Forward to the future? The democratic peace after the Cold War

Abstract

This paper helps resolve the ongoing debate concerning whether the democratic peace is limited to the Cold War period. Some critics have attributed the democratic peace to a set of common interests among democracies that uniquely existed during the Cold War. This study is the first direct test of this proposition. I use a new measure of Cold War preferences to test if the effect of joint democracy is rendered statistically insignificant either during or after the Cold War as critics contend. I also test, as some democratic peace proponents have suggested, whether the pacifying effect of democracy is strengthened in the post-Cold War era. Evidence from period-specific logit analyses suggests that joint democracy promotes peace independently of Cold War security interests. Indeed, the democratic peace exists in the post-Cold War era.

Keywords

Causes of peace Cold War common interests democratic peace post-Cold War

Introduction

Is there evidence of the democratic peace in the post-Cold War world? Several critics, most notably Gowa (2011), suggest that the peace between democracies was merely a spurious correlation arising from temporary shared interests. Exploring this puzzle is important to the current world, where more democracies are observed than ever before. Although the democratic peace is often referred to as a law-like regularity (Levy 1988), whether the post-Cold War world remains more peaceful for the democratic community is unclear.

Extant empirical research is mixed. Gowa (1999) provides evidence that the democratic peace only holds for the Cold War era.¹ She also finds that democracies tended to ally with each other during the Cold War. These findings, however, do not confirm that the democratic peace is an artifact of Cold War interests, as Gowa has not provided a direct measure of security interests. To convincingly test the claim of democratic spuriousness, alliance may be a proxy, as Gowa (1999) suggests, but this measure is found to be sensitive to research design and specification choices (e.g. Bennett and Stam, 2000b).

For the 1950–1985 period, Russett and Oneal (2001: Chap. 6) test their Kantian peace thesis against Gartzke’s (1998) affinity variable, a broader measure of common interests. They find that affinity does not nullify the effect of democracy, while democracy significantly influences affinity. Since that study, data have become available for a more extended time period.² Over the 1885–2001 period, Oneal and Russett (2005) report the significance of joint democracy, even controlling for alliance. These two studies, however, do not offer direct counter-evidence against Gowa’s security interests argument for two reasons. First, alliance and affinity are not measures specific to Cold War interests. Second, the 1885–2001 data are pooled. As Russett and Oneal (2001: 230) acknowledge, the democratic peace as “a product of Cold War tensions cannot be rejected without analyzing international relations in other [particular] historical periods”, especially the post-Cold War era.

Most recently, Gowa (2011) revisits this issue with extended data. She finds that the dispute rate of democratic dyads is indistinguishable from that of mixed and non-democratic dyads between 1992 and 2001. If Gowa’s new analysis is sound, the near law-like status of the democratic peace needs to be amended and contextualized. However, several flaws in Gowa’s (2011) design create doubts about the validity of the reported results.³ The most important of these include: (1) omission of a direct measure for common interests as in her previous work; (2) failure to differentiate mature democracies from fledgling democracies prevalent in the post-Cold War era; and (3) ignoring of the possible lagged effect of democracy on conflict (Oneal et al., 2003; Russett and Oneal, 2001).

In this study, I take Gowa’s finding seriously to probe whether the democratic peace was a by-product of the Cold War and thus has disappeared since 1989. I use a research design similar to Gowa (2011) but provide a more direct measure of Cold War interests. A set of period-specific logit analyses show that democratic dyads are more peaceful than other combinations of regime types, even for the post-Cold War period.

Democracy vs common interests

Broadly speaking, proponents explain the democratic peace based on norms and institutions. First, democratic norms teach politicians how to compete for power without violence and place a greater emphasis on the rule of law (Simmons, 1999). The democratic process socializes leaders to resolve conflict by peaceful means such as mediation, negotiation and compromise at the domestic level (Dixon, 1994; Raymond, 1994). Second, when a conflicting interest arises between two democracies, democratic institutions allow time for the disputants to find peaceful solutions (Maoz and Russett, 1993) and provide them with the ability to communicate clearly (Fearon, 1994; Schultz, 2001).

Gowa (1995, 1999, 2011; see also Farber and Gowa, 1995, 1997) has criticized both normative and institutional explanations as unpersuasive. She shares concerns with two outstanding strands of critiques.⁴ With realist critics, she emphasizes system-level polarity and alliance patterns as the source of conflict dynamics between states (Layne, 1994; Mearsheimer, 1990; Rosato, 2003; Spiro, 1994). Like those who focus on state preferences as drivers of foreign policy (e.g. Gartzke, 1998, 2007; Werner, 2000; see also Signorino and Ritter, 1999), she associates the democratic peace with common interests. Gowa has argued that the East–West divide induced strong common or conflicting security interests among members of the international system according to which bloc they belonged.⁵ For Gowa, the prime factor explaining the observed peace among democracies is not democracy itself but security interests induced by the bipolar system. Most recently, Gowa (2011) finds that the democratic peace is limited to the Cold War era and does not extend to the post-Cold War era, corroborating her previous claim and finding. According to some researchers (e.g. Gelpi and Griesdorf, 2001; Tomz and Weeks, 2011), Gowa has provided a more plausible challenge than other critics over the past 18 years.⁶

However, Gowa’s (2011) evidence has three important shortcomings. First, as in previous research (Gowa, 1999), the new analysis lacks a direct measure for Cold War interests to test the hypothesized relationship. Without such a measure, the null finding from the decade-long data from the post-Cold War era cannot confirm the claim that the democratic peace is a Cold War phenomenon. As noted in Gowa (2011: 158), “There is no guarantee, of course, that the years between 1992 and 2001 will be typical of those that follow”. If the democratic peace ought to be considered a Cold War artifact, it would be better to consider states’ alliance behaviors in regard to the Cold War superpowers (i.e. the United States and the USSR) as a measure for common or conflicting interests of the Cold War. This is one of the main tasks of this study.

The second shortcoming in Gowa’s (2011) design is its failure to take into account the significant inflow of fledgling democracies to the post-Cold War system. It measures democracy with an arbitrary threshold (i.e. Polity score = 7). The use of a categorical measure may not be problematic per se, and some even regard it as an improvement (Alvarez et al., 1996). However, the conventional threshold, like a Polity score of 6 or 7, when used for the post-Cold War era, obscures the difference in the pacifying ability between dyads of well-established democracies and those including fledgling ones. This difference is theoretically expected and empirically observed. According to the logic of Kantian learning (Cederman, 2001; Kant, 1957[1795], 1970[1784]), newly independent democracies are not immediately distinguishable from their old regimes and other entrenched autocracies in foreign policy interactions owing to the lack of time for democratic learning and the possible instabilities associated with the regime transition. There is a scholarly consensus that regime transition tends to produce external conflicts, although it is unclear whether this trend is more apparent for democratization or autocratization (see Braumoeller, 2004; Enterline, 1998; Mansfield and Snyder, 1995; Narang and Nelson, 2009; Ward and Gleditsch, 1998). Many new democracies appeared soon after 1989, and the Polity project assigns 6, 7 or 8 to most of these new regimes. It is hard to imagine that the improved polity scores of these new democracies (e.g. Haiti’s polity score 7 in 1990 and Zambia’s polity score 6 in 1991) instantly pacify interstate relations. Additionally, former communist countries and Soviet republics accounted for the majority of new democracies around 1989, according to the Polity 4D data.⁷

Relatedly, the third problem is the lack of any lag structure in Gowa’s (2011) design. This is problematic for two reasons. First, the possible endogeneity between conflict and the independent variables may create biased inferences. For instance, militarized conflict can affect relative power, alliance status, level of democracy and the like. That is why many statistical studies lag independent variables of interests, usually by one year, although this is not a perfect solution (e.g. Dafoe, 2011; Gartzke, 2007; Russett and Oneal, 2001).⁸ Second, and more importantly to the post-Cold War analysis, there are few (if any) reasons to expect newly democratized countries to have a pacifying effect right after their transition. Therefore, putting the contemporaneous dummy term for joint democracy in the post-Cold War equation, as Gowa (2011) does, may not capture the underlying effect of democracy on interstate conflict. By a similar token, a thorough investigation by Oneal et al. (2003) into the dynamic structure of the statistical conflict model shows that past values of the Kantian factors, including democracy, are more useful predictors of military conflict as compared with contemporaneous measures.

However, each of these three shortcomings can be improved upon. As Gartzke (2007: 173, fn. 36) notes and Gowa (2011: 156) acknowledges, given the increase in number and heterogeneity of democracies, the post-Cold War era is “the most obvious place” that provides “the hardest test” of the democratic peace. If the post-Cold War data, with a corrected design that addresses the shortcomings in Gowa (2011), reveal the continued effect of joint democracy, it should provide strong support for the democratic peace.

Hypotheses: The democratic peace beyond the Cold War

In some sense, the post-Cold War era may be a fairer period to test democratic pacifism against the Cold War effect for both skeptics and proponents. The end of the Cold War came with a remarkable democratization trend over the globe. The increased proportion of democracies in the post-Cold War era ensures that democracies have democratic neighbors, reducing the concern that the rarity of inter-democratic conflict is a function of the sparse dispersion of democracies during most of the 19th and 20th centuries (Mearsheimer, 1990; Spiro, 1994). For proponents, the increased proportion means a more democratic world system. Enough democracies are required for their peace-inducing quality to be fully effective (Cederman, 2001; Kant, 1957 [1795]). Taken together, I test the following hypothesis:

Hypothesis. Democratic dyads will be less conflict-prone than other types of dyads during and after the Cold War.

In what follows, I discuss different expectations about the above hypothesis between common-interests critics and democratic-peace proponents.

Critics

It is plausible that the Cold War generated unique security interest patterns among countries. From Gowa’s (2011) perspective, it is the East–West rivalry that has produced a large number of disputes of democracies against non-democracies and/or non-democracies against non-democracies. In fact, the Cold War was a long-lasting conflict between the Western and Eastern blocs. At the pinnacle of each bloc there were two superpowers, the United States and the USSR. Whether countries had conflicting or shared security preferences with each other largely depended on which bloc they separately or jointly belonged to. During the Cold War, most of the democracies belonged to the Western Bloc, and few (if any) were democracies in the Eastern Bloc. This suggests that controlling for Cold War interests renders democracy insignificant for the Cold War era and that peace will no longer be uniquely observed for inter-democratic relationships in the post-Cold War era.

Some realist critics have provided similar expectations. For instance, Mearsheimer (1990: 11, 13) argues that systemic factors like bipolarity rather than democracy produced the unprecedented stability and peace in the entirety of Europe, a region central to the international system. He predicts that the demise of the Cold War bipolar order will increase the chance of wars and major crises in Europe. Recently, Rosato (2003) asserts that the democratic peace is an imperial peace based on American power placing an overriding emphasis on peace in Western Europe and the Americas. Certainly, the presence of the Soviet threat accounts for much of that emphasis. Now, the United States may not have the particularistic incentives to control these regions to the same extent as it did in the Cold War era.⁹

Proponents

It is important to note that proponents of the democratic peace do not dispute the swaying influence of the East–West conflict in the Cold War system. Certainly, many regional disputes were driven by the superpower competition to protect and maintain interests in their preferred regions. However, proponents have their own reasons for believing that democracy is a significant pacifier independent of the systemic imperative. As mentioned, democratic norms and institutions have been proposed as those that underpin the dyadic pacifism (Russett and Oneal, 2001). Therefore, democracy should remain significant in the post-Cold War equation and in the Cold War equation regardless of the Cold War control.

One difficulty with this expectation is that the effect of democracy is found to be inconsistent in the pre-World War I period (Gowa, 1999; Maoz and Abdolali, 1989). Some studies may account for this time inconsistency. For example, Cederman (2001) finds that the conflict propensities of democratic dyads have steadily fallen from the late 1830s to the early 1990s and that this pacifying effect is strengthened as their democracy matures.¹⁰

This finding is consistent with Kant’s vision. In reinterpreting the democratic peace, Cederman highlights Kant’s emphasis on human learning processes in liberal democracies. Kant reasoned that attaining perpetual peace is a progressive learning process. Over time citizens will learn that war is disastrous and immoral. A monadic learning process, however, is not sufficient for the perpetual peace if there are many autocratic neighbors. There must be a gradual diffusion of the democratic learning process from a powerful, enlightened state to others, thus building a peaceful association of states.¹¹ Progress through learning is a dynamic and dialectical process interrupted by several setbacks. In the long run, it is reasonable to observe a time-enhancing pacific effect as existing democracies mature and more democracies appear over time. In the short run, however, it may not be surprising to observe an insignificant effect of democracy in the earlier periods or within several years right after an important system change. Therefore, if the post-Cold War data over a decade reveal that democracy continues to pacify state interactions, it should attest to the significance of the democratic peace.

Research design

I employ the standard conflict model in the literature used by both critics and proponents (e.g. Choi, 2011; Dafoe, 2011; Gartzke, 2007; Gowa, 2011; Oneal and Russett, 2005). They commonly use the onset of a militarized interstate dispute (MID) of any severity as the dependent variable, take dyad years as the units of analysis, and measure democracy based on the Polity index of the Polity IV project. When there are differences in design between skeptics and proponents, I try to maintain a fair balance by adopting the established practices.

My sample covers all dyad years from 1816 to 2001.¹² I define the Cold War as spanning from 1950 to 1989 and the post-Cold War from 1990 to 2001. This way of disaggregating the data is in line with Gowa’s (2011) method. However, the decision for the end year of the Cold War is slightly different. While Gowa (2011) chooses 1991, I regard 1989 as the end year.¹³ The Cold War came to an end with a set of dramatic events in 1989. For example, its demise was manifest with the completion of Soviet troop withdrawal from Afghanistan, the fall of the Berlin Wall, and the collapse of the Communist Party dictatorship in Eastern Europe. Also, in 1989, Gorbachev publically and frequently declared to the Western world that “the postwar period is over” (Associated Press, 15 July 1989). Furthermore, the USSR and the United States in 1989 signed an agreement to avoid or limit accidental military encounters growing into serious confrontations (Schmemann, 1989).¹⁴ A year later, the two Cold War rivals became partners in the coalition forces in the Gulf War against Iraq.

I use EUGene (version 3.2) to generate data for both dependent and independent variables (Bennett and Stam, 2000a). For the dependent variable, I use the MID data (version 3.1) of the Correlates of War (COW) project. My dependent variable is the onset of an MID, defined as the threat, display and/or actual use of armed force by one state against another. If there is an MID onset between two countries in a given year, it is coded 1, otherwise 0. Following Bennett and Stam’s (2000a) advice, I drop all ongoing years and joiner dyad years unless a new, direct MID occurs. Below, I discuss how I address the three problems, previously identified in Gowa’s (2011) research design.

Measuring common interests

Gowa (2011) does not directly control for security interests, although she has suggested that formal alliance is indicative of common interests (Gowa, 1999, 2011). I utilize the formal alliance data (version 3.0) of the COW project to construct a measure of Cold War bipolarity. Yet, I do not simply employ the usual alliance variable that codes 1 for any dyads sharing an alliance (i.e. defense pact, neutrality or entente) and 0 otherwise.¹⁵ Concerning the Cold War effects, we should take into account the contrasting pattern of alliance-making behaviors between two countries in a given dyad. It is not necessary that allies fight each other less frequently than do non-allies. Rather, two countries with a high chance of military fighting sometimes need to jointly sign ententes, mutual neutrality or defense pacts to prevent a possible full-scale clash. In addition, even hostile rivals can be allied with one another when a third party with increasing military capabilities begins to pose a serious threat to the security of both countries.

Non-alliance between two states does not indicate the presence of conflicting security interests. We need to differentiate conflictual unallied dyads from usual unallied dyads and cooperative allied dyads from less cooperative dyads. Given that the term “Cold War” refers to the competition between the Western and Eastern blocs, I employ a categorical variable that dissects the kinds of allies and non-allies in terms of the superpower rivalry within the Cold War. The first category, “Cold War Rival”, captures the presence of conflicting interests between two unallied countries in a dyad.¹⁶ The coding process is as follows. I first look at whether two countries are unallied and then whether they belong to different blocs through a formal alliance with either of the Cold War superpowers, the United States and the USSR, respectively. Cold War Rival is coded 1 if state A is allied with one superpower while state B is with the other superpower, and 0 otherwise. A zero indicates the absence of conflicting Cold War interests within the dyad. More specifically, if one country in a dyad was allied with the United States but not with the USSR, while the other was allied with the USSR but not with the United States, this dyad constitutes a Cold War rival. Examples include such dyads as the UK vs East Germany (1956–1989), Spain vs Iraq (1972–1989), and Denmark vs Poland (1949–1989). When an unallied dyad includes one of the superpowers, I regard it as a Cold War rival if the non-superpower partner belonged to the other bloc. For example, Cold War Rival is coded 1 for the United States–North Korea dyad and the USSR–South Korea dyad, since North Korea was allied not with United States but with the USSR, and South Korea, not with the USSR, but with the United States. In the Cold War sample, Cold War rivals constitute less than 7.5% of all dyad years but account for more than 15% of all MIDs.

As for allied dyads, I divide them into three categories. First, some realist critics point out that most democracies were allies with each other as part of the US superpower alliance structure (e.g. Mearsheimer, 1990; Rosato, 2003). This suggests a category that indicates whether two allies shared the United States as an ally. I call this category “US-Bloc Ally”. I code 1 for these dyads and 0 otherwise. Second, allied countries that shared the Soviet as an ally are likely to share similar security interests (Mearsheimer, 1990). I call this dummy variable “Soviet-Bloc Ally”.¹⁷ Alliances observed in any dyads including the United States are also coded as a US-Bloc Ally, and the same is applied to dyads including the Soviet Union.¹⁸ Lastly, I code 1 for purely dyadic allies that were not under one of the superpower alliance structures and 0 otherwise. I call this category “Dyadic Ally”. Taken together, the results for all these categories are interpreted in relation to the excluded reference category of unallied dyads with no Cold War rivalry.

Measuring democracy

I measure the joint level of democracy using the Polity IV data set. The Polity2 score is used to facilitate the time series analyses of this study. This score ranges from 10 (full democracy) to −10 (full autocracy). I consider two measures. First, following Gowa (2011), who uses a Polity score of 7 as the cutoff for democracy, I trichotomize the type of dyads: democratic (democracy–democracy), mixed (democracy–non-democracy), and non-democratic dyads (non-democracy–non- democracy). The first two categories will be included in the equation against the reference category (i.e. non-democratic dyads). I call these two “Democratic Dyad” and “Mixed Dyad”, respectively. Second, based on the weak link assumption (Russett and Oneal, 2001), I take the score of the less democratic state in every dyad as a continuous dyadic measure for joint democracy. I call this variable “Lower Democracy”. Following Oneal and Russett (2005), I also include “Higher Democracy” to account for the “cat-and-dog” type relationship between democracies and autocracies.

Each measure has its own pros and cons, but the continuous index is preferable to the trichotomous measure for the following reasons. First, it would be fairer for a balanced analysis to use the proponents’ measurement than the critics’ unless there are critical disadvantages. More importantly, while setting an arbitrary cutoff point is unavoidable for a nominal measure of joint democracy (Jaggers and Gurr, 1995), this arbitrariness may lead to critically different results for the post-Cold War equation given the prevalent influx of new democracies into the interstate system, as already discussed. Although it cannot settle all difficulties regarding new democracies, the continuous measure at least partly recognizes the difference between “fledgling” democracies that might score a 7 on the Polity scale and “fully-fledged” ones that might score a 9 or 10. However, as mentioned above, I do not simply throw away the trichotomous measure. Opting for a higher threshold like 8, rather than 7, would help differentiate well-established democracies from new democracies in the post-Cold War period.

Lag structure

As discussed, Gowa’s (2011) research design lacks any lag structure that helps account for the endogeneity problem and the underlying lagged effect of democracy. For this reason, I incorporate a lag structure in relevant statistical models. As is the standard (Dafoe, 2011; Gartzke, 2007; Russett and Oneal, 2002), I lag all independent variables by one year in relevant models.

Control variables

I employ a set of control variables to account for the effects of relative power, geographic proximity, power status and the duration of peace. The operationalization of these concepts follows the best available practices. Using the Composite Index of National Capabilities (version 3.02) of the COW project, I measure relative power as the natural logarithm of the higher/lower ratio between states A and B. I label this variable as “Capability Ratio”. For geographic proximity, I consider two related but different factors, contiguity and distance between two states as in Oneal and Russett (2005) and Gowa (2011). The first measure is “Noncontiguity”, which equals 1 unless two states share a land boundary or are separated by less than 150 miles of water. I also employ the natural logarithm of the great circle distance in miles between the capitals of the two countries. I label this variable as “Distance”.¹⁹ The propensity for distance to constrain militarized options may be lessened for powerful countries with the land, sea or air capability to exercise military power over a long distance. Instead of the popular COW major power designation subject to a retrospective bias (Hegre et al., 2010), I use the natural logarithm of the larger CINC score in a given dyad. I label this variable as “Higher Power”.²⁰ Like Gowa (2011), I use the cubic spline peace year correction with three knots at 1, 4 and 7 years in order to test and solve for deterministic temporal dependence, as suggested by Beck et al. (1998). Also, I use robust standard errors to account for clustering on dyad.

Analyses and results

In order to compare the effects of joint democracy across the two different eras, I run a set of period-specific logit models for the Cold War (1950–1989) period and the post-Cold War periods (1990–2001).²¹ I also estimate a model for the entire period (1816–2001).²²

The Cold War Era (1950–1989)

Does Cold War bipolarity explain the democratic peace? In Table 1, the inclusion of the Cold War control does not obviate the significance of joint democracy. Joint democracy lessens the likelihood of MID onset as the coefficient for both Democratic Dyad (Models 1 and 2) and Lower Democracy (Models 3 and 4) is negative and statistically significant. Both Mixed Dyad (Models 1 and 2) and Higher Democracy (Models 3 and 4) have a significant positive coefficient, suggesting that democratic dyads, followed by non-democratic dyads, are the most peaceful while mixed dyads are at greater risk of hostile interstate relations.

Table 1.

The Cold War equation (1950–1989): the effect of joint democracy on MID onset.

	Trichotomy e.g. Gowa (2011)		Weak link e.g. Oneal and Russett (2005)
	Model 1	Model 2	Model 3	Model 4
Democratic Dyad	−0.5239* (0.2497)	−0.6542** (0.2579)
Mixed Dyad	0.6310*** (0.1235)	0.4971*** (0.1314)
Higher Democracy			0.0526*** (0.0088)	0.0426*** (0.0096)
Lower Democracy			−0.0581*** (0.0119)	−0.0554*** (0.0121)
Cold War Rival		0.3068 (0.2252)		0.2217 (0.2290)
US-Bloc Ally		0.0090 (0.2536)		−0.0291 (0.2654)
Soviet-Bloc Ally		−1.8509*** (0.3554)		−1.7843*** (0.3573)
Dyadic Ally		−0.2098 (0.1818)		−0.1802 (0.1861)
Capability Ratio	−0.2994*** (0.0463)	−0.2854*** (0.0499)	−0.2930*** (0.0461)	−0.2821*** (0.0479)
Noncontiguity	−2.5732*** (0.1815)	−2.6376*** (0.1822)	−2.6208*** (0.1796)	−2.6706*** (0.1816)
Distance	−0.4930*** (0.0643)	−0.5039*** (0.0701)	−0.4826*** (0.0628)	−0.4952*** (0.0682)
Higher Power	15.5255*** (1.1854)	15.9291*** (1.2711)	15.4194*** (1.1871)	15.9235*** (1.2780)
Peace Years	−0.6002*** (0.0994)	−0.5936*** (0.1009)	−0.6047*** (0.1033)	−0.5977*** (0.1041)
Spline 1	0.0022 (0.0189)	0.0002 (0.0193)	0.0011 (0.0194)	0.0003 (0.0198)
Spline 2	−0.0169** (0.0071)	−0.0157* (0.0072)	−0.0164* (0.0072)	−0.0156* (0.0073)
Spline 3	0.0093*** (0.0014)	0.0089*** (0.0015)	0.0092*** (0.0015)	0.0089*** (0.0015)
Constant	2.1134*** (0.4377)	2.2355*** (0.4988)	1.8210*** (0.4240)	1.9377*** (0.4928)
Pseudo R²	0.3909	0.3956	0.3916	0.3955
N	290,740	290,740	290,740	290,740

Notes: significance levels are indicated as follows: * p <0.05; ** p < 0.01; *** p <0.001 (one-tailed). Robust standard errors clustered on dyad are in parentheses.

Security interests, as measured through Cold War alliances, do not have a robust influence on conflict propensities, although interesting patterns do emerge. The positive sign for Cold War Rival suggests a higher conflict rate when two states belong to different blocs. It is marginally significant at the 0.1 level (one-tailed) in Model 2 and insignificant in Model 4. Common interests within the same block appear to reduce conflict only within the Soviet side. Soviet-bloc is statistically significant at the 0.001 level in both of Models 2 and 4. These results reveal that the democratic peace was not driven solely by Cold War bipolarity, unlike the claim made by Gowa (1999) and Mearsheimer (1990). Also, in contrast to Rosato’s (2003) hypothesis, not only does democracy remain a useful predictor of conflict, but common security interests only appreciably dampened conflict on the less democratic side of the Cold War.

The results for the control variables are in the expected directions. Skewed capability ratio, non-contiguity, distance, absence of a powerful state and duration of peace reduce conflict. These results hold for the post-Cold War analysis.

The post-Cold War era (1990–2001)

The post-Cold War analysis addresses the inconsistent effect of joint democracy reported by Gowa (2011). Model 1 in panel A of Table 2 produces results similar to the main results in Gowa (2011: 162, column 1, Table 2).²³ The insignificant positive sign for Democratic Dyad suggests that there is little difference in conflict proclivity between democratic dyads and non-democratic dyads. Even the difference in dispute propensity between democratic dyads and mixed ones is only marginally significant (p-value = 0.095, one-tailed).²⁴ Yet this difference is not significant when the Cold War control is included (Model 1 in panel B). This is the type of evidence that Gowa (2011) uses to dismiss the democratic peace in the post-Cold War era.

Table 2.

The post-Cold War equation (1990–2001): the effect of joint democracy on MID onset.

	Trichotomy Gowa (2011)	Weak link Oneal and Russett (2005)	Trichotomy (cutoff = 8)	Trichotomy_t−1 (cutoff = 7)
	Model 1 (1990–2000)	Model 2 (1990–2000)	Model 3 (1990–2000)	Model 4 (1990-2001)
(A) The models without the common interest control
Democratic Dyad	0.01927 (0.2407)		−0.9004** (0.3197)	−0.6370* (0.3285)
Mixed Dyad	0.3044* (0.1518)		0.2839267* (0.1528)	0.4313** (0.1532)
Higher Democracy		0.0338* (0.0164)
Lower Democracy		−0.0358** (0.0149)
Pseudo R²	0.3569	0.3581	0.3645	0.3512
N	144,861	144,861	144,861	141,764
AIC	3529.080	3522.630	3487.855	N/A
BIC′	−1827.675	−1834.126	−1868.900	N/A
(B) The models with the common interest control
Democratic Dyad	0.1477 (0.2419)		−0.9579** (0.3741)	−0.6416* (0.3065)
Mixed Dyad	0.2662 (0.1705)		0.2627 (0.1661)	0.3900** (0.1662)
Higher Democracy		0.0263 (0.0168)
Lower Democracy		−0.0280* (0.0147)
Cold War Rival	1.0376*** (0.3333)	1.0965*** (0.3373)	0.9516** (0.3363)	0.8445** (0.3167)
US-Bloc Ally	−0.2475 (0.3732)	−0.0682 (0.3795)	0.0697 (0.3318)	0.0857 (0.3410)
Soviet-Bloc Ally	−0.2639 (0.4495)	−0.2821 (0.4420)	−0.2304 (0.4404)	−0.3211 (0.4513)
Dyadic Ally	0.3221 (0.2772)	0.2721 (0.2725)	0.2393 (0.2842)	0.1781 (0.2727)
Pseudo R²	0.3535	0.3542	0.3585	0.3540
N	132,631	132,631	132,631	141,764
AIC	3291.031	3287.622	3266.228	N/A
BIC′	−1618.275	−1621.684	−1643.078	N/A

Notes: all of the control variables (i.e. Capability Ratio, Noncontiguity, Distance, Higher Power, Peace Years) and splines are included in all models but not reported owing to space constraints. Significance levels are indicated as follows: * p < 0.05; ** p < 0.01; *** p < 0.001 (one-tailed). Robust standard errors clustered on dyad are in parentheses. The observation number differs between (A) and (B) except for Model 4, where all the independent variables are lagged by one year. This is because the COW alliance information is not available for 2001.The lower the AIC is, the better the fit, and the more negative the BIC′ is, the better the fit. The conventional guideline for the strength of evidence favoring one model against the other suggests an absolute difference of >10 in BIC′. Models 1 and 4 are not comparable by AIC and BIC because these criteria require the same observations to be used across models.

The insignificance of Gowa’s trichotomous measure, however, may be due to the fledgling democracies that appeared with the winding down of the Cold War. I address this caveat through Models 2–4 in panel A. First, the weak link measure, Lower Democracy, may capture at least partially the differences between well-established democracies and inchoate ones. Lower Democracy is statistically significant at the 0.01 level in Model 2. Second, I increase the threshold for Gowa’s trichotomous measure from Polity score 7 to 8. In Model 2, Democratic Dyad has a significant negative coefficient. The model fit statistics based on the Akaike (AIC) and Bayesian (BIC) information criteria strongly support the models with these alternative measures over the one with Gowa’s measure. Lastly, in Model 4, I use Gowa’s measure but lag it by one year with all other independent variables. Democratic Dyad has a negative and significant coefficient.²⁵ Qualitatively identical results are produced with the inclusion of the Cold War control in panel B. These results suggest that Gowa’s insignificant finding for democracy after the Cold War is associated with new democracies. A democratic dyad may not enjoy the benefits of peace immediately if at least one country is a new democracy. Even with one consideration made in this regard—the weak link measure, the higher threshold or one-year lag—I produce the inference that joint democracy continues to pacify interstate relations in the post-Cold War era. The balance of the evidence is consistent with the democratic peace operating in the Cold War period.

Testing across eras

One remaining task is to compare the pacifying effects of joint democracy across the two periods. I conduct the analysis of cross-era comparison in terms of both log-odds ratio and predicted probability. I focus on Lower Democracy, as it is of the primary importance in this study. Owing to space constraints, I only report the results for the models without the Cold War control (i.e. Model 3 in Table 1 and Model 2 in panel A of Table 2).²⁶

Figure 1 plots the coefficients of Lower Democracy along with the 90% confidence intervals. The first plot shows that the effect of joint democracy is smaller in the post-Cold War era than the Cold War era. This difference in the estimated coefficients, however, is not statistically significant as its 90% confidence interval includes 0.²⁷ A similar pattern is found in the first plot of Figure 2 that traces the predicted probabilities of conflict along with the 90% confidence intervals over the values of Lower Democracy.²⁸ The line tracing the mean probability at each point of Lower Democracy is steeper for the Cold War equation than in the post-Cold War equation. Yet there is a considerable overlap in the confidence intervals of the mean probabilities between the two equations. As for the change in the predicted probability of conflict onset (Table 3), Lower Democracy decreases the likelihood of conflict by 67.6% (27.6%) in the Cold War era and 50.8% (20.3%) in the post-Cold War ear, as it moves from its minimum to maximum value (from −1/2 standard deviation to +1/2 standard deviation around its mean).

Figure 1.

Dot plot of Lower Democracy coefficients and their difference across the Cold War and post-Cold War periods. The first row is for the Cold War era and the second row for post-Cold War era. The third row represents the difference in the coefficients between the two periods. The first column is for the estimates for Lower Democracy. The second column is for Lower Democracy_t−1. The bars represent the 90% confidence intervals. The dots represent the mean coefficients (rows 1 and 2) or the mean coefficient differences (row 3).

Figure 2.

The predicted probabilities of MID onset estimated over the values of Lower Democracy (A) and Lower Democracy _t−1 (B). All continuous variables are held at their median while the dummies are set to be conflict-prone. The solid lines and bars represent the mean probabilities and the 90% confidence intervals for the Cold War era. The dashed lines and bars are for the post-Cold War era.

Table 3.

Probability changes in risk for annual MID onset according to joint democracy.

Specification	Change in X	% Δ of Prob. of MID onset
		Cold War era	post-Cold War
Lower Democracy	min to max	−67.6%	−50.8%
	±1/2 std.	−27.6%	−20.3%
Lower Democracy _t−1	min to max	−68.4%	−77.0%
	±1/2 std.	−28.1%	−37.3%

The analysis is redone with the one-year lag specification. This alternative approach produces more favorable results for the Kantian thesis. The second plot of Figure 1 shows that the negative coefficient for Lower Democracy is greater in the post-Cold War era than the Cold War era. This difference, however, is insignificant as 0 falls within its 90% confidence interval. Likewise, no stark difference in pattern is found between the two mean probability lines in the second plot of Figure 2. Regarding the change in predicted probability, the probability of MID onset decreases by 68.4% (28.1%) in the Cold War era and 77.0% (37.3%) in the post-Cold War era as Lower Democracy moves from its minimum to maximum value (from −1/2 to +1/2 standard deviation around its mean).

In sum, the results suggest that the substantive pacifying effect of joint democracy continues to exist beyond the Cold War. Depending upon the specification, the effect is estimated to be larger or smaller in the post-Cold War period.

Conclusion

My analysis reveals that the ongoing skepticism of the democratic peace is not supported by the available evidence. It is hard to deny that the Cold War imposed particular security interests upon interstate relations, but controlling for Cold War interests does not render democracy irrelevant. Rather, the democratic peace continues to have a statistically discernible and substantively important effect in the post-Cold War era across each specification that either takes into account the nuances of democratic coding or lags the effect of democracy on peace.

These findings provide some guidance for future research regarding the deepening and maturing aspects of the democratic peace.²⁹ Additional studies are needed that account for the maturity of democratic dyads and democratic neighborhoods in testing the democracy–peace relationship. Building on the work of Gleditsch (2002) and Cederman (2001), it will be useful to move beyond linear testing and allow for flexible relationships between the maturation process of democracies, their neighborhoods and the respective probabilities of peace between pairs of states. A more dynamic and flexible approach may help improve our understanding of the multifaceted nature of the democratic peace, while also possibly spurring new critiques and boundary conditions.

Footnotes

Acknowledgements

I am deeply grateful to Michael Colaresi and Valentina Bali for their time, insight and encouragements throughout the entire process of this project. My special thanks are due to Petra Hendrickson for repeatedly proofreading this paper at its various stages. I also thank Cristina Bodea, Allan Dafoe, Patrick James, Steven Kautz, John Oneal, Bruce Russett, Michael Tomz and Matt Zierler for their helpful suggestions and comments. Joanne Gowa kindly shared her data that I used for the replication analysis reported in the web Appendix. I appreciate Glenn Palmer’s excellent editorial suggestions and the four anonymous reviewers’ useful comments. The web Appendix and replication files are available at

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

Alvarez

Cheibub

Limongi

Przeworski

(1996) Classifying political regimes. Studies in Comparative International Development 31(2): 3–36.

Beck

Katz

Tucker

(1998) Taking time seriously: Time-series-cross-section analysis with a binary dependent variable. American Journal of Political Science 42(4): 1260–1288.

Bennett

Stam

(2000a) EUGene: A conceptual manual. International Interactions 26(2): 179–204.

Bennett

Stam

(2000b) Research design and estimator choices in the analysis of interstate dyads: When decisions matter. Journal of Conflict Resolution 44(5): 653–685.

Box-Steffensmeier

Reiter

Zorn

(2003) Temporal dynamics and heterogeneity in the quantitative study of international conflict. In: Mansfield

Pollins

(ed.) Economic Interdependence and International Conflict: New Perspectives on an Enduring Debate. Ann Arbor, MI: University of Michigan Press, pp. 273–289.

Braumoeller

(2004) Hypothesis testing and multiplicative interaction terms. International Organization 58(4): 807–820.

Cederman

L-E

(2001) Back to Kant: Reinterpreting the democratic peace as a macrohistorical learning process. American Political Science Review 95(1): 15–31.

Choi

S-W

(2011) Re-evaluating capitalist and democratic peace models. International Studies Quarterly 55(3): 759–769.

Dafoe

(2011) Statistical critiques of the democratic peace: Caveat emptor. American Journal of Political Science 55(2): 247–262.

10.

Dixon

(1994) Democracy and the peaceful settlement of international conflict. American Political Science Review 88(1): 14–32.

11.

Enterline

(1998) Regime changes and interstate conflict, 1816–1992. Political Research Quarterly 51(2): 385–410.

12.

Farber

Gowa

(1995) Polities and peace. International Security 20(2): 123–146.

13.

Farber

Gowa

(1997) Common interests or common polities? Reinterpreting the democratic peace. Journal of Politics 59(2): 393–417.

14.

Fearon

(1994) Domestic political audiences and the escalation of international disputes. American Political Science Review 88(3): 577–92.

15.

Gartzke

(1998) Kant we all just get along? Motive, opportunity, and the origins of the democratic peace. American Journal of Political Science 42(1): 1–27.

16.

Gartzke

(2007) The capitalist peace. American Journal of Political Science 51(1): 166–191.

17.

Gelpi

Griesdorf

(2001) Winners or losers? Democracies in international crisis, 1918–1994. American Political Science Review 95(3): 633–647.

18.

Ghosn

Palmer

Bremer

(2004) The MID3 data set, 1993–2001: Procedures, coding rules, and description. Conflict Management and Peace Science 21(2): 133–154.

19.

Gibler

(2007) Bordering on peace: Democracy, territorial issues, and conflict. International Studies Quarterly 51(3): 509–532.

20.

Gleditsch

(2002) All Politics is Local: The Diffusion of Conflict, Integration, and Democratization. Ann Arbor, MI: University of Michigan Press.

21.

Gowa

(1995) Democratic states and international disputes. International Organization 49(3): 511–522.

22.

Gowa

(1999) Ballots and Bullets: The Elusive Democratic Peace. Princeton, NJ: Princeton University Press.

23.

Gowa

(2011) The democratic peace after the Cold War. Economics and Politics 23(2): 153–171.

24.

Gowa

Mansfield

(1993) Power politics and international trade. American Political Science Review 87(2): 408–420.

25.

Hegre

Oneal

Russett

(2010) Trade does promote peace: New simultaneous estimates of the reciprocal effects of trade and conflict. Journal of Peace Research 47(6): 763–774.

26.

Huth

Allee

(2002) The Democratic Peace and Territorial Conflict in the Twentieth Century. New York: Cambridge University Press.

27.

Jaggers

Gurr

(1995) Tracking democracy’s third wave with the Polity III data Journal of Peace Research 32(4): 469–482.

28.

James

Park

Choi

S-W

(2006) Democracy and conflict management: Territorial issue claims in the western hemisphere revisited. International Studies Quarterly 50(4): 803–817.

29.

Kant

(1957 [1795]) Perpetual Peace, Beck

(ed.). New York: Bobbs-Merrill.

30.

Kant

(1970 [1784]) Idea for a universal history with a cosmopolitan purpose. In: Reiss

(ed.) and Nisbet

(trans.) Kant’s Political Writings. Cambridge: Cambridge University Press, pp. 93–130.

31.

King

Tomz

Wittenberg

(2000) Making the most of statistical analyses: Improving interpretation and presentation. American Journal of Political Science 44(2): 347–361.

32.

Layne

(1994) Kant or cant: The myth of the democratic peace. International Security 19(2): 5–49.

33.

Levy

(1988) Domestic politics and war. Journal of Interdisciplinary History 18(4): 653–673.

34.

Mansfield

Snyder

(1995) Democratization and the danger of war. International Security 20(1): 5–38.

35.

Maoz

(2005) Dyadic MID Dataset (version 2.0). Available at: psfaculty.ucdavis.edu/zmaoz/dyadmid.html.

36.

Maoz

Abdolali

(1989) Regime type and international conflict: 1816–1976. Journal of Conflict Resolution 33(1): 3–35.

37.

Maoz

Russett

(1993) Normative and structural causes of the democratic peace. American Political Science Review 87(3): 624–638.

38.

Mearsheimer

(1990) Back to the future: Instability in Europe after the Cold War. International Security 15(1): 5–56.

39.

Mitchell

(2002) A Kantian system: Democracy and third-party conflict resolution. American Journal of Political Science 46(4): 749–759.

40.

Mitchell

Prins

(1999) Beyond territorial contiguity: Issues at stake in democratic militarized disputes. International Studies Quarterly 43(1): 771–792.

41.

Mitchell

Kadera

Crescenzi

(2008) Practicing democratic community norms: Third party conflict management and successful settlements. In: Bercovitch

Gartner

(eds) International Conflict Mediation: New Approaches and Findings. Abingdon: Routledge, pp. 243–264.

42.

Narang

Nelson

(2009) Who are these belligerent democratizers? Reassessing the impact of democratization on war. International Organization 63(2): 357–379.

43.

Oneal

Russett

(2005) Rule of three, let it be? When more is really better. Conflict Management and Peace Science 22(4): 293–310.

44.

Oneal

Russett

Berbaum

(2003) Causes of peace: Democracy, interdependence, and international organizations, 1985–1992. International Studies Quarterly 47(3): 371–394.

45.

Park

(2011) Exploring the second wave of democratic peace critiques: Commonality, territoriality, and selectivity. PhD Dissertation. Michigan State University.

46.

Park

Colaresi

(2013) Safe across the border: The continued significance of the democratic peace when controlling for stable borders. International Studies Quarterly, in press.

47.

Park

James

(2012) Democracy, territory, and armed conflict, 1919–1995. Unpublished manuscript.

48.

Ray

(2005) Constructing multivariate analyses (of dangerous dyads). Conflict Management and Peace Science 22(4): 277–292.

49.

Raymond

(1994) Democracies, disputes, and third-party intermediaries. Journal of Conflict Resolution 38(1): 24–42.

50.

Rosato

(2003) The flawed logic of democratic peace theory. American Political Science Review 97(4): 585–602.

51.

Russett

Oneal

(2001) Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: Norton.

52.

Schmemann

(1989). Gorbachev, in Bonn, sees postwar hostility ending. The New York Times, 13 June.

53.

Schultz

(2001) Democracy and Coercive Diplomacy. New York: Cambridge University Press.

54.

Signorino

Ritter

(1999) Tau-b or not Tau-b: Measuring the similarity of foreign policy positions. International Studies Quarterly 43(1): 115–144.

55.

Simmons

(1999) See you in court? The appeal to quasi-judicial legal processes in the settlement of territorial disputes. In: Diehl

(ed.) A Road Map to War: Territorial Dimensions of International Conflict. Nashville, TN: Vanderbilt University Press, pp. 205–237.

56.

Spiro

(1994) The insignificance of the liberal peace. International Security 19(2): 50–86.

57.

Tomz

Weeks

(2011) An experimental investigation of the democratic peace. Unpublished manuscript.

58.

Ward

Gleditsch

(1998) Democratizing for peace. American Political Science Review 92(1): 51–61.

59.

Weesie

(1999) Seemingly unrrelated estimation and the cluster-adjusted sandwich estimator. STATA Technical Bulletin November: 34–47.

60.

Werner

(2000) The effects of political similarity on the onset of militarized disputes, 1816–1985. Political Research Quarterly 53(2): 343–374.