Abstract
Countries interested in the promotion of political development often provide aid in the form of democracy assistance. However, some regimes resist these attempts to promote democracy, introducing repressive measures to counteract their effectiveness. Hence, democracy assistance sometimes has the unintended consequence of curtailing democracy. This article explains how the size of the targeted regime’s military determines the effectiveness of democracy assistance and why it can sometimes result in lower levels of political freedom. Large militaries, often holding a privileged position in authoritarian regimes, will be threatened by political liberalization and its associated redistribution of resources. They will thus work with the regime to limit the effect of democracy assistance, while their size makes this repression more feasible. In states with smaller militaries, regimes have less incentive and capacity for repression, and democracy assistance is more successful at empowering democratic opposition. Cross-national statistical analysis of the United States Agency for International Development democracy assistance supports the argument.
Keywords
Democracy assistance—foreign aid focused on promoting democratic development—is a promising form of intervention for improving political and social development around the world. Such aid aims to foster open politics and societies by increasing the level of political competition and improving the institutional framework through which competition occurs (Carothers 1997, 112-15). The United States Agency for International Development (USAID) devotes substantial amounts of aid to promote rule of law, develop civil society, improve elections, and foster decentralization and better governance, hoping to increase democracy in the recipient states. These efforts have often been considered successful; cross-national studies have found that such aid is associated with an increased level of democracy (Finkel, Pérez-Liñán, and Seligson 2007; Scott and Steele 2011). However, there have been notable cases where democracy assistance has triggered a backlash from the recipient regime (Carothers 2006). While democracy assistance can be a positive force for political development, it is not always well received by the targeted state. It is unclear why democracy assistance may have a positive effect on democratic governance on average while in some states the result is less freedom.
Backlashes against democracy assistance efforts curtail democratic freedoms (Carothers 2006). In September 2012 (Elder, 2012; Herszenhorn & Ellen, 2012), the Russian government announced that it would no longer allow the USAID to continue its programs and operate in Russia. The USAID had funded democracy promotion programs in the country since the collapse of the Soviet Union (Mendelson 2001). Even prior to the expulsion of the USAID, Russia had restricted the freedom of civil society groups that received external support (Ambrosio 2007, 239). Russia’s behavior is not unique, many countries such as Venezuela, Egypt, Belarus, Uzbekistan, and Zimbabwe have placed restrictions on nongovernmental organizations (NGOs) receiving foreign funding (Gershman and Allen 2006). In these cases, democracy assistance unintentionally resulted in increased repression, representing wasted resources. Understanding the circumstances that lead to these backlashes could enable more effective democracy assistance.
The incentives and capacity of the military may play significant roles in how democracy assistance is received. The military is often an important actor within a regime, with the ability to influence the level of political liberalization (Huntington 1981; Bellin 2004; Albertus and Menaldo 2012). When a military is large, it has more reason to fear the redistributive consequences of increased liberalization. Furthermore, large militaries make it easier for regimes to introduce repressive measures and diminish the effect of democracy assistance. For these reasons, regimes in states with large militaries may respond to democracy assistance with increased repression.
I argue that the effect of democracy assistance is conditional on the size of the military in the recipient state. As the military gets larger relative to the population or economy of the recipient state, democracy assistance will be less effective or even counterproductive. I use a time-series panel data set containing observations from 141 developing countries from 1990 to 2004 to test this argument. The findings show that democracy assistance that has been directed at states with large militaries was associated with a negative or limited effect on democracy, while the effect in countries with small militaries has been larger than previously estimated.
The Effectiveness of Democracy Assistance
The effect of foreign aid in general on democracy is unclear (Knack 2004; Djankov, Montalvo, and Reynal-Querol 2008; Wright 2009). However, various studies have found a positive association between democracy assistance and democracy (Finkel, Pérez-Liñán, and Seligson 2007; Scott and Steele 2005, 2011). Finkel, Pérez-Liñán, and Seligson (2007) show how democracy assistance influences the overall level of democracy, finding ten million dollars in democracy assistance raises the level of democracy by a quarter of a point on the Freedom House scale. This is a modest, but substantively interesting increase in the level of democracy. They argue that democracy assistance can affect the short-term or micro-level factors that influence democracy. Democracy aid helps constituencies to develop democratic reforms. By developing the strength and size of these constituencies, the prospects of political liberalization are improved.
While cross-nationally there seems to be a small increase in the average level of democracy in states associated with democracy assistance, other studies—both qualitative and quantitative—have found such aid to be ineffective or detrimental to the level of democracy in certain states (Gershman and Allen 2006; Bosin 2012). For example, states can respond to democracy aid by limiting freedoms; this is particularly true in hybrid regimes that have incentives to allow aid but where leaders deviate from democratic practices to hold on to power (Gershman and Allen 2006). Other cross-national work has shown that regime type might influence the effect of democracy assistance, as single-party regimes are more sure of survival and have longer time horizons than others (Cornell 2013). In former-Soviet countries, democracy assistance has been largely ineffective due to the common historical and institutional legacies of communism within the former Soviet Union that reduce the incentives of leaders to adopt democratic reforms (Bosin 2012). This suggests that democracy assistance may be conditionally effective. However, there is little clear and systematic explanation of why democracy assistance works in some instances and not others. This study contributes to the literature exploring the conditional effects of democracy assistance.
The effects of development policies and aid are dependent on the institutions and characteristics of the states receiving the aid or enacting the recommended policies (Djankov et al. 2003; Rodrik 2008). States with greater social capital can adopt less dictatorial policies (Djankov et al. 2003). Aid is more effective combined with orthodox economic policies (Burnside and Dollar 2000). Authoritarian leaders with longer-time horizons adopt policies that result in growth when combined with aid (Wright 2008). There is no reason to expect that democracy assistance will be any different. If democracy assistance works, it will do so only in some institutional and political contexts. The USAID, the leading provider of democracy assistance, is becoming more aware of these problems. In a 2008 report, the USAID claimed that Selecting targets for DG [Democracy and Governance] programming requires a theory, or at least a hypothesis, about relationships among different institutions and processes and how they contributed to shaping overall trajectories toward democracy and governance. (Goldstone et al. 2008, 43)
Democracy Assistance and Military Size
The following argument uses two findings from the democratization literature to explain the disparate effects of democracy assistance. First, it considers the important redistributive consequences of political liberalization. Second, the theory incorporates the role of the military as part of a ruling coalition and the influence of the military on a regime’s choices. This provides a more complete basis for understanding a regime’s incentives and choices when faced with democracy assistance.
Regimes vary along a continuum from democracy to autocracy. Democracy is a political system where both political contestation and political participation are free (Dahl 1972). Regimes can constrain contestation and participation in many ways. They may place institutional restrictions, limiting legal rights and free participation in the political process (Silitski 2009). Regimes can use violence and intimidation to exclude opposition actors from the political sphere or limit their effectiveness (Davenport 2007). Legal restrictions or violence and harassment will reduce political participation and the level of democracy. Common measures of democracy such as Freedom House and Polity include indicators of violence and repression. Democracy can rise and fall with the level of harassment, as well as formal institutional changes.
Constraints on contestation allow the ruling coalition to extract rents, and increased democracy can threaten access to these rents (Boix 2003; Acemoglu and Robinson 2005; Benhabib and Przeworski 2006). Democracy assistance increases the strength of pro-democracy groups and opposition to the regime (Finkel, Pérez-Liñán, and Seligson 2007). By strengthening potential opposition, democracy aid increases the costs of maintaining a closed political system. The more utility powerful actors get from limiting political participation and contestation, the more likely it is that the regime leaders will engage in increased repression to preserve the regime (Acemoglu and Robinson 2000). Where these incentives are lower or repression is too costly, the increased strength of the pro-democracy forces within the country should result in the regime liberalizing.
The military can determine how regimes respond strategically to threats from society. The military must be endowed with enough resources and power to protect the regime from domestic and external threats, but in doing so, the military is empowered to threaten the regime itself (Feaver 1999, 214). While other actors and institutions, such as strong parties, may influence regime decisions and stability, the military because of its control over much of the state’s coercive capacity will nearly always be in a position to influence outcomes. Regime leaders and opposition must consider the military’s interests and preferences when making choices regarding repression and liberalization or opposition.
The larger a regime’s military in terms of the population employed by the military or the proportion of the economy devoted to the military, the more important it is likely to be as a coalition member or constituent to the regime (Brooks 2013). Increased size relative to other domestic actors means increased strength. Further, a larger military in these terms is more likely to benefit from the political system. Svolik (2012, 2013) has shown that providing a larger amount of the regime’s resources to the military makes the military more likely to support the regime when faced with domestic threats, while the increased power of the military can also force the regime to represent the military’s interests. Increased size makes the military more capable of and interested in influencing outcomes and strengthens its ties to the regime.
The military often benefits from restrictions to political freedom. Closed political systems provide the military with avenues to influence political outcomes (Bellin 2004, 145). Less democratic regimes rely on the support of the military, providing the military with opportunities to extract rents (Kimenyi and Mbaku 1996; Gupta, De Mello, and Sharan 2001). In a more democratic regime, the military competes with more actors, reducing their influence and power (Davenport 1995, 123). One study has estimated that democratic transitions can lead to military spending being reduced by 2 percent of gross domestic product (GDP); (Collier and Hoeffler 2002). As a probable target for redistribution when a regime liberalizes, a large military has strong incentives to preserve the existing system or something similar.
The scope of this argument extends beyond autocracies to countries where democracy is present but is not fully or safely institutionalized. Often in new or unconsolidated democracies, militaries have negotiated or preserved privileges from previous regimes (Pion-Berlin 1992). Consolidation of civilian control over the military is often a long and difficult process and failure to consolidate civilian control can lead to democratic back sliding (Pion-Berlin 1992; Serra 2010; Barany 2012). A legacy of military rule is associated with a failure of democracies to consolidate (Cheibub 2007; Svolik 2008). Moreover, it is not clear that democracy limits repression and violence at the traditional cutoff points used to separate regime types (Davenport and Armstrong 2004). The theory should apply both to autocracies and democracies in developing states where norms and institutions of civilian control are not fully established. In contrast, where the military has a strong organizational identity that is in tune with democratic principles, intervention into politics is unlikely (Huntington 1981). Additionally, states with developed, long-established democracies or developed economies are not eligible for democracy assistance and hence are beyond the scope of this argument.
Faced with democracy assistance, militaries can affect the regime’s choice of action; and if they are larger, they have incentives to limit the effects of such aid. Militaries can discipline actors within the regime who might be tempted to liberalize (Davenport 1995; Albertus and Menaldo 2012). They can intervene directly into politics if their mission is not well defined or their organizational interests are threatened (Huntington 1981). When the military is powerful enough, the regime may act in accordance with military interests to reduce the military’s incentives to act against the regime (Svolik 2013). The regime could attempt to curtail the newly empowered societal actors by introducing institutional restrictions.
The military can be a tool of the regime, providing a means of efficiently engaging in repression (Albertus and Menaldo 2012). Less democratic regimes can use the threat of the military to limit freedoms in society and repress those actors who organize to promote democracy. A larger military provides regimes with a greater and more credible threat to clamp down on opposition. While the military is not commonly used to engage in day-to-day repression, it will often support the goal of regime survival in the face of large-scale unrest (Svolik 2011, 7). On the other hand, where militaries feel that they are losing out or being materially slighted by the regime, they are unlikely to intervene to prevent democratization (Pion-Berlin, Esparza, and Grisham 2014). The military refusing to back the regime is often an essential part of democratic transitions (Lee 2009). If the regime believes that a military will support it, this should result in increased willingness of the regime to try to preserve their control. It should also work to deter democratic opposition, even if they are receiving democracy assistance.
Military size could influence a regime’s bargaining position with donor states. 1 A large military may make the regime a more valuable ally to the donor states or simply less dependent on the donors. The regime could demand “tamer,” less effective forms of democracy assistance (Bush 2015). This would result in less effective democracy assistance when the military is large. However, it would not predict a backlash.
The final mechanism raises the question of why regimes allow democracy assistance at all. Hyde (2011) has demonstrated that regimes have an incentive to signal their commitment to democracy. Leaders receive a “democracy premium” if the international community believes they are committed reformers. Leaders who are not committed to democracy pool with committed democratizers by adopting policies that can increase the costs of holding onto power in order to gain access to the premium. Informing a donor state that they do not want democracy assistance may be a more obvious signal of the regime’s lack of commitment than altering domestic conditions afterward. This pooling behavior could explain why an opposition engages in behavior that results in repression rather than being deterred. By accepting democracy assistance, the regime not only influences the belief of the international community, they may also increase the belief of domestic actors that the regime lacks the resolution to defend the current system.
This process is compounded by information asymmetries that exist between the military and civilians (Feaver 2009, 68-72). It is unlikely that a military’s jurisdiction extends to negotiating aid packages with donor states. The USAID admitted to dealing with civilians and engaging with the military in only a very limited manner (Forhman and Welch 1998, 5). In general, militaries also possess limited policy expertise, which often causes them to misjudge government policy, leading them to act politically (Svolik 2013, 768). Poor information sharing between civilians and the armed forces can lead to differences in risk assessment and result in serious strategic miscalculations (Brooks 2008). Militaries may have limited knowledge about and influence over the regime’s initial decision to accept democracy assistance, forcing them to rely on post hoc measures. These asymmetries partly explain the military’s failure to prevent the acceptance of democracy assistance in the first place.
The military played a role in one of the most notable successes of democracy assistance, the Georgian Rose Revolution (Mitchell 2006). In 2003, opposition coalesced to the authoritarian regime of Shevardnadze following a rigged election. The leaders of this movement had engaged with and received support from US democracy promotion agencies (Mitchell 2009). The President of Georgia, Eduard Shevardnadze, approached the military to help repress the revolution. The military’s refusal to back Shevardnadze ensured that the Revolution was successful (Mitchell 2004, 7). A pro-democracy activist, Giorgi Kandelaki (2006, 11), stated that the realization that the military would not intervene was a crucial part of mobilizing support for the Rose Revolution. The military’s decision influenced the strategic considerations of both the opposition and regime leaders.
The Georgian military resented its lack of resources. In 1998, 2001, 2002, and 2003, parts of the Georgian military staged mutinies in response to poor conditions (Jones 2013, 264). The commander of the mutinous battalion in 2001 Colonel Georgi Krialashvili stated that “I feel ashamed and I can’t look into the eyes of my soldiers and officers. The soldiers serve practically barefoot, the service conditions are very poor. Officers haven’t received their salaries for more than a year” (British Broadcasting Corporation [BBC] 2001). In 2003, shortly before the Rose Revolution, there was another mutiny among veterans possibly due to poor conditions (BBC 2003). Georgia had a small military, with less than 1 percent of its labor force in the armed forces and devoting just over 1.3 percent of its GDP to military spending, well below the global average of 2.3 percent for the year. Shevardnadze had done little over the course of his tenure to guarantee military loyalty.
The new leader Saakashvili emphasized his support for the military. Saakashvili discussed strengthening the armed forces in his inaugural address (Fairbanks 2007). Saakashvili lived up to his promise to improve the conditions of the military. For example, the government increased the wages of officers from US$48.50 a month in 2003 to US$605 a month in 2007 (Jones 2013, 266). The new regime saw the importance of building the military as part of its state-building agenda.
Such processes can also help explain a backlash. Armenia received democracy assistance in all the years under observation in the sample. The military was large compared to the successful case described earlier. The government has close ties to the military and military veterans, who have both been central in repressing pro-democracy forces (Way 2008). One human rights activist describes the Armenian military in these terms: “The pervasive nature of corruption is also disturbing: there is no oversight mechanism regarding soldiers’ food, clothing and fuel expenses. The lack of monitoring level is tremendous. Great resources are provided to military sphere and pilfering and stealing is simply very seductive for commanders and other military officials. From the viewpoint of elections, the Army is made a wonderful resource to serve for political purposes. The list of voters are kept confidential, which are under control of their commanders and the soldiers are told by their commanders who to vote for” (Sakunts 2013). The Armenian military has a clear economic incentive to preserve the regime.
The military contributed to regime stability in Armenia. In 2008, as in Georgia a few years earlier, protesters went to the streets motivated by a perception that an election was stolen. The role of the military and strategic uncertainty in triggering a backlash is also apparent in this episode. Lev Ter-Petrosyan, the former President and leader of the protest movement, believed and reassured his supporters that the military would not get involved repressing the opposition (Radio Free Europe/Radio Liberty [RFE/RL] 2008). Another leading opposition politician Alexander Arzoumanian when asked if the regime would use force stated, “They are cowards. They don’t have the courage to do such a thing. They would never have the courage to come and attack such a powerful crowd” (Collin 2008). Along with the police, the military used force to repress the protesters, killing eight people (Human Rights Watch [HRW] 2009). Political liberties as measured by Freedom House declined after these events.
Democracy assistance increases the strength of domestic pro-democracy forces. A stronger military increases a regime’s incentives to counter the effects of democracy assistance, while increasing the feasibility of doing so if there is also a greater threat due to democracy assistance increasing the strength of pro-democracy groups. Absent a threat, there may be no change associated with military size. These two effects of military size combined with the effect of democracy assistance lead to the following hypotheses:
Democracy assistance will increase the strength of opposition groups. When there is little incentive for regimes to repress, which is more likely when the military is small, democracy assistance stands a better chance of being effective. Increasing military size will limit the effects of democracy assistance, as the incentives of the military and regime to engage in repressive behavior are larger. Larger militaries will increase repression or deter democratic opposition, reducing the effectiveness of democracy assistance. Alternatively, larger militaries may allow the regime to demand less-effective democracy aid. The positive effects of democracy assistance will be counteracted when the military is large.
While large militaries could deter the opposition from using the resources they gain from democracy assistance, it is also possible that they will prompt a backlash from the regime, resulting in reduced democracy. Uncertainty is one source of repression. If there is uncertainty regarding how strong or resolved the regime is relative to the opposition, there are incentives for the regime to engage in repression to prove their strength (Acemoglu and Robinson 2000). As argued earlier, regimes who are not committed to democracy may allow democracy assistance in an attempt to convince others that they are, such a purpose could produce uncertainty. If there is uncertainty regarding democracy assistance’s effects on the regime’s relative power, then repression could result, not just deterrence.
Further, democracy assistance by altering the balance of power between the regime and opposition in the long term may make credible commitments impossible. Institutional stability depends on actors not having incentives to deviate from institutional arrangements and agreements. The inability of opposition to commit to moderate demands can create incentives for the regime to repress (Kalyvas 2000). Changes in the distribution of power have been shown to undermine credible commitments and lead to violence (Walter 1999). Absent repression, democracy assistance could upset the future distribution of power, strengthening pro-democracy forces and leading them to make large demands in the future. If repression is cheaper before opposition groups are empowered, prompted by a powerful military, there may be regimes that can repress immediately but not after democracy assistance has had time to work. In this situation, the regime cannot rely on deterrence to preserve their privileges. This inability to commit to repression to overcome opposition in the future forces regimes to reduce democracy in countries with large enough militaries that they do not concede immediately.
Data and Methods
The effect of democracy assistance interacted with military size on democracy was assessed using a time-series cross-sectional data set including 141 countries from 1990 to 2004. The four criteria established by Finkel, Pérez-Liñán, and Seligson (2007) to determine eligibility were as follows: (1) if a state received any democracy assistance during the period under observation; (2) if a state was defined as low or middle income by the World Bank; (3) if a state was defined as partially free or unfree by Freedom House on average over the period of 1974 to 2004; or (4) if a state was newly independent. These four criteria define a population of theoretical interest, as they are likely to correspond with poorly institutionalized civil–military relations.
While the United States is not the only country interested in democracy promotion, it is arguably the most prominent actor in this area. Democracy promotion has been one of the United States’ foreign policy goals for much of its existence (Cox, Ikenberry, and Inoguchi 2000; Smith 2012). The use of the USAID data is appropriate, as the dominant position of the United States in the international system makes it an important empirical object of study. From a theoretical perspective, its preeminent global position means the United States is the state that will be most influential in terms of altering other states’ preferences. 2
Focus on a single donor state ensures consistency in measurement. Different states have different definitions of democracy or governance assistance; for instance, Britain focuses heavily on good governance issues rather than democracy (Carothers 1997, 110). This makes it difficult to assess the effects of democracy assistance using overall levels of aid from such sources as the Organization for Economic and Cooperation and Development. Therefore, focusing on the USAID’s efforts leads to greater confidence that the aid in question is aid directed at democracy promotion as opposed to good governance projects more generally.
Dependent Variables
I used two different dependent variables to measure the aggregate level of democracy in a country: Polity and Freedom House. Aggregate measures of democracy are needed because there are many legal or extralegal, formal or informal techniques open to autocrats to counter democracy assistance. 3 An aggregated measure that incorporates many different factors will pick up variation that may be missed by more fine-grained measures. The two dependent variables represent different aspects of democracy and should capture any potential backlash.
Polity is more concerned with political contestation and institutions. This variable is taken from the Polity IV data set (Marshall and Jaggers 2013). Polity is the revised Polity score designed for use in time-series analysis. The variable ranges from −10 to 10, with higher values of the variable indicating a higher level of democracy. Freedom House focuses on civil and personal liberties to a greater extent. Following the procedure of past studies of democracy assistance, this is an aggregated variable consisting of both Freedom House’s Political and Civil Liberties indices (Finkel, Pérez-Liñán, and Seligson 2007; Scott and Steele 2011). These variables range from 1 to 7, with 1 being the highest level of freedom and 7 being the lowest. For the purpose of this study, the two indices were combined and reduced by one so that the variable ranged from 1 to 13. For ease of interpretation, the scores were reversed so that 13 is now the highest level of democracy.
Independent Variables
Democracy assistance measured all aid, measured in millions of constant US$2,000, directed by the USAID to four different areas: civil society promotion, elections and political processes, rule of law, and governance. This represents all the aid that is believed to increase democracy levels. In practice, as aid appropriated for one year was often spent the following, a two-year moving average of appropriations better reflected spending practice on democracy promotion. These data were taken from Finkel, Perez-Linan and Seligson. 4 As in past studies, the analysis was conducted using the absolute spending per country (Finkel, Pérez-Liñán, and Seligson 2007; Scott and Steele 2011). However, to account for the relative size and wealth of the country, the analyses were also run using the democracy assistance variable normalized by GDP, Democracy assistance (%GDP).
Two variables were used to measure military size. Percentage labor force in military measured the percentage of the labor force that was a member of the armed services. This measure was taken from the World Bank Development Indicators, which record the number of people over the age of fifteen employed by the military. 5 The larger the percentage of the population that is a member of the military, the stronger the military and the greater coercive capacity of the state. It also reflects the proportion of citizens who are reliant on the military for their economic welfare. Percentage labor force in military thus proxies for the size of the constituency invested in the military’s well-being. However, this variable may not measure the benefit that the military is extracting from the regime, and hence their interest in preserving the regime. To capture this, military spending as a ratio of GDP was used. Military spending data were taken from the Stockholm International Peace Research Institute.
Alternative Explanations and Control Variables
Higher levels of development are associated with democracy (Lipset 1959; Boix and Stokes 2003). GDP per capita, the natural log of GDP per capita in constant US$2,000, was included from the World Bank’s World Development Indicators (WDI). Furthermore, GDP per capita is associated with the size of the military (Cappelen, Gleditsch, and Bjerkholt 1984). Higher levels of development allow for more resources to be assigned to military spending. Theoretically, military size could proxy for state capacity. GDP per capita, however, has been used to proxy for state capacity, and therefore its inclusion addresses these concerns (Fearon and Laitin 2003). 6
Oil revenue measured the oil revenue of the state per capita. The resource curse is well known (Ross 2001). The rents derived from resources could be used to expand the military spending of the state. The data were taken from Haber and Menaldo (2011).
Growth measured the annual percentage change in GDP, drawn from the WDI. The level of economic growth may be associated with democratization (Bermeo 1990; Przeworski and Limongi 1997). Growth has been shown to be associated with military spending (Cappelen, Gleditsch, and Bjerkholt 1984; Chowdhury 1991).
Gini from the Standardized World Income Inequality Database was included (Solt 2009). Inequality has been associated with regime change (Boix 2003; Ansell and Samuels 2010). Inequality is also correlated with military structure and military spending (Henderson 1998; Caverley 2014).
Keeping allies in power may sometimes be a higher priority than democracy promotion (Mckoy and Miller 2012). To control for this, I included Affinity, a variable measuring affinity with the United States using UN General Assembly roll call votes (Gartzke 1998). This variable measures how close a state’s voting matches that of the United States and can be used as a proxy for shared interests. To further control for the strategic interests of the United States, Military assistance, US military assistance as a percentage of GDP, was also included as a control. The United States is likely to direct higher levels of military assistance to allies and to states it deems important for security. This may affect a military’s incentives in a variety of ways; it could act as a constraint on military intervention or bestow them with more freedom by signaling their value as an ally.
Incentives to transition may vary with regime type (Geddes 2003). Previous studies have found somewhat contradictory effects of military dictatorship. One study found that military dictatorships are less invested in holding on to power and consequently less inclined to resist democratic transitions when faced with domestic pressure (Geddes 1999). Military leaders can be held accountable by other officers in the regime, meaning they cannot act against the interests of the military too easily (Weeks 2008, 46). However, other studies have found that military regimes form narrower coalitions and as a result are less influenced by international pressure and public opinion (Wright 2009; Escribà-Folch and Wright 2010). A dummy variable Military regime was included coding whether or not a regime was military dictatorship. In addition, party-based regimes have greater institutional capacity to manage conflict. Party-based regime variable from Geddes, Wright, and Frantz (2014) was included.
International crisis controlled for the effects of external threat, which can create incentives for leaders to concentrate their power and reduce the level of democracy (Thompson 1996). External threat can also create incentives to build up military size. A dummy variable coded whether or not a state was involved in an international crisis in the previous five years using the Correlates of War (COW) militarized interstate dispute data set (Ghosn, Palmer, and Bremer 2004). As larger states are likely to feel more secure than smaller ones, Population was also included in the model. Finally, the variable Instability was used to control for domestic threats. This variable was created using the Political Instability Task Force data set and indicates whether or not a state was undergoing a revolutionary war, an ethnic war, or genocide or politicide in that year (Goldstone et al. 2010). Adverse regimes changes were excluded, as they measure changes in the level of democracy.
Analysis and Results
The theory posits that the effect of democracy assistance is conditional on the size of the military of the recipient state. To assess this argument, interactions between the variables representing democracy assistance and military size were examined. 7 Models were run with fixed effects for both country and year to address unobserved country level influences, and a lagged dependent variable was used to address potential autocorrelation. 8 Civil–military relations are influenced by unobservable, slow moving factors. Historical legacies and path-dependent processes influence military political involvement (Feaver 1999; Pion-Berlin and Trinkunas 2007; Barany 2012). For example, institutional conditions in Pakistan and the international environment shortly after independence created a military culture that favored political intervention (Staniland 2008). Elsewhere, the legacies of communism have increased civilian control over the military (Bunce 2003). Such historical processes make including fixed effects important for estimation.
Fixed effects models estimate the effect as independent variables change relative to the mean for that country. This means that what is being estimated is the effect of military size and democracy assistance as they change relative to that country’s mean level. By controlling for other dynamic institutional factors related to the regime, it should be possible to isolate the effects of military power and democracy assistance. However, while fixed effects may reduce bias, variance is increased.
Additionally, combining fixed effects with a lagged dependent variable can result in Nickell bias, where the centered lagged dependent variable is correlated with the centered error term, when the time-series component of the panel is short. In this case, the series was probably long enough that these concerns were outweighed by the need to include a lagged dependent variable to deal with autocorrelation (Beck and Katz 2011). However, some concern may still exist. For this reason, a generalized method of moments model was implemented. The model is useful for obtaining efficient estimators that account for serial correlation and unobserved effects (Wooldridge 2001). 9
Polity was regressed on the various interaction terms (Table 1, models 1–4). In three of the four models, the interaction showed a significant and negative effect. However, the coefficient of an interaction term provides only limited information about the marginal effects. Marginal effects of an interaction need to be calculated across the range of the variables (Braumoeller 2004; Brambor, Clark, and Golder 2006). The marginal effects are shown in Figures 1 and 2.
Effect of Interaction between Democracy Assistance and Military Size on Democracy.
Note: Cluster robust standard errors within parentheses. FH = Freedom House; GDP = gross domestic product; MID = militarized interstate dispute.
*p < .10. **p < .05. ***p < .01.
Marginal effect of democracy assistance on polity as percentage labor force in military changes. Dashed lines represent 95 percent confidence intervals.
Marginal effect of democracy assistance (% gross domestic product [GDP]) on polity as military spending changes. Dashed lines represent 95 percent confidence intervals.
The marginal effect of democracy assistance for a country with 0 percent of its labor force in the military was an increase in democracy of .1245 on the Polity scale for each million spent on democracy promotion. At 3.43 percent of the population in the military, the positive effect was gone, and beyond that the effect was negative (see Figure 1a, model 1). The results for democracy assistance (%GDP) were similar. While there was a large positive effect when the military was small, when the military made up slightly more than 3.4 percent of the labor force the marginal effects of democracy assistance were zero (see Figure 1b and Table 1, model 2).
Interacting democracy assistance and democracy assistance (%GDP) with military spending gave marginal effects that similarly supported the hypothesis. When a small percentage of GDP was devoted to the military, democracy assistance was associated with a comparatively large increase in democracy as measured by Polity. As military spending increased, the marginal effect diminished and then became negative at around 3.5 percent of GDP devoted to military spending for democracy assistance and 4 percent for democracy assistance (%GDP). The interaction term was significant in both models at the 1 and 5 percent levels (see Figure 2 and Table 1, models 3 and 4).
All else being equal, the interaction between military size and democracy assistance was associated with a reduction in aggregate levels of democracy as military size increased. When the military was roughly 3.4 percent of the labor force, democracy aid was wasted or even had a negative effect on the overall level of democracy. The marginal effect of democracy assistance was zero or negative when military spending was a little over 3.4 or 4 percent of GDP. All of the models supported Hypothesis 1 that democracy assistance will work when military size is small. Three of the four models supported Hypothesis 2 that a large military interacted with democracy assistance can result in a backlash, at the 5 percent level (see Figures 1 and 2).
The second measure of democracy, Freedom House, was then regressed on the interactions of the independent variables (see Table 1, models 4–8). The results again provided support for the first hypothesis (Figures 3 and 4). Three of the models showed that democracy assistance had a positive association with democracy only when the military size was low. The interaction term was significant in two of the four models as well, providing support for Hypothesis 2 (see Table 2, models 5 and 8). Substantively, the effects are similar too. When military size either in terms of spending or labor force is a little over 4.4 percent, respectively, the marginal effect of democracy assistance is zero.
Marginal effect of democracy assistance on Freedom House as percentage labor force in military changes. Dashed lines represent 95 percent confidence intervals.
Marginal effect of democracy assistance (%gross domestic product [GDP]) on Freedom House as military spending changes. Dashed lines represent 95 percent confidence intervals.
Effect of Interaction between Democracy Assistance and Military Size on Democracy (Alternative Measures of Military Size).
Note: Cluster robust standard errors within parentheses. All control variables, country and year fixed effects, and lagged dependent variable included. FH = Freedom House; GDP = gross domestic product.
*p < .10. **p < .05. ***p < .01.
The interaction between democracy assistance and military size is more strongly related to Polity than Freedom House. Freedom House measures democracy using a very broad array of indicators, including the quality of the legal system and factors measuring economic freedom. Even when disaggregated, Freedom House’s political freedoms variable includes indicators concerning corruption and transparency. Polity measures more exclusively political factors. It may be that the smaller effects associated with Freedom House are due to this difference. If regimes are preventing opposition, organization, and contestation, they may adopt political strategies that show up more readily in Polity than the broader measure of Freedom House. 10
Robustness
Three alternative measures of military size: military spending (%Government expenditure), total personnel, and armed forces (%Pop) were used to test robustness. Military spending (%Government expenditure) was taken from the WDI. Total personnel was taken from the COW National Capabilities data set and measured the total number of military personnel. Calculating the marginal effects of the interactions, these models also showed that democracy assistance was most effective when military size was low. Nine of the twelve models showed support for the claim that democracy assistance was associated with increased levels of democracy when the military is small (see Table 3), while five showed that democracy assistance was associated with reduced democracy when the military size was large (see Table 2).
Effect of Interaction between Democracy Assistance and Military Size on Democracy (Random Effects).
Note: Standard errors within parentheses. All control variables, year dummies, and lagged dependent variable included. AIC = Akaike information criterion; FH = Freedom House; GDP = gross domestic product.
*p < .10. **p < .05. ***p < .01.
Fixed effects only account for within-unit variation. To test cross-national variation while still accounting for unobserved factors, random effects models, which provide a matrix-weighted average of between and within variation, were estimated using restricted maximum likelihood. 11 Despite Hausmann tests indicating that fixed effects are preferable, seven of the eight models show that democracy assistance has a significant effect only when military size is low. 12 Four of these models support the backlash hypothesis that military size could induce leaders to reduce democracy when combined with democracy assistance (see Table 3). 13
Discussion
The results show that democracy assistance does not have a uniform effect. Democracy assistance at lower levels of military size is correlated with higher levels of democracy, while as the military gets larger, the effects diminish and then become negative. Consistent with previous studies of democracy assistance, the results are small but still substantively significant. The finding has both negative and positive implications.
In some contexts, democracy assistance is not going to be effective and may even make things worse. Roughly 15 percent of democracy aid spending during the period from 1990 to 2004 was in states with militaries of sufficient percentage labor force in the military such that spending was ineffective or even detrimental to attempts to promote democracy. The findings are even more sobering when military spending is used as the measure of military size. Over 28 percent of democracy assistance spending was targeted at regimes spending 4 percent of GDP or more on the military. This represents over US$1.5 billion wasted in the constant US$2,000.
To put this in other terms, out of a possible 2,064 country-year observations in the data set, there were 681 observations where military spending was greater than 4 percent of GDP and 440 observations where more than 3.5 percent of the labor force was in the military. This amounts to, out of 141 countries, fifty-two states that at different times had spending above 4 percent and thirty-seven states that had a labor force with the military above 3.5 percent. In other words, a substantial number of the observed units had militaries of sufficient size that democracy assistance could potentially be detrimental or ineffective.
Despite this gloomy outlook for spending in states with large militaries, in states with smaller militaries, the effect of democracy assistance is larger than previously estimated. Indeed, the effect of spending for states with only 1 percent of their labor force in the military is twice as large that estimated by Finkel, Pérez-Liñán, and Seligson (2007). Properly targeted democracy assistance can be effective.
Due to the inclusion of fixed effects, which estimate within unit changes, there may be concern that the effects are due to military size proxying for a changing institutional environment within a country. However, this interpretation cannot explain insignificant coefficient for military size, which shows the effect of changes in military strength absent democracy assistance. If changes in military size were simply measuring concurrent changes in democracy, then this should be true absent democracy assistance. The theory presented here can explain both results; a stronger military increases repression when faced with increased domestic threat from democracy assistance but not otherwise.
Qualitative accounts have shown that regimes use a variety of different tools to undermine democracy promotion efforts. Venezuela used arbitrary tax audits of NGOs, while Zimbabwe outright banned some foreign NGOs and groups from operating within the country, and the Singaporean government used defamation cases to bankrupt opposition politicians (Gershman and Allen 2006). This is consistent with the observation that the interaction between military size and democracy assistance is statistically associated with a decline in both formal political competition and legal and civil rights. It seems regimes use a variety of different methods to limit the effectiveness of democracy assistance.
While the USAID has considered the role of the military in political development, they have mainly focused on the relationship between civilians and the military (Forhman and Welch 1998). The above mentioned analysis shows that the problem may be more fundamental. The issue is not ensuring that the military defers to civilians but that the military has no interest in maintaining the regime when it is threatened by pro-democracy forces. The issue is not one of civil–military relations as such, but rather the military’s incentives to preserve an authoritarian regime.
Conclusion
Political institutions play a central role in economic and social development. Democratic institutions reduce rent seeking and corruption by fostering political openness and political competition (North, Wallis, and Weingast 2009; Acemoglu and Robinson 2012). This in turn improves the prospects for economic development. So while there are direct benefits to promoting democracy and political freedom, there may also be benefits that extend to other realms such as economic growth and social development. The findings presented here demonstrate that democracy assistance has the potential to achieve these aims, but only in certain contexts.
Past studies of democracy assistance have found a positive association between democracy assistance and democracy. However, it is also clear that some states have adopted more authoritarian measures to limit the effectiveness of democracy assistance. This article has demonstrated that the effect of democracy assistance is conditional on the size of the military. Regimes that control large militaries respond to democracy assistance by increasing restrictions on political and civil society. Regimes that control small militaries liberalize when faced with democracy assistance.
This article proposed three mechanisms to explain the relationship. A stronger military could motivate the regime to repress. A large military could be a tool of the regime used against the domestic opposition. Powerful militaries could boost regimes’ bargaining power relative to donor states resulting in tamer democracy assistance. Distinguishing between these mechanisms should be the subject for future research. In addition, explaining and predicting when regimes resort to different strategies such as repression or legal restrictions is an avenue for exploration.
Recent studies of development and aid have increasingly focused on how development policies and aid work only conditionally. The findings of this article are consistent with this work. Democracy assistance is likely to be effective in some cases and ineffective or even counterproductive in others. States that wish to use democracy assistance to promote democracy need to carefully consider where they target their resources. Alternatively, they need to develop different forms of democracy assistance for different contexts. Assuming that effects will be constant across countries is a recipe for poorly allocated resources. As the USAID considers the future role of their governance and democracy programs, these issues should be taken into account.
Footnotes
Acknowledgments
I would like to thank Hendrik Spruyt, Jonathan Caverley, Susan Hyde, Andrew Little, Fredrik Sjoberg, Andrew Kelly, and Tsveta Petrova. I would also like to thank the two anonymous reviewers and the editors for their feedback and comments.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
