Oil, Non-Tax Revenue, and Regime Stability: The Political Resource Curse Reexamined

Transcription

1 Oil, Non-Tax Revenue, and Regime Stability: The Political Resource Curse Reexamined Kevin Morrison Department of Political Science Duke University Comments most welcome. Prepared for presentation at the Harvard University Comparative Political Economy Workshop October 2005 Abstract: Building on theories of regime change that focus on redistributional dynamics, this paper generates hypotheses regarding non-tax revenue and regime stability. Nontax revenue, the importance of which has been largely ignored in theories of the public sector and regime change, is argued to include foreign aid as well as the majority of the oil revenue that accrues to governments. This non-tax revenue is hypothesized to affect redistribution in dictatorships but not in democracies, and to stabilize dictatorships but not democracies. These hypotheses are supported in cross-sectional time-series analyses of all developing countries over a period of Acknowledgments: I am grateful to Dylan Fagan for excellent research assistance, to John Doces, Marcela González Rivas, Robert Keohane, and Michael Ross for comments on an earlier version, and to Karen Remmer for comments and overall guidance. All errors are my own. An earlier version of this paper was presented at the Annual Meeting of the American Political Science Association in Washington, DC, in September I gratefully acknowledge financial support from a National Science Foundation Graduate Research Fellowship, a James B. Duke Fellowship, and a Vertical Integration Grant from the Duke University Graduate School. Any opinions, findings, conclusions, or recommendations expressed in this paper are those of the author and do not necessarily reflect the views of the National Science Foundation or any entity of Duke University.

2 Oil, Revenue, and Regime Stability: The Political Resource Curse Reexamined What is so peculiar about oil? The association of oil with authoritarian regimes has been the focus of a large case-study literature (e.g. Beblawi and Luciani 1987b; Chaudhry 1997; Karl 1997), as well as recent statistical work on cross-national time-series datasets (Ross 2001; Smith 2004). Overall, there seems to be something going on, but after so much work, we still remain unclear as to the mechanisms by which oil affects political regimes. The confusion is illustrated by the fact that the two most rigorous statistical analyses by Ross (2001) and Smith (2004) come to different conclusions about oil. Ross (2001) argues that oil leads to authoritarianism, and by implication is destabilizing for democratic regimes. Smith (2004) argues that oil leads to regime stability, implying that oil wealth would stabilize democracies. These empirical differences are evidence of a broader problem: a lack of theoretical underpinning for the effects of oil wealth in different regimes. The causal mechanisms between oil and democratic transition (or lack thereof) have varied in the literature. The central argument, however, is that oil revenues provide central governments with discretionary income that can be used to buttress their political position, protecting them for pressures for democratic transition. As Jensen and Wantchekon state, The key mechanism linking authoritarian rule and resource dependence, both in democratic transition and democratic consolidation, is the incumbent s discretion over the distribution of natural resources (2004, 821). Similarly, Smith notes, While scholars approach the political economy of oil from diverse methodological origins, the theoretical arguments about the structures and nature of the rentier state flow from the state s access to externally obtained revenues from the sale of oil (2004, 233, emphasis added). For one group of scholars (e.g. 1

3 Acemoglu, et al. 2004; Bazresch and Levy 1991; Chaudhry 1997; Crystal 1989; Entelis 1976; Karl 1997; Vandewalle 1998), the key issue is that oil revenues allow governments to buy off political consensus (Beblawi and Luciani 1987a, 7). By this argument, the soft budget constraints that oil revenues provide enable governments to reduce the pressures for regime change (Anderson 1995; Luciani 1987). A second group of scholars (e.g. Bellin 1994; Chaudhry 1994; Clark 1997; Moore 1976; Ross 2001; Shambayati 1994; Skocpol 1982) have pointed to the association of oil revenues with repression of various social groups and citizens. The result of such repression is the same as buying off those groups: less pressure for regime change. This paper takes off from two points that can be made with regard to this literature. First, the fact that two quite different causal mechanisms cooptation and repression have been offered to explain oil s relationship with authoritarianism indicates that there is possibly nothing inherent in these resources that makes regimes act in certain ways. And if this is the case, then it may make far more sense to think of these types of resources as simply revenue entering a particular political economy, rather than as a resource with antidemocratic properties (Ross 2001, 325). That they are used in authoritarian regimes for repression may be nothing more than a reflection of those regimes preferences over the use of state finances. Presumably democratic regimes would use these resources in a different way, but we have no theories to account for how. Second, if the key mechanism at work here is the state s access to externally obtained resources, we should not expect that oil revenues are particularly unique (though they might make up a large share of such resources). In fact, there may be a variety of such resources, whose key characteristic is that they are not derived from taxation, but rather available mainly 2

4 as windfalls to the government. For example, some scholars have begun to feel that the literature on oil revenues has relevance for another fungible, external resource: foreign aid (Bräutigam 2000; Moore 1998; Therkildsen 2002). 1 This paper explores the effect of this broader set of revenues what I term non-tax revenues on regime stability. Oil revenues by and large accrue to states through stateowned enterprises, so the revenue does not come through taxation. As will be discussed below, such non-tax revenues which include foreign aid can make up substantial amounts of government revenue, and ignoring them in theories of regime stability has likely come at a cost. I first review the rather sparse literature on revenue and regime stability and then develop a theoretical framework based on recent work on the redistributional dynamics inherent in regime change. This theoretical framework allows me to make predictions about the effects of these resources. Next, I discuss data issues and give some descriptive statistics about non-tax revenue. I then test the effects of non-tax resources on regime change in a cross-national time-series dataset, comparing their effects to those of oil in both dictatorships and democracies. It turns out that when non-tax revenue is taken into account, oil does not affect regime stability in a significant way in either type of regime. It is not oil, per se, that leads to regime stability, but government ownership of that oil and in fact, other kinds of non-tax revenue have similar properties. Furthermore, these revenues only lead to increased stability in dictatorships, not in democracies. 1 It is interesting to note in this context that Mahdavy s definition of a rentier state was a state that received substantial rents from foreign individuals, concerns or governments (1970, 428, cited in Ross 2001). 3

5 Previous Literature and Theoretical Framework The literature on revenue and regime stability is surprisingly sparse, despite the emphasis on the centrality of revenue to the nature of regimes in landmark works by Schumpeter (1918 (1954)), Musgrave (1959), Brennan and Buchanan (1980), and Levi (1988). 2 The work that does exist is usefully divided into that focused on tax revenue and that focused on non-tax revenue (there is no work that systematically addresses them both). The work on tax revenue and regime stability has generally focused on a proposed link between taxation and representation in the transition from autocracies to democracies. The principal argument is that important western democracies arose as a result of a bargain: rulers in need of resources were forced to grant representation in exchange for taxes (Bates and Lien 1985; Levi 1988; North and Weingast 1989; Tilly 1990). As Ross (2004) points out, there are two different versions of this argument. One, seemingly supported by scholars like Brennan and Buchanan (1980) and Huntington (1991), is that citizens demand representation in exchange for higher levels of taxes. The other argument, seemingly supported by the likes of Bates and Lien (1985), is more conditional: citizens demand representation if the ratio of government services to taxes falls below a certain threshold. Ross provides cross-national statistical tests for these theses, finding support for the latter but not the former. There is no work I know of that focuses specifically on non-tax revenue, as such, and its relation to regime stability. However, certain kinds of non-tax revenue have been studied on their own. Oil revenues are, in general, one example. The majority of oil revenue for governments comes not through taxes on foreign companies but rather through state-owned 2 It should be noted that this paper is focused on regime stability in the sense of authoritarian regimes switching to democratic ones, and vice versa. The literature on this topic has existed parallel to a literature on regime stability in the sense of political regimes succumbing to civil war. For a recent work that examines revenue in the context of this latter kind of regime stability, see Snyder and Bhavani (2005). Ideally, these literatures would coincide theoretically and empirically more than they do now. 4

6 companies. Seventy-five percent of the world s oil production, and 90 percent of its reserves are in the hands of such state-owned companies (Ivanhoe 2000). Ross (1999) has suggested that this state ownership of oil companies may be an underlying factor in the association of oil wealth with poor economic performance, but to my knowledge this has not been tested. More importantly for the purposes at hand, state ownership of oil companies may also be important for the association of oil wealth with regime stability. Another kind of non-tax revenue that has been studied on its own is foreign aid. Research has indicated that foreign aid is a highly fungible resource (Feyzioglu, et al. 1998) and acts similarly to oil in the sense that it provides extra resources the government can use to distribute to its key constituencies without taxation (e.g. Bratton and van de Walle 1997). For example, van de Walle has argued that democratization in Africa was encouraged by a fiscal crisis resulting from, among other things, an increased willingness on the part of donors to restrict aid to countries that did not respect human rights: With fewer resources at their disposal and an increasingly decrepit state apparatus, leaders found it harder to sustain critical clientelist networks, with the result that the old political aristocracy was more likely to fractionalize (2001, 240). There are, then, several different revenue sources that one might consider to be important in a theory of regime stability and revenue. However, we lack a theory that could help us understand how these different kinds of revenue interact with each other to affect regime stability. Such a theory would not only have to address both kinds of revenue, but crucially also address the demand side: the demand from society for expenditures. A good place to start constructing such a theory and the way in which hypotheses will be generated for this paper is to build off of works that focus on redistributional 5

7 conflicts as central to regime change. The reason these works are helpful is that they begin to address both the demand and supply sides of revenue, though they tend to ignore non-tax sources of revenue. The redistributional approach to regime change has a distinguished history, including the landmark works of Moore (1966) and Rueschemeyer, Stephens, and Stephens (1992). These works have argued that democratization tends to be most resisted by the rich. If one considers that political regimes are in essence a way of allocating resources in society (Kitschelt 1992), it is intuitive to expect that richer citizens would prefer not to have poorer citizens playing a role in such allocations, and vice versa. The most important recent contributions in this body of work are by Boix (2003) and Acemoglu and Robinson (2005; 2001), who have used methods and theories largely developed in the social choice literature. An important aspect of these recent works is that they build off work on the size of the public sector that sees public revenue and expenditures as principally driven by redistributional concerns. For example, Boix (2003) and Acemoglu and Robinson (2005; 2001) use formal models that employ the benchmark result of Meltzer and Richard (1981), who argued that more unequal democracies will have larger governments than more egalitarian democracies. As a result, these recent works on regime change suffer from a problem shared by most theoretical and empirical works on government size: the possibility of non-tax revenue is largely ignored. The Meltzer-Richard prediction, for example, is generated solely by preferences over redistribution through taxation. Thus, while a large literature has considered the determinants of the size of the public sector, (e.g. Boix 2001; Persson and Tabellini 2003; Rodrik 1998), few scholars have studied the effects of non-tax sources of revenue on government size (an exception being Remmer 2004). 6

8 To understand how non-tax revenues might affect regime change in a redistributional approach, it is necessary to understand how they might affect redistribution. And therefore it is necessary to understand why and how much redistribution takes place in democracies and dictatorships. If we define democracies as those regimes that have functioning elections, it is generally safe to say that democracies give a more important say in politics to the poorer masses than do dictatorships, where a far smaller group almost always more well-off than the rest makes decisions over resource allocations. The result of this greater say in politics is that democracies are usually thought to be better for the poorer members of society than dictatorships. 3 The recent work of Lindert (2004), for example, has carefully documented how the expansion of the share of the population with political voice in western countries at the end of the 19 th and beginning of the 20 th centuries (an expansion to the poorer elements of society) led to an expansion in both social insurance and public education provision. This assertion is also supported by the work of Bueno de Mesquita, et al. (2003), who demonstrate that as the size of the winning coalition in a regime increases, so do important education and health indicators. The extent of redistribution in democracies is clearly dependent on certain constraints, as a large literature has investigated. Some of these constraints that have been analyzed in the literature include the political mobilization of the rich, electoral funding laws, and the mobility of capital. And in fact, these constraints play a critical role in theories of regime change based on redistribution. In the Acemoglu and Robinson framework, for example, the reason democracies fall is because the threat of redistribution is too much for the rich, who launch a coup. In the Boix framework, the mobility of capital makes democracy more likely, 3 See, for example, Sen (1999) and Zweifel and Navia (2000). For an argument that there is no difference between the regimes once adjustments are made for missing data, see Ross (2005). 7

9 because the rich able to move their money abroad are not as afraid of the redistributive power of democracies. For these reasons, democracies in these frameworks maximize redistribution to a certain point, at which the rich are just indifferent between being in a democracy and launching a coup. Dictatorships redistribute for a different reason. If one assumes that those in power in a dictatorship are generally the more well-off in a society, then it is a quick jump to understand that they will want as little redistribution as possible. However, like democracies, dictatorships also work under certain constraints. One of the most important of these is the potential for revolution from the large portion of society not in power the poorer portion. While there is far less empirical work studying redistribution in dictatorships than in democracies, many scholars (e.g. Acemoglu and Robinson 2005; Acemoglu, et al. 2004; Grossman 1991; 1994; Huntington 1968) have theorized that an important reason that dictatorships redistribute is to avoid such revolution. Rigolini (2003) provides some empirical support of this hypothesis. Building on this thinking, the argument of Acemoglu and Robinson (2005; 2000) is that the elite will redistribute just enough to keep the poor from revolting. And when they cannot credibly promise enough redistribution to do this, they democratize, because democracy is preferable to them over an outright revolution. If these redistributional dynamics are correct, then non-tax revenues should have very different effects on redistribution through taxation and spending in dictatorships and democracies. In dictatorships, the elite in power only want enough resources funneled to the poor to keep them from revolting. In the absence of non-tax revenue, these resources will have to come from the pockets of the rich. However, non-tax revenues give the rich some flexibility. In the absence of an increased threat, they can simply redistribute less of their own 8

10 resources to satiate the masses, while spending on the poor would stay basically the same. It may be noted that at the extreme, this kind of dynamic would lead a purely distributional state (Delacroix 1980). In the presence of an increased threat from the poor, the additional resources enable the rich to top-up spending on the poor. On the other hand, in democracies, non-tax revenues accruing to the state should have no effect on taxation of the rich, since the median voter, being less well-off, will still want to maximize redistribution from the rich. Instead, non-tax revenues will simply add to the amount of funds that can be spent on the poor. Table 1 summarizes these predictions. Table 1: Predictions for the effects of rises in non-tax revenues Taxation of rich Spending on poor Dictatorships Decreases, or stays the same Stays the same, or increases in the face of increased threat in the face of increased threat Democracies Stays the same Increases Accordingly, these redistribution dynamics imply that non-tax revenues will have different effects on the stability of dictatorships and democracies. In redistributional approaches to regime stability, put rather simply, dictatorships fall because they spend too little on the poor, and democracies fall because they take too much from the rich. Because non-tax revenue should not have an effect on taxation of the rich in democracies, there is no reason to suspect that such revenue will have any effect on regime stability either. In dictatorships, however, the influx of extra resources can enable stressed dictatorships to funnel more resources to the poor, thereby obviating the need to democratize. In other words, non-tax revenue should have a stabilizing effect on dictatorships, but no effect on democracies. The following sections explore these hypotheses empirically. 9

11 Some Details on Non-tax Revenue As the conception of non-tax revenue advanced above is relatively new, it is useful to begin with a discussion of descriptive statistics. The best available data on non-tax revenues is from the International Monetary Fund s Government Finance Statistics. Unfortunately for researchers interested in this revenue over a long time period, the IMF (2001) recently changed the way they categorize revenue. Revenue has been coded in the new way only from 1990 to present, but the data for the previous kind of coding by the IMF (1986) is available over a time period of Therefore, to attain a longer time-series, I have used the 1986 coding. The IMF s variable for non-tax revenue can be found in the World Bank s (2004) World Development Indicators (WDI). The World Bank notes that the indicator consists of not compulsory, nonrepayable payments for public purposes, such as fines, administrative fees, or entrepreneurial income from government ownership of property. Proceeds of grants and borrowing, funds arising from the repayment of previous lending by governments, incurrence of liabilities, and proceeds from the sale of capital assets are not included. As the description implies, this indicator does not include receipts from foreign aid, which are a crucial part of my definition of non-tax revenue. I have therefore created my own variable that sums the value of non-tax revenue (as defined by the IMF) and net aid receipts (also from WDI), both as a share of GDP. The resulting variable is quite close to the theoretical concept outlined above revenue the government receives that does not derive from taxation. 4 My sample consists of all non-oecd countries for which data is available. There are Where relevant, I made similar adjustments to the IMF s indicator for total revenue, adding foreign aid to them because the IMF does not. 10

12 observations of non-tax revenue (henceforth, non-tax revenue will refer to my variable that includes foreign aid), comprising 129 countries. The first point to make is that, for many countries in the world, this type of revenue is quite significant. In the 2225 observations, non-tax revenue makes up an average of 35 percent of government revenue. Table 2 lists some of the countries for which non-tax revenue is crucial. In countries such as the Democratic Republic of Congo, Kuwait, and the United Arab Emirates, non-tax revenue makes up virtually all of government revenue in certain years. The table also shows that Middle Eastern countries tend to get their revenue from nontax revenue, indicating that theories explaining authoritarianism in oil-rich countries on the basis of tax-based arguments (e.g. Boix 2003) are incomplete. And finally, the table indicates that the importance of these revenue sources crosses regional boundaries. All of this points to the importance of studying this kind of revenue, which has been largely ignored in the political science literature. These non-tax revenues are also large relative to the size of the economy. Table 3 lists the components of revenue tax and non-tax as share of GDP by region. It also lists the regional averages of oil exports as a share of GDP, a common indicator of oil dependence. Of note is the fact that among the six regions, sub-saharan Africa and South Asia rank fourth and sixth in terms of their oil exports as a share of GDP. Yet they rank second (tied) and fourth respectively in terms of their non-tax revenues, as a result of large aid inflows. The Middle East, as one might expect, is ranked first in both variables. 11

13 Table 2: The Importance of Non-tax Revenue in Countries Revenue Streams Percentage of total revenue that is non-tax revenue (1 = 100 percent) Observations Mean Std. Dev. Min Max Bahrain Bangladesh Bhutan Bolivia Botswana Burkina Faso Congo, Dem. Rep Congo, Rep Iran Jordan Kuwait Madagascar Maldives Mali Nepal Nicaragua Oman Sierra Leone Syria Uganda United Arab Emirates Table 3: The Importance of Non-tax Revenue, in economic terms All variables are as a percent of GDP (100 = 100 percent). Data are for the 1511 observations in which all variables are available. Oil exports Tax revenues Non-tax revenues excluding aid Aid Total non-tax revenues* Middle East and North Africa East Asia and Pacific Sub-Saharan Africa South Asia Europe and Central Asia Latin America and Caribbean *Total non-tax revenue is the sum of the previous two columns (differences are due to rounding). Regional breakdowns are as defined by the World Bank. 12

14 Analysis of the Redistribution Hypotheses I now present the results for the first set of hypotheses, regarding the effects of non-tax revenue on redistribution in democracies and dictatorships. For my indicator for political regime, I have chosen to use the definition of Przeworski et al. (2000), who define democracies as regimes with functioning elections. Specifically, a regime is coded as democratic if the chief executive is elected, the legislature is elected, there is more than one party, and incumbents lose elections. If all of these characteristics are not present, the regime is a dictatorship. There are therefore no in-between regimes either a regime is a democracy or a dictatorship. This binary coding of Przeworski et al., and their focus on elections, matches well with the theoretical approach outlined above. Recent redistributional approaches to regime change have also considered regimes as dichotomous, and the defining characteristic of regimes is the ability (or lack thereof) for large numbers of citizens to vote. Boix, for example, defines democracy as a regime in which all individuals vote (or may vote). In a dictatorship, only the preferences of part of society are taken into account to decide the final allocation of assets (2003, 10). The distinction Acemoglu and Robinson make is quite similar: In a democracy the majority of the population is allowed to vote. [N]ondemocratic regimes share one common element: instead of representing the wishes of the population at large, they represent the preferences of a subgroup of the population, the elite (2005, 16). I use two dependent variables in the analysis in this section. As discussed above, redistribution is made up of the taking away from the rich and the giving to the poor. For the former I use income tax revenue as a percent of GDP, since income tax is considered a 13

15 progressive tax, and for the latter I use the sum of health, education, and welfare spending as a share of GDP, since these types of spending are generally considered to be progressive. The statistical estimations here are based on an error-correction model of the type utilized in other recent research on government spending and revenue (Iverson and Cusack 2000; Remmer 2004; Rodden 2003). The model is of the form: Y i,t = α + X i,t β + Φ(Y i,t-1 X i,t-1 γ) + ε i,t in which Y i,t is the revenue or spending variable in country i in time t, and X is a matrix of independent variables. Therefore the dependent variable is the change in revenue or spending as a share of GDP from one year to the next. The independent variables include both the annual rate of change and the lagged values of the independent variables, as well as the lagged value of dependent variable. As Remmer writes about this type of model, The underlying theoretical assumption is that the relationship between the variables in the model resembles a moving equilibrium in which the dependent variable may not only fluctuate in response to short-run changes in the independent variables but also assume over the long run levels consistent with those of the independent variables. The central advantage of the model is thus that it makes it possible to distinguish between the short- and long-term relationship of X and Y. (2004, 82) The equation actually estimated is as follows: Y i,t = β 0 + Y i,t-1 β 1 + X i,t β 2 + X i,t-1 β 3 + ε i,t, in which β 1 estimates Φ in the error correction model, β 2 estimates β (thus representing the short-term relationship between X and Y), and β 3 estimates Φγ (and thus the long-term relationship between X and Y). As is standard in the literature, the equation was estimated using ordinary least squares with panel-corrected standard errors, to accommodate the problems that plague cross-sectional time series research designs, notably heteroskedasticity 14

16 and contemporaneous correlation of errors across countries (Beck and Katz 1995). The estimation also accounts for panel-specific patterns of first-order autocorrelation. While the principal variable of interest in the regression is non-tax revenue, it is important to control for other variables that might be driving changes in income tax revenue or social spending. For both sets of regressions, building off recent work on government size, I include five additional variables in the model. First, I control for population (in log form), since prior research (e.g. Shadbegian 1996) has indicated that larger countries are more efficient at public good provision because of economies of scale. Second, I control for the percentage of the population that is 65 years and older, since this tends to drive pensions (e.g. Perotti 1996), an important part of public spending. Third, I control for trade dependence, measured as exports plus imports as a percent of GDP, building off work that asserts a relationship between trade openness and the public sector (e.g. Rodrik 1998). Fourth, I control for per capita GDP, to account for the effect of economic development on the public sector (e.g. Boix 2001). In the income tax regressions, the fifth variable is central government expenditure as a share of GDP, to account for the potential impact of changes in spending requirements in a country (e.g. Remmer 2004). Likewise, in the spending regressions, I include tax revenue as a share of GDP, in account for the potential impact of changes in tax revenue. Finally, in all of the regressions I include dummy variables for the 1970s and 1980s decades, as well as dummy variables for geographic region. All of these variables, as well as the dependent variable, are attained from the World Bank (2004). Table 4 presents the regression results for income tax, where dictatorships and democracies are as defined above and for which the codings are taken from Cheibub and Gandhi (2004), who use the previously discussed regime definitions of Przeworski et al. 15

17 Table 4: Determinants of Changes in Income Tax Revenue as a Share of GDP (log) Dictatorships Democracies Coefficient Standard Error Coefficient Standard Error Non-tax revenues as a share of GDP (log) t ** GDP per capita(log) t * *** Population(log) t * * Share of population over 65 t ** Trade/GDP t *** Government Expenditures as a share of GDP(log) t Non-tax revenues as a share of GDP (log) GDP per capita (log) 0.487*** Population (log) 3.120* Share of population over Trade/GDP 0.003*** *** Gov t Expenditures as a share of GDP (log) 0.131* ** Income tax revenue as a share of GDP(log) t *** *** Constant ** ** s * s Africa Middle East Latin America Europe and Central Asia * South Asia ** Observations Countries R Wald χ * 0.10 ** 0.05 ***

18 (2000). As predicted, there is a significant negative relationship between non-tax revenues and changes in tax revenue in dictatorships, but not in democracies. There also seems to be an interesting pattern with regard to GDP per capita. As dictatorships get richer, they tend to tax the rich less, while richer democracies tend to tax the rich more. I now turn to the spending regressions, which are represented in Table 5. The results for democracy support the hypotheses above, with non-tax revenues positively and significantly associated with increases in social spending in democracies. Interestingly, nontax revenue is also positively and significantly associated with social spending in dictatorships. This seems to indicate that, as hypothesized above, increases in non-tax revenue have occurred in stressed dictatorships, which then funnel the revenue into spending on the poor. The results in this section indicate that, as theorized above, non-tax revenues will be stabilizing in dictatorships but not in democracies. To reiterate the theory, dictatorships fall because the poor receive too little, and it has been shown here that non-tax revenues enable the rich in dictatorships to fund social spending while diminishing redistribution of their own income. As such, the hypothesis that non-tax revenues will stabilize dictatorial regimes still seems sound. On the democratic side, it was theorized that democracies fall because they tax the rich too much. Since non-tax revenues have been shown here to have little effect on such taxation instead, non-tax revenues in democracies tend to increase social spending these revenues should not have any significant effect on democratic regime stability. With some empirical support for this causal mechanism, I now move on to testing the regime change variable directly. 17

19 Table 5: Determinants of Changes in total Social Spending (Health, Education, and Welfare) as a Share of GDP (log) Dictatorships Democracies Coefficient Standard Error Coefficient Standard Error Non-tax revenues as a share of GDP (log) t *** ** GDP per capita(log) t ** Population(log) t Share of population over 65 t Trade/GDP t Tax Revenue as a share of GDP(log) t Non-tax revenues as a share of GDP (log) 0.063** GDP per capita (log) *** Population (log) ** ** Share of population over Trade/GDP ** Tax Revenue as a share of GDP (log) 0.084** *** Social Spending as a share of GDP(log) t *** *** Constant s *** s *** Africa ** Middle East dropped Latin America Europe and Central Asia South Asia Observations Countries R Wald χ * 0.10 ** 0.05 ***

20 Analysis of the Regime Change Hypothesis In order to study regime change, the dependent variable in these regressions is a binary variable that takes a value of 1 if there is a change from one year to the next in the regime coding by Cheibub and Gandhi (2004). It takes a value of 0 if there is no change. I use a statistical model identical to the one underlying the revenue and spending regressions above, both so the results are comparable and also for underlying theoretical reasons. As discussed above, an error correction model enables the study of both long-term and short-term relationships between variables. Many theories of regime change imply that certain independent variables could have both long-term and short-term effects. For example, there are hypothesized effects on regime change of both the level of income in a country (Lipset 1959) and economic growth and decline (Remmer 1991). For this reason, many scholars (e.g. Przeworski, et al. 2000) include both level and growth of GDP per capita in their regressions. However, few scholars do this with other variables they include in the regressions. A close reading of the resource curse literature, however, implies that a similar approach should be taken to oil variables. It is often unclear whether scholars are talking about the effects of levels or changes in oil revenue, but one can certainly generate hypotheses about both based on the literature. For example, I read the following hypothesis as one about the level of oil: resource wealth retards democratization by enabling governments to boost their funding for internal security (Ross 2001, 328). And I read this next one as being a hypothesis about the change in oil revenues: during [oil] booms politicians are likely to flood the domestic economy with revenues, spending unwisely and spurring destabilizing inflation 19

21 (Smith 2004, 233). It is possible that both levels and changes have effects, and the errorcorrection model enables the study of both. Again, the principal variable of interest is the non-tax revenue variable. However, in this regression I am also particularly interested in how non-tax revenue compares to the variable that scholars have used to measure oil dependence: oil exports as a share of GDP. This variable can be constructed using the World Bank s World Development Indicators, and as Smith writes, it highlights both the role of oil as a source of export revenues and its importance in the domestic economy (2004, 236). I also control for other variables that might affect regime stability. First, I control for levels and changes in GDP per capita, for the reasons mentioned above. Second, I control for ethnolinguistic fractionalization (ELF), since many scholars (e.g. Horowitz 1985) have argued that social fragmentation increases instability in certain regimes. The measure I use is the probability that two randomly chosen individuals in a country do not speak the same language. Roeder (2001) provides observations of this variable for both 1961 and For all observations prior to and in 1980, I used the 1961 measure, and for all subsequent years I used the 1985 measure. Finally, like Przeworski et al. (2000) and Smith (2004), I control for past instability in a country, measured simply as the number of all past regime changes in that country in the sample. The results are estimated by logistic regression with unaltered standard errors, as in Smith (2004). Table 5 reports the results. As predicted, non-tax revenues have a significant stabilizing impact on dictatorships but not in democracies. The oil exports variable generally fails to reach standard levels of significance in either of the regressions, though the change variable is significant at the 0.10 level in the dictatorships regression. Interestingly, ELF is 20

22 Table 5: Determinants of Regime Instability Dependent variable is dichotomous, coded as 1 if the regime changes in the current year and 0 if there is no change (therefore a negative coefficient is a stabilizing effect). Observations were included in the dictatorships (democracies) regression if they were coded as dictatorships (democracies) in year t-1. Dictatorships Democracies Coefficient Standard Error Coefficient Standard Error Non-tax revenues as a share of GDP(log) t *** Oil exports as a share of GDP(log) t GDP per capita(log) t * Non-tax revenues as a share of GDP(log) ** Oil exports as a share of GDP(log) * GDP per capita(log) ** Ethnolinguistic Fractionalization 4.656*** Sum of past regime failures t * Regime transition t Constant *** s *** ** s Africa Middle East dropped Latin America Europe and Central Asia dropped South Asia Observations Countries Log Likelihood * 0.10 ** 0.05 ***

23 only destabilizing to dictatorships, and GDP per capita both its level and change are only significant in the democracy regression. Because past works have indicated that oil does have impacts on transitions to and from democracies, we want to be confident that the results here are not driven simply by the sample of observations for which non-tax revenue is available. The principal difference between the oil and non-tax variables is that there are no observations of the non-tax variable prior to 1970, whereas observations for the oil variable stretch back into the 1960s. Once one drops these years prior to 1970, the sample of countries for which only the oil variable is available is very similar in important respects to the sample in which it and the non-tax variable are both available. Table 6 reports a comparison of the two samples. As can be seen, they share very similar characteristics in terms of their overall oil exports-to-gdp ratio, their regime failures observed, and their existing political regimes. They are also quite similar in terms of their temporal, spatial, and economic characteristics. While we of course cannot rule out that the results here are driven by the sample, this comparison of samples gives increased confidence. The result of these regression analyses is therefore an inability to reject the hypotheses advanced above in the theoretical discussion. It seems indeed to be the case that non-tax revenues have stabilizing effects of dictatorships, but not in democracies, for the redistributional reasons outlined above. 22

24 Table 6: Sample properties for oil and non-tax variables Sample in which oil as a percent of GDP is observed Sample in which both the oil and non-tax variables are observed Observations Mean of oil exports as a % of 7 7 GDP % in which regime failure is 3 3 observed % Dictatorships % 1970s % 1980s % 1990s % 2000s 11 6 Sub-Saharan Africa Middle East and North Africa Latin America and Caribbean Europe and Central Asia South Asia 5 7 East Asia and Pacific High Income (non-oecd) 9 12 Upper Middle Income Lower Middle Income Low Income Region and income codings are by the World Bank. Conclusion A fundamental question in the study of democratization is whether theories developed in one part of the world apply to other parts of the world (Bunce 2000). The recent statistical works on the relationship between oil and political regime have been excellent examples of the importance of testing the broader validity of an area study theory. This paper will hopefully be seen in this tradition as well. However, instead of focusing only on the area studies literature focused on oil (and therefore mainly the Middle East), this paper also builds on literature that has examined the impact of foreign aid on regime change a literature that 23

25 studies different regions of the world. My underlying hypothesis has been that these types of revenue sources have similarities that cause them to have certain effects on different types of political regimes. In examining this underlying hypothesis, this paper has made four principal contributions. First, it has pointed to the fact that a sizeable portion of government revenue non-tax revenue remains largely ignored in political science, despite its overall importance in the revenue strains of many governments. In fact, in the sample of countries and years studied here, it accounted for an average of more than a third of government revenue. Our theories of government size have very little to say about this portion of government revenue, and this clearly is a gap in our understanding of the public sector. Second, this paper has begun to flesh out a theory of revenue and regime stability, based on approaches to political regimes that focus on the regime s role in resource allocation and re-allocation. As the important recent works in this tradition build off work on the size of the public sector which have, as just mentioned, ignored the difference between tax and non-tax revenue it is not surprising that these theories have ignored the potential role of non-tax revenue. My analysis has indicated that since the dynamics driving redistribution in dictatorships and democracies are different, the effect of non-tax revenue on these regimes should also be different. Specifically, non-tax revenue should not affect redistribution and regime stability in democracies, but it should diminish redistribution and increase regime stability in dictatorships. Third, the preceding analysis indicates that the answer to the question that opened this paper What is so peculiar about oil? may be that so much of it is produced by successful state-owned enterprises, particularly in dictatorships. This is certainly not to argue that 24

26 theories that link oil-wealth to authoritarianism based on taxation dynamics (e.g. Boix 2003) are not useful. However, it may be the case that taxation dynamics are more relevant to the effects of privately held oil resources. And since so much of the world s oil resources are in public hands (as discussed above), we need to consider the effects of this public ownership on the effects of oil in different political regimes. Finally, while the focus of the discussion in this paper has been on oil, it should also be noted that this paper has implications for the effect of foreign aid on regime stability, a topic that has received surprisingly little attention in the literature. The results that we do have are inconsistent. Goldsmith (2001) finds a small but significant positive correlation between level of democracy in Africa and aid as a percentage of GNP, and Knack (2004) finds no correlation between improvements in level of democracy and aid as a percentage of either GNP or government spending. The analysis here indicates that aid s effects are conditional on the political regime in place. While the results presented here are illuminating, much work remains to be done on this topic. First, a formalization of the argument presented in the second section of this paper would enable a useful direct comparison of the formal treatments of regime change in Boix (2003) and Acemoglu and Robinson (2005). 5 Second, a broader range of tests of the effects of non-tax revenue and redistribution and regime stability would be illuminating. And third, some in-depth case studies might improve our on-the-ground understanding of the linkages between redistribution and regime stability, which remain rather slim. Enlightening case studies do exist, such as those by Boix (2003) and Rueschemeyer et al. (1992), but they remain rather few. And none of them (that I know of) examine the effect of non-tax revenues on this relationship. Of course, as discussed above, case studies on certain types of non-tax 5 I have formalized this argument for dictatorships in Morrison (2005). 25

27 revenues and regime stability do exist, but none of these have studied these revenues in an overarching framework, which would hopefully shed light on the similarities between oil income and other types of non-tax revenue. References Acemoglu, Daron, and James A. Robinson. Economic Origins of Dictatorship and Democracy. Cambridge: Cambridge University Press, "A Theory of Political Transitions." American Economic Review 91 (2001): "Why Did the West Extend the Franchise? Democracy, Inequality, and Growth in Historical Perspective." The Quarterly Journal of Economics 115, no. 4 (2000): Acemoglu, Daron, James A. Robinson, and Thierry Verdier. "Kleptocracy and Divide-and- Rule: A Model of Personal Rule." Journal of the European Economic Association 2, no. 2-3 (2004): Anderson, Lisa. "Peace and Democracy in the Middle East: The Constraints of Soft Budgets." Journal of International Affairs 49, no. 1 (1995): Bates, Robert H., and Da-Hsiang Donald Lien. "A Note on Taxation, Development, and Representative Government." Politics and Society 14, no. 1 (1985): Bazresch, Carlos, and Santiago Levy. "Populism and Economic Policy in Mexico, " In The Macroeconomics of Populism in Latin America, edited by Rudiger Dornbusch and Sebastian Edwards. Chicago: University of Chicago Press, Beblawi, Hazem, and Giacomo Luciani. "Introduction." In The Rentier State, edited by Hazem Beblawi and Giacomo Luciani, London: Croom Helm, 1987a.. The Rentier State. London: Croom Helm, 1987b. Beck, Nathaniel, and Jonathan N. Katz. "What to Do (and Not to Do) with Time-Series Cross- Section Data." American Political Science Review 89, no. September (1995): Bellin, Eva. "The Politics of Profit in Tunisia: Utility of the Rentier Paradigm?" World Development 22, no. 3 (1994): Boix, Carles. Democracy and Redistribution. Cambridge: Cambridge University Press, "Democracy, Development, and the Public Sector." American Journal of Political Science 45, no. 1 (2001): Bratton, Michael, and Nicolas van de Walle. Democratic Experiments in Africa: Regime Transitions in Comparative Perspective. Cambridge: Cambridge University Press, Bräutigam, Deborah. Aid Dependence and Governance. Stockholm: Almqvist & Wiksell International, Brennan, Geoffrey, and James M. Buchanan. The Power to Tax: Analytical Foundations of a Fiscal Constitution. Cambridge: Cambridge University Press, Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D. Morrow. The Logic of Political Survival. Cambridge, MA: MIT Press, Bunce, Valerie. "Comparative Democratization: Big and Bounded Generalizations." Comparative Political Studies 33, no. 6/7 (2000):