Poe et al. (2000) found that the repressive response to domestic threats was dependent not only upon the level of the threat but also upon the prior level of repression. Generally the greater the level of the threat, the more likely a regime was to engage in coercive force, although some regimes did in fact engage in repression even when they were not seriously threatened. However, where repression levels were already high, nonviolent protest and nonviolent rebellion had little or no effect, but these threats did have some effect on repression in states that had exercised only low to moderate levels of repression in the past. On the other hand, violent opposition or rebellion increased the level of repression regardless of the prior level of repression, unless repression was already quite high. The analyses also showed that regimes responded in most cases by lessening the repression once the threat ceased; this result suggests that regimes do weigh the potential cost of engaging in political repression, perhaps in fear of their supporters’ defecting if repression continues once the threat diminishes or perhaps in response to the economic costs in terms of resources or the loss of support of international actors (Poe et al. 2000, 58). Regan and Henderson (2002) also found that the level of threat was positively associated with repression, and that the impact of threat was greater than that of regime type. They also found that when controlling for threat, less-developed states with intermediate levels of democracy had the greatest odds of using coercive force. Political scientists now have multiple measures capturing a fuller range of repressive tools that regimes have at their disposal, as well as multiple measures capturing the most significant dimensions of threat. In Chapter 6 I examine more fully the theoretical expectations in regard to both internal and external threats, as well as formal regulations of regime behavior during times of threat.
REGIME CHARACTERISTICS
Democracy: As I noted above, political democracy has proven to be one of the more potent and consistent explanators in political repression models; almost four decades of research support the expectation that higher levels of democracy are associated with lower levels of repression, across both categories of repression (for example, Hibbs 1973; Ziegenhagen 1986; Mitchell and McCormick 1988; Henderson 1991; Poe and Tate 1994; Davenport 1995c, 1999, 2007a; Poe, Tate, and Keith 1999; Richards 1999; Zanger 2000; Apodaca 2001; Keith 2002a; Bueno de Mesquita et al. 2005). As discussed above, more recent studies of democracy’s influence on repression have urged us to look beyond a simple linear relationship and have indeed demonstrated that the relationship is not as simple as we have generally theorized. Gartner and Regan (1996) and Regan and Henderson (2002) demonstrated that the impact of democracy on human rights abuse is curvilinear when controlling for the level of threat, with semi-democracies more likely to repress than full democracies and autocracies. Their work was, however, limited to only less-developed countries.
Davenport and Armstrong (2004) demonstrate that we have probably misspecified democracy’s role in our models, specifically showing that democratization tends to have no impact until it reaches a critical threshold. Using the Polity measure of democracy, they find that at the lower levels of the scale (below 8) there is no impact, but as states progress into the intermediate level there is some negative effect on repression, and as states reach the highest level of democracy there is a strong negative effect (548). Other scholars have perceived democracy as a multidimensional concept and have sought to understand which dimensions of democracy are more likely to lessen the use of political repression (Gleditsch and Ward 1997; Keith 2002a; Bueno de Mesquita et al. 2005). Gleditsch and Ward (1997) examined empirically the components of the Polity measure and demonstrated that the measure is primarily driven by its executive constraint dimension and that the other dimensions, which measure patterns of executive recruitment and the extent and competitiveness of participation, were “not especially powerful in determining the degree of democracy” (380). They conclude that to use the data “summarily to classify modern polities as democracies directs attention away from the actual data that have been collected on authority patterns” (380). Thus, they urged scholars to move beyond discrete classification of “democracy” or “autocracy” and to focus more on the sub-dimensions of this measure. Studies of political repression do not use the Polity data as a dependent variable, nor do we typically use the data to divide countries into two discrete categories; rather, we use the index to control for the level of institutional democracy attained in the country.
Nonetheless, Keith (2002a) followed Gleditsch and Ward’s suggestion and tested the four components of the Polity democracy measure separately; two of the components produced statistically significant relationships: constraint on the chief executive and the competitiveness of political participation. While Ward and Gleditsch’s analysis revealed that the Polity democracy measure was largely driven by executive constraint, Keith found the competitiveness of political participation to be a more powerful explanator. Not surprisingly, the combined maximum impact of the two individual components was approximately equal to that achieved by the composite measure. Regardless of the construction of the democracy measure, its impact was one of the largest in the analysis. Bueno de Mesquita et al. (2005) confirmed this result, and also demonstrated that improvement of the state’s level of democracy did not make a difference in human rights until the state achieved full democracy, supporting Davenport and Armstrong’s analysis. Additionally, they demonstrated that the significant components of democracy had a sharp threshold effect, and they concluded that “real improvements in human rights do not occur smoothly but reflect a discontinuous step function achieved only when a society becomes fully democratic” (453). Ultimately, they concluded that accountability through a competitive party system is a critical feature of democracy, but for it to be effective in reducing human rights abuses, institutional foundations must first be put into place. In Chapter 3 I engage in a thorough examination of democracy, its various components, and effects.
Types of Autocracy: Currently, Davenport’s work (2007b) represents the most comprehensive examination of autocratic regimes’ use of political repression, in particular demonstrating that there is variation among types of autocratic regimes in their use of repression. He uses Geddes’ (1999) categorization of autocratic regimes, which distinguishes two extreme types on this dimension: personalist systems and single-party systems, with some regime types such as military regimes and military hybrid systems fitting in between. His work demonstrates that only military regimes influence the level of restrictions, and they do so in a negative direction. When he adds interactions for the end of the Cold War, the military regime’s effect remains, but then he finds that both personalist and single-party regimes are less repressive. In regard to personal integrity rights, the military regime’s effect disappears, and only the single-party autocracy measure influences the level of repression, and it does so in a negative direction. Controlling for the various types of autocracy removes the effect of the leftist regime variable, which as discussed above has performed somewhat inconsistently in previous work.
Marxist-Leninist Regimes: The expectation that leftist regimes will be more likely to repress than non-Marxist regimes has not been consistently supported, even though it has been considered a part of the standard model. While Mitchell and McCormick (1988) did find that leftist regimes were more repressive than non-Marxist authoritarian regimes, Poe and Tate (1994) found such a link only when employing measures of human rights based on Department of State reports, which suggested quite plausibly a bias in the Department of State reports of the 1980s. Subsequent analyses that expanded the time frame beyond the 1980s have failed to support the hypothesis that leftist regimes are more repressive; in fact, the analyses have suggested the opposite (Poe, Tate, and Keith 1999; Keith 2002a, Keith, Tate, and Poe 2009). Thus, past repression in leftist regimes may actually reduce the need for future repression or the need for more severe forms of repression, such as those measured by abuse of personal integrity rights. Interestingly, Davenport (2007b) found that that a leftist regime is more likely to increase civil liberties restrictions, but, as with the studies above, that regime type decreases the odds of personal integrity abuses. The latter finding is not unexpected, since he uses the same data set (Poe, Tate, and Keith 1999; Keith 2002a); however, it is interesting that when he adds the autocracy types to his model, the effect of a leftist regime disappears altogether. I explore the Marxist/Marxist-Leninist influence