examining the influence on both the broader category of civil liberties restrictions and the more severe personal integrity abuse. I continue to compare the measures based on Amnesty International reports with those based on Department of State reports to control for possible ideological biases.
Military Control: Despite the strong theoretical arguments that would lead us to expect that military regimes will be more repressive than nonmilitary regimes, empirical evidence has been rather mixed. Poe and Tate (1994) found no such effect; however, subsequent studies of longer periods yielded the expected effect, although its impact remained somewhat minor (Poe, Tate, and Keith 1999; Keith, Tate, and Poe 2009). These studies have employed Madani’s (1992) classification, which follows McKinlay and Cohen (1975), where military regimes are defined as those that have come to power “as a consequence of a successful coup d’état, led by the army, navy or air force, that remained in power with a military person as the chief executive, for at least six months in a given year” (Madani 1992, 61; see McKinlay and Cohen 1975, 1). This operationalization also includes a small number of mixed regimes “with either a civilian as the chief executive and several military persons in the cabinet or a military head of government who nominated a civilian as the head of government and himself worked behind the scenes” (Madani 1992, 61).
Davenport (2007b), who utilizes a different categorization of military regimes, also fails to find an effect for military or hybrid military regimes on personal integrity abuse but does find evidence that military regimes are less likely to restrict the broader civil liberties category of repression (as defined and measured by Freedom House). Davenport uses Geddes’ (1999) somewhat broader definition, which delineates military regimes as those in which “a group of officers decides who rules and influences policy” (Davenport 2007b, 493; Geddes 1999, 4). He also includes in his analysis a military-personalist hybrid measure that fails to demonstrate any effect on either type of repression. Thus, as a whole the empirical analysis suggests that military regimes are not more likely to engage in personal integrity abuses such as torture, killing, and disappearances, and, even more interesting, that military regimes may be less likely to engage in restrictions on civil liberties. In explaining these different results Davenport notes: “It seems reasonable to suggest that these political systems reduce civil liberties restrictions because they tend to avoid involving themselves with the political processes which are normally responsible for these repressive activities (i.e., legislatures and courts) … and more inclined to use repressive techniques which are more directly within the realm of their expertise—physical violence” (500).
While I find Davenport’s explanation plausible, there is evidence of military regimes’ curbing freedom of the press, detaining protestors, and engaging in lower forms of repression. The nature of the Freedom House civil liberties measure also makes it somewhat problematic to assume that it does not capture physical violence, as its survey on civil liberties does include questions about political terror, physical harm, and violence. Also, as I read Davenport’s table, the military-regime measure did not have a statistical effect on any of the personal integrity models, a result that parallels previous findings (Poe, Tate, and Keith 1999). I continue to explore this perplexing relationship in each of the following chapters.
THE DOMESTIC SOCIOECONOMIC ENVIRONMENT
Economic Development and Growth: As I noted above, one of the earliest-identified and most consistent influences on political repression is the effect of the level of economic development or wealth (McKinlay and Cohen 1975, 1976; Mitchell and McCormick 1988; Henderson 1991; Poe and Tate 1994; Keith 1999, 2002a; Poe, Tate, and Keith 1999; Richards 1999). Most recent studies have consistently used per-capita gross national product or gross domestic product (GDP) to measure economic development and growth, and most have supported the hypothesis that higher levels of economic development do lead to less state repression, regardless of the category of repression. However, these studies have demonstrated that the size of the impact is typically not substantively significant; coefficients on average approach zero, especially when the models incorporate factors beyond the limited standard model, many of which represent a more viable policy alternative to improve human rights conditions than the level of increase in economic development that would be required to produce even a modest improvement (for example, Keith 2002a; Keith, Tate, and Poe 2009). Over time economic growth has been dropped from many models because of its repeated failure to achieve statistical significance; nonetheless, I continue to test it here, as the time frame of my current dataset represents the largest period under study thus far. In the section on international factors that follows I discuss the impact of other important economic factors that have a transnational character and are more controversial in regard to their influence.
Population Size and Growth: As with the domestic economic environment, population size and level of growth are seen as conditions associated with popular unrest stemming from scarcity within the state. Population size, typically logged, has consistently been shown to positively increase levels of repression, but, as with economic development, population growth has proven to be insignificant as studies of repression have extended over time and controlled for additional influences. I continue to examine these dimensions of population in subsequent chapters.
Colonial Experience: Over time the inclusion of a control for British colonial experience has become standard in explanatory models of state repression (for example, Davenport and Armstrong 2004; Keith 1999, 2002a; Poe and Tate 1994; Poe, Tate, and Keith 1999; Howard and Carey 2004). Interestingly, though, the results have been somewhat inconsistent. For example, the variable was not statistically significant in Poe and Tate’s early work, which examined repression in the 1980s, but the variable was statistically significant in all four of their models in their 1999 study (Poe, Tate, and Keith), which expanded the data back to 1976 and forward to 1993. However, Keith (2002a) and Keith, Tate, and Poe (2009), which extended the analyses forward to 1996 and added measures of constitutional protections and institutional provisions, found the measure to be statistically insignificant. Keith (2004) found some evidence of colonial influences, in that rights-related constitutional provisions performed better in countries with Iberian colonial experience; and Keith and Ogundele (2007) found that in Sub-Saharan Africa former francophone colonies had better human rights protection than anglophone ones and that constitutional protections and institutional provisions of former francophone colonies outperformed those of anglophone ones. In my work here, I examine these findings fully, along with other colonial legacies such as the type of legal system.
Domestic Law and the Judiciary
A growing body of empirical studies has examined the question of whether constitutional provisions for rights and an independent judiciary constrain the likelihood that states will repress their own citizens. Most of these studies have examined the role of specific rights typically embedded in bills of rights, and to a lesser extent institutional constraints such as provisions for an independent judiciary, state of emergency clauses, or provisions for federalism. The earliest analyses (Boli-Bennett 1976; Pritchard 1986) found evidence that not only were the constitutional provisions not associated with improved rights behavior, but also the associations were in the opposite direction, suggesting a harmful effect. However, Blasi and Cingranelli (1996) found some evidence of a weak direct effect. These early studies tended to utilize simple bivariate analysis of the association between constitutions and various rights measures, and additionally they tended to be limited to a single year; thus the generalizability of the results is significantly limited. Over time the analyses have become more statistically sophisticated, and the depth of analysis has extended to longer periods and a broader set of countries, beginning with Davenport (1996), which presents a rigorous and comprehensive analysis of the impact of constitutional provisions on state use of negative sanctions on its citizens. He, too, finds limited effects: only three constitutional indicators (out of fourteen) demonstrate a statistically significant effect on state repression, and several other provisions appeared to increase sanctions, although their coefficients did not achieve statistical significance. As Davenport notes, his sample is limited and not representative of the world and his time frame ends in 1982, well before the development of constitutions in the Third Wave of democratization and the post-Communist era. Nonetheless, the initial skepticism generated by his analysis continues across subsequent studies. Cross (1999) finds no statistically significant association between constitutional provisions for reasonable searches and human rights protection, but he does