998 to the present (see Figure 2).

Figure 2.

Regime trends in Azerbaijan and Georgia, 1991–2015Source: Marshall and Jaggers (2002)

In contrast, Georgia's trajectory has been one of more successful democratization. Following independence, Georgia's first democratically elected leader, Zviad Gamsakhurdia, was deposed in 1991. A new constitution was adopted in 1994, and Eduard Shevardnadze, another former Soviet elite, was elected president. Like Aliyev, Shevardnadze attempted to consolidate power and extend his rule. Unlike Aliyev, he was unsuccessful. In 2003, following troubled legislative elections, mass nonviolent protests culminating in the Rose Revolution, and international condemnation, Shevardnadze resigned. The supreme court annulled the electoral results and new elections were scheduled for 2004. Though far from perfect, checks and balances and mass nonviolent mobilization in the Georgian context prevented a former Soviet-era elite from consolidating personalist rule. Since 2004, Georgia has been considered a democratic regime, with Polity scores ranging from 6 to 7 (Figure 2).

Why did Aliyev succeed in reconsolidating power where Shevardnadze failed? One plausible answer is oil income allowed Aliyev to continue to support patronage networks established during the Soviet era (Jones Luong and Weinthal 2001). Azerbaijan began its independence deeply dependent on oil exports, essentially the country's only source of revenue, but also in need of infrastructure modernization to be competitive on the global market (Jones Luong and Weinthal 2001). Instead of privatizing Azerbaijan's oil industry, the government entered into partnerships between the State Oil Company of Azerbaijan Republic (SOCAR) and Western oil firms, circumventing Azerbaijan's nascent business and investor class and providing the Aliyev regime with significant autonomy vis-à-vis Azerbaijani society (Jones Luong and Weithal 2010). The resulting oil rents have allowed successive Azerbaijani leaders to resist democratic pressures through patronage spending and repression (Bunce and Wolchik 2006a; Guliyev 2013). Moreover, these oil resources have significantly curtailed Western governments’ ability to promote democratization there, as there were comparatively few economic carrots or sticks Western powers could use (Bunce and Wolchik 2006a).³¹ In Azerbaijan, massive oil wealth has obviated the need for the Azerbaijani government to defer to the policy and institutional preferences of the society it governs and provided the means for it to resist international attempts to promote democratization.

In contrast, Georgia does not have significant oil wealth. Though oil makes up a significant share of exports, this oil merely transits Georgia from Azerbaijan to the Black Sea via pipelines. Georgia's economy is much more diversified—with major exports including copper ore, nuts, and chemicals—and generates significantly fewer government rents. At the end of the Cold War, Georgia's government was comparatively rent-poor, requiring it to depend much more on the resources of the governed and conditional Western aid.

These joint dependences—on the local economy and taxpayers, and on Western donors and financial institutions—shaped the trajectory of the Rose Revolution in three key ways. First, shortfalls in domestic revenue generation contributed to the economic crisis that spurred initial protests. Beginning in 1998, domestic tax receipts began falling short of projections due in part to weakness in tax collection and to an economic downturn in Russia, Georgia's main trading partner (Papava 2006). These budget shortfalls resulted in unpaid government salaries and pensions, fueling antiregime sentiment.

Second, unlike Azerbaijan, whose primary export is traded on a global market, Georgia's economy was more susceptible to regional demand shocks in its major trading partners and more dependent on fickle foreign investors. After the Russian financial crisis in 1998, Georgia saw foreign direct investment begin to dry up, exports shrink, and current account deficits grow, further exacerbating economic grievances (Papava 2006). Meanwhile, though Azerbaijan temporarily suspended currency trading and Russia temporarily suspended oil imports therefrom due to inability to pay, the Azerbaijani economy was not similarly affected. Georgia's economic growth contracted from 10 percent in 1997 to 3 percent in 1998, while Azerbaijan's growth rate nearly doubled (6 percent to 10 percent).

Third, these economic woes left the Shevardnadze government increasingly at the mercy of foreign donors—principally the United States, after the International Monetary Fund suspended its loan programs in 2002—who used this leverage as a means of encouraging a negotiated settlement to the mass protests and new elections (Papava 2006). Additionally, private Western donors, including George Soros, provided democracy assistance funds and bankrolled several prodemocracy organizations and the training of nonviolent activists abroad (Fairbanks 2004). The Shevardnadze government had little economic wherewithal to sustain itself against domestic activists and foreign pressure.

Aside from oil wealth, what other factors might plausibly explain differences in outcomes across the two cases? The most significant structural difference between the two countries is the majority religion: Islam in Azerbaijan, Orthodox Christianity in Georgia. Djankov and Hauck (2016) argue religious cultures have shaped the post-Soviet trajectories of democratization in the former Soviet bloc. They group the twenty-nine postcommunist countries according to dominant religious traditions (Protestant-Catholic, Orthodox, and Islam) and demonstrate majority Protestant-Catholic states have democratized the most, majority-Muslim states the least, and majority-Orthodox states have democratized partially.

While this competing political-cultural explanation cannot be completely dismissed, there are reasons to doubt it is a persuasive counterargument. First, Azerbaijan is perhaps the most secular majority-Muslim country: a recent Pew Research Center survey of Muslims in thirty-nine countries revealed Azerbaijan as having the least support for Sharia law (only 8 percent of respondents) and the least support for religious leaders having some or large influence on politics (14 percent) of any country in the sample (Lugo et al. 2013). Second, Azerbaijan and Georgia have, at least according to commonly used measures, similar political-social value systems, appearing adjacent to one another in the Inglehart-Welzel cultural map in terms of traditional versus secular/rational values and survival versus self-expression values considered a precondition for democratic consolidation (Inglehart and Welzel 2005).³² On these bases, it is difficult to accept the premise that such divergent democratic outcomes in Azerbaijan and Georgia would be attributable to Islam's effect on Azerbaijani political culture. Indeed, one of the lasting legacies of Soviet rule was to diminish the political relevance of religious identity—Christian and Muslim alike (Dragadze 1993).

Another difference is the degree of foreign intervention: Georgia was invaded by Russia, Azerbaijan was not. In 2008, Russian forces quickly routed Georgian armed forces, established military bases in Abkhazia and South Ossetia, and expelled ethnic Georgians from both territories. Yet this difference is not a plausible cause of democratic divergence in the two cases for several reasons. First, Russian intent in the invasion of Georgia was not to alter its political institutions but rather establish territorial control over regions with large ethnic Russian minorities. If anything, this kind of external security threat—neighboring a major power with irredentist ambitions regarding its territory—biases countries against democratizing (Gibler and Sewell 2006). Second, foreign interventions tend not to alter political institutions in countries that were already democratic, as Georgia was in 2008 (Pickering and Kisangani 2006). Thus, neither religion nor foreign intervention provide compelling alternative explanations for differing democratic trajectories across the two cases.

While dynamics in both countries obviously were more complicated than can be fully detailed here, the post-independence trajectories of Azerbaijan and Georgia are consistent with my argument. The preceding discussion provides support for the five observable implications of the theory elaborated earlier in this section: oil wealth allowed Azerbaijan's post-Soviet elites to reconsolidate authoritarian rule, while Georgian elites, not similarly endowed with rents, lacked the economic resources to stave off democratization.

Discussion

This article presents strong evidence, both quantitative and case-based, for a negative but temporally bounded association between oil and democracy. During the Cold War period, oil had either a null or a weakly positive association with democracy. Since the end of the Cold War, the association has been negative, significant, and significantly larger than previously estimated. Obviously, this is not the first study to identify a temporal break in the relationship between oil and democracy. It is the first, however, to argue the break had more to do with global geopolitics and the constellation of interests in the international system than with the post-oil shock price environment or national control of oil rents. The analysis presented here suggests the strongest negative effects of oil on democracy did not emerge until the early 1990s, which is roughly a decade later than the shifts in oil prices and ownership discussed by Andersen and Ross (2014). I argue the timing of the oil curse is a function of the recession of the United States and the Soviet Union from antidemocratic meddling due to the waning of the Cold War. The results might be consistent with other competing causal mechanisms. However, two prominent alternatives can probably be ruled out: higher oil incomes in the post–Cold War era and a greater degree of state ownership of the oil sector.

High prices and therefore higher oil incomes are unlikely culprits. One of the main explanations offered for the 1980 structural break is that it came on the heels of a sea change in the price environment, with successive price shocks sending prices and therefore oil incomes to unprecedented levels. However, the 1990s and early 2000s were characterized by comparatively low oil incomes, with average levels hovering around pre-1973 levels (Table 5). The 1990s saw the lowest oil incomes per capita in the postwar era. If prices in the preceding decade are the most significant determinant of the oil curse, the 1990 break must be reconciled with the fact that oil incomes per capita during the 1980s were less than half of what they were during the 1970s. Moreover, the immediate post–Cold War era witnessed a slight retrenchment of the state's role as owner of oil resources, with privatization occurring most markedly in the former Soviet Union. If the oil curse accelerated in the post–Cold War environment, it did so despite comparatively low prices and a reduced role for state ownership of oil resources (Jones Luong and Weinthal 2010).³³

Table 5.

Oil income per capita (1000s constant USD) by decade, 1951–2006

Period	N	Average oil income PC	Standard dev.	Low	High
1951–1960	951	0.53	3.34	0.00	34.41
1961–1970	1282	0.42	2.40	0.00	25.96
1971–1980	1391	1.33	6.71	0.00	78.59
1981–1990	1421	0.66	2.99	0.00	48.52
1991–2000	1613	0.32	1.39	0.00	14.21
2001–2006	982	0.57	2.04	0.00	18.11

Alternatively, perhaps the effects of oil income in the 1970s and 1980s, operating through domestic spending and repression effects, did not become manifest until the 1990s. However, this possibility seems unlikely. Even for the states of the GCC, which combine massive oil and fiscal reserves with relatively small populations, oil price fluctuations cause large year-to-year changes in government spending, and fiscal break-even points track closely with oil prices (El Anshasy and Bradley 2012; Aissaoui 2013). Additionally, it might be the case that the global increases in trade, investment, and financialization that followed the Cold War allowed the effects of oil income in previous periods to compound via reinvestment and diversification, making oil-rich regimes less dependent on current oil income over time. There is some evidence consistent with this conjecture. Figure 3 plots the annual bivariate correlation between oil income per capita (logged) and Haber and Menaldo's fiscal reliance variable, which is the percentage of government revenues from oil, gas, or minerals for eighteen resource-dependent economies from 1960 to 2006 (2011, 4).³⁴ The correlation was, on average, much stronger for the Cold War (1960–1990) period than the post–Cold War period (0.77 vs. 0.44, t = 10.82), indicating lessened fiscal reliance on resource rents over time. Direct resource rents are accounting for a smaller share of government expenditures, even as oil, gas, and mineral exports continue to make up a significant share of exports and GDP in each case. Future work should investigate this potential mechanism in more detail.

Figure 3.

Annual correlation between total oil income (logged) and fiscal reliance for 18 resource-rich countries over time Notes: Fiscal reliance is the share of government expenditure coming from oil, natural gas, or minerals. Source: Haber and Menaldo (2011)

I find weak evidence of a conditional oil blessing: during the Cold War, oil wealth is associated with increases in—or at least no retrenchment of—democracy. Though this conditional correlation has been replicated and discussed, thus far the literature has been relatively silent on the mechanisms by which this blessing might operate. One might be a spending/redistribution effect, similar to that posited by Dunning (2008); because oil rents often accrue directly to the state, regimes can engage in social spending that is not redistributive and therefore not threatening to entrench economic interests. However, this would imply an unconditional effect, rather than one confined to the Cold War. It may be that oil-fueled social spending helped mollify some of the grievances that could be capitalized upon by communist parties, thus affecting both the likelihood of communist authoritarian takeover and, indirectly, of Western support for anticommunist authoritarian regimes. Further statistical tests and case narratives will be necessary to parse these mechanisms and are beyond the scope of this study.

Today, whether oil is a blessing or a curse to countries that discover it is a topic of significant importance. The early twenty-first century commodity boom, during which real prices for most globally traded commodities more than doubled, catalyzed a gold rush–like frenzy of exploration efforts, resulting in radical upward revisions of proven oil reserves not only in many legacy exporters (Iraq, Saudi Arabia, and Venezuela, but also the United States and Canada), but also in much smaller nonlegacy exporters. Countries across West Africa (Côte d'Ivoire, Ghana, and Liberia), East and Southern Africa (Kenya, Mozambique, and Uganda), and Southeast Asia (Cambodia, Myanmar, and Vietnam) have seen their proven oil reserves increase significantly.³⁵ The analysis presented here suggests these discoveries auger poorly for democratization in newly resource-rich authoritarian regimes.

Notably, this boom is occurring in the context of China's rise as a major energy consumer and military power. Klare (2009) argued that a resource scramble has supplanted the capitalist-communist ideological struggle as the main source of conflict in the post–Cold War era. This conflict is playing out both within resource-exporting states and as part of a larger pattern of geopolitical competition between the world's two largest importers—the United States and China—over access to the oil reserves of the Middle East, Africa, and Central Asia. To the extent the relationship between the United States and China assumes Cold War–like properties—a contentious claim in its own right—the oil-democracy relationship may be headed toward another structural break. Thus far, however, there is little evidence that Chinese oil imports are resulting in more cursed dynamics in the sense that they cause authoritarian retrenchment or the curtailing of civil liberties (Bader and Daxecker 2015). Also, as Klare (2017) recently argued, emerging alternative fuel technologies and the shale revolution may reduce the geostrategic significance of oil as global markets enter a period of seemingly perpetual glut. Still, China's extensive engagement around extractive industries in general, and oil in particular, is already altering the geopolitical landscape in Africa, but its broader effects warrant further study (Brautigam 2009; Hendrix and Noland 2014).

Extant scholarship on the link between oil and democracy treats the relationship as an almost exclusively domestic phenomenon, with international dynamics only appearing as either drivers of prices—and thus rents—or as control variables. That is, scholarship on the oil-democracy link has treated oil as if it were simply a large source of unearned income. It has not accounted for systemic influences on patterns of democratization nor the way major powers have sought to either promote or inhibit democratization at different times. In doing so, research on the oil curse has played a variant of the “reductionist gamble,” analyzing the domestic politics of oil largely in isolation from broader international dynamics (Oatley 2011, 311). Recent scholarship has begun to uncover strong effects of oil wealth on interstate behavior, including conflict and participation in intergovernmental organizations (Colgan 2013; Hendrix 2017; Ross and Voeten 2016). Thus far, it has not investigated whether conditions in the international system—apart from prices in international markets—affect internal political dynamics. These findings indicate the effect of oil wealth on democracy is shaped not just by the international price environment, but geopolitical dynamics, as well.

Supplementary Information

Supplementary information is available at the Journal of Global Security Studies data archive.

Footnotes

1

However, evidence for this modest blessing only emerges in models using instrumental variables that are estimated on smaller sample sizes than their “naïve” (uninstrumented) models (with 13 percent to 16 percent of observations dropped). Brooks and Kurtz (2016) do not address the possibility of temporal breaks in their model. Their naïve models return a null effect.

2

Viewing the 1973 oil embargo as an exogenous source of increased revenue for non-Gulf states—states for whom the embargo was an external shock—Liou and Musgrave (2014) find little evidence this large resource windfall hindered democratization.

3

Based on author's calculations from the Oil & Gas Journal Top 200/100.

4

Like many concepts in the social sciences, democracy can be measured a variety of different ways, with various operationalizations reflecting different theoretical constructions of the concept (Coppedge et al. 2011). Most contributions to this literature have relied on the Polity scale (Marshall and Jaggers 2002), which measures the competitiveness of executive recruitment, checks on executive authority, and the degree of political competition allowed in a state on a scale that ranges from –10 (unconstrained hereditary monarchies, like Qatar) to +10 (consolidated democracies, like the United Kingdom and Sweden). The focus on procedure, rather than on outcomes or degree of participation, has led to some criticism, but the Polity scale is the most widely used comparative indicator of democracy.

5

The percentage of democracy decline is calculated as the fall in Polity score over the sample mean Polity score of 45.7. If 1984 is used as the structural break point, the effect is a 10 percent diminution of democracy.

6

Aslaksen departs from the norm by operationalizing democracy using the Freedom House Political Rights Index.

7

Following the example of the British navy, which transitioned to oil-powered ships in the early twentieth century, all major naval powers had primarily oil-powered fleets by WWII.

8

Prior to the oil conversions, similar concerns around coal shaped major-power foreign policies, though most major powers could source coal domestically (Schulman 2015).

9

The United States’ import dependence through the latter half of the twentieth century was due both to burgeoning demand and strategic choice on the part of US policymakers. The Foreign Petroleum Policy of the United States (United States State Department 1944) encouraged a shift from domestic production and export to conservation of Western Hemisphere petroleum and the supplying of demand domestically and in Europe via exports from the Middle East. Following the oil shocks of the 1970s, the United States enacted the Energy Policy and Conservation Act of 1975 and the Export Administration Act of 1979, both of which significantly curtailed the ability of US oil producers to export their products. Due in large part to the shale revolution, US policymakers are revisiting the ban on exports, beginning with an August 2015 decision to engage in crude swaps with Mexico.

10

Angola won independence from a leftist military caretaker government that oversaw Portugal's transition to democracy; the Communist Congolese Party of Labor ousted military dictator Alfred Raoul in the Republic of Congo.

11

In practice, communist regimes often allowed multiple parties (for example, East Germany's National Front, which was composed of five different political parties), though a central party authority controlled these parties.

12

Examples include efforts to either install and support a new government such as US involvement in Mobutu's ouster of Patrice Lumumba in Zaire (DRC) or support for an existing government such as US support for South Korea under Syngman Rhee.

13

Those country-years with Polity scores ≤ –6.

14

Paler (2013) details an experiment in which Indonesian citizens responded to prompts about their tax contributions with increased effort in monitoring regional government actions and anti-incumbent sanctions. Conversely, prompts about windfalls—either natural resource revenue or central government transfers to regional governments—did not elicit a similar response.

15

Bermeo (2016) assesses the differential impacts of aid and oil income on democratic transitions in the post–Cold War era among aid recipient countries and finds evidence of a negative post–Cold War association between oil wealth and democratic transition. However, her results should not be interpreted as speaking directly to the oil-democracy link: “It would be a mistake to draw lessons from this analysis for the debate regarding the relationship between oil revenue and democratization. For example, see Ross 2012; and Haber and Menaldo 2011. This analysis is restricted to aid recipient countries because aid is the primary variable of interest . . . This makes it inappropriate to draw lessons for the overall relationship between oil revenue and democratic change from this group of countries; oil has made some countries rich and therefore ineligible for official development assistance” (Bermeo 2016, n58).

16

These cases included some Gulf states (pre-revolution Iran and Oman, and Saudi Arabia in some years) but also then long-standing democracies Trinidad and Tobago and Venezuela. Temporary support for two sub-Saharan African countries, Republic of Congo and Gabon, was part of a strategy by the United States to court regional support for anticommunist forces in nearby Angola.

17

On Saudi support for Afghan mujahedeen, see Coll (2004). On Saudi support for North Yemen and Nicaragua, see New York Times (1987).

18

These calculations exclude countries for which the Eisenhower Doctrine's security guarantee was the only form of US or Soviet influence. For these cases, US influence consisted entirely of an implicit security guarantee, promulgated by President Dwight Eisenhower in 1957, to help Middle Eastern countries (Bahrain, Kuwait, Oman, Qatar, and United Arab Emirates) resist Communist influence. Berger et al. (2013) do not identify particular instances of influence exertion or intervention in their description of these cases, and other instances of implicit security guarantees, such as the Kennedy or Johnson Doctrines covering Latin America, were not coded similarly. The data exclude cases of ultimately unsuccessful intervention, such as US support for National Union for the Total Independence of Angola rebels. However, if interventions in oil-rich countries were numerous and were underrepresented in the data because they were less likely to have succeeded, this would constitute supportive evidence that oil-rich countries were systematically more impervious to external meddling.

19

Because all models include the lagged Polity score, using either Δ Polity or Polity does not alter the results, save for the coefficient on the lagged Polity score.

20

Calculated as “barrels produced, divided by population, multiplied by the real-world price, expressed in thousands of 2007 dollars” (Haber and Menaldo 2011, 5).

21

Taking a value of 0 for years 1800–1980, 1 for 1981–2006.

22

Taking a value of 0 for years 1800–1990, 1 for 1991–2006.

23

This specification corresponds to those reported in in Table 5, column 1 in Haber and Menaldo (2011),Table 2, columns 1 and 2 in Andersen and Ross (2014).

24

Models with interactions should also include all constituent terms of the interactions in their uninteracted form. Andersen and Ross omitted one of the constituent variables from their original analysis.

25

In the appendix, I explore whether these findings are recovered using five-year versus one-year lags and a more conventional dynamic fixed effects estimation strategy (Tables A1–A3; see supplementary files). In each case, the results are similar—and often of larger magnitude—than those reported here; oil is (weakly) positively associated with democracy during the Cold War period, but negatively associated with democracy in the post–Cold War period.

26

Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and United Arab Emirates.

27

Excluding these cases obviates also the question of whether the results are dependent on a particular year (1990 vs. 1991) being chosen as the end of the Cold War.

28

Armenia, Azerbaijan, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russia, Tajikistan, Turkmenistan, Ukraine, and Uzbekistan.

29

The TOI_t–1 × post-1990 coefficient is negative, though standard errors are large; joint significance test for TOI_t–1 × post-1990 and TOI_t–1 × post-1980, however, is significant (p < 0.01).

30

Total battle deaths in Azerbaijan amounted to 0.07 percent of the 1991 population; in Georgia, the share was 0.05 percent. Calculations based on the Uppsala Conflict Database Battle-Related Deaths Dataset (Melander, Petterson, and Themnér 2016).

31

Natural Resource Governance Institute. “Azerbaijan Overview.” Accessed September 10, 2017. http://www.resourcegovernance.org/countries/eurasia/azerbaijan/overview

32

Bunce and Wolchik make this point bluntly: “It is far easier for the United States to support democratic change in Georgia, for example, than in Azerbaijan, Russia, Uzbekistan, or Kazakhstan. Part of the reason is oil, and the other part is the geopolitical importance of these countries in the war on terror” (2006a, 300). While the antiterror issue is a potential confounder, that element of US foreign policy does not begin in earnest until 2001. As such, the war on terror does not provide a compelling explanation for US ineffectiveness in stopping authoritarian reconsolidation in Azerbaijan, which occurred by 1998.

33

Whether the Inglehart and Welzel model of democratic consolidation is in fact accurate is irrelevant for the question of whether their measures assess the political cultures in the two countries to be similar.

34

Weinthal and Jones Luong (2006) analyze the privatization of the oil sector in Russia in the 1990s, noting that privatization had the effect of both making the Russian oil sector more competitive and establishing independent economic actors, like former Yukos head Mikhail Khodorkovsky, who could press the Russian government for continued economic reforms and greater political transparency. Khodorkovsky's subsequent prosecution and incarceration, the bankrupting of Yukos, and the reconsolidation of the Russian state's position in the Russian oil market (controlling half the country's crude output and 45 percent of domestic refining, according to the Fitch Group [2013]) show how privatization in the extractive sector can be undone.

35

Mexico, Venezuela, Ecuador, Trinidad and Tobago, Nigeria, Angola, Indonesia, Iran, Algeria, Bahrain, Equatorial Guinea, Gabon, Yemen, Oman, Kuwait, Norway, Chile, and Zambia.

Acknowledgements

The author thanks two anonymous reviewers, the Journal of Global Security Studies editors, and Maria Lotito for editorial assistance. The author is grateful for feedback on earlier drafts from panel participants at the 2015 American Political Science Association annual meeting and seminar attendees at Colorado School of Mines, McGill University, the Peterson Institute for International Economics, and the University of Amsterdam. Hussein Amery, C. Fred Bergsten, Ursula Daxecker, Caroline Freund, Simon Johnson, Stephan Haggard, Marcus Noland, and Krzysztof Pelc provided helpful comments or suggestions on earlier versions. Aleksandra Egorova and Jonathan Pinckney provided excellent research assistance. All errors are my own. This research was supported by the Smith-Richardson Foundation via grant no. 2008–7432.