1
Explaining democratization
Ideally, an explanatory theory of democratization should fulfill several criteria.1 It should be sufficiently general to encompass the regularities in patterns of regime change across both time and space, but without sacrificing concreteness and the ability to account for complex causal mechanisms at work in singular instances. It should transcend the structure and agency divide by specifying how and when structural constraints affect the desires and beliefs of social actors, while at the same time providing some logic for understanding why these actors make decisions the way they do, and how the interaction of these decisions produces an outcome in terms of regime change. This ideal theory should be capable of explaining both short-term and long-term dynamics in regime trajectories over time, and both why autocracies turn into democracies and why some democracies become autocracies again. It should be probabilistic by nature, and thus leave the possibility of perfectly determined outcomes as a limiting case under extreme circumstances rather than as an inherent quality of the theory itself. It should be integrated and internally consistent: if consisting of more than one explanatory factor (which seems likely), these should not be mutually contradictory. Finally, this theory should be empirically corroborated, that is, it should be supported by as broad a body of empirical evidence as possible.
Perhaps not very surprisingly, all theories of democratization to date fall short of these standards. Social scientists have employed various theoretical and methodological approaches in order to explain democratization. The territory has been charted by sociologists, political scientists and economists alike, drawing on in-depth case study knowledge, comparative-historical analysis and statistics. Most work has
' On idea] criteria for explanatory theories in general, see King et al. (1994, pp. 99-114), and Gcrring(2001, chap. 5); on such criteria in the case of explaining democratization, see Coppedge, forthcoming, chap. 3.
16
The structural approach
17
been based on informal theory, largely developed inductively, but more recently formal, game-theoretic and deductive approaches have come to the fore. Beneath these conspicuous differences, however, one may discern three intellectual traditions that have forged the landscape of democratization studies. These are the structural and strategic approaches and the social forces tradition. To this we must now also add the more recent economic approach to explaining democratization. In this chapter I shall review these four approaches in turn, assessing their respective theoretical strengths and weaknesses. Equipped with these insights, I conclude by defending the particular approach to explaining democratization that will be adopted in this book.
The structural approach
Undoubtedly one of the most well-known propositions in comparative politics is Lipset's (1959) claim that countries having undergone a more extensive process of societal modernization are more likely to be democratic. Partly drawing on Daniel Lerner (1958), Lipset sustained the idea that countries tend to undergo a large number of more or less parallel and simultaneous processes, most notably industrialization, urbanization, increasing levels of education, rising national income and a continued spread of communications technologies - all of which may be viewed as different aspects of modernization. Democracy, then, is but the tip of an iceberg composed of these larger processes at work in society.
Lipset's hypothesis posits what is still one of the most well-established, and yet - probably for that very reason - most controversial relationships within studies of comparative democratization (see, e.g., Przeworski etal. 2000; Boix and Stokes 2003; Epstein et al. 2006). But Lipset's work is also paradigmatic in the way it has inspired an outburst of structural accounts of what explains democratization. The last decades have seen tremendous growth in the number of structural factors added to the list of democracy's hypothesized determinants. Pertaining to aspects of the domestic economy other than modernization, these include state involvement in the economy (Brunk et al. 1987; Burkhart 2000), income inequality (Muller 1988; Burkhart 1997), economic crises (Gasiorowski 1995; Bernard etal. 2001; 2003), and natural resource abundance (Ross 2001; Dunning 200S). Conjectured non-economic but still domestic structural influences on democratization include country size (Dahl and
rift"
«12 0" B
Explaining democratization
The
strategic approach
19
Tufte 1973), religious composition (Lipset 1994), societal fractionalsza-tion (Dahl 1971; Rabushka and Shepsle 1972; Horowitz 1985), colonial heritage (Bernard et al. 2004), social capital (Paxton 2002), and mass political culture (Inglehart 1997; Inglehart and Welzel 2005). There are also structural accounts that suggest various international determinants of democratization, including economic dependency (Bollen 1983) -nowadays rephrased in terms of globalization (Li and Reuveny 2003; 2009; Rudra 2005) - drn.H.tutic Cit'ii-ion iSn-r i^I: Brinks and Coppedge 2006) and regional organizations (Pevehouse 2005).
What distinguishes these theories is that they locate the most significant triggers of democratic advancement in social or economic structure, that is, beyond the immediate reach of human agents. In many instances there are not even any actors or agents specified within these explanatory models. The causal process conveyed is largely mechanical: a structural shift occurs in the "environment" that precipitates change in the political regime. True, this mechanical view rarely, if ever, reflects an onrological position. Most theorists within this camp - definitely including Lipset himself - would agree that only through the behavior of individual and collective actors could the process of regime change actually occur. But in terms of the explanatory properties they specify in their theories, the role of human agency is "black boxed." The distinguishing features of structural theories of democratization, then, are (a) that structural factors are given causal primacy, and (b) that questions on how, why and which social actors are motivated to produce specific regime outcomes are left underspecified.
Much the same could be argued about studies on the institutional determinants of democracy. These include a renewed interest for some of the classic debates in how varying constitutional frameworks, such as forms of government and electoral systems, may affect democratic stability (Gasiorowski and Power 1998; 1998; Bernard et al. 2001; Cheibub 2007; Norris 2008). As of late there has also been a growing interest in how different authoritarian institutions may affect the prospects for democratization (Bratton and van de Walk 1997; Snyder and Mahoney 1999; Hadenius and Teorell 2007; Magaloni 2008; Brownlee 2009). Although institutional explanations by and large are more sensitive to problems of agency (in particular, seeGeddes 1999; 2003), they still bear a strong resemblance to the structural theories of democratization by locating causal primacy in institutional conditions exogenous to human agency.
The structural approach to explaining democratization contains both strengths and weaknesses. On the positive side, most structural theories are highly general and, as a result, have been subject to extensive empirical testing, mostly from within the probabilistic worldview of statistics. On the downside, however, the palette of suggested structural determinants of democratization has grown out of proportion and is in great need of theoretical synthesis and integration. The black-boxing of causa! mechanisms also means that most structural theories of democratization lack micro-foundations. In Samuel Huntington's (1991, p. 107) famous words, the structural approach is thus in great need of moving from "causes" to "causers" of democratization.
The strategic approach
In a widely cited article Dankwart Rustow (1970, p. 340) criticized Lipset and the wider structural literature on democratization for neglecting the "genetic question of how a democracy comes into being." To address this neglect Rustow furnished a process model of democratization, highlighting certain "phases" - the preparatory, decision and habituation phases - through which all countries pass on their way from authoritarian to democratic rule. It would take another 26 years, with the publication of O'Donnell and Schmitter (1986), before Rustow's model (1970) would reappear in a form that literally transformed the study of comparative democratization. The key components of this emerging "transition paradigm" (Carothers 2002) were, however, already present in his original contribution. Most notably, these include the idea of democratization as a process occurring according to a certain sequence of "phases," the notion that (almost) no structural prerequisites exist for the initiation of this process, and the key part played by agency and strategic decision making, particularly among political elites.
In O'Donnell and Schmitter's (1986) terminology, which is now the most well known, Rustow's (1970) "preparatory" phase is called liberalization, the "decision" phase is democratization, and the "habituation" phase is consolidation.' The installation of a democratic regime is
2 Although O'Donnell and Schmitter (1986, pp. 11-13) originally termed this third phase "socialization," they also used the word ''consolidation'" at various places in their book, and this was the term that stuck (see, e.g., Schedler 1998).
20
Explaining democratization
Xhe strategic approach
21
again largely explained through a process of elite interaction. The actors involved, drawing a highly influential distinction, are primarily the "hardliners" and the "softliners" of the incumbent regime, although the opposition is also to some extent taken into account. The result of these elite interactions is conditioned to a high degree by the bargaining skill of the actors involved. In addition a number of random conditions and unexpected events have the potential to determine the outcome. Alrhou,-.!- O'H..;!,:-,:; K'-mvv.i-r . I^Sb. 4-5) acknowledged the possibility that long-run trajectories of regime change were shaped by structural forces, they stressed the indeterminacy of the short-term dynamics. Whereas both virtu and fortuna played an influential part of their theory, they thus gave short shrift to Machiavelli's third determinant of events: necessity.
Similar to Rustow, O'Donnell and Schmitter (1986) discerned otic common denominator of all democratization processes (although a different one): "that there is no transition whose beginning is not the consequence - direct or indirect - of important divisions within the authoritarian regime itself." In other words, no transition to democracy could be forced solely by an opposition facing a cohesive, undivided authoritarian regime (1986, pp. 19-21). Democracy, in other words, is installed through an elite-driven process from above, with other segments of society playing at most an "ephemeral" role (1986, p. 55). In essence, this means that the process of transition itself determines its outcome. Apart from this proposition, the most widely held generalization from the transitions project is that there is no generalization about the "prerequisites" of democratization (see, e.g., Karl 1990; Karl and Schmitter 1991; Shin 1994). Democracy may crop up under extremely varying historical, institutional and structural conditions.3
The strategic approach to explaining democratization has had an enormous impact, both within and outside the scholarly community. Bringing actors and agency to the analytical fore was a substantial contribution at a point in time when the structural approach was perceived as presenting a deterministic stranglehold on the prospects for democratization, not least for real-life pro-democratic forces in
3 Considering the centrality of elite actors and snategic decision making in this approach to explaining democratization, it is no surprise that same of its insights have heen amenable to game-theoretic interpretations (Przeworski 1991; Gates and Humes 1997, chap. 5; Weingasr 1997; Crescenzi 1999; Swaminathan 1999; Colomer 2000; Sutter 2000).
countries under authoritarian rule (see Przeworski 1991, pp. 97-98). The work of Juan Linz (1978) and others has shown that the opposite process of democratic breakdown is also amenable to similarly volun-taristic theorizing (Cohen 1994; Alexander 2002). Moreover, the generalization that democratic transitions always originate from above through a split in the authoritarian regime is still high on the research agenda (Geddes 1999; Bratton and van de Walle 1997; McFaul 2002). Even the "no prerequisites" proposition has received empirical support in a large-scale statistical analysis showing that whereas the survival of democracy appears to be largely affected by structural forces, transitions from autocracy to democracy occur almost haphazardly (Przeworski et al. 2000) - although that finding is contested (Boix and Stokes 2003; Epstein et al 2006).
Nevertheless, the strategic approach suffers from several theoretical shortcomings. It concentrates on the short-term calculations of a narrow set of actors during a condensed period of time. The approach almost by definition thereby excludes the possibility of long-term forces shaping the outcome. In addition, it has never adequately addressed the question of what it is that conditions the presence of certain sets of actors in certain circumstances, and what determines their preferences, interests and beliefs.4 For this reason, the explanatory factors that enter the strategic model are usually very proximate to the outcome they should explain. Explaining democratization with reference to characteristics of the process leading to that outcome borders on tautology.5
4 Notable attempts to overcome this voluntaristic bias include Snyder (1998) and Brownlee (2002).
5 A clear example of these shortcomings comes from a comparative study of regime outcomes in no fewer than twenty-four transition processes across the globe between 1973 and 1990 (Casper and Taylor 1996). Despite an unusually sophisticated research design compared to other work within the strategic tradition, the nature of the strategic approach severely affects the quality of Gretchen Casper and Michelle Taylor's conclusions. One of their main "findings," for example, is that when incumbents acquiesce to the democratic demands of the mass public, the probability of a successful transition to democracy increases. But this conclusion is simply a way of answering the question of what explains democratization by posing a new one: what explains why incumbents sometimes acquiesce to popular demands, and sometimes not? Casper and Taylor's (1996) analysis clearly shows how little explanatory leverage is gained by moving
the explanatory factors too close to the outcome that is to be explained. This demonstrates that the basic flaw in the strategic approach is theoretical and cannot be solved through a more elaborate research design (cf. Kitschelt 1992, p. 1,028; Rueschemeyer et al. 1992, pp. 32-33).
22
Explaining democratization
Xhe- social forces tradition
23
The social forces tradition
i
Alongside the structural and strategic approaches another vibrant tra- |
dition has evolved that instead seeks the origins of democratic rule in the 1
characteristics of and relationships among social classes in society. As ;
with the former approaches, its roots can be traced back to one seminal £
contribution, this time by Barrington Moore (1966). Moore's main 1
concern was the role of the landed upper classes and the peasantry in |
explaining why the transformation from agrarian to industrial societies f ended in democracy in some countries and in left-wing or right-wing
dictatorships in others. Ironically, however, Moore's fame has not i-
primarily been based on his analysis of these two social classes, but on |
his remarks on another more peripheral, collective actor in his book: the *
middle class. Moore's dictum "No bourgeois, no democracy" {1966, t, p. 418) is still among the most widely cited phrases within the entire literature on democratization.
Although empirically challenged,5 Moore's way of theorizing the societal conditions that bring democracy about has attracted many followers.
Key to this approach is a focus on class-based definitions of collective f"
actors, primarily driven by material interests. These collective actors are '
thus "most likely to champion democracy when their economic interests r
put them at odds with the authoritarian state" (Bellin 2000, p. 177). ! Democracy is conceived of as forged from below, through a power
struggle among social forces with competing economic interests. While \ this approach is actor-centric, in sharp contrast to the structural
approach, it paradoxically lacks a theory of agency. Psychological per- \
ception or calculation is never an integral part of its explanatory scheme, L
which of course makesit quite distinct from the strategic approach. I
All of these ingredients are clearly present in the most renowned f
contribution to this tradition after Moore: the book by Dietrich \
Rueschemeyer et al. (1992). Foreshadowed by the insightful observations 1 of Goran TEerbofn (1977), Rueschemeyer et al. argue that the working
rather than the middle class, or bourgeoisie, is the foremost champion of I
democracy. On the other extreme stands the landlord, who - following '
6 For an excellent review of rhe mixed and conditional support received by the
Moore thesis in later comparative-historical work, see Mahoney (2003). Two j:;
important contributions that in various ways complement and refine Moore's -|
approach and findings are Luebbert (1991) and Downing (1992). e
the logic of Moore - resists democracy by any means. The bourgeoisie in Rueschemeyer et #/.'s account takes a more ambiguous stance toward democracy. With this set of class actors in place, and their stance toward democracy defined, the approach to explaining democratization becomes a model of "relative class power." What largely explains the rise and survival of democracy is conjectured to be a set of twin factors: the .strengthening in size and density of the organization of the working class, and the weakening in size and power of the large landowners. According to Rueschemeyer et al. (1992), this pattern also explains the correlations between indicators of socioeconomic development and democracy unraveled by the modernization school (1992, in particular pp. 46-61).
Incorporating actors into the structural approach, tying the strength of these actors to structural conditions, and including some general assumptions on the interests that motivate these actors, are probably the main achievements of the social forces tradition. These notwithstanding, there are several blind spots or other weaknesses in need of elaboration. First, this theoretical approach still awaffs~fiesting based on systematic evidence collected through both space and time. Its empirical foundations are hitherto dominated by case studies, and primarily from Western Europe and Latin America. Second, this approach smacks of determinism, By tracing the roots of democratic development to the presence or absence of a more or less complex series of "conditions" -most notably the strength or weakness of this or that social class or class alliance - adherents of the social forces tradition come close to providing deterministic explanations of regime outcomes.7
Third, despite efforts to relate the basic class interests to characteristics of the regime, the social forces tradition has problems in clearly specifying under what conditions different classes support or oppose democracy. Rueschemeyer et al. (1992) move in the right direction by stressing that labor's embrace of democracy is socially constructed - not predetermined by class position in the class structure. However, they fail to point out the same predicament for other social classes, and they lack a theory for explaining the process of preference formation even for labor (cf. Bellin 2000). Another problem in the social forces tradition also indicates a
7 Nowhere is this tendency more evident than in Gregory Luebbert's (1991, p. 306) assertive claim that "leadership and meaningful choice played no role in the outcome."
24                                                    Explaining democratization :
need to dovetail strategic and class-based theories of democratization: the Í lack of a theory of how organized class action from below interacts with
elite decisions from above to bring about regime change or stability. An j;
important attempt in this direction is Ruth Berins Collier's (1999) bnd- t ging approach, in which both class and elite actors, protest activity and negotiation, moves by excluded groups and responses by regime incumbents, are incorporated into the same analytical framework.8
A fifth weakness of the social forces tradition is its relative neglect of i non-class collective actors or of non-material collective identities more f; generally. Social forces other than economic class actors, such as university students, human rights activists, church leaders and regional elites, t which have been professed to play a part in the popular mobilization t surges behind recent democratization elsewhere in the world (Bratton Í and van de Walle 1997, 107), do not fit easily into a class theory of f collective action (Slater 2009). Although traditionally not committed to f the study of democratization, the literature on social movements has > recently made some promising steps toward such a broader theory of * collective identities, and their links to organized protest and democratiza- I tion (Markoff 1996; Tilly 2004). Moreover, the broader study of "civil i society" could contribute to the social forces tradition on this score (Gill . 2000; Bermeo 2003a; Kopecký and Mudde 2003).
ř
The economic approach
The latest approach to explaining democratization uses the tools of f
economics to understand regime transitions and democratic stability. f
There are three key components of this approach. First is the incorpora- f
tion of a wider set of actors than merely regime elites and opposition j-
groups. Basically, the preferences of the entire population matter for f
regime outcomes, and, critically, non-elites may exert important influ- j,
ence over regime outcomes through the organization of protest action. [
Second, the economic approach purports to explain the preferences í of these actors in terms of structural preconditions and material
resources.9 The third component is the use of deductive formal \
.:}§
8 In a similar vein, Nancy Bermeo (1997) discusses how elites calculate the risks of sustained mobilization by marginalized groups.
9 Both these features set the economic approach apart from the simpler game-theoretic accounts of democratic transitions following in the tradition of i
The economic approach 25
modeling: explicitly stated assumptions and mathematically deducted corollaries from these assumptions in the form of predictions derived from equilibrium conditions.
Although they lack this third component, Stephan Haggard and Robert Kaufman (1995; 1997) may be seen as an early precursor to this emerging literature. While retaining its actor-centric essence, they explicitly criticize the transition paradigm for failing "to address the factors that shape actors' political preferences, the conditions under which these preferences change, and even the identity of pivotal actors" {Haggard and Kaufman 1995, pp. 5-6). The first more fully blown contribution, however, is that of Boix (2003). Firmly based in the tradition of formal economic theory, Boix assumes that people only care about their income, and hence evaluate their preferences for democracy or autocracy in terms of this. By implication, the fundamental struggle over democracy occurs between the rich and the poor. Based on the median voter theorem and the assumption that the median voter is poor, it follows that under democracy the poor set a positive tax rate in order to redistribute income (in the tradition of Meltzer and Richard 1981). Under (right-wing) dictatorship, by contrast, policies are not determined by the median voter but by the rich themselves, who choose a zero tax rate and no redistribution. Thus, the poor generally prefer democracy whereas the rich prefer (right-wing) dictatorship.
Two fundamental parameters may, however, alter this scheme of things. The first is income inequality: the more equally distributed the level of income is to begin with, the less the rich have to fear from conceding democracy to the poor.10 The second is capital mobility, or asset specificity: the less productive an asset is at home relative to abroad, the lower the tax rate will have to be in order to avoid capital flight. This means that the cost of democracy to the rich decreases as asset specificity decreases.11 From these simple assumptions, Boix
Przeworski (1991), where preferences are taken as given and only elite actors are being analyzed.
10 Peter Rosendorff (2001) presents a model similar to Boix's (2003) in structure and outcome with respect to the relationship between income distribution and democracy, but also derives predictions for the effects of the size of capital stock and size of the workforce.
11 Interestingly these two fundamental parameters parallel Bellin's (2000) discussion of two factors that determine the stance toward democracy among the capital class: fear (of redistribution) and state dependence (the latter, among other things, leading to low capital mobility). The same two factors
26                                                        Explaining democratization «
(2003) develops a simple static game-theoretic model in which the rich f
may choose to repress (sustain autocracy) at a certain cost or not repress •
(allow democracy), and the poor may choose to revolt (mobilize against £
the regime) or acquiesce. The two most important observable implica- J tions from the model concern income inequality and asset specificity: as
they decrease, the likelihood of democracy increases. |
Although in large part based on the same fundamentals, Daron \
Accmug'.t. r.:c Tai;:es RoH'Xvn ;2;i|i<;i make another contribution to S
this emerging literature. In a setting basically made up of the same actors t
(rich and poor), and with the same preferences for policies (tax rate) and j,
regime outcomes (dictatorship and democracy) as in Boix's model, \
Acemoglu and Robinson ask why it is that the elite would concede 1
democracy under the threat of revolution and not simply switch to a "
more redistributfve policy. Their explanation stresses the importance of \
democracy as a solution to a commitment problem. As long as those £
wielding political power (the rich) are not themselves the beneficiaries of J
redistributive policies, they will not make a credible commitment to \
honour in the future the promise of such policies made in the past ,
under the threat of revolution. With democracy, however, this problem j-
is solved since the wielders of power and the beneficiaries of redistribution *
are the very same actors (the poor). Following this logic, Acemoglu and |
Robinson (2006) propose a dynamic model of democratization much j
richer in subtle technical detail than the simple static setup of Boix (2003). r
There are two key virtues of this new and emerging theoretical \
approach to democratization. The first is consistency and integration. t
The deductive tools of formal theory ensure that these theories in \
themselves are not self-contradictory. When different causal factors -f»
are discussed in tandem, their inter-linkages are fully explored. But f
even more importantly, the economic approach integrates all of the f
three previous approaches to democratization. This approach provides ^
structural conditions (such as the level of income inequality), it relates r
these to the preferences (such as the demand for redistribution) and >-
actions (such as popular mobilization) of "social forces," and it models |
how these conditions affect the strategic choices made by political elites \.
(such as the decision to extend the suffrage). This integrative effort links |
to the second virtue: transcendence of the structure and agency divide. |
|
also appear to have shaped the shift in interests of business elites in El Salvador <
and South Africa prior to democratization ("Wood 2000), {
fhe economic approach 27
gy virtue of their appeal to micro-foundations of macro-phenomena, these theories satisfy all three criteria Jon Elster (1983, p. 86) once stipulated as ideals for a "general sociological theory":
(i) the explanation of individual action in terms of individual desires and beliefs, (ii) the explanation of macro-states in terms of individual actions, and (iii) the explanation of desires and beliefs in terms of macro-states.
The greatest drawback of the economic approach, to date, is its relative lack of empirical corroboration. Following Rebecca Morton (1999), we may distinguish among three types of tests of formal models: evaluation of predictions, assumptions and alternative models. The economic theories of democratization have hitherto been almost exclusively evaluated in terms of their empirical predictions. The most systematic effort in this regard is Boix's large-« and historical analyses (2003, chap. 2-3), whereas Acemoglu and Robinson (2006) almost exclusively provide anecdotal evidence. Neither, however, provides much evidence for the Meltzer-Richard theorem on which their models rely - namely, that tax rates increase with inequality under democracy, and that tax revenues in democracies are really used to redistribute income to the poor. There is even ample evidence that these predictions are not quite as robustly borne out as expected (Perotti 1996; Gradstein and Milanovic 2004; Mulligan etal. 2004; Ross 2006).
With respect to assumptions, there have been far fewer, if any, efforts toward empirical corroboration. The most crucial assumptions in the economic approach to democratization are the ones concerning information and preference formation. The world of Acemoglu and Robinson's (2006) model is populated by individuals that are fully informed: of the preferences of all other individuals, of the (assumed) redistributive consequences of dictatorship and democracy, of the costs of repression, of the share of society's economic resources that would be destroyed by a revolution, and so on. These are of course extremely unrealistic assumptions.12 With respect to what determines individuals' preferences over outcomes, people in both Boix (2003) and Acemoglu and Robinson (2006) are only concerned with their personal income. They thus evaluate their preferences regarding democracy vs. dictatorship solely in terms of
L True, they are relaxed in Boix's (2003) model, but only with respect to one piece of information: citizens are assumed to be uncertain about the elite's repression costs.
28 Explaining democratizatin
the net materia] benefit that would accrue to them from living under each type of institution. To the best of my knowledge, however, there have been no attempts to empirically validate these assumptions.13
The third type of empirical evaluation - testing alternative theories -has similarly not been paid much attention in the economic literature on democratization. Few, if any, attempts have been made to pit more than one formal model against another. The greatest need for this arises
when.....dir'rrfPt i.'i;:;!cls roll dillcv.ir s:t>rcs about the underlying
mechanisms responsible for the same empirical regularity, such as the correlation between democracy and economic modernization (see, e.g.,. Chen and Feng 1999; Feng and Zak 1999; Zak and Feng 2003; and Glaeser etal. 2007).
The present approach: theoretical eclecticism, methods combined
In sum, the field of comparative democratization studies is inchoate, diversified and incoherent. Neither the structural, strategic nor social forces traditions have succeeded in stating a generally acceptable theory of democratization. And although the more recent economic approach contains the seeds of a much-needed theoretical integration, it has thus far relied too much on unverified assumptions and empirical predictions. On top of this, methodological divides have even further widened the existing gulfs in our theoretical understanding of democratization. Statistical methods applied to a large number of countries have dominated the structural approach from its inception, and in so far as they have been systematically assessed, this goes for the economic models of democratization as well. The Transitions from Authoritarian Rule edition that epitomized the strategic approach (O'Donnell etal. 1986), by contrast, was empirically designed as single-country studies, whereas the social forces tradition has primarily relied on historical-comparative analysis of country trajectories over long stretches of history. Studies crossing these boundaries are, however, a rare species. In other words,
13 Using World Value Survey data, Torsten Perssonand Guido Tabellini (2009) find that the cumulative experience of democracy at home and abroad (termed "democratic capital") is positively related to aggregate support for democracy in a cross-section of countries. This lends support to one of their assumptions about preference formation, but not to the larger assumptions of Boix (2003) and Acemoglu and Robinson (2006).
The present approach
29
the extant literature on comparative democratization is deeply divided between tables of causal effect estimates and close-up case narratives that rarely speak to each other.14
Given the relative merits and drawbacks of each methodological approach, this is of course no ideal state of affairs. The general strength of statistics include claims of generality, probabilistic estimates of causal -effect magnitudes, and the ability to incorporate multiple putative causes simultaneously, thereby also controlling for confounding factors behind each cause. Case studies, whether comparative or single-case, are at a relative disadvantage in these regards. They tend to be concentrated in time and space, thereby both impairing generality and the range of confounding factors that can be held constant across cases compared. Case studies instead thrive on sensitivity to temporality, agency and process. Whereas statistical analysis, with data over time, may provide insights into temporal priority, the results tend to be sensitive to untested assumptions about the length and equality of causal lags. Statistical approaches are also not capable of tracing causal mechanisms in terms of intervening variables, but these mechanisms tend to be devoid of actors, action and processes of interaction.
Departing from this depiction of the current state of affairs, in this study I will deploy two strategies to further the field of comparative democratization: theoretical eclecticism, and a combination of statistical with case study methods. The theoretical ambiguity as to what forces drive democratization implies that no particular approach may be elevated above the others in terms of testing priority. Thus, I will not test any singular explanatory approach, but rather a vast range of theoretical predictions stemming from several approaches. This eclectic strategy will, as a first step in each of the following five substantive
An exception has in recent years evolved among area specialists studying democratization within different world regions, such as Sub-Saharan Africa (Bratton and van de Walle 1997), the post-communist countries (Fish 1998; Bunce 2000; McFaul 2002), and Latin America (Mainwaring and Perez-iinan 2003; 2005a). The virtue of this regional specialization is that it combines deep contextual knowledge from comparative case studies with systematic statistical analysis. What these studies never make quite clear, however, is that as long as they rely on cross-country variation they can only uncover factors that explain variations in democracy among countries within specific regions. What they fail to assess is the extent to which these same factors have affected democracy in the region as a whole. This latter question can only by addressed by making comparison between regions - that is, more or less global analyses.
fylod&'ri^10™' performance and resource abundance
55
\ The poiver of prosperity: economic I determinants
It only takes a quick glance at the world to reach the conclusion that democracy and prosperity are related. Richer countries are no doubt on average more democratic. But is this relationship causal? More specifically, were prosperous or socioeconomically more devel- : oped countries during the third wave more likely to democratize? What then about short-term fluctuations in economic performance: was that a prosperity-enhancing factor moving in the same direction? It takes no 1 more than a second quick glance at the world to find an obvious exception to the relationship between prosperity and democracy: the oil-rich autocracies in the Middle East. Were oil or other sources of natural resource abundance a more general impediment to democratization during this time period?
In this chapter I will address these questions in an attempt to systematically assess a series of economic determinants of democratization. These are divided into two sections: one on economic determinants observed in my full estimation sample of 165 countries; the other on determinants observed for a more restricted sample. I find statistical : support for four economic determinants: the level of socioeconomic modernization, short-term economic growth, oil abundance, and freedom from state incursion in the economy. Through both large-» analysis and case study evidence I then assess the mechanisms that underlie . the first two of them.
Modernization, performance and resource abundance
Since Lipset's (1959) seminal argument, summarized in Chapter 1, there have been countless studies confirming that one of tire most stable determinants of democracy across the globe is the level of socioeconomic modernization. That said, Lipset's thesis still faces several unad-dressed challenges. To begin with, the empirical support for his proposition has for the most part been based on measures of
modernization in terms of economic development, such as energy consumption and GDP per capita. This pertains both to earlier cross-sectional studies (for an overview, see Diamond 1992) and to the more recent tests based on pooled time-series data (Burkhardt and tewis-Beck 1994; Helliwell 1994; Londregan and Poole 1996; Gasiorowski and Power 1998; Barro 1999; Przeworski et al. 2000; Ecix a'11' Stokes 2003; Epstein et al. 2006). However, in Lipset's (1959) original account, as well as in the early studies following in its wake (Cutright 1963; Neubauer 1967; Olsen 1968; Winham 1970), a much wider range of indicators of socioeconomic development was employed. Apart from national income they included industrialization, education, urbanization and communications. According to modernization theorists these developmental processes should be viewed as parts of one underlying syndrome, "socioeconomic modernization," which eventually enhances democratic development (Lerner 1958; Deutsch 1961). This broader theoretical underpinning for the Lipset hypothesis has received surprisingly little attention in the more recent comparative democratization literature. I will try to remedy this situation below by treating socioeconomic modernization as a coherent syndrome with multiple observable indicators.
Secondly, there is the widely cited finding by Przeworski etal. (2000) that socioeconomic modernization does not trigger transitions to democracy, but instead helps to sustain democracies once installed. This finding has been both confirmed (Persson and Tabellini 2009) and criticized (Boix and Stokes 2003; Epstein etal. 2006) on empirical grounds, but mostly based on a dichotomous measure of democracy.1 By separating the effects on upturns from those on downturns, in this chapter I will perform a systematic test of this finding using a graded democracy scale,
A third challenge comes from recent work purporting to show that the effect on democratization, of both economic income and education only applies across countries at a given point in time, but not within countries over time (Acemoglu et al. 2005; 2007; 2008; Robinson 2006). The authors of this series of papers instead argue that different societies  may   have  embarked   on  divergent political-economic
The one exception I am aware of is Acemoglu et al. (2007), who (in tables 3-4) make use of the same technique as I in order to separate the effects of transition toward and away from democracy.
54
5(i
The power of prosperity
Modernization, performance and resource abundance
57
development paths at certain critical junctures some 500 years ago, with some countries following a high income-high democracy path, others a low income-low democracy path, but without the two ever being causally related. In the robustness tests supplied in Appendix C, I address also this challenge, again based on ray broader measure of socioeconomic modernization.
A theoretical argument developed alongside the modernization hypothesis has been concerned with the impact of economic performance. Both autocracies and democracies, the argument goes, are more likely to break down when facing temporary performance crises, since this means "a reduction in the resources available to political elites for sustaining bases of support" (Haggard and Kaufman 1995, p. 29; 1997). The large-ra empirical support for this contention has mostly been based on yearly growth rates as the measure of performance, and again mostly on dichotomous conceptions of the dependent variable, simply indicating whether regimes are authoritarian or democratic. Two findings have been predominant. On the one hand that growth is negatively related to transitions from authoritarian to democratic rule -. or, inversely, that authoritarian regimes fall under the pressure of economic crisis (Gasiorowski 1995; Remmer 1996; Przeworski et al, 2000). On the other hand growth has been shown to positively affect democratic survival, implying that democracies too are vulnerable to economic crises (Przeworski et al. 2000; Bernard et al. 2001; 2003). These results do not translate easily into contexts where graded measures of democracy are being used. They could, however, imply that the coefficients for economic performance should be differently signed depending on the direction of change in the democracy scale, which might explain why the few studies that have tested them on graded measures have produced weak and inconsistent results.
A more robust finding appears to be the anti-democratic effect of natural resource abundance. In a set of regressions predicting the development of democracy over time, Ross (2001) found that both the abundance of oil and of other non-fuel minerals as the primary sources of national exports had a markedly negative effect on the prospects for
1 Using the same democracy index (Polity), but different controls, John Londregan and Keith Poole (1996) found a negative but small short-term impact of growth on democratization, whereas Quan Li and Rafael Reuveny (2003) found no effect of growth bur a positive effect of inflation that decreased over time.
democratization. Earlier studies claiming to show the negative impact of oil had only made cursory remarks on the poor performance of democratic governance in a few oil-producing countries on the Arabian Peninsula (Helliwell 1994; Barro 1997; 1999). Ross (2001), by contrast, was able to show that the effect occurred on a global scale, and pertained to other sources of strongly profitable materials. According to Ross the relationship is due to the development of a "rentier state" in countries rich in natural resource wealth. Regimes that are predominantly reliant on such vast resources are capable of using both the carrot (tax cuts and patronage) and the stick (repression) to hold contestation at bay.3
Turning to my empirical results in Table 3.1, I first replicate the finding from some 50 years of comparative research on the positive relationship between socioeconomic modernization and democratization. As already noted, this result is based on a composite measure of the entire process of modernization, not simply on one of its macroeco-nomic subcomponents. Interestingly, moreover, during the third wave the effect of modernization should according to my estimates not be interpreted as a tendency among modernizing countries to advance toward democracy. Rather, there was a tendency among less modernized countries to revert toward authoritarianism. True, this tendency is not very strong. A standard deviation change in the modernization index - which is approximately equivalent to a move from the level of Somalia (at the very bottom) to the level of Namibia, or from EI Salvador (at the mean) to the level of Ireland - according to these estimates results in an expected decrease of .061 in the propensity of democracy to backslide toward authoritarianism the following year. The impact on (the absence of) downturns is, however, statistically significant, unlike the effect on upturns.
This pattern bears a striking resemblance to the finding by Przeworski et al. (2000) and Persson and Tabellini (2009) that socioeconomic modernization does not effect transitions to democracy, but hinders reversals from authoritarianism. Moreover, the finding is highly robust to alternative specifications. As Appendix C makes clear, even the
! Thad Dunning (2008), in a recent contribution, adds an important caveat to Ross' (2001) argument by observing that when non-resource wealth is unequally distributed, resource abundance could actually foster democratization (or hinder de-democratizarion) by mitigating conflicts over redistribution.
58
The power of prosperity
idernization, performance and resource abundance
59
Table 3.1 Domestic Economic Determinants
Short-run
	General	Upturns	Downturns	Long-run
Socioeconomic modernization	.038 •*	-.023	.061*"	.472"
	(.021)	(.019)	(.015)	(.242)
Growth..............................-.................------------	-.IM	=.540*»»	.356»*	-2.30'
	(.259)	(.195)	(-171)	(3.210) ■
Inflation	-.002	-.001	-.001	-.025
	(.004)	(.003)	(.002)	(.051)
Fuels	-.228»	-.189*»*	-.039	-2.85**-
	(.120)	(.073)	(.080)	(1.447)
Minerals	-.034	.053	-.087	-.424 :
	(.230)	(.156)	(.182)	(2.872)
* significant at the .10-level, ** significant at the .05-level, "**significant at i
the .01-ievel. |
No. of observations = 3,795; no. of countries - 165; mean years observed per |
country - 23.0. |
Note: Entries are unstandardized regression coefficients with panel-corrected standard "i errors in parentheses. All models also include two lags of the dependent variable and
the determinants in Table 2.1,4.1 and 5.1 as controls. Ail explanatory variables in ihi, s
table have been lagged one year. .
within-country evidence (applying country-fixed effects) supports the view that modernization during the third wave decreased the expected level of backsliding toward authoritarianism. This, in turn, means th: u I do not find support for the argument by Acemoglu et al. (2005; 2007; 2008) that modernization only influences the cross-country correlations.4
When looking at short-term macroeconomic performance, I find no -effect of inflationary crisis on regime change. Growth, however, negatively affected upturns toward democracy during the third wave. Since growth rates are measured in fractions, the coefficient of -.540 implies that for a country whose economy contracted by, say, 10 percentage
4 True, when I apply both country- and year-fixed effects (robustness model 6 of Appendix C), the coefficient for modernization fails to reach statistical significance. This is, however, an extremely demanding test for a slow-moving characteristic such as socioeconomic modernization.
oints, the propensity to democratize the following year increased by 054. This finding, although substantively small, turns out to be highly robust to alternative specifications,5 and is consistent with Haggard and {(aufaan s (3995; 1997) argument that authoritarian regimes are more likely to democratize when suffering from recessionary crisis. The coefficient for downturns, however, while appearing to be statistically significant, hinges on the inclusion of two extremely influential outliers.6 I thus find no robust support for the supplementary view that economic growth kept democracies from backsliding. But the fact that the impact of growth on downturns is positively signed explains why there is no effect of growth on democratization in the general model, nor any discernable impact on the long-run equilibrium democracy level.
Despite the fact that I control for a much larger set of determinants, my results confirm Ross' (2001) findings on the anti-democratic effect of oil (or fuels more generally). According to my estimates the discovery of oil, increasing the export share of oil from 0 to 100 percent of GDP, would have led during the third wave to an expected decrease of .228 in the level of democracy the following year, and to a downward shift of 2.85 in the long-run equilibrium level of democracy. This effect is primarily caused by a tendency among oil-dependent countries to resist increases in their level of democracy (or, by implication, a democratizing tendency among countries not dependent on oil). Probably due to the fact that so few oil-rich countries have reached higher levels of democracy, however, the effect of oil as a trigger of downturns is weak and insignificant.
Whereas I thus confirm Ross' (2001) primary finding with respect to the effect of natural resource wealth, there are some qualifications. The fuels variable is to begin with not robust to all alternative specifications (see Appendix C). Most notably, oil dependency cannot explain movements in the level of democracy within countries (that is, taking into account country-fixed effects). Nor does it exert a significant impact in the imputed sample of all 196 countries across the globe. I thus cannot exclude the possibility that the oil effect is sample dependent. Finally,
5 The only exception is that growth only affects upturns in political rights (according to Freedom House) and the Polity scores, not in civil liberties (see Appendix C). There is also a relatively extreme influential outlier, Panama in 1989, but the impact of growth on upturns is robust to excluding that case. These are Nigeria in 1984 and Suriname in 1980, two cases of devastating military coups staged the year after a serious economic recession hit each country.
60 The power of prosperity t
contrary to Ross, I do not find any significant negative impact of ni -fuel metals and ores.
Economic inequality and freedom |
s ■
I now turn to two tests where I accept a somewhat smaller estimation sample due to the theoretical significance of the propositions involved, j A key prediction from the economic models of democratization, pre- I sented in Chapter 1, is that democratization should be negatively linked I to income disparity. More unequally distributed income means mi r. pressure for redistribution, which autocratic elites are inclined to avoid by use of repression. Similarly, under high levels of income disparity the i rich elites have a larger incentive to stage a coup against democracy to -avoid future redistribution. We should thus expect income inequality to '. be negatively related to movements toward democracy, but positively   1 : related to movements away from democracy (Boix 2003; Acemoglu and Robinson 2006).
Most previous studies on the democratizing effects of inequality have been inconclusive, mostly due to scarce and unreliable income data (src, e.g., Bollen and Jackman 1985; 1995; Muller 1988; 1995a; 1995b). The most ambitious attempt on cross-sectional time-series data is Boix ; (2003), who finds that income inequality negatively impacts on the   -; probability of transitions to democracy and positively impacts on ] democratic stability. Both these results appear to be highly sensitive to specification, however, and only apply to a sample of at best some 1,000    ' '. observations from 50 countries. Due to a novel dataset on economic    * . inequality (measured through the Gini coefficient) I am in Table 3.2 able to test its impact on democratization in a sample about twice as large (and on a graded measure of democracy).
It turns out, however, that no coefficients for income inequality are statistically significant. This also applies to the lower panel of the table, where I allow for curvilinear effects. Ross Burkhart (1997) finds a nonmonotonic relationship between income distribution and democracy, --where the most democratic countries are located at intermediate levels of income inequality. This is actually in line with Acemoglu and   - : Robinson's (2006) more nuanced theoretical prediction, according to   . : which the poor have less incentive to stage a popular revolt against dictatorship in highly equal societies. Together with the elite effect (that fear of redistribution should be decreasing in the level of economic _
Economic inequality and freedom 61
Table 3.2 Income Inequality
Short-run
General      Upturns       Downturns Long-run
linear:
Income inequality	.271	.171	.100	2.36
	(.612)	(.446)	(.352)	(5.34)
Curvilinear:				
Income inequality	2.644	3.229"	-.585	22.8
	(3.2.38)	(1.949)	(2.288)	(28.2)
Income inequality2	-2.745	-3.538	.793	-23.7
	(3.930)	(2.288)	(2.798)	(34.1)
^ significant at the ,10-ievel,    significant at the ,05-level, significant at the .01-levd.
No. of observations - 1,829; no. of countries = 114; mean years observed per country = 16.0.
Note: Entries are unstandardized regression coefficients with panel-corrected standard errors in parentheses. All models also include two lags of the dependent variable and the determinants in Table 2.1, 3.1,4,1 and 5.1 as controls. Ail explanatory variables in the table have been lagged one year.
equality) a curvilinear effect would be the expected pattern. Adding the square of income inequality to my model, however, again produces only insignificant results. In sum, despite the dieoretical elegance of the economic models of democratization, income inequality appears not to have been a significant determinant of democratization during the third wave.
This result also runs against the first of Bellin's (2000) twin factors affecting capitalist elites' calculations regarding the net gains and losses of a democratic regime. This first factor is fear of redistribution, which, again, should be positively related to the level of economic inequality. There is, however, also a second factor in Bellin's theory, which similarly should affect the calculations of labor, namely economic dependency on the state. The more economic actors (capital or labor) depend on state subsidies and sector-specific regulations, the argument goes, the less inclined will they be to oppose an authoritarian regime. This factor also fit into Boix's (2003) framework in terms of capital mobility, since heavy state subsidies should increase the costs of moving capital abroad.
62
The power of prosperity
Socioeconomic
modernization
63
There is a tradition of studies related to this proposition that inquites into the relationship between democracy and "economic freedom," that is, freedom from state incursion in the economy. Previous studies have found intermediate levels of economic freedom to be most conducive to democracy {Brunk et al. 1987; Burkhart 2000) and that economic liberalization triggers democratization (Fish and Choudhry 2007) -although the direction of causality between economic liberalization and democratization may be questioned (Giavazzi and Tabellini 2005).
In Table 3.3 I assess the extent to which state dependence affects democratization through a measure of economic freedom ranging from 0 to 10. Although I again draw on data with widest possible coverage, this is the second test where I must accept a reduced estimation sample (this time restricted to 119 countries). But this time I do find an impact in the expected direction. As the coefficient for downturns in the upper panel indicates, countries with less state incursion in economic activity were during the thitd wave significantly less likely to backslide toward authoritarianism. The effect magnitude implies that by comparing a relatively state-dependent economy such as Zimbabwe, with an average of 4.74 on the economic freedom index during the estimation period, to economically freer Botswana, with an average of 7.19, leads to an average decrease in the tendency to backslide the following year by .042*(8.51-3.85), or approximately .099. Although there are some extreme outliers influencing this finding, it is robust to their exclusion.7 As indicated in Appendix C, moreover, the economic freedom effect passes all robustness checks save one (it does not hold in the imputed sample of 196 countries). This, however, docs not apply for the estimates of a curvilinear effect reported in the lower panel of Table 3.3. Although there is a tendency that downturns are most effectively avoided at intermediate levels of economic freedom, this finding is highly sensitive to a small number of extremely influential outliers.8
This includes three extremely influential confirming outliers (Ghana 1981, Argentina 1976, and Peru 1992), as well as one extremely influential disconfirmrng oudier (Haiti 1991).
The marginally significant effect of the squared index on upturns falters once a single influential outlier, Zambia in 1991, is excluded from the estimations sample. Excluding two relatively influential outliers, Ghana in 1981 and Argentina in 1976, renders the impact on downturns marginally significant.
Table 3.3 Economic Freedom
Short-i
General    Upturns     Downturns Long-run
linear:				
Economic freedom index	.024	-.018	.042*"	.252
	(.025)	(.017)	(.015)	(.245
Curvilinear:				
Economic freedom index	.083	-.142**	.225*'**	.856
	(.100)	(.071)	(.070)	(1.04)
Economic freedom index'	-.005	.012*	-.017***	-.057
	(.008)	(.006)	(.006)	(.089
'significant at the ,10-level, "'significant at the .0J-level, ""significant at the ,01-level.
No. of observations =2,827; no. of countries - 119; mean years observed per country = 23.8.
Hote: Entries are unstandardized regression coefficients with panel-corrected standard errors in parentheses. All models also include two lags of the dependent variable and the determinants in Table 2.1,3.1,4.1 and5.1 as controls. All explanatory variables in the table have been lagged one year.
Socioeconomic modernization and democratic recalcitrance
I will now turn to an assessment of the causal mechanisms underlying two of the economic factors that were linked to democratization during the third wave: socioeconomic modernization and economic crises. There are two reasons motivating the choice of these particular two. The first is theoretical. The putative mechanisms explaining why natural resource abundance hurts and economic freedom helps democracy are relatively unambiguous and well developed theoretically, a point that I shall return to in the concluding chapter. This is, however, not the case for socioeconomic modernization and economic performance, where there is instead a multitude of potential and not fully fleshed-out explanations alluded to in the literature. The second reason is empirical. For oil and minerals, Ross (2001) has already performed some empirical tests of hypothesized mechanisms. To the best of my knowledge, however, there have been no similar tests performed for the mechanisms of socioeconomic modernization or economic crisis. For economic performance, moreover, the most influential study
64
The power of prosperity
socioeconomic
modernization
65
uncovering mechanisms is based on case studies (Haggard and Kaufman 1995), the connection to large-?? studies of which are largely unknown. Both theoretical indeterminacy and empirical shortage thus speak in favor of an attempt to unravel the mechanisms explaining these-particular economic determinants.
So why then are socioeconomically modernized countries less likely to de-democratize? While the literature is replete with suggestions as to.i what explains the general connection between modernization and" democracy, there have been few attempts to address this particular question.9 Drawing on the more general literature, however, I may still work out some testable propositions.
Five ideas have dominated this literature, four of which were actually already present in Lipset's (1959) seminal article. First, according to Lipset development affects the nature of the class struggle in society. . Most importantly, it helps develop a vibrant middle class, turning "the shape of the stratification structure so that it shifts from an elongated pyramid, with a large lower-class base, to a diamond with a growing : middle class. A large middle class plays a mitigating role in moderating conflict since it is able to reward moderate and democratic parties and penalize extremist groups" (1959, p. 83). Although it has been argued that the working class rather than the middle class are the staunchest :■ champions of democracy (Rueschemeyer ei al. 1992; Collier 1999), the idea that modernization affects democracy through its impact on socioeconomic inequality has been widely accepted in the literature, and as 1 of late incorporated in the newest formal models of democratization (Boix 2003; Acemoglu and Robinson 2006).
Second, Lipset argued that economic development helps foster intermediary organizations in society - or, to use the catchword that has popularized this theme, to strengthen civil society. A society with a multitude of voluntary associations, the argument goes, helps check
? Among the rare exceptions are Przeworski (2005), Benhabib and Przeworski (2005|, and Bueno de Mesqmta et al. (2003, pp. 388-92), all of whom develop formal models based on the idea that wealthier societies are more likely to succeed in developing redistribution schemes acceptable to all parries, Another novel interpretation is Persson and Tabellini's (2009) "endogenous selection" interpretation: that income affects the survival of democracy more strongly than the transition from autocracy because countries more productive as democracies are over-represented among democracies. This latter theory does not easily extend to the non-economic components of the modernization syndroine that 1 study, however, and the former theories are not easily amenable to empirical testing.
aiid balance the power of the state, "serve to train men in the skills of politics" and "help increase the level of interest and participation in politics" (1959, p. 84). This theme has also been picked up and eXpanded more recently by, for example, Graham Gill (2000).
Third, Lipset (1959, pp. 83-84) argued that development helps foster democracy by enhancing certain political values, that is, by strengthening political culture. These values, or this culture, are according to Lipset primarily characterized by tolerance toward opposing views and, in general, a more moderate, restrained and rational style with respect to politics and political opposition. Piecemeal reform is preferred over radical change and social revolution. A recent book-length treatment by Ronald Inglehart and Christian Welzel (2005) shows that this tradition is also very much alive in the current debate on how to explain the modernization effect.
Fourth, and related to this, Lipset (1959, pp. 79-80) claimed that one component of the general modernization syndrome has special prominence, namely education. Mostly through its effect on the aforementioned democratic values, education raises the prospects for democracy. This idea is echoed in a recent economic theory arguing that human capital raises the benefits of political participation and draws relatively more people to actively support democracy (Glaeser et al. 2007).10
The one notion that is not easily traced back to Lipset's work, and thus is more genuinely new, is Boix's (2003) theory of asset specificity, and Acemoglu and Robinson's (2006) narrower but related concept of the capital-to-land ratio. Both accounts operate through the expected tax level under dictatorship and democracy. As society modernizes, in Boix's theory, assets become less specific and hence less easily taxed, which weakens the fear of democracy, or the preference for autocracy, among the economic elites. This facilitates both the emergence and survival of democracy. Acemoglu and Robinson (2006, chap. 9) model a similar dimension of the structure of the economy - the capital-to-land ratio -and find that as this ratio increases (e.g., with industrialization), the costs
Yi Feng and Paul Zak (1999) make the similar argument that increasing wealth makes more people take political action in favor of democracy {cf. Zak and Feng 2003). Persson and Tabellini (2009), in turn, present a theory where per capita income positively affects democratization and democratic survival by influencing how ordinary people internalize the true economic benefit of being in a democracy (which is assumed to be positive), which in turn makes them more likely to rebel against dictatorship and defend democracy against coups.
66
The power of prosperity
Socioeconomic modernization
67
of repression and coups increase, the tax rates under democracy decline, and the size and affluence of the middle class increases, which in turn means that the likelihood of democratization and democratic survival increases. This is thus their "modernization story," similar to but nar-. rower than Boix's (2003) notion of asset specificity.
Obviously all of these five partially overlapping propositions invoke ■ phenomena that are difficultto observe empirically in a large sample of countries. Nevertheless, the first two of them have been partially■■: addressed already. There is, to begin with, no support in my data for the contention that income inequality drives democratization (or de-democratization), nor that this would explain the modernization effect.12 To the extent that popular mobilization proxies for the vitality of civil society - an assumption that of course could be questioned (see, e.g., Kopecky and Mudde 2003) - I may further reject the second proposition. The cultural proposition I have dealt with elsewhere (Hadenius and Teorell 2005a; Teorell and Hadenius 2006; cf. Welzel and Inglehart 2006). Suffice to say here that the survey data hitherto gathered to assess this thesis is too scattered across both time and space to allow for any robust conclusions.
What remains then are the fourth and fifth propositions, both of: which at least partly are incorporated as components in my composite modernization index. What happens if I disaggregate this index and these components are pitted against each other in a single model? Although one should expect no clear-cut results from such an exercise considering their high level of inter-correlation, the fourth proposition should lead me to expect that education trumps the other modernization indicators when it comes to explaining downturns. The fifth is proxied by several components, but should (apart from education) lead me to expect industrialization as the primary force hindering de-democratization.
Baizhu Chen and Yi Feng (1999) instead suggest a model where economic development leads to democracy through the choice of redistribution policy. This idea is similar to Przeworski (2005) and Benhabib and Przeworski (2005), although the latter two only pay attention to why democracies survive at high levels of per capita income.
True, the nature of the class structure could tap into other dimensions than ; economic inequality. Since I lack systematic cross-country data on class structure, I can, however, not test this idea more directly. Moreover, Renske Doorenspleet :■ (2005) finds no significant relationship between transitions to democracy and measures of the size of the middle vs. working class in a country.
Table 3.4 Explaining the Modernization Effect (downturns)
Model
(1)
(2)
(3)
(4)
(5)
Industrialization
-Education
Urbanization
Health conditions
Media proliferation
Log(GDPZcapita)
Socioeconomic modernization No. of successful coups
No. of observations No. of countries
-.001 (.002) -.000 (.001) -.001 (.001) -.005 (.030)
.042" .065*»* .046**
(.017)   (.013) (.010)
.016
(.022)
2,485 2,88' 143 150
-1.68* (.105) 2,884 150
.061** (.015)
3,792 165
.033**
(.012)
-1.72*
(.095)
3,792
165
'significant at the ,10-level, "significant at the ,05-level, ""significant at the .01-fevel.
Note: Entries are unstandardized regression coefficients with panel-corrected standard errors in parentheses. All models use downturns in the level of democracy as the dependent variable, and also include two lags of the level of democracy and the determinants in Table 2.1, 3,1, 4.1 and 5.1 as controls. All explanatory variables in the table except the no. of successful coups have been lagged one year.
As model (1) of Table 3.4 makes clear, however, neither of diese expectations is borne out in the data.13 As can be seen, neither the level of education, nor industrialization, urbanization or even national income affects the tendency to avoid downturns, when all other
13 I have in this model made two groupings of the nine modernization indicators making up the overall index (see Appendix A), in order to introduce some order to the interpretations, First, the two health indicators (life expectancy and infant mortality) and second, the three indicators of media proliferation (the number of TVs, radios and newspapers per capita) are made into two separate factor indices.
68 The power of prosperity
modernization indicators are simultaneously being controlled for. But one indicator does exert a significant impact, even under these extremely tough controls: the level of media proliferation. Of course, the degree of multicollinearity (the extent to which the independent variables covary) is extreme. But it is as extreme for the media proliferation component as for the other modernization indicators. It is thus a noteworthy finding that this component comes out as the sole significant determinant.
This is a truly unexpected finding, to the best of my knowledge not anticipated in any of the literature purporting to explain the modernization effect.14 Why would it be the case that media proliferation helps obstruct democratic backsliding? Although this is not the time and place to delve into such a theory-building effort, two observations on the special nature of the media are in place. First, there is some evidence that television, in particular, is the most forceful check and balance underpinning democracy. This evidence comes from systematic analysis of the so-called "vladivideos," that is, the recorded corrupt transactions by Vladimir Montesinos, Peru's secret-police chief under President Alberto Fujimori in the mid-1990s. These transactions reveal that the typical bribe paid to a television-channel owner was much larger than that paid to an opposition politician or to a judge. If the size of the bribe price indicates how much Montesinos was willing to pay to buy off those who could have checked his and Fujimori's authoritarian rule, television was in other words democracy's staunchest guardian (MacMillan and Zoido 2004).
Second, as opposed to industrialization or education, the hypothesized effects of which are not dependent on the current regime, media proliferation should be expected to have a democratizing impact that increases with an already achieved level of democracy. Media proliferation does not lead to democratization in an authoritarian context because widespread media outlets could be a forceful tool in the hands of an authoritarian regime if media freedom is not protected (Norris 2008, p. 189). As media freedom increases, however, the potential for widely disbursed communications technologies to safeguard democracy is unleashed. In other words, media proliferation can hardly
In Upset's (1959) original argument, measures of media proliferation were simply entered as proxies for national wealth, without any further theoretical justification.
Socioeconomic modernization 69
be expected to foster democratization, but, if free from state control, it should be able to hinder de-democratization.
Unfortunately, this theory of an interaction between media freedom and proliferation is not easily amenable to empirical testing. I would instead like to add another piece of empirical evidence to help address this puzzle in the future. More specifically, how does a counrry backslide into lower levels of democracy? Although there are alternatives, including electoral misappropriation by the ruling party, one prominent path should be that of the coup d'etat. If media proliferation helps in staving off backsliding by preventing government coups (either from happening or from succeeding), one should expect to see a reduction of its effect on downturns once successful coups are being controlled for. This is exactly what I observe in models (2) and (3) of Table 3.4. The impact of media proliferation, which is here entered alone as my most prominent modernization indicator, drops from .065 to .046 when successful coups are controlled for. This makes for a reduction of some 30 percent of its effect.
Two caveats should be borne in mind when interpreting this result. First, media proliferation significantly affects downturns even regardless of coups, which means that there are other important pathways to be explored. Second, I am not in a position to claim that I have "explained" 30 percent of the effect of media proliferation this way. The reason is that the impact of coups on downturns is truly tautological. To say that democracies fall when successful coups are instigated is not to claim much.
What this finding does indicate, however, is that a fruitful avenue for further research on this topic would be to explore the ways in which media proliferation may defer coup attempts, or make them less likely to succeed. It has for example been observed that one of the first objectives of a military seizure is to grasp control of the key broadcasting installations, such as national TV and radio stations (see, e.g., Nordlinger 1977, p. 103). To the extent that media proliferation proxies for a broad array of media outlets, such seizure may be more difficult to carry out. Alternatively, "scandal-hungry journalists" may be "swift to scent out any new coup conspiracy" (Randall 1993, p. 644). To develop and test such assertions, a fruitful avenue for future research could thus be to study failed or aborted coups under conditions of high media proliferation.
Of course, as a final word, one should perhaps not take the prominence of media proliferation among the modernization indicators too
70
The power of prosperity
Economic crises and authoritarian vulnerability
71
literally. To the extent that one wishes to continue treating socioeconomic modernization as one coherent syndrome, however, a similar logic for further research suggests itself. As models (4) and (5) of Table 3.4 show, the effect of this general index also decreases substantially once successful coups are controlled for. In effect, almost half the impact of socioeconomic modernization on downturns is due to its ability to deter or derail coups. To explore why this might be the case_ is definitely worth further inquiry.
Economic crises and authoritarian vulnerability
My mechanism-oriented account of why economic crises trigger democratic upturns is based on the case studies of Haggard and Kaufmann (1995). As already noted in Chapter 1, an explicit ambition of their theoretical model is to bridge the structure and agency divide by linking actors' preferences and action to exogenous economic conditions. The key presumption is that authoritarian regimes, while lacking the kind of diffuse legitimacy which may be bestowed by competitive elections, still rest on "authoritarian bargains" with specific support groups. More precisely, they mention three such pivotal sets of actors: "private-sector business groups," "middle-class and popular-sector organizations," and "military and political elites who control the state and the main instruments of coercion." Declining economic conditions and corresponding pressures for policy adjustment potentially disrupt the authoritarian bargains forged with all three, thus creating a more hospitable environment for democratization (1995, p. 29).
First, tough stabilization programs are likely to threaten several bases on which private-sector business groups tend to cooperate with the regime, such as the protection of private property, state-led development policies that favor certain sectors, or more particularistic favors and rents. When the authoritarian government's ability to deliver in these areas is weakened due to economic crisis, the chances are that business groups increasingly perceive democracy as a preferable alternative. As a result, they are more likely to defect from the authoritarian bargain.
Second, authoritarian regimes often deflect opposition from popular sector groups through public employment, public works projects antl consumer subsidies. Again, as an economic crisis hollows out state resources, the government's ability to sustain these material rewards
weakens. The primary weapon of disaffected popular sector groups is then the mobilization of protest. These protests, especially large-scale strikes and mass demonstrations, become a dual weapon against the incumbent regime, by both fueling deteriorating economic performance itself, and by turning economic grievances into political demands for regime change.
Third, and perhaps most importantly, economic crisis exacerbates divisions within the authoritarian regime itself. Haggard and Kaufman (1995) explicitly follow O'Donnell and Schmitter (1986) and Przeworski (1991) in arguing that "the most proximate cause for the exit of authoritarian regimes can almost always be found in splits within the elite." They add, however, that a downturn of the economy tends to widen the gulf between hardliners and softliners. To begin with, the military, which tends to be well represented among the hardliners, is an exclusive popular sector group itself, offering its allegiance only in return for material favors. The regime's weakened capacity to deliver these favors when tough adjustment programs are required may drive a deeper wedge between the hardliner military establishment and softliner groups among regime elites. Moreover, the combination of the first two components of Haggard and Kaufman's theory may themselves boost the strategic importance of regime divisions:
[F]rom the perspective of the authoritarian leadership, the defection of private-sector groups and the widening of popular-sector protest increase both the cost of coercion and the risk that it will prove ineffective. It is precisely under such conditions that the splits we have noted within the regime begin to have strategic importance for the transition process. "Softliners" begin to calculate that the corporate interests of the ruling elite are best guarded by conciliation, rather than further repression. (1995, pp. 29-32)
In other words, economic crisis makes private-sector defection, popular protest and elite divisions interact to make democratization more likely.
Since the second of these three mechanisms, that of popular protest, is a determinant in itself to which I will return in Chapter 5, it will not be considered here. Having found that economic crisis propels democratic upturns even when the level of protest is being held constant, I want in this section only to find the mechanisms that explain the effect of economic crises other than through this particular pathway.
The case studies Haggard and Kaufman provide in order to substantiate their theory are based on the transitions to democracy in Peru in 1980,
72
The power of prosperity
Bolivia in 1982, Argentina in 1983, Brazil and Uruguay in 1985, and the Philippines in 1986 (1995). Three of these fit my concept of pathway cases very well, one relatively well, whereas two cases do not fit at all. The two non-fitting cases are Peru in f 980 and Brazil in f985, both of which only experienced inflationary, but no real recessionary crises. Since inflation is not systematically related to democratization in my analyses, I do
lot co-.iAiJt- ilev cises :.irlhf-. __________
Figure 3.1 displays how the four remaining cases are located in the partial regression plot of growth and democratic upturns. The plot itself is simply a multivariate analog to the well-known scatter diagram. What it shows is the relationship between growth (lagged one year) and upturns after taking all other determinants into account. The negative regression slope corresponds exactly to the short-term impact of -.540 that growth according to Table 3.1 exerts on upturns. Within this context, Argentina in 1983, Uruguay in 1985 and the Philippines in 1986 all fit very well to the concept of pathway cases. As indicated within brackets, they are ranked 10th, 11th and 12th among all 3,795 country years in terms of the pathway criterion.15 Bolivia is somewhat more problematic with a pathway rank of 40 in 1982, but if I shift to the year of 1981 its pathway rank is 31.16 Taken as a whole, then, 1 believe
Figure 3.1 also highlights some other features of my pathway criterion. The first is that most pathway cases are fairly extreme outliers in the sense that there is a large amount of change in their level of democracy that is unaccounted for even considering their degree of economic crisis (i.e., they have large residua's). As argued in Appendix D, 1 see no reason why this should disqualify them as pathway cases for assessing the mechanisms responsible for this particular determinant. Second, by Gerring and Seawright's (2007) account, most pathway cases experience little or no change m their level of democracy. The top-ranked pathway case according to their criterion, for example, is Gabon in 1975, positioned right on the regression line to the very left of the figure (not labeled). Gabon in 1975, however, experienced zero change on the democracy scale, and thus would provide very weak evidence of potential mechanisms linking economic performance to democratization.
This shift may seem somewhat surprising given that it was not until in 19S2 that the military in Bolivia eased repression and finally surrendered power to the congress elected two years earlier (Collier 1999, p. 148). The explanation comes from an oddity in the timing of the Freedom House scores, which lack a separate coding for the particular year of 1982. Instead, the 1981 scores incorporate the events up until August 1982, which (according to Freedom House's judgment) appears to cover the most substantial regime changes. Since Bolivia was more severely hit by recessionary crisis in 1980 than in 1981, this peculiarity in timing means that the year of 1981 appears to fit better than 1982.
Economic crises and authoritarian vulnerability
73
£ 2
-2 -
Uruguay 19S5 [11] »
Argentina 19B3 [10]
Philippines 1986 [12]  a <->
f •<Boiivia'l982[40] »Bb|Svia:1981 [31]
-A
Growth at (- 1 (conditional)
Figure 3.1 Case Studies of the Impact of Economic Crises on Upturns Note: The graph is a partial regression (or added-variable) plot of the conditional relationship between economic growth, measured in fractions and lagged one year, and upturns {n=3,795). The figures within brackets are the rank order of each case in tetms of the pathway criterion.
the case of Bolivia in the early 1980s fits the pathway criterion sufficiently well for my purposes.
The origins of the economic crises of these four countries in the early 1980s were a combination of past domestic policy mistakes and the international debt crisis. With the withdrawal of external lending, governments turned to the inflation tax to finance their fiscal and current-accounts deficits. After considerable hesitation adjustment measures were taken, such as devaluations and trade or exchange controls. The postponement of necessary policies, however, made their consequences even more disruptive (Haggard and Kaufman 1995, pp. 45-53).
The resulting defection of business elites was triggered by somewhat different grievances among the cases in question, depending on the nature of the original authoritarian bargain. In Bolivia and the Philippines, write Haggard and Kaufman, "business opposition crystallized against networks of favoritism that excluded significant private-sector interests" (1995, p. 56). The intricacies of the Bolivian transition
74 The power of prosperity
from a military dictatorship headed by General Banzer to the restoration of democracy in 1977-82, wrought by coups, counter-coups and elections, cannot be assessed in detail here (see, e.g., Whitehead 1986; Collier 1999, pp. 143-49). Suffice to say that the Banzer regime was generally considered good for business. However, the advent of the extremely corrupt transitory military administration headed by Garcia Meza in 198.0-81 in particular, led the n^g,qf_pri'atetsectPf..QrgaiiizaT ... tion, the Bolivian Federation of Private Entrepreneurs (CEPB), to change stance. The business elite became convinced that the military had a "statist bent" and could not handle the economic crisis, and that the public sector was "becoming parasitic." Thus the CEPB already in August 1981 called for the restoration of the Congress and president elected prior to the Meza government (Haggard and Kaufman 1995, P-57).
In the Philippines, the authoritarian regime was headed by Ferdinand Marcos, who as the then-elected president had seized power through a self-coup in 1972 by declaring martial law. As in Bolivia, private-sector elites in the Philippines had by and large endorsed martial law, but already in the early 1980s this support was beginning to crack under the pressure of economic decline. Several prominent members of the Makati Business Club, the main vehicle for the Philippine's non-crony private sector, started to speak out against the cronyism and corruption of Marcos' regime. Private-sector defection was, however, most importantly triggered by the assassination of Benigno Aquino, Marcos' longtime political opponent and himself from a prominent business-class family, in August 1983. After this event "the private sector played a crucial role in forging a centrist anti-Marcos coalition that included opposition politicians, academics, the Church, and the middle classes" (Haggard and Kaufman 1995, p. 56), which - as we shall see in Chapter 5 - eventually helped ensure the democratization of 1986-87.
In Argentina and Uruguay, usually considered as prominent examples of what O'Donnell (1979) termed "bureaucratic-authoritarian" regimes, the crisis dynamics unraveled somewhat differently. Here the grievance of private-sector groups was not primarily due to the erratic nature of corruption and cronyism, but rather to the inadequacies of technocratic control over economic decision making. Following the devaluations and tightened macroeconomic policies, serious strains between government and business began to surface. Argentina's major industrial association, The Industrial Union of Argentina, thus came out
Economic crises and authoritarian vulnerability 75
as outspoken critics of the military government after the bankruptcies of several industrial and financial firms in 1981 and 1982. In Uruguay, more or less the entire private business environment had already been alienated by the regime in the early 1980s. In Haggard and Kaufman's own words:
With economic problems mounting, business elites began to reevaluate the costs and benefits of the technocratic decision-making style that characterized authoritarian rule... The private sector's gradual disaffection did not reflect a democratic epiphany, but a pragmatic response to changing circumstances. With authoritarian governments increasingly unable to deliver their side of the bargain, "voice" began to appear increasingly important to business groups. (1995, pp. 58-60)
The way in which all these four authoritarian regimes ended, finally, was significantly affected by how declining economic conditions widened intra-regime divisions. In the Latin American cases, these splits mostly occurred along the lines of hardliners and soffliners within the military establishment itself. In Argentina and Bolivia, this led to a contracted process of coups and counter-coups, in the Argentinean case not ending until the disastrous Malvinas invasion had failed in 1982. In Uruguay, the transition was more peaceful as the military government gradually negotiated itself out of power, and the military establishment managed to remain more cohesive. The negotiations were, however, clearly affected by the economic difficulties, which weakened the hands of the diehards and promoted the acceptance of democracy among soffliners. In the Philippines the declining economy deepened the rift between the civilian-led regime under Marcos and the military establishment, in the end leading to attempted mutiny in the wake of Marcos' attempt to steal the snap 1986 election (1995, pp. 66-71).
In sum, my statistical findings and Haggard and Kaufman's narratives of the four pathway cases combine as telling evidence that economic crisis is a truly causal determinant of democratization. That said, Haggard and Kaufman's account will not be the last theory on the consequences of economic crises for democratization. Nor can their case studies be considered the final word on the causal mechanisms linking these phenomena. Most importantly, the role of private-sector defection is of course conditioned on the existence of a private sector, which delimits the scope of this particular mechanism to the set of non-socialist economies (cf. 1995, pp. 371-74). The theory thus needs to be
76 The power of prosperity
broadened to include other forms of authoritarian bargains with other support groups, and case study evidence from a broader array of economies needs to be mustered.
Another unresolved issue is why, assuming that Haggard and Kaufman's theory is correct, only recessionary and not inflationary, crises act as significant determinants of democratization in my analyses. RoLtled '.<> '.his, one cock! .i4i vvhciic: k is the e-ouoruc o '.self or the particular policy response to crisis that is most critical. Timing is also of critical import. Although I for the sake of simplicity have imposed a uniform one-year lag in my statistical analyses of the effect of economic crisis on democracy, the case studies reveal a much more complicated pattern. In most instances, the economic crises were ongoing for several years before they precipitated regime change, but the time it took also varied by country. Future statistical work on the democratizing effects of economic crises must be made more sensitive to this issue.
Conclusion
In this chapter I have found that socioeconomic modernization affected the third wave of democratization by hindering backsliding toward authoritarianism. The most pertinent component of this underlying syndrome was media proliferation, which helps deter or derail coups. : Recessionary crises, by contrast, triggered democratization, according to the case study evidence by provoking private-sector defection and intra-regime splits. Oil dependence instead blocked democratization, whereas freedom from state incursion in the economy hindered authoritarian reversals. Income inequality, however, did not exert any significant impact on the prospects for democratization during the third wave.
The power of economic prosperity is thus a double-edged sword. Whereas natural resource abundance and short-term economic growth keep democracy at bay, prior democratic achievements are better sustained at higher levels of socioeconomic modernization.
4 ^ The impetus from abroad: ' international determinants
Politics has of course never been carried out in fully closed domestic arenas. Yet most observers agree that globalization during recent decades has been an important trend in the system of international relations. Countries are to begin with economically interdependent, and increasingly so. The expansion of international broadcast media, and the growing importance of intergovernmental organizations, also implies that national systems of government are becoming increasingly interdependent politically. Global ideas diffuse across country borders. Decisionmaking power in previously domestic affairs is being transferred to new government bodies in the international arena. Last, but not least, countries intervene intentionally in the domestic politics of one another.
There are thus various reasons to believe that international determinants played a key role in the third wave of democratization. But did they? In this chapter, I address this question after reviewing the extant literature on international determinants of democratization. My results speak in favor of there being three non-domestic forces at work: international trade, neighbor diffusion and pressure from regional international organizations. By again consulting both statistical and case study evidence, I then aim to uncover the mechanisms responsible for these results.
The literature on international determinants
There is a large and growing literature on factors impeding or enhancing democratization at the international level. An old school of thought in this regard comprises the so-called world system position and dependency theorists (for an overview see Bollen 1983, pp. 469-71; Hadenius 1992, pp. 91-98). They claimed that international capitalist exchange involving trade and investments favored the wealthy international "core" at the expense of the poor "periphery," which was exploited. In order to maintain this system of exploitation democratic rule in peripheral countries needs to be stifled, according to dependency
77
78
The impetus from abroad
79
theorists, since authoritarian leaders are supposedly more receptive to the interests of international economic centers.
However, most of the early cross-sectional tests of the dependency predictions produced weak or inconsistent support (Bollen 1983; Bollen and Jackman 1985; Gasiorowski 1988; Gonick and Rosh 1988). In a more recent account - although couched in the language of "globalization," presently more in vogue - Li and Reuveny (2003; 2009) have tested some of the old predictions in a cross-sectional time-series setting. Interestingly,, their results by and large confirm dependency theory. According to their findings, both the volume of trade and portfolio investment inflows negatively affect democratization. And while foreign direct investment inflows - their third indicator of globalization - had a positive impact, it has weakened over time. Li and Reuveny conclude by stating that "the economic aspects of integration into the world economy are beginning to cause a decline in national democratic governance" (2003, p. 53).
That trade volume impacts negatively on democracy was also recently found by Roberto Rigobon and Dani Rodtik (2005). Li and Reuveny (2003), however, found a positive effect of another facet of international dependence: the spread of democratic ideas across countries, or what is usually referred to as democratic diffusion. To systematically assess such external diffusion or demonsttation effects with large-re data is a fairly novel enterprise in this field. Yet hitherto the evidence has by and large confirmed expectations. Diffusion has been shown to affect democratization both at the most proximate level of neighbor states, at the level of world regions, and at the global level (Starr 1991; O'Loughlin etal. 19.98; Kopstein and Reilly 2000; Starr and Lindborg 2003; Brinks and Coppedge 2006; Gleditsch and Ward 2006).
In a recent book, Pevehouse (2005) suggests another potential non-domestic determinant of democratization: regional international organizations. With a mixture of case study and statistical evidence, Pevehouse purports to show that democratic regional organizations can pressurize authoritarian member states to undertake democratic reforms, socialize military and economic elites into accepting democratic procedures, and bind newly elected elites in fledgling democracies to these reforms once committed. In this way, membership in democratic regional organizations, according to Pevehouse (2005), both precipitates movements toward democracy and enhances democratic survival.
Finally, a more drastic form of international determinant that has climbed onto the international agenda, particularly since the US-led
invasions of Afghanistan and Iraq, is direct foreign intervention. States attempting to improve the level of democracy in another country may, for example, impose economic sanctions or stage a foreign military intervention. Recent comparative case and large-ra studies, however, cast some doubt ovet the general efficiency of these strategies, noting that such impositions from abroad sometimes work and sometimes fail ■'liueno de Mesquita and Downs 2006; Pickering and Peceny 2006; Grimm 2008; Hufbauer et al. 2009, pp. 67-69, 158-59; cf. Pickering and Kisangani 2006).
Most of these studies of international determinants have, however, not assessed the impact of globalization, diffusion, regional organizations and foreign interventions net of all other domestic influences of democratization. As should be evident, what for example appears to be a diffusion linkage between two countries could disappear once possible confounding factors simultaneously affecting democracy in both countries are taken into account. Basically the same goes for economic dependence, shared membership in regional international organizations, economic sanctions or foreign military intervention. I shall now try to remedy this by assessing international effects in the context of more fully specified models.
Empirical results
Turning then to Table 4.1, my results partly confirm the finding by Li and Reuveny (2003) and Rigobon and Rodrik (2005) thattrade volume impeded democratization during the third wave. For this time period there was a statistically significant and negative, although substantively not very large, impact on upturns, implying that a 100 percent increase in a country's trade volume (relative to its GDP) led to an estimated .076 decrease in its propensity to democratize the following year. At face value this finding seems to confirm the old prediction by dependency theory that largely trade-dependent countries are hindered from democratizing, although this is an interpretation I shall return to later in this chapter. The effect is fairly robust to alternative specifications, but not quite to the same extent as the other determinants are. Most importantly, I find no effect of trade when only the within-country variation is retained, and rhe impact is sensitive to the choice of democracy measure. Moreover, in the fully imputed global sample the coefficient, while similar in magnitude, is only marginally significant (see Appendix C).
The impetus from abroad
fynpinc
■^1 results
Table 4.1 International Determinants
Short-i
	General	Upturns	Downturns	Long-run"
Trade volume	-.086*	-.076"	-.011	-1.08-
	(.051)	(.036)	(.026)	(.604)
Democratization among-......-—	— .!!.•••	—;G7?-«-*	.....v039"	1.45"*
neighbors	(.031)	(.024)	(.019)	(.488)
Prior level of democracy	.008	-.003	.Oil**	.097 :
among neighbors	(.007)	(.005)	(.005)	(.089)'
No neighbors	-.017	-.086"	.070	-.208
	(.069)	(.044)	(.044)	(.867)
Regional level of democracy	.019"	.008	.012'	.242**
	(.010)	(.007)	(.006)	(.118)
Global level of democracy	.010	.002	.009	.126
	(.017)	(.014)	(.011)	(.208) ■
Democratization of regional	.030	.029	.001	.373
organization	(.028)	(.021)	(.018)	(.367)
Prior level of democracy of	.028"	.035 s"	-.007	.348"
regional organization	(.013)	(.007)	(.009)	(.143)
No regional organization	.214"	.203*"	.011	2.67"
	(.095)	(.062)	(.056)	(1.07)
* significant at the .10-level,significant at the .05-level," * * significant at the .01-levef. No. of observations = 3,795; no. of countries - 165; mean years observed pet country = 23.0.
Note: Entries ate unstandardized regression coefficients with panel-corrected standard-errors in parentheses. AH models also include two lags of the dependent variable and the determinants in Table 2.1, 3.1 and 5.1 as controls. Trade volume and democratization among neighbors and of regional organization have been lagged one year, whereas the remaining variables are Jagged two years.
As shown in Table 4.2, for a more restricted sample of countries (due to limited data availability), larger integration into the world economy in terms of international capital flows according to my estimates did not impact on democratization during the third wave. Apart from being inconsistent with dependency theory, this contradicts Li and Reuveny's (2003; 2009) findings for both foreign direct investment and portfolio ~ flows.
Turning back to Table 4.1, my next set of international determinants aim at capturing geographical diffusion effects - the spread of democracy
Table 4.2 International Capital Flows
Short-
— Portfolio inve
General	Upturns	Downturns	Long-
-.046	-.088	.042	-.615
(.093)	(.070)	(.047) '	(1.27)
.143	.177	-.034	1.92
(.154)	(.122)	(.075)	(2.10)
* significant at the. 10-)evel, * * significant at the .05-level, ^ '** significant at the .01-level. No. of observations = 2,577; no. of countries - 152; mean years observed per country = 17.0.
Note: Entries are unstandardized regression coefficients with panel-corrected standard errors in parentheses. All models also include rwo lags of the dependent variable and the determinants in Table 2.1, 3.1,4.1 and 5.1 as controls. All explanatory variables in the table have been lagged one year.
or autocracy from one country to another. The measurement strategy here follows the larger literature by including the average level of democracy at three spatial levels: globally, among countries belonging to the same world region, and among neighboring countries. Of the three spatial levels included, only the most geographically proximate appears to have had an effect. If the mean level of democracy among neighboring countries was shifted upward one unit between time t-2 and t-1, the net expected change in democracy at time t was .116. The long-run equilibrium level of democracy was accordingly increased by 1.45 on the 0-10 democracy scale. This implies a fairly tight long-run adjustment of the levels of democracy among neighboring states. As the relative magnitudes of the impact on upturns versus downturns indicate, moreover, this neighbor diffusion effect is primarily driven by countries moving in the upwards direction. The effect on upturns is robust to most alternative specifications (see Appendix C),1 whereas the impact on downturns depends on some relatively extreme influential observations.2
1 For reasons unclear to me, neighbor diffusion is not a statistically significant determinant of upturns when only the Freedom House measure of civil liberties is used as the dependent variable, and only marginally significant on Freedom House political rights. More research is needed to explain why the diffusion effect is thus primarily driven by the Polity measure of democracy. These observations are Gambia in 1994, Ghana in 1981 and Thailand in 1976. If any of these three cases are omitted from the analysis, neighbor diffusion is no longer a statistically significant determinant of downturns.
82 The impetus from abrc>,- i
At first glance, Table 4.1 appears to indicate that diffusion effects also ~ occurred at the level of world regions, but this result falters once two H extremely influential outliers are excluded.3 In other words, during the .' third wave there seem to have been no diffusion effects at work (net of - I other influences) either at the regional or global level. In this regard my | results differ from the existing literature on diffusion effects, the probable reason being my more fully specified explanatory model. I
Next, I test Pevehouse's (2005) argument on the importance of reg . nal organizations. I follow his own measurement strategy, which .',; implies that the level of democracy for a specific military, economic or   .! political intra-regional organization is defined as the average degree of democracy among the countries belonging to the same regional organi- ..■ zation. For countries belonging to more than one regional organization, [_ only the score for the most democratic regional organization is : included. The results confirm one key prediction from Pevehouse's theory: During the third wave membership in relatively democratic regional organizations precipitated upturns in the level of democracy of a country - a result that turns out to be extremely robust to alter- :■. native specifications (see Appendix C).4 As should be clear, it was 11-■-current level of democracy of a regional organization that mattered, n-the change. I may further conclude that the statistical pattern is t\i -driven by countries joining democratic regional organizations, since the same results apply when controlling for the number of regional organ1 zations to which a country belongs, as well as change in that number (results not shown). The key finding is thus that countries belonging : to a regional organization in which all member countries were fully democratic (earning a score of 10) on average increased their democracy scores by .28 as compared to countries belonging to regional organizations where all member states were fully authoritarian (i.e.. with a score of 0), implying a long-term upward shift in level of demot. ■ racy by 3.48.
3 These outliers are Turkey in 19S0 and Fiji in 1987, where military coups brouy^r down democracy to a level more in tune with their respective regional averages at the time. Once these two outliers are excluded, regional diffusion exerts no significant impact on democratization.
4 The result is insensitive to the exclusion of four relatively influential outliers: Turkey in 1.983, Guyana in 1975, Guinea-Bissau in 1994 and Portugal in 197 I
I also get similar results when using the mean of the level of democracy among ihc ... regional organizations to which a country belongs rather than the maximum.
empirical results 83 Table 4.3 Foreign Interventions
Shorr-run
General Upturns Downturns Long-run
Economic sanctions	-.068	.146	-.213**	-.720
	(.143)	(.099)	(-104)	(1.53)
fjo.-of-observations	2,630	2,630	2,630	2,630
Ho. of countries	158	158	158	158
International military intervention	.119	.408*	-.289	1.19
	(.330)	(.238)	(.210)	(3.28)
No. of observations	2,181	2,181	2,181	2,181
No. of countries	124	124	124	124
Foreign intervention index	-.045	.197"	-.242'»	-.481
	(.136)	(.096)	(.101)	(1.45)
No. of observations	3,035	3,035	3,035	3,035
No. of countries	159	159	159	159
* significant at the .10-level, * * significant at the .05-level, * * * significant at the .OTlevel. Hote: Entries are unstandardized regression coefficients with panel-corrected standard errors in parentheses. All models also include two lags of the dependent variable and the . determinants in Table 2.1, 3.1, 4.1 and 5.1 as controls. All explanatory variables in the table have been lagged one year.
I find no support, however, for the flip side of Pevehouse's argument: that regional organizations also help democracies survive. In terms of my empirical strategy, that is, I find no effect of regional organizations on downturns.
Turning, finally, to the impact of foreign intervention, I perform three simple tests in Table 4.3 on a more restricted sample of countries. Economic sanctions mostly occur in the form of cut-offs of finance, aid and/or trade (see Hufbauer et al. 2009). As the uppermost panel indicates, such sanctions during the third wave had a statistically significant but negative impact on downturns. This implies that economic sanctions imposed by democratic countries on average led to more backsliding toward authoritarianism. Hostile military interventions here involve actions where troops on the ground move into the territory of another country in order to oppose the target government or support rebel groups (Pickering and Peceny 2006). In the middle panel, I find that military intervention hostile to the incumbent regime had a marginally significant but positive impact on upturns. In neither case, however,
84
The impetus from abn
were there any general average effects on democratization, which is explained by the fact that the effects of both types of foreign interventions are reversely signed for upturns as compared to downturns.
This more general finding comes out even more clearly in the lowermost panel, where I collapse the measures of economic sanctions and military interventions into one "index," simply indicating whether any ' form of intervention occurred. This substantially increases the number of interventions included, and also allows me to cover a longer time span (in effect, from 1974 to 2001). As can be seen, this index exerts ■ significant and positive impact on upturns, and a significant but tive impact on downturns. There is no general effect, however, neither in the short- nor in the long-run. In line with the previous literature, I thus find that foreign interventions imposed by democratic com i ■ ir ■. ,i \. a mixed blessing. They are almost as likely to improve as to deteriorate the conditions of democracy in the target country.
In sum, I find three robust international determinants of democrari-: zation during the third wave: trade volume, with a negative impact well as neighbor diffusion and membership in democratic regional organizations, both with a positive impact. I now turn to an exploration of what mechanisms may explain the three.
Trade dependency and stalled democratization
As already noted, that countries dependent on the external world in terms of trade should be less likely to democratize is an idea dating back to dependency and world system theories in the 1970s. Although most,; of these writings concentrated on growth and global inequality, they also yielded predictions on the prospects for democracy. Most imp tantly, these scholars assumed that the world could be divided into a set; of wealthy and powerful democracies at the "core," and a set of poor and marginalized countries at the "periphery." Most simply put, the notion was that the economic and political elites in these two sets of;: countries worked in tandem to suppress and limit the suffrage of the general populace in the periphery. In terms of trade, the argument goc the core countries penetrate the peripheral countries economically, and an authoritarian government in these countries is necessary to sustain this exploitation (see, e.g., Bollen 1983; Hadenius 1992, pp. 91-98).
In Table 4.4, I test whether the negative impact of trade volume on upturns could be explained in terms of with whom one trades. Since I
"Trade dependency and stalled democratization
Table 4.4 Explaining the Effect of Trade Volume (upturns)
85
Model
: -rjjde volume Trade with US, UK and France Trade with China and USSR/Russia
(1)	(2)	(3)	(4)
-.087	-.111*	-.081	-.106 s
(.054)	(.057)	'(.054)	(.057)
	.104		.101
	(.070)		(.069)
		-.140	-.117
		(.206)	(.204)
'significant at the .104evel, * * significant at the .05-level, * * * significant at the .01-level. No. of observations = 2,961; no. of countries - 153; mean years observed per country = . 19.3.
Note: Encries are unscatidardized regression coefficients with panel-corrected standard errors in parentheses. All models use upturns in the level of democracy as the dependent variable, and also include two lags of the level of democracy and the determinants in Table 2.1, 3.1,4.1 and 5.1 as controls. All explanatory variables in the table have been lagged one year.
have only information on the latter for a subset of countries and years (data is from Gleditsch 2002), I start in model (1) by rerunning the model with all determinants on this more restricted sample. As can be seen, the negative coefficient for trade on this sample is of approximately equal size, even slightly stronger (as compared to Table 4.1), but not statistically significant. Although this is unfortunate, it does not completely obviate the logic of testing whether the effect decreases when potential mechanisms are being controlled for. In model (2), I thus test the expectation from dependency theory that the negative effect of trade is due to trade with the "core" economies of the world. Although one could quibble about what countries more exactly form this core, few would deny that the United States, Great Britain and France are among them. When controlling for the share of trade with these particular countries, however, the negative coefficient on general trade volume is unaffected (it even increases somewhat). Equally important, trading with these core economies in itself does not hurt democracy. This finding goes directly against the grain of dependency theory.
By an alternative version of this theory, however, it could be that what matters is being dependent on trade with a powerful authoritarian state whose security interests favor imposing autocracy in the partner countries
86
The impetus from abroad'
as well. Two suggestions for such "regional hegemons" that directly come to mind are of course China and the USSR/Russia. Controlling for bilateral5 trade volume with these two countries in model (3) again produces insig-■« nificant results, however, and does not diminish the negative effect of trade ' volume. Nor are the results different if I control for trade with both the capitalist core and hegemonic authoritarian countries in model (4). In sum," the negative trade effect is a finding in want of a theoretical explanation.
What explains neighbor diffusion?
The growing literature on democratic diffusion has suggested severah causal mechanisms through which the process of neighbor diffusion may occur. Two of the most plausible rival explanations are that neighbor diffusion may occur through imposition or emulation. In the first case, countries that move toward democracy themselves try to promote democratization among their neighbors (Brinks and Coppedge 2006, p. 467; Gleditsch and Ward 2006, p. 919). Their reasons for doing so may vary, but one likely possibility is that they would do so in order to enhance internal security. If democracies are less likely to fight each other, as democratic peace theory would have it, then fledgling democracies have a security interest in helping the democratic opposition topple incumbent regimes among its neighbors (Pevehouse 2005, p. 18). In the case of emulation, by contrast, the driving force of diffusion comes from within the neighboring countries themselves. By emulating the successful-example of the neighbor that first installs democracy, by discovering . "that it can be done" and learning "how it can be done" (Huntington 1991, p. 101), the democratic opposition may raise its chances of succeeding in overthrowing its own autocratic incumbent. Also the incumbent elites may be affected through emulation, since "reluctant leaders in autocracies may be more willing to initiate difficult reforms if the experi-: ences of other states suggest that the costs and consequences of reforms may not be as bad as they had feared" (Gledistch and Ward 2006, p. 920). The two mechanisms of imposition and emulation are of course not mutually exclusive. In the words of Beissinger, it may not "simply be:: a matter of the pull of example," but "also in part a matter of the push by those who have already succeeded" (2007, p. 266).
Unfortunately, the data at my disposal does not allow me to fully, disentangle these two mechanisms of diffusion. Indirectly, however, I may invoke evidence brought to bear on both of them. To begin with, I
\Vhat explains neighbor diffusion}
Table 4.5 Explaining the Neighbor Diffusion Effect (upturns)
87
Model
(1)
(2)
(3)
(4)
Democratization among neighbors .081**
(.025)
Democratization of regional organization
inflation among neighbors Growth among neighbors Demonstrations among neighbors Riots among neighbors Strikes among neighbors
.077**
(.024)
.029
(.021)
(.027)
.027
(.021)
-.001
(.004)
-.272
(.290)
No. of observations No. of countries
3,795 165
3,795 165
3,706 162
.088** (.027) .030 (.022)
-.249
(1.005)
.243
(1.004)
.046
(.033)
3,719
163
* significant at the .10-level, * * significant at the .05-level, * * * significant at the .01-ievel. Note: Entries are unstandardized regression coefficients with panel-corrected standard errors in parentheses. All models use upturns in the level of democracy as the dependent variable, and also include two lags of the level of democracy and the determinants in Table 2.1, 3.1,4.1 and 5.1 as controls. The neighbor democratization and democratization of regional organization variables have been lagged one year, whereas the remaining variables in the table are lagged two years.
have already in Table 4.1 controlled lor one plausibly powerful mechanism of imposition: the influence of regional organizations. Since regional organizations, by their nature, bring together countries located in close geographical proximity, the process of neighbor imposition may of course work through the channels of a regional organization. What then is the effect of neighbor diffusion when this mechanism is not included in the model? Model (1) of Table 4.5 provides the answer. Without controlling for regional organizations, the effect of neighbor democratization is .081. The reduction of this effect to .077 in model (2) is, however, miniscule. I thus find no support for this particular process of neighbor imposition, namely that it would work through die channels of a common regional organization.
88 The impetus from abroad
In models (3) and (4), I try instead to proxy for some of the possible processes through which neighbor emulation may occur. Imagine a hypothetical country A, surrounded by countries B, C and D. The underlying idea is to control for events in the neighboring countries IV C and D that may first have triggered democratization in these countries and then caused democratization through emulation in A. To the extent that I may uncover such events, the direct effect of democratization in B, C and D on democratization in A should dissipate. In model (3), following this line of thought, I include indicators of economic crises among neighboring countries. To the extent that authoritarian regimes were toppled by economic crisis, as my results in the previous chapter tend to support, one might plausibly believe that the spread of economic crisis to neighboring countries masquerades as neighbor diffusion. This is, however, not the case. The positive effect of neighbor democratization on upturns is unaffected by the extent of economic crisis in neighboring countries, whereas the latter has no independent effect in itself. ■
In model (4), finally, I perform a similar exercise, but by introducing indicators of popular mobilization among neighbors. Recent work on the so-called "colored revolutions" that swept through Serbia, Georgia,: Ukraine and Kyrgyzstan in the first years of the twenty-first century has. uncovered important diffusion processes at work through the spread of collective protest (see, e.g. Beissinger 2007; Tucker 2007). If these mechanisms travel back in time, and to other geographical places, perhaps what drove neighbor diffusion during the third wave was the spread of popular protest from one country to the other? Again, however, expectations falter when countered with empirical evidence. As : model (4) shows, there is no reduction in the effect of neighbor democratization. And popular mobilization among neighboring countries does not affect democratization, net of other influences.
Further work thus needs to be done in order to pinpoint the causal mechanisms through which neighbor diffusion works. A key contribution in this regard would be to bring in more systematic case study evidence to assess whether the rough proxies for neighbor diffusion really pick up real-world processes on the ground. An indication that this might not be quite so is that most of the effect of this variable, according to my estimates, seems to have taken place in a part of the world with, to the best of my knowledge, comparatively little case, evidence of neighbor imposition or emulation: Sub-Saharan Africa. There is, for example, substantial anecdotal evidence to the effect that -
j)o regional organizations foster democratization? 89
experiences in the near abroad affected the mindset of pro-democratic forces in the sequential fall of communist regimes in Eastern Europe. In the words of Timothy Garton Ash, who was a local observer at most of these dramatic transitions, with strong personal connections to the democratic opposition: "Everyone knew, from their neighbours' experiences, that it could be done. More than that, their neighbours had given them a few ideas about how it should be done" (1990, p. 127). It has also been stated that Latin American democratization during the third wave was largely a process of regional diffusion (Mainwaring and Perez-Linan 2005a; 2005b).
Whereas the neighbor diffusion effect is highly robust to the exclusion of Eastern European or Latin American regimes (or any other of the world regions), however, it falters completely if I exclude Sub-Saharan African countries from the estimation sample. Similarly, among the twenty most fitting cases according to the pathway criterion, only one is located in Eastern Europe (Armenia in 1998), and six in Latin America, whereas ten are Sub-Saharan African countties (most prominently the Central African Republic in 1993, and Zambia and Angola in 1991). A critical test for the neighbor diffusion hypothesis in the future, at least in the form it is now being proxied for in statistical studies, would thus be to assess the events more closely in these particular cases.
Do regional organizations foster democratization?
I will now perform such an assessment, but for the third and final international determinant: membership in democratic regional organizations. The argument for neighbor diffusion - that the measure of the influence of regional organization is very approximate in nature -applies equally well here. Recall that, following Pevehouse (2005), I measure the level of democracy of a regional organization as the mean level of democracy among its member states, and then for each country insert the level of democracy of the most democratic regional organization to which that particular country belongs (without taking the level of democracy in the country itself into account). The fact that I in Table 4.1 find a statistically significant effect of the level of this variable on upturns, net of all other determinants, is of course not necessarily evidence of a causal relationship. It could simply reflect the fact that countries belonging to the same regional organizations democratize in tandem (albeit with some time lag), or that they tend to approach the
90
The impetus from abroad
same long-run equilibrium level of democracy due to other exogenous reasons that have not been controlled for. Case study evidence is thus needed to test whether this measurement strategy really picks up a reai-world impact of regional organization. And, at least indirectly, case study evidence for such an effect could affect our belief in the neighbor diffusion variable as well.
Pevehouse hypothesizes three distinct causal mechanisms through^ which membership in a democratic regional organization may further democratization in a country. The first he simply calls pressure. Through open verbal condemnation and threats of sanctions or other punishments, such as membership suspension, a regional organization may hurt or at least threaten to hurt the economy of an authoritarian regime, and even help to delegitimize it domestically: "If allies and institutional partners treat the regime as a pariah state, this can impact: on public and elite perceptions within the state. These pressures can help : to weaken an authoritarian regime's grip on power" (2005, pp. 17,19).
The second mechanism is termed the acquiescence effect. Much as Haggard and Kaufman (1995), Pevehouse assumes that an authoritarian regime depends on the support of certain critical elite groups that may veto an attempted move toward democracy. Involvement in a democratically committed regional organization, the theory goes, may dampen fears of democracy among these elite groups and thereby help them acquiesce to the transition. Pevehouse provides two more specific examples of such groups, and thus of how this mechanism may operate. The first is credible commitment to liberal-economic protection of property-right infringements through membership in regional economic and/or trade organizations. This particular acquiescence is thus mostly crucial for the private-sector economic elite, whose preferences for democracy may be "locked in" through the ties to a regional organization. The second example is socialization of the military by cooperating with other more democratic countries in regional security organizations or military alliances. By interacting with military elires from these other countries, the military may be persuaded that "the role of the military is not that of an internal police force involved in domestic politics, but rather to protect the state from external enemies" (2005, pp. 20-25).
Finally, regional organizations may promote democratization by legitimizing interim governments, that is, caretaker governments in power between the breakdown of autocracy and before the holding of founding elections. Membership in democratic regional organizations
po regional organizations foster democratization?
91
£ 2
-2 -
■"S"P»u200qi933«
8       t* i.
*- Turkey 1983[1] • s Hungary 1990 [20]
IQCKl MOOT
-5 0 5
Level of democracy of regional organization at f-2 (conditional)
Figure 4.1 Case Studies of the Impact of Regional Organizations on Upturns Note: The graph is a partial regression (or added-variable) plot of the conditional relationship between the level of democracy of a country's most democratic regional organization, lagged two years, and upturns [n - 3,795). The figures within brackets are the rank ordet of each case in terms of the pathway criterion.
may help these fragile governments to credibly signal their commitment to complete the move to democracy (2005, pp. 25-27).
The drree case studies for which Pevehouse traces evidence of these causal mechanisms are highlighted in Figure 4.1. The redemocratization of Turkey in 1983 turns out to be the best fitting of all 3,795 cases in terms of the pathway criterion. The transition to democracy in Hungary in the late 1980s also fits relatively well, with the "liberalization" of 1989 ranked 18th and the move to more "full democracy" in 1990 ranked 20th. The Peruvian cases, however, do not fit as well, being ranked 93rd and 128th, respectively. I shall return to an interpretation of this below and now instead turn to the case study evidence for each of these three countries.
Turkey
The military intervention in Turkey on September 12, 1980 was the third time the military had unseated a civilian government since the
92 The impetus from abn
inauguration of multiparty politics in 1950. Under the auspices of r i, National Security Council, chaired by General Evren, the military dis- ' ■ solved the government and Parliament, banned all political activities monitored the press and installed martial law. Arrests and harsh repi ■. sion of tens of thousands of political activists and journalists follow^ The return to multiparty competition was initiated in 1982 through the drafting of a new constitution, approved in a nationwide referendum and completed in November 1983, when the one party not headed bv a military was able to win a majority of seats in national elections (Ev i 1994, pp. 23-26; Dagi 1996, pp. 125-26; Pevehouse 2005, p. 140).
Although General Evren already at his first press conference after the takeover promised to reinstall a new democratically elected civilian government in due course (Dagi 1996, p. 125), Pevehouse argues that : ■ the actions of primarily two regional organizations in the area helped to shape the timetable of this transition back toward democracy. The first of these was the European Economic Community (EEC), with which Turkey had signed an association agreement in 1970. The EEC mostly -put economic pressure on the military regime in Turkey by first increas- .: ing its financial assistance to Turkey in 1981, and then withdrawing ". that aid conditional on improvements in human rights and the return of democracy (Pevehouse 2005, p. 146; Dagi 1996, pp. 129-30, 137-38).
Turkey not being a member state of the EEC, however, the relevance of this case evidence for explaining the statistical relationship between regional organizations and democracy could be questioned. In effect, the EEC never enters my (or Pevehouse's) data for the Turkish case. . ■ Instead, there are three political, economic or military regional organi-  ■ ■ zations of which Turkey was a member in 1981 (two years before the significant upturn in 1983): the Council of Europe (COE), NATO and the Organization for Security and Cooperation in Europe (OSCE). Fortunately, the COE is the second organization for which Pevehouse ' (2005) is able to trace case study evidence for Turkey.5 The COE was able to put psychological pressure on the military regime by exploiting
5 Strictly speaking, the COE is not the regional organization whose democracy score enters my (and thus Pevehouse's) data, since Turkey's most democratic regional organization was NATO. The difference in democracy level between NATO and the OSCE in 1981 is miniscule, however (being 9.58 for the OSCE and 9.76 for NATO), and, as argued before, the effect of regional organizations is robust to the replacement of the maximum by the mean level of democracy among all regional organizations.
po regional organizations foster democratization? 93
the great symbolic importance membership in the council had for Turkey. Already in September 1980, the Parliamentary Assembly of the COE condemned the events in Turkey and called for an immediate return to democracy. The issue of suspending Turkey's membership was ilso discussed on several occasions in the years to come, although that threat was never realized (Dagi 1996, pp. 131-34),, The COE sent multiple fact-finding missions to Turkey to investigate allegations of human rights abuses, the most crucial of which arrived in late 1981, only a month after the military government had completed its stranglehold on the opposition by banning all political parties. Turkey tried to reject this visit by a 25-member delegation but was pressured to admit it. Only a week before the delegation's arrival, moreover, General Evren announced an approximate date for new general elections, a move that at least partly was motivated by an anxiety regarding how the report from the delegation would affect Turkey's future status in the Council (Dagi 1996, pp. 137-38; Pevehouse 2005, p. 147).
Pressure from the Council of Europe thus constitutes the major causal mechanism present in the Turkish case. True, membership in NATO could have provided room for an acquiescence effect, but NATO appeared to have been very soft on the Turkish military regime in the early 1980s. Although there are some allegations of it in the literature, even Pevehouse (2005, p. 147) concedes that "NATO's contribution to the redemocratization of Turkey flowed from its socialization of officers in the Turkish military is fairly weak."
Hungary
The critical democratizing reforms in Hungary were initiated in early 1989, when the Communist Party, which had been in power since after the end of World War n, legalized independent political parties. Round-table talks in September/October that same year ended in an agreement to hold genuine multiparty elections, as well as in several other constitutional amendments that were approved by a national referendum in November. I'ree and fair parliamentary elections were held in the spring of 1990, which resulted in the formation of a coalition government comprising opposition political parties under Prime Minister Antall (Pevehouse 2005, pp. 116-17; also see Gill 2000, pp. 196-97; Saxonberg 2001, pp. 8-9).
The case study evidence for an effect of regional organizations in Hungary is, however, somewhat strained. Apart from the problem that
94
The impetus from abro.nl ""
to regional organizations foster democratization?
95
Pevehouse (2005) mostly discusses the impact of regional organizations to which Hungary was not a member in 1989-90 (the EEC and NATO), or one that it did not join until 1990 (the COE), there is in this case also a more general temporal disjunction. Namely that most of the activities any regional organization directed toward Hungary took place after 1990. The most important effects, according to Pevehouse, were exerted by the EEC and the COE. Already in July 1989 the EEC had initiated an economic restructuring program, including both grants and loans, geared toward Hungary and Poland (implying a causal mechanism of direct economic benefits, not anticipated by Pevehouse's theoretical framework). The COE, after granting membetship to Hungary in November 1990, also provided technical assistance, and together with the EEC helped legitimize the Antall government and to signal its commitment to continued reforms (Pevehouse 2005, pp, 121-26). By that time, however, Hungary had already completed its move from 4.1 in 1988 to 9.16 in 1990, accordmg to the democracy scale used throughout this book.1' Its later progression to the highest score of 10.0, achieved only in 2004, was gradual and does not drive my result on the impact of regional organizations on democratization.
In 1987-88 (applying the two-year lag from my statistical model), Hungary was a member of three regional organizations: the Council of Mutual Economic Assistance, the Warsaw Pact and the OSCE. Of these three, the OSCE is of course the most democratic (with a democracy score of around 7.80, excluding Hungary), and thus the one that should play the democracy-enhancing role posited by Pevehouse (2005) in order to explain the statistical impact of regional organizations on.. upturns. However, most of the influence the OSCE appears to have exerted again occurs after 1990. Pevehouse primarily discusses the OSCE's importance for an acquiescence effect, which allegedly would have helped to socialize the Hungarian military into accepting civilian supremacy. Although NATO's role in this regard probably overshadows the OSCE through the Partnership for Peace agreement (launched in 1994!), the OSCE also provided retraining of Hungarian military leaders in order to encourage Western-style civil-military relations
6 As a matter of fact, Hungary "completed" its "transition to democracy" in 1990 also according to the operational criterion Pevehouse himself uses to establish this, that is, receiving a polity score of at least +6.
(2005, p- 119)- There is, however, no evidence that this occurred as early as in the late 1980s, nor that the military ever posed a threat to the transition to democracy in Hungary. If anything, the acquiescence of the Hungarian mu'tar>r might have- played a role in discouraging a later coup, but that would be part of a mechanism linking regional organizations to the absence of downturns (a relationship my statistical analysis does ..not support).
What the OSCE might have provided that fits the timing of Hungary as a pathway case is more direct pressure on the Communist regime. Although there is some evidence suggesting that the 1975 Helsinki Final Act (on which the OSCE is based) promoted a general climate of support for oppositional movements, particularly in Poland and Czechoslovakia (2005, p. 118; Whitehead 1997, pp. 44-47), there is no direct support for this in the Hungarian case. And while there seems to be a temporal proximity between the OSCE's adoption of the Vienna Concluding Document in January 1989, where it made "great strides forward in the sensitive area of human rights" (Hyde-Price 1994, p. 237), and the legalization of opposition parties in Hungary in the February, I have not been able to find any confirmation of a direct link between these two events (nor has Pevehouse). In sum, then, the Hungarian case provides only limited support for a democratizing effect of the one democratic regional organization it belonged to in the late 1980s. There is more to suggest that the EEC played a critical role by extending economic assistance conditional on a completed transition to democracy, but, then again, that is not evidence in favor of the importance of organizational membership.
Peru
Ten years after the reestablishment of democracy in Peru in 1980, the political outsider Alberto Fujimori was elected president with an overwhelming majority. After increased tensions with the legislative and judiciary branches of government, Fujimore instituted a "self-coup" (autogolpe) on his own government by dissolving congress and suspending the judiciary on April, 5, 1992 (Pevehouse 2005, pp. 128-29). Pevehouse follows two events - one in 1993, the other in 2000 - which he deems critical to the return of democracy to Peru. I shall follow this subdivision in order to check the alleged importance of regional organizations for democratization in the Peruvian case.
The impetus from abroad   T       p0 regional organizations foster democratization?
97
In both events, the Organization of American States (OAS) was the external progenitor exerting pressure on Peru.7 In 1991 the OAS mem- - -ber states approved the Santiago Declaration (Resolution 1080), which mandated that the OAS must convene an immediate emergency meeting in the event of an interruption of democracy in any member state. Thjs resolution was immediately invoked in the wake of Fujimori's auto-golpe, leading the OAS to a swift condemnation of his actions and. calling an emergency meeting in the Bahamas in May. After a series of coordinated visits by OAS ambassadors, Fujimori decided to atti-i, Amidst allegations of the threat of sanctions, Fujimori addressed the OAS assembly concerning his actions, and promised a more swift return of democracy than he had initially anticipated. Instead of holding a plebiscite to legitimize his rule, which Fujimori had initially planned, this led to rescheduled elections to a new congress, which would .i <,<■■ serve as a constitutional assembly drafting a new constitution, hek. i November 1992 (Pevehouse 2005, pp. 128-29). After a year of deliberation, the constituent assembly approved a new draft constitution that passed by a small margin in a national referendum in October 1993 (Mauceri 1997, p. 902; Cameron 1998, p. 225).
Was this a democratic achievement accomplished by the OAS. .-■ Pevehouse argues? There are reasons to be skeptical about that interpretation. In fact, what Fujimori seems to have accomplished v. as mostly to untie his hands and dictate a solution to a crisis his own ■. autogolpe had spawned (Levitt 2006, p. 104). In the words of one astute observer:
Fujimori's forces controlled the entire process of rewriting the constitutjem and restoring Congress, from start to finish ... Fujimori and allies manipu ■■ lated the process for their own ends with little serious external resistance, hi fact, by sending election monitoring teams to observe the November vote, the OAS and the US inadvertently gave their seal of approval ro an electoral
7 In 1991 Peru was a member of four regional military, economic or political organizations: the Latin American Integration Association (LA1A), the Agency for die Prohibirion of Nuclear Weapons in Latin America, the Latin American Economic System, and the OAS. Although the OAS was strictly speaking the next - ■ . most democratic of these organizations (the most democratic being LAIA), rhe difference between rhem is inconsequential (the democracy score was 7.8 for the OAS, 8.0 for the LAIA, excluding Peru itself). In 1998, when Peru had also joined the Asia-Pacific Economic Organization, the OAS was the most democratic of Peru's regional organizations.
cess that, on the outside, appeared fair and transparent but that had been :.....(je5igned to exclude Peru's mainstream political parties. The Fujimori government emerged more powerful and consolidated than prior to the coup. (Legler 2003, p- 63)
It can thus hardly be argued that the OAS should be complimented for "restoring democracy" in Peru in 1993.
-Given that the 1993 event in any case does not fit the pathway criterion very well, we might instead turn to the events culminating in Fujimori's downfall in 2000. The actions of the OAS this time no doubt appear more critical. Beginning in the spring, the OAS monitored the presidential elections in which Fujimori defeated the opposition's frontrunner
: Toledo, but without gaining the required majority of votes. After increased suspicions of electoral fraud, the OAS mission left Peru, denouncing the run-off elections to be held in May as illegitimate. This made Toledo withdraw his candidacy, and Fujimori could safely win the contest. As a response to these events, however, the OAS submitted a high-level mission to Peru with the purpose of "strengthening democracy." This mission came to play a key role in the Peruvian case by establishing a round-table which brought together government, opposi-
........tion as well as civil society representatives. True, the event that in the
end triggered Fujimori's downfall was the disclosure on September 14 of a videotape which showed the leader of the government intelligence service, Vladimir Montesinos, bribing an opposition member of congress. In the dramatic events that followed, however, the round-table "served as an institutional forum through which domestic and international forces could pressure Fujimori during the crisis." The round-table ptessured the president to announce new elections to be held in 2001, and blocked military interventions in the process. When in November Fujimori resigned and fled the country, unable to escape from the massive corruption scandal that unraveled, the round-table eased the transition to an interim government (Pevehouse 2005, pp. 135-37; also see Cooper and Legler 2001; McClintock 2001; Legler 2003).
Summary
To sum up, the case study evidence in support of an effect of democratic fegional organizations on upturns is fraught with problems. A first thing to note is that Pevehouse's (2005) theoretical argument draws heavily on psychological phenomena. Perceived threats, socialization
98 The impetus from abroad
and legitimization are all highly cognitive mechanisms. The case stud evidence, however, rarely touches this fine-grained level of analyse _ With respect to the COE's role in Turkey, the potential influence .|< the OSCE in Hungary and the OAS in Peru, open pressure is the one of Pevehouse's mechanisms that is most well documented. Yet there ., precious little insight into how this pressure was experienced from t|..j perspective of the actors themselves._____
Second, as I have stressed in both the Turkish and Hungarian cases, there is a mismatch between the particular regional organization that enters my statistical analyses and the ones for which Pevehouse provic., case study evidence. Given that his strategy for gauging the level o-democracy among regional organizations should be seen as a pro -, rather than as a direct measure, this may seem less problematic at general level. After all, the fact that some regional organizations actually exerted important influences in the cases that fit the pathw n criterion could be interpreted as indirect evidence that the measurement strategy works. Yet one would prefer this strategy not only to work, but also to work for the right reasons. Moreover, in the Hungarian case the " problem is not just about what regional organizations did the work, but when they did it (most of the evidence indicates that this occurred after the transition to democracy).
Third, and related to this, Pevehouse's case study evidence cannot fully dispel the suspicion that the one regional organization that really ■ -has had power to influence domestic processes during the third wave of democratization is the EEC (later the EU). Perhaps there is no general ■ effect of democratic regional organizations, but what really works . perhaps the promise of future membership in this extremely rare case of an unusually powerful organization. True, the role of the OAS in Peru in 2000 is an important exception to this, but not even this case fits the pathway criterion very well. And perhaps the role of the OAS in Peru is a very unique exception? A recent review of the democracy-enhancing role of the OAS in Latin America states that "Peru is perhaps the closest thing to a successful case of OAS intervention," and more generally that the importance of the OAS for democracy promotion seems to be in . decline in the early twenty-first century (Boniface 2007, pp. 42-43, 57).::
The case study evidence in favor of a democratizing effect of regional organizations is thus mixed. There seems to have been an effect w.lii respect to the Council of Europe in Turkey in 1983, and the ■ Organization of American States with respect to Peru in 2000. The
Conclusion 99
support for a similar mechanism operating in Hungary from the Organization for Security and Cooperation in Europe, and the OAS in Peru in 1993, is, however, more strained.
Conclusion
J have in this chapter found statistical evidence indicating that trade volume impeded democratization during the third wave, but why this would be the case remains unclear. Moreover, neighbor diffusion is robustly related to democratization, but again without any observable causal mechanisms discovered, and mostly in a region (Sub-Saharan Africa) where it has rarely been documented. Membership in democratic regional organizations also appears to promote democratization, but the case study evidence in support of this contention is mixed. Foreign intervention, finally, sometimes works, sometimes not. In sum, the impetus from abroad at first seems obvious. But on closer inspection its inner workings appear elusive.