The Spuriousness Critiques

The first set of critiques is that the observed correlation between dyadic democracy and peace is spurious. More informally, the critique is that the observed peace between democracies is caused by factors other than democracy, and not by democracy itself. More rigorously, consonant with the definition of a law provided above, the observed peace between democracies would not support the counterfactual that taking a pair of democracies and rendering one of them non-democratic would make their relationship less peaceful. Whether the democratic peace is spurious or causal is not merely a semantic quibble. Scientists across the social and natural sciences maintain a deep interest in determining whether an observed correlation is causal or merely spurious. Identifying causation is critically important in translating scientific findings into policy recommendations, in areas such as dietary guidelines, poverty reduction, education, fighting disease, and others.

Scholarship making the claim that the democratic peace is spurious frequently takes the following form. On the theoretical side, an alternative explanation for the causes of peace is provided. On the empirical side, a critique will present a previously published multivariate regression analysis showing support for the dyadic democratic peace, and then show that adding to this regression analysis an additional independent variable that measures the new, alternative explanation will cause the dyadic democratic peace variable to become statistically insignificant. Adding the new variable is justified from a methodological point of view as a means of improving the model by reducing what is referred to as “omitted variable bias.” The critique then draws the inference that because the inclusion of this additional variable (or variables) renders the democracy variable statistically insignificant, the initial result was flawed because of omitted variable bias. In turn, the inference is that the initial observed correlation between democracy and peace is spurious rather than causal, and that as a causal hypothesis the democratic peace proposition is not supported. A further implication is that because the democracy-peace relationship is spurious rather than causal, policy-makers should avoid concluding that spreading democracy will in turn cause the world to be more peaceful.⁵

Scholars have made a number of arguments about the spuriousness of the democratic peace, that is, they have pointed to a series of different variables that if included in multivariate regressions render the democracy-peace correlation statistically insignificant. The oldest and perhaps most central proposition of this type is the realist argument that common national interests rather than joint democracy explain peace. As indicated above, realism proposes that international relations are fundamentally driven by national interests, and not by domestic politics or institutions. Further, realism places no faith in the ability of public opinion coupled with democratic institutions to be a force for peace, because public opinion is not necessarily rational or peaceful; and because elected and other leaders can circumvent the constraints of public opinion through secrecy and other forms of manipulation (e.g., Mearsheimer, 2011; Rosato, 2003; Schuessler, 2015). Historically, the collapse of the international order in the interwar period made realist critics such as E. H. Carr, Walter Lippmann, and Hans Morgenthau deeply skeptical of Wilson’s vision that the spread of democracy could support global peace. Waltz (1959; 1993, p. 78) from the 1950s through the 1990s was also critical of the Kantian hope that democracy would bolster peace, proposing that the brutally competitive nature of the anarchic international system forces different types of political regimes to adopt converging foreign policies in order to survive. Realists in turn proposed that any observed correlation between democracy and peace must be spurious, and in turn that the observed peace between democracies was caused by commonalities in interest and/or by a functioning balance of power rather than by regime type (see Layne, 1994; Mearsheimer, 2014; Rosato, 2003).

Several quantitative studies have endeavored to demonstrate that decisions for war and peace are caused by realist factors such as national interests and the balance of power, and not by regime type. In the 1990s, realist critics took note that the first wave of rigorous quantitative democratic peace studies focused on the 1950–1985 time period, suggesting that especially during this Cold War period democracies were unwilling to fight each other not because of institutions or norms, but because North American, East Asian, West European, and South Pacific democracies needed to balance together against a common Communist threat. A variant of this argument is that peace among democracies during the Cold War was maintained by American hegemony, that a democratic America managed conflict among states within the democratic, anti-Communist bloc to solidify its global power position.

These studies took different approaches to demonstrating this point. Gowa (1999) argued that the democratic peace was a temporal phenomenon; that pairs of democracies were indeed less likely to become involved in militarized interstate disputes or wars after 1945; were less likely than other pairs of states to become involved in MIDs but not wars from 1919–1938; and were as likely to become involved in wars and MIDs before World War I. That is, she measured the presence of common interest indirectly by comparing political eras, arguing that democracies shared common interests after 1945, confronting the Communist threat, and therefore unsurprisingly were less likely to fight each other. Before World War II, she argued, when there were fewer common interests among democracies, the observed correlation between democracy and peace disappears.

Gartzke (1998) took a more direct approach toward testing the same theoretical supposition. He also proposed that common interest rather than joint democracy was the true cause of the observed peace between democracies, especially in the post-1945 period. Rather than comparing eras as Gowa did, he analyzed the post-1945 period, but included in his regression analysis a variable of common interest, measuring how similar were the United Nations General Assembly voting patterns of two states. He found that this variable was statistically significantly related to dyadic peace, and that inclusion of this variable rendered the joint democracy variable statistically insignificant as an explanation of dyadic peace. Some observers have also suggested that the observed peace between democracies is caused by geographic factors rather than regime type (Worley, 2012).

An additional cut on the national interests argument is that conflicts are caused by interstate disputes over contested issues, like territory, and not by regime type. Gibler (2012) focused on territorial disagreements between states. He proposed that territorial disagreements are the fundamental cause of conflict between states, and that inclusion of variables that measure the stability of borders, and therefore the absence of territorial disagreement, rendered the joint democracy independent variable to be statistically insignificant as a cause of peace.

A second cluster of spuriousness critiques focuses on economic rather than political factors. A perhaps more limited version of this critique is that there is a peace among democracies, but only in the developed world and not in developing areas such as sub-Saharan Africa (Henderson, 2008). A more ambitious form of this critique is that development and markets are the true causes of peace, and that democracy is uncorrelated with peace when these factors are accounted for. There are some variants of this observation. Gartzke (2007) focused on higher levels of economic development, proposing that more developed states enjoy lower marginal gain from winning a war over economic assets, and in turn are less likely to become embroiled in war. Conceptually, there is a related strand of research that war has become obsolete as states’ economies have become more advanced and rely more on trade and the global market (Rosecrance, 1986). This point is also related to the more popular assessments of a “McDonald’s Peace,” the observation that countries with McDonald’s restaurants have never fought, McDonald’s being a sign of development (Friedman, 2000, ch. 21), or the “greens peace,” the observation that nations in which golf is sufficiently popular (again, a sign of development) never fight each other (Plotz, 2000). Mousseau (forthcoming; 2009) made a different argument, proposing that only some forms of economic development nurture peace. He proposed that market-based societies place a cultural emphasis on contracts and the law. In turn, this cultural emphasis on law percolates into foreign policy preferences, pushing such states to prefer nonviolent means of conflict resolution. Mousseau proposed that inclusion of a variable measuring this emphasis on contracts and law, what he termed to be “contractualism,” renders the joint democracy variable statistically insignificant.

A third critique focuses on gender. One of the central questions asked in the study of gender and politics is the relationship between gender and war, with many arguing that biological sex and/or cultural constructions of gender are critical factors affecting the onset of political violence. Further, some scholars have used gendered perspectives to critique the proposition that democracy causes peace. Allison (2001) used Kant’s framework of perpetual peace to propose that the key cause of international peace is a feminine perspective on interpersonal care rather than joint democracy. Wisotzki (2015) suggested that gender equality encourages both democracy and peace, though she stopped short of proposing that there was no causal relationship between democracy and peace. Hudson and colleagues (2012) used new data on the physical security of women and political violence, finding that lower physical security of women makes political violence more likely, and that the inclusion of gender equality in the analysis renders democracy an insignificant determinant of peace. Notably, Hudson and colleagues’ unit of analysis is the state rather than a pair of states, and their measure of violence incorporates many types of violence, including intrastate violence. One of Hudson’s coauthors in earlier studies of the causes of interstate violence found in multivariate analyses mixed evidence that both measures of gender inequality and democracy were statistically significant causes of peace (Caprioli, 2000; Caprioli & Boyer, 2001).

Another possible critique, not quite leveled explicitly by any critics, is that common culture and common identity, rather than democracy, cause peace. This is perhaps an implication of Huntington’s (1996) “Clash of Civilizations” thesis: that differences in civilization or culture rather than regime type determine conflict between states.

Discussion

To the great benefit of the broader field, these democratic peace critiques have enjoyed intensive scholarly debate, with both supporters and critics of the democratic peace successfully pushing each other to refine and improve their theoretical arguments and research designs. Regarding the critique that the democratic peace was purely a Cold War phenomenon, Russett and Oneal (2001), Maoz (1998), and Thompson and Tucker (1997) demonstrated that democracies were less likely to fight each other in the interwar and pre-World War I periods as well as in the post-1945 period. Russett (1993) also presented evidence of a democratic peace in ancient Greece and in pre-modern societies, and Park (2013) demonstrated that the democratic peace existed in the post-Cold War period, as well. Cederman (2001) took a different angle in addressing this question of the democratic peace being confined to the post-1945 time period. He agreed that the peaceful tendencies of democracies had strengthened over time, but he proposed that such a dynamic reflects the kinds of macrohistorical learning process that Kant himself predicted would happen.

Supporters of the democratic peace have also published analyses showing that inclusion of UN voting records does not render the democracy variable statistically insignificant (Oneal & Russett, 1999), these claims in turn attracting response (Gartzke, 2000). The proposal that capitalism rather than democracy causes peace has also attracted critiques, mostly focusing on issues of research design to show that inclusion of capitalism variables does not render democracy variables insignificant (Choi, 2011; Dafoe, 2011). Regarding the possibility of common culture or civilization rather than democracy causing peace, observational data and survey experimental studies have found that the inclusion of culture or civilization as an independent variable does not moot the effects of joint democracy (Bolks & Stoll, 2003; Henderson, 1998; Johns & Davies, 2012; Lacina & Lee, 2013). Regarding whether inclusion of factors such as trade or geography render the democratic peace to be insignificant, Russett and Oneal (2001) openly claim that both joint democracy and bilateral trade cause peace, as part of the “Kantian triangle.” They demonstrate that inclusion of trade and geography variables does not render the democratic peace relationship insignificant. The observation that contractualism and not democracy causes peace has been critiqued (Dafoe et al., 2013), and that critique has in turn attracted response (Mousseau, forthcoming).

The proposal that at least during the Cold War American hegemony rather than democracy itself fostered peace between democracies has also attracted scholarly debate. On a simple level, studies finding evidence of democratic peace routinely include dyadic alliance as an independent variable, including in the post-1945 period (Russett & Oneal, 2001). A broader question is whether or not the United States at least during the Cold War used its power to maintain both democracy and peace within its sphere of influence. In general, the United States supported a variety of anti-Communist states, including democracies like Japan and France and non-democracies like South Vietnam and South Korea (Reiter, 2001).

The critique that conflict is caused by territorial dispute rather than regime type has also experienced rigorous debate (e.g., Gibler, 2014; Park & Colaresi, 2014). Huth and Allee (2002) found that democracy played an important role in affecting whether or not territorial rivalries escalated to violence. Other studies looked at the related issue of rivalry between states, territorial dispute being one type of rivalry. Hensel and colleagues (2000; see also Rasler & Thompson, 2005) found that democratic dyads are less likely to experience rivalries; and that among rivals, the presence of democracy makes the onset of violent conflict less likely.

The gender critique has also enjoyed some scholarly exchange. The most aggressive gender-based critique of the democratic peace (Hudson et al., 2012) uses a monadic design, and a dependent variable inclusive of a wide array of forms of violence including intrastate and interstate conflict. As noted, some other work that focuses on interstate conflict has included gender as an independent variable, and still shown that democracy has a pacifying effect. Using a dyadic research design, Regan and Paskeviciuti (2003) found that both gender and joint democracy affect the likelihood of interstate violence.

In short, the variety of critiques arguing that inclusion of additional variables in multivariate regressions of observational data renders democracy variables to be insignificant have each enjoyed rigorous debate. A larger question to consider is whether there are other ways of testing causation beyond this approach of testing the possibility of spuriousness by adding variables to a regression. There is ongoing debate within political science and the social sciences more generally as to the general utility of whether the addition of more independent variables always makes a regression model “better” by generating a net reduction in bias (Achen, 2005; Clarke, 2005; Pearl, 2011). There is a broader debate as to the general utility of multivariate regression of observational data as a means of assessing causation, given that this approach, sometimes called a quasi-experiment, requires the nonrandom assignment of the treatment condition (the independent variable).

This is not to take a maximalist position that quasi-experiments add nothing, or that adding variables is never advised. It does suggest, however, considering other means of assessing causation, in addition to the conventional approach of seeing if adding plausible exogenous variables renders the democracy-peace correlation to be statistically insignificant. Scholars have explored other means of assessing causation in the democratic peace, and have amassed three other types of evidence that support the conclusion that democracy causes peace: evidence demonstrating support for other empirical patterns suggested by democratic peace theory; evidence produced using experimental methods; and evidence produced using case studies.

The first type of evidence explores for the existence of other empirical patterns predicted by democratic peace theory. If a theory predicts the existence of a variety of empirical patterns and these patterns are demonstrated through tests, we can be more confident in the validity of the theory, and in turn that observed correlations are causal and not just spurious. And, indeed, there is a wide array of quantitative empirical studies that provide support for various assumptions or implications of democratic peace theory, especially for institutionalist accounts of the democratic peace. Perhaps the central institutionalist explanation of the democratic peace proposes that elected leaders are motivated to avoid fighting wars, because the costs of wars will incite popular discontent in turn threatening their hold on power. Studies have demonstrated a number of empirical patterns consistent with this view. Democracies fight shorter wars (Reiter & Stam, 2002, ch. 7). Democracies suffer fewer casualties when they fight wars (Valentino et al., 2010), and when they fight, popular support for the leadership declines as casualties escalate (Mueller, 1973). The benefits of victorious wars may sometimes push democratic publics to accept the costs of war when they are confident of victory, and accordingly democracies almost never start wars they go on to lose (Reiter & Stam, 2002). During war, public support erodes as the perceived likelihood of victory declines (Gelpi et al., 2009). As the institutional explanation of the democratic peace would predict, variations of institutional and leadership form within democracies also affects conflict behavior, as in general more constrained states are less conflict prone (Reiter & Tillman, 2002). Consistent with the audience costs explanation, democracies can more effectively signal their resolve than at least some kinds of autocratic states (Schultz, 2001; Weeks, 2014). There are also some studies supporting elements of the normative explanation. For example, some studies have found that democracies are especially likely to use mediation or binding arbitration to resolve interstate disputes (Dixon, 1993; Raymond, 1994, 1996). In total, though there are certainly scholarly debates about some of these observed patterns,⁶ this collection of studies improves our confidence that democracy is causing peace in the manners described by democratic peace theory.

The second type of evidence uses experimental methods. Some have proposed that experimental methods enjoy critical advantages over the analysis of observational data in assessing causation. Experimental methods are able largely to skirt some of the biggest causal inference problems associated with quasi-experimental methods, such as biased samples and nonrandom assignments of treatment. That said, the limitation of experimental methods is that, especially in international relations, they can only be used to test some arguments, or some components of arguments. For example, regarding the democratic peace an experimenter cannot take a set of states and then randomly assign some to be democratic and others to be non-democratic.

That said, scholars have thus far been able to conduct survey and laboratory experiments that have tested some elements of the democratic peace. A number of surveys have found support for one of the core assertions of dyadic democratic peace theory: that citizens of democracies are significantly less likely to support the use of force against democracies as compared to using force against non-democracies (Geva et al., 1993; Johns & Davies 2012; Lacina & Lee, 2013; Mintz & Nehemia, 1993; Rousseau, 2005, pp. 219–232; Tomz & Weeks, 2013) Other experiments have tested elements of the audience costs variant of the democratic peace, showing that the public does inflict audience costs on leaders who back down in a crisis (Horowitz & Levendusky, 2012; Tomz, 2007; Trager & Vavreck, 2011).

A third empirical means of demonstrating causation is to engage in process tracing through case studies. Scholars have presented several individual case studies of the democratic peace in events such as 19th-century American diplomatic crises, the 1898 Fashoda Crisis, the onset of World War II, the Spanish-American War, and many others (see Elman, 1997; Owen, 1997; Ray, 1995; Risse-Kappen, 1995; Rousseau, 2005; Schultz, 2001; for case studies presenting evidence against the democratic peace, see Layne, 1994). Some of these case studies demonstrate specific parts of the causal logic of the democratic peace, such as the ability of democracies to signal more effectively through invoking greater audience costs (Schultz, 2001), or the inability of elected leaders to manipulate public opinion or secretly drag their nations into wars the public would otherwise avoid (Reiter, 2012a). Perhaps the most striking case study of democratic peace dynamics is the pacification of Western Europe after World War II, democracy helping to dissolve immediately and completely one of the most violent interstate conflicts in modern history, the France-Germany rivalry (Russett & Oneal, 2001).

The Causal Arrow Critique

A different cluster of arguments critiquing the claim that democracy causes peace focuses on the direction of the causal arrow, proposing that peace causes democracy, but that democracy does not cause peace. The central claim of the “peace causes democracy” element of this claim is that threatening international environments motivate governments to improve their abilities to balance against these threats, and that part of that response is to expand the power of the state and to weaken democratic political institutions. Lasswell (1941) described how a sense of international threat can encourage the emergence of the “garrison state,” a government that empowers military leaders, emphasizes the collective over the individual, and prioritizes coercion over bargaining, as part of a process that eventually destroys democracy. Sometimes states will take internal actions to improve their ability to confront these threats, and these internal actions can in turn bolster autocracy and undermine democracy. One course of action is to strengthen the state itself, recalling Tilly’s (1975, p. 42) claim that “war made the state, and the state made war.” Greater external threat may require a government in the short term to increase defense spending (Nordhaus et al., 2012) and revenue collection and total spending more broadly (Lektzian & Prins, 2008), expanding its control of the national economy.

A related point is that higher levels of external threat push states to reduce individual liberties, including political competition. Poe and Tate (1994; see also Davenport & Armstrong, 2004) demonstrated that participation in interstate wars was negatively correlated with a measure of the percentage of the adult population who vote and with respect for human rights. Conversely, autocracies may be more willing to undergo democratization in a peaceful international context as compared with a more threatening environment (Thompson, 1996).

Importantly, there are two variants of the peace causes democracy argument. The more moderate form proposes that the causal arrow runs both ways, that democracy causes peace and peace causes democracy (Crescenzi & Enterline 1999; Gleditsch, 2002; Midlarsky, 1995; Rasler & Thompson, 2005; Reuveny & Li, 2003; Thompson, 1996). Russett and Oneal (2001) framed this as part of a Kantian positive feedback loop in which peace, economic interdependence, democracy, and international organization all mutually and positively reinforce each other.

The more ambitious form of the argument contends that peace causes democracy but not the reverse. Authors making this claim present different types of quantitative tests to show support. James and colleagues (1999) is perhaps the most sophisticated, using a multiple equations model to demonstrate that peace causes democracy, but that the effect of democracy on peace is statistically significant but substantively negligible. Gibler (2012; Gibler & Tir, 2014) coupled his critique that border stability and not democracy causes peace with the observation that border stability in turn causes democracy. Mitchell and colleagues (1999) presented an interesting variant, albeit using a system level of analysis. Using Kalman filter analysis, they found that though democracy spreads peace, war itself can cause democracy, because democracies often win wars and victors in war pursue regime change.

Discussion

As with the spuriousness critiques, the claim that peace causes democracy but that democracy does not cause peace has attracted scholarly debate. Regarding the garrison state thesis, some have proposed that levels of threat may cause states to grow stronger, but that need not come at the expense of undermining democracy (Friedberg, 2000; see also Zakaria, 1998). The James and colleagues (1999) article described above attracted scholarly exchanges (Oneal & Russett, 2000; James et al., 2000), though in a later article James himself (Enia & James, 2015) moved to the more moderate position of bidirectional causality, that peace and democracy both cause each other. Some scholars have been skeptical of the proposition that democracy causes peace, presenting quantitative evidence that peaceful international environments do not make democratic transitions more likely, or strengthen democratic regimes (Reiter, 2001). Pevehouse (2005) found only mixed evidence that regional conflict impeded democratization, and little evidence that regional conflict impeded democratic consolidation. Mousseau and Shi (1999) used interrupted time series models to make the point that peace did not cause democracy but democracy does cause peace. Kadera and colleagues (2003; see also Gleditsch, 2002) critiqued the view that the level of conflict in the system affects a democracy’s ability to survive, finding instead that democratic survival is more strongly affected by the number of other democracies in the system. Regarding the thesis that democracies spread democracy after winning wars, more recent analysis paints a different picture, finding that foreign imposed regime change often fails to implant either stability or democracy (Bueno de Mesquita & Downs, 2006; Peic & Reiter, 2011).

There are important methodological dimensions of the debate around the causal arrow critique, mostly concerning the difficulties of constructing a research design that would definitively conclude that peace caused democracy but democracy did not cause peace. One issue concerns the level of analysis. The dyadic democratic peace focuses on the pair of states as its unit of analysis, whereas the peace causing democracy thesis focuses generally on the single state, that a peaceful international environment will affect a single state’s likelihood of becoming democratic. The difference in level of analysis presents difficulties in constructing an integrated estimation strategy that could test satisfactorily both causal arrows (Reuveny & Li, 2003, perhaps come closest to tackling this thorny problem).

Methodological issues aside, there is some theoretical inconsistency within the claim that peace causes democracy but not the reverse. The spuriousness critiques make theoretically straightforward claims: Peace is caused by factors other than regime type, such as national interests, and therefore democracy does not cause peace. In contrast, the causal arrow critique is less straightforward. It proposes that a threatening international environment undermines democracy because leaders perceive that democracy would impede a state’s ability to fight war, because a weak state would be insufficiently agile or powerful to mobilize quickly for war, and/or because individuals empowered by democratic political institutions might resist mobilization and/or a decision for war.⁷ And yet if these assumptions are true, then they are in turn of course reasons why democracies are more peaceful in their international relations, the heart of the democratic peace thesis. That is, if states or leaders shun democracy because they fear democratic systems are better suited for peaceful rather than violent international environments, then the implication is that democracies ought to be more peaceful in their foreign policies. Peace would cause democracy because democracy causes peace.