1. Introduction
Comparative authoritarianism scholars often note that personalist regimes—those where power is concentrated in a single leader—are more dependent on natural resource wealth and international aid than their party-based or military counterparts (Callaghy Reference Callaghy1984; Acemoglu et al. Reference Acemoglu, Robinson and Verdier2004; Wright Reference Wright2008). This connection also underpins the prominent argument that personalist regimes are uniquely susceptible to collapse when economic challenges deplete their ability to finance patronage (Bratton and van de Walle Reference Bratton and van de walle1994; Geddes Reference Geddes1999; Escribà-Folch and Wright Reference Escribà -Folch and Wright2010).
Systematic cross-national evidence of this connection, however, is relatively limited. Wright (Reference Wright2008, 325) offers the clearest demonstration that personalist autocracies are more dependent on “unearned” income than other autocracies, though this point is a precursor to his larger argument that legislatures in such regimes have less impact on investment and economic growth. Beyond this, the connection between personal rule and external rents is largely a matter of faith (Guliyev Reference Guliyev2014).
Even more troubling, competing theories offer very different explanations as to why this connection exists. The studies cited above tend to argue that autocratic leaders use these rents to placate regime insiders through patronage, as opposed to the more challenging task of creating political parties and legislatures. In an early example, Chebabi and Linz (Reference Chebabi, Linz, Chehabi and Linz1998) argue that “easily exploitable natural resources whose production is in the hands of one or only a few enterprises with high profits” can provide the resources for personalist—or what they caused “sultanistic”—regimes (27). More recently, Wright et al. (Reference Wright, Frantz and Geddes2015) suggest that oil wealth facilitates co-optation by the regime, including “distributing more rents to discontented regime insiders, leaders of the opposition, and/or security forces” (289). These perspectives imply that such rents are an underlying cause of personalist authoritarian rule and that as these rents increase in size, so does the level of personalism in an autocracy.
By sharp contrast, other scholars argue that personalist rulers are more likely to extract and ultimately depend on these unearned revenues. Menaldo (Reference Menaldo2016) suggests that resource dependence is not exogenous but rather a purposeful choice when states lack the capacity to generate sufficient operating revenues through conventional taxation. Though his focus is on non-democratic regimes broadly, it seems likely that personalist regimes suffer from capacity deficits vis-à-vis other models of autocracy. Acemoglu and Robinson (Reference Acemoglu and Robinson2006) and Dunning (Reference Dunning2005) suggest more general versions of this relationship, where autocrats resist diversifying resource-dominant economies for fear of empowering societal actors who may mobilize against them. If such arguments are accurate, the above interpretation—that greater rents allow personalist rule to emerge and consolidate—reflects reverse causality.
Resolving the debate between these perspectives is critical. Personalism is expanding globally (Kendall-Taylor et al. Reference Kendall-Taylor, Frantz and Wright2017), so understanding its roots is increasingly important. Moreover, this focus may also resolve considerable gaps in the broader political resource curse literature, which typically argues that high levels of resource wealth (most often oil) are an impediment to democratization. Increasingly robust empirical evidence supports this conclusion, but the precise causal mechanism at work remains the subject of much debate.Footnote 1 Oil may finance societal benefits which reduce mass demands for democracy (Ross Reference Ross2012), co-opt regime insiders (Wright et al. Reference Wright, Frantz and Geddes2015), or pay for coercive tools to crush protests (Girod et al. Reference Girod, Stewart and Walters2018) to name only a few plausible linkages. Understanding the connection between oil income and personalism holds some promise to further clarify the political resource curse. For instance, personalist regimes are less likely to successfully transition to democracy than are other models of dictatorships (Chehabi and Linz Reference Chebabi, Linz, Chehabi and Linz1998; Frantz and Kendall-Taylor Reference Frantz and Kendall-Taylor2016). If oil income leads to personalism, this can clarify one causal pathway through which oil impedes democracy. The remainder of this note explores this in detail.
2. Personalism and oil: a look at the data
Prior scholarship on personalism relies overwhelmingly on the categorical measures of authoritarian regime type created by Geddes (Reference Geddes1999) and expanded by Geddes et al. (Reference Geddes, Wright and Frantz2014). Though influential, this approach leaves the core theoretical debate described above unresolved, since it is unable to capture the hypothesized changes in the level of personalism, no matter if it is a consequence or cause of increased external rents. This categorical approach is also problematic conceptually since personalism is a trait that exists in varying levels across all dictatorships (Weeks Reference Weeks2014, 38; Svolik Reference Svolik2012, 30).
Recent work from Geddes et al. (Reference Geddes, Wright and Frantz2018; see also Wright Reference Wright2017) addresses these shortcomings by developing a finely-grained measure of personalism emphasizing the degree to which political power is concentrated in the hands of individual leaders. As in their prior work, the unit of analysis is the autocratic regime-year from 1960 to 2010. Their specific measure uses eight observable indicators of personalism and an item-response theory approach to construct a latent variable measure for each autocratic regime-year in the data. In general, the indicators capture whether or not institutions exist that can check the leaders power, the degree to which leaders control appointments to political office, and whether leaders take steps to ensure the loyalty of the military and security forces. The IRT process transforms these indicators into a continuous variable, where higher values indicate a greater degree of personalism. I use this latent variable in the remainder of this paper.
Figure 1 presents an initial look at the relationship between personalism and the presence of resource rents, specifically oil. I focus on oil rents because of their unique size and the ease by which governments can appropriate them (Ross Reference Ross2012), which is a key theoretical underpinning of the arguments described above. The dashed lines display across-time trends in the average level of personalism in autocratic regimes with significant oil wealth (defined as those averaging at least $100 in Ross and Mahdavi's (Reference Ross and Mahdavi2013) variable oil income per capita over the regime's duration) and autocracies without.Footnote 2
Personalism and Oil Price in Autocracies, 1960–2010.
Though only a preliminary test, Figure 1 suggests that oil income is associated with greater personalism in autocracies, but that this effect is largely limited to the post-Cold War era. There is no meaningful difference between autocracies with and without oil wealth in terms of the level of personalism until the last few years of the 1980s. More importantly, the gap increases once the Cold War ends. This pattern is consistent with an argument that oil-wealthy autocracies were able to withstand post-Cold War pressures for political liberalization without diminishing their personalized character.
I test this argument statistically in the analysis that follows, though there are several reasons to believe that oil has this limited temporal impact. Andersen and Ross (Reference Andersen and Ross2014) demonstrate that oil's political impacts only began appearing after 1980 when a wave of nationalizations in the oil industry ushered in the “Big Oil Change.” Figure 1 is consistent with this argument; there is a lag between the start of the Big Oil era and the first observed differences in the level of personalism across autocracies.
This is also consistent with Hendrix's (Reference Hendrix2018) recent analysis emphasizing the end of the Cold War as the critical temporal break in the relationship between oil and political regimes. Specifically, Hendrix argues that support from the US, the Soviet Union, and their allies during the Cold War constituted a source of external rents than enabled regimes to resist pressures to democratize. When these rents diminished after 1990, however, only oil-wealthy regimes had the resources to ignore increasing demands for political liberalization. His reanalysis of several prominent studies provides quantitative evidence that oil wealth only impedes democratization in the post-Cold War period (10).
Volatility in global oil prices, displayed on the right-side axis in Figure 1, also suggests that the impact on personalism is limited to the post-Cold War era. Oil prices reached their high mark in 1980 at a value of nearly $80 USD per barrel (in constant $2000), but followed this with a decade-long slide. Even if increases in oil income facilitated the personalization of autocracies, the long decline in prices through the 1980s suggests there were few autocracies that would experience these changes. By contrast, the post-Cold War era began with a decade of relatively stable oil prices before starting a long ascent in the 2000s.
3. Empirical approach
I employ two-way fixed effect regressions to more formally evaluate the impact of oil rents on the level of personalism in autocracies during the post-Cold War era. This strategy includes individual fixed effects for years and for each autocratic regime. This limits the analysis to the within-regime relationship between a change in the level of oil rents and the subsequent change in the level of personalism. All standard errors are clustered on individual regimes.
The key independent variable measuring oil rents are Ross and Mahdavi's (Reference Ross and Mahdavi2013) oil income per capita, logged to address the highly skewed distribution in the untransformed version. The regressions also include measures of (logged) income per capita and population in millions. Following Geddes et al. (Reference Geddes, Wright and Frantz2018), I also control for whether or not the current leader was the first leader of the regime, as they are more likely to consolidate power personally, and the (logged) length of the current leader's time in office (regime leader duration). As the dependent variable reports the level of personalism as of January 1 for a given year, the oil income variable and all economic measures are lagged by two periods. Other controls are lagged by one period. As noted above, the sample includes all autocratic regimes from the years 1991 to 2010.Footnote 3
There are several ways in which statistical endogeneity might impact this analysis. First, some unobserved factor may jointly determine the levels of oil income and personalism, though the year and regime fixed effects help mitigate against this. Second, there may be some previously identified factor that impacts the relationship between oil income and personalism. No existing work makes this specific claim, though Menaldo (Reference Menaldo2016) argues that low levels of state capacity encourage the extraction of more oil while simultaneously influencing the level of democracy, the extent of non-resource taxation, and institutional quality. To the extent that low state capacity may be linked to greater levels of personalization in autocracies, one should control for this factor. I follow Menaldo's (Reference Menaldo2016) approach and use a measure of tax revenue as a share of GDP, though I rely on data from the IMF's World Revenue Longitudinal Dataset (2016).
Third, it remains possible that personalist leaders could increase oil production themselves, which would be the clearest example of reverse causality. The measure of oil income, along with the appropriate temporal lag, offers a partial guard against this issue (see Ross Reference Ross2012, 15–19) but is unable to fully dismiss this possibility.
Several prior studies employ instrumental variables measuring oil endowments or new discoveries to deal with similar issues (Tsui Reference Tsui2011; Menaldo Reference Menaldo2016). These approaches, however, are unable to solve the problem. At issue is the fact that, as Smith (Reference Smith2017, 601) recently notes, virtually all measures of oil income are endogenous to politics. Discovery of new fields, accurate reporting of reserves, and production levels all depend on the main actors—in particular the foreign multinationals that dominate oil production—judging the political environment as stable enough to generate a return on their investment. In methodological terms, such instruments do not meet the exclusion restriction.
Instead, this paper follows Smith's (Reference Smith2017, 601) recommendation that scholars are “better off thinking about how our measures are endogenous, and accounting for it, rather than trying to find others that are not.” I do this by testing the relationship while excluding large oil producers, since they are better able to manipulate oil production for political gain, and further supplementing the baseline specification with a measure of openness to international capital since more open regimes might have an advantage in endogenously raising oil production. I measure this openness as FDI inflows as a share of GDP, using data from the World Development Indicators (World Bank 2016).
4. Results and discussion
Table 1 reports the results of the two-way fixed effects regressions. I report the results of two model specifications (a baseline model and one with the additional control variables) with and without large oil producers, defined as the top 10 percent of observations in terms of oil income per capita in the prior specification.Footnote 4
Two-way fixed effect regression results, 1991–2010
Note: Columns report two-way fixed effect regression coefficients with standard errors clustered on regimes reported in parentheses. Dependent variable is Geddes et al. (Reference Geddes, Wright and Frantz2018) latent measure of personalism. Regressions include a constant, which is not reported to save space. ***p < 0.01, **p < 0.05, *p < 0.1.
Across all four models, the reported coefficient for oil income per capita is positive and statistically significant; an increase in oil wealth is associated with more personalized autocratic regimes. As I demonstrate below, the impact is relatively large in substantive terms. This result holds even when controlling for factors that may encourage leaders to extract more oil income (state capacity, measured as the tax revenue share of GDP) or enable them to do so (financial openness, measured as FDI inflows as a share of GDP), both in the full sample and while excluding large producers. Longer-serving leaders and new leaders are associated with increases in personalism as well, though the impact of the later is only statistically significant in the models including additional control variables. These results are also robust to modeling the uncertainty in the latent dependent variable following Schnakenberg and Fariss' approach (Reference Schnakenberg and Fariss2014; see also Crabtree and Fariss Reference Crabtree and Fariss2015) and to clustering on countries or individual leaders.Footnote 5
Figure 2 depicts the marginal effects of three key explanatory variables (oil income, leader duration, and new leader), expressed as the predicted change in the measure of personalism per an inter-quartile range increase in the independent variable, following the specifications in Models 2–4. For each simulation, all other variables are held to their mean values.Footnote 6 Inter-quartile range increases are based on the distribution of values in each estimation sample. For Model 2, oil income per capita (logged) is measured as an increase from 0 to 6.1, or from $0 to approximately $490 per capita, about the value of oil income in Angola in the late 1990s. For leader tenure, this is an increase from 5 to 18 years in power. For the variable new leader, the IQR increase is from 0 to 1, or simply the marginal effect of a regime's first leader. Models 3 and 4 employ very similar IQR ranges.
Estimates of Parameters of Interest from Models 2 to 4.
In most specifications, the impact of oil is on par with that of regime's first leaders, and between two and three times larger than the marginal effect of increases in a leader's time in office. Given the range of the dependent variable, a marginal effect of 0.4 for oil income would represent an increase of nearly 13 percent.
5. Conclusion
Important theories of autocratic politics rest upon the premise that resource rents help authoritarian leaders centralize their political power by funding patronage and enabling them to resist pressures to create power-sharing institutions with regime insiders. This note employs newly published data on personalism to demonstrate the cross-national quantitative basis of such claims. Over the last two decades, increases in oil income are associated with greater levels of personalism in authoritarian regimes.
Still, this result should be explored further. These findings connect increases in rents to greater personalization, but do not identify the specific ways in which leaders use these resources to consolidate their personal hold on power. For instance, are these revenues used to subsidize consumer goods that reduce the demands for more institutionalized means of influence for mass publics and elites? Fails (Reference Failsforthcoming) suggests that more generous fuel subsidies prevent democratization. Might similar spending choices enable personalist leaders to resist creating meaningful powersharing institutions? These are important avenues for future research.
These findings also contribute to the broad literature on the political resource curse. Specifically, the results suggest that one important channel connecting oil revenues to lower likelihoods of democratization runs through the consolidation of personalist regimes. Increases in oil income over the past two decades have facilitated the emergence of a uniquely problematic model of authoritarian politics, at least in terms of the prospects of successful democratization. These results are particularly important given that recent work strongly suggests that oil's main political effect has been to insulate producers from otherwise global trends toward more political liberalization since the early 1990s (Hendrix Reference Hendrix2018; Ross Reference Ross2012, 75–7). By pointing toward one particular channel, these findings provide important evidence as to exactly why oil impedes democracy.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2019.14
Acknowledgements
The author would like to thank Erica Frantz, Nicole Mathew, Elizabeth Wilhelmi and the Editors and reviewers for their helpful comments.
