Level II: Regime Type and Conflict
This week’s main claim is that the cause sits at Level II, in the internal makeup of states. The democratic-peace tradition claims that liberal or democratic institutions and norms keep similar states from fighting one another; the critics reply that the peace may be an artifact of alliance structure, hegemony, or hierarchy rather than of regime type itself, and that “autocracy” is far too coarse a category to support a causal argument.
Reading list
- Doyle. 1983. “Kant, Liberal Legacies, and Foreign Affairs,” Part I. Philosophy and Public Affairs.
- Maoz and Russett. 1993. “Normative and Structural Causes of the Democratic Peace, 1946–1986.” American Political Science Review.
- Rosato. 2003. “The Flawed Logic of Democratic Peace Theory.” American Political Science Review.
- Weeks. 2014. Dictators at War and Peace, Ch. 1.
- Bell and Quek. 2018. “Authoritarian Public Opinion and the Democratic Peace.” International Organization.
- Application: Colby and Kaplan. 2020. “The Ideology Delusion: America’s Competition with China Is Not About Doctrine.” Foreign Affairs.
Doyle (1983): Kant, Liberal Legacies, and Foreign Affairs
Philosophy and Public Affairs · the early formulation of the liberal “separate peace”
Doyle treats liberalism as both an ideology and a set of institutions that shape how states see and conduct foreign affairs. Its normative core is the freedom of the individual and the duty to treat persons as ends, which generates a threefold set of rights: negative freedoms (freedom from arbitrary authority, conscience, press, speech, equality under law, secure property); positive freedoms (rights that protect and promote the capacity for freedom, education, health care, employment); and democratic participation/representation, needed to guarantee the first two so that legislation expresses citizens’ will. He pairs this with an institutional profile of four essentials: juridical equality and civic rights; representative legislatures as effective sovereigns; private property including the means of production; and economies guided predominantly by supply and demand. (It is worth holding Doyle’s “negative/positive” next to Isaiah Berlin’s Two Concepts of Liberty: negative liberty asks “how far am I free from interference?” and positive liberty asks “who is my master?”, Berlin warns that “positive liberty” can turn into paternalism, “forcing people to be free.”) Doyle proceeds top-down: he marries Kant’s normative arguments to historical pattern-claims (a “separate peace” among liberal states) and a regime-type classification, anchoring his mechanisms in republican caution, mutual respect/publicity, and commercial ties, with qualitative, list-based evidence. His liberalism is close to classical liberalism. Moravcsik, by contrast, is bottom-up: he builds micro-foundations from domestic society, societal preferences, representation inside states, and interdependence, to produce a positive theory. Where Doyle asks “which regime types and Kantian mechanisms explain a dyadic peace?”, Moravcsik asks “which societal preferences, filtered through institutions, produce state preferences and bargaining outcomes?” As for Table 1, I find several inclusions strained against Doyle’s own criteria: Singapore (1965–) and Malaysia (1963–) kept tight constraints on opposition and media; South Korea (1963–72) and the Philippines (1946–72) had strong executive/military influence that strains “effective representative sovereignty”; El Salvador, Nigeria, and Turkey had short, coup-prone democratic interludes. Doyle himself calls the list “approximate,” but several entries would not pass a contemporary Polity/V-Dem liberal-democracy cutoff; a transparent coding rule (how much media freedom? how to treat military tutelage?) would help. DPT here is the claim that constitutionally secure liberal states form a “separate peace” with one another while remaining capable of conflict with non-liberal states, a steadily expanding pacification among liberal societies. The mechanism is Kantian, but the evidence is historical: a long chronological war list (Table 2) used to show that liberal–liberal wars are essentially absent and that conflicts overwhelmingly involve at least one non-liberal side; historical vignettes such as Italy in 1914–15 joining the alliance with France and Britain so as not to fight other liberal states; and “durable pacification” language in which anarchy is “tamed” among similarly liberal states.
Doyle argues realism explains general insecurity (sovereign states in anarchy, fear, preventive war) but cannot predict a dyadic exception, why liberal–liberal pairs are pacified while others are not, so he adds a Kantian supplement. I disagree that realism is helpless once we drop his assumption of a flat anarchy. Reread through the colonial empire–colony relationship, the nineteenth- and twentieth-century system was not truly anarchic but hierarchical and imperial. The “liberal world” was the core of that order, and liberal states kept peace among themselves not from shared morality but because they had already reached a balance of power and a stable division of spoils inside the imperial system. Realism also explains their cohesion through common external threats: colonial expansion against the “uncivilized world,” then the Cold War against communism, supplied shared enemies that pushed alliance and coordination. And the internal peace of the liberal core rested on the externalization of violence, colonial wars, exploitation, and coerced trade exported conflict to the periphery, so the core could enjoy peace at home; cooperating to exploit the colonies together simply paid better than fighting one another. So the “Pacific Union” looks less like a triumph of liberal norms and more like a hegemonic peace sustained by imperial hierarchy, built on shared strategic interests, external threats, and colonial order, not on republican institutions alone. Doyle extracts three Kantian mechanisms. (1) Republican constitutions (caution): representation makes leaders internalize war’s costs and citizens resist risky wars. (2) A “pacific union” of liberal states (respect): a treaty-like, steadily expanding association that prevents wars among members through mutual legitimacy and publicity. (3) Cosmopolitan law/hospitality (interest via commerce): rights of peaceful visitation enable trade, and cross-border interdependence disincentivizes war. The first restrains a state’s internal impulse; the second externalizes that restraint into cooperation through mutual trust; the third converts peace into a stable structure of material interest. But Doyle gives a functional, institutional version of Kant, stripping out the moral and transcendental dimensions. In Kant, the republican constitution is the outward expression of moral autonomy (a “kingdom of ends”), the pacific federation is an “idea of reason” rather than a mere treaty, and cosmopolitan right rests on the equal dignity of all persons as co-inhabitants of the Earth. Doyle “operationalizes” these: the constitution becomes a domestic cost-constraint, the federation becomes the empirical community of liberal states (roughly the liberal international order), and cosmopolitan right becomes economic interdependence.
Maoz & Russett (1993): Normative and Structural Causes of the Democratic Peace
American Political Science Review · turning Doyle’s claim into testable propositions, 1946–1986
They define the democratic peace as two jointly true empirical regularities: democracies are about as conflict-prone overall as nondemocracies, but democracies rarely fight each other, they almost never escalate disputes with other democracies to violent conflict. Their normative model traces explicitly back to Kant, Wilson, and “such modern scholars as Doyle,” sharing his claim that liberal norms of compromise are internalized and extended outward. The key differences: where Doyle inferred a regularity from history through an interpretive, philosophical reading of Kant’s pacific union, Maoz & Russett turn the claim into falsifiable propositions testable on a large cross-national dataset, and they decompose what Doyle treated as interwoven, republican institutions and liberal norms, into two analytically separable, competing models (normative and structural). Normative model: states externalize their domestic political norms into foreign relations; in anarchy, when democratic and nondemocratic norms clash, the nondemocratic (tougher) norms tend to dominate. So when two democracies dispute, shared norms of compromise and nonviolent resolution lower the incentive to escalate; against a nondemocracy, a democracy may adapt to harsher norms to avoid exploitation. Observable implication: democratic dyads settle disputes peacefully and reciprocally. Structural model: democratic policymaking is fragmented and slow, with many veto points; citizens bear the costs and elites are accountable, so leaders anticipate punishment for costly wars; transparency (free media, public debate) reduces surprise and aids signaling; secrecy and surprise are harder, so mobilization is visible and escalation constrained; and third-party constraints (courts, opposition, coalitions) further inhibit escalation. Plausibility: both are credible, but the normative logic is harder to observe independently of outcomes, while the structural logic risks being conditional and manipulable (leaders shape opinion). They test three hypotheses, joint democratization, norms, and institutional constraints each lowering dispute onset and escalation. Unit of analysis is dyad-years among “politically relevant” dyads, contiguous neighbors, or pairs with at least one major power, or indirectly allied, over 1946–1986, on the order of ~19,000 dyad-years. Restricting to relevant dyads excludes pairs that could never realistically fight, avoiding zero-inflation and making “joint democracy → peace” closer to a real causal claim than a by-product of geography or power; the question is not whether any two states fight, but whether regime type still matters when interaction is actually possible. They use two datasets and two stages: onset (a Militarized Interstate Dispute, threat, display, or use of force, on a 0–4 scale, or an ICB crisis) and escalation (to war); using both MID and ICB checks that the pattern is not an artifact of one coding scheme. Regime type comes from Polity II, both as a continuous score (REG, and the dyadic JOINREG similarity index based on the absolute difference between the two regime scores) and as a categorical dummy (democracy ≥ +30; anocracy in between; autocracy below). The continuous measure tests a marginal “more democratic → more peaceful” effect; the dummy tests a threshold effect; both point the same way. Norms are proxied by political stability, executions, and domestic violence; structural constraints by an Institutional Constraints Index from Polity (executive constraints), validated against Morgan & Campbell. Controls, wealth, growth, alliances, contiguity, capability ratio, guard against spuriousness. Estimation is pooled dyad-year logit. Takeaway: joint democracy is strongly negative and significant for both onset and escalation, in both MID and ICB, even after the controls; contiguity and capability asymmetry behave as expected.
They consider wealth/trade (rich, trading states have more to lose), growth (fast-growers avoid costly wars with each other), and alliances (many democracies are directly or indirectly allied), plus contiguity and capability ratio as potent predictors to be controlled. All four matter, and building them into the design so the democracy effect is not spurious is a major strength. But a couple of plausible alternatives weren’t the focus: common IGO/UN membership or joint trade dependence (interdependence) could dampen conflict independently of regime, and regional security orders (e.g., NATO’s institutions beyond the alliance tie per se) bundle norms and enforcement in ways that overlap with but are not identical to regime type. Both mechanisms get support, but the normative effects are more robust and consistent: the structural (institutional-constraint) channel is especially visible in limiting escalation, while norms also reduce the chance disputes emerge at all. I am only partly convinced, because the design measures only direct interstate conflict, it cannot capture the outsourcing of violence. During the Cold War, liberal democracies rarely fought one another not necessarily because they internalized pacific norms, but because conflict was displaced onto the periphery through proxy wars and regime interventions; a dyad-year model by construction misses covert operations and sponsored civil conflicts. Critics (Layne 1994, Farber and Gowa 1997, Barkawi and Laffey 1999, Rosato 2003) read the democratic peace as a feature of the Cold War alliance system and global hierarchy rather than democracy per se, an “imperial peace” sustained by externalized violence. Our class discussion pushed further: even if “democracies rarely fight each other” is a real, scope-limited finding (the theory only ever claimed to explain why democracies don’t fight one another, not that they are more peaceful overall), DPT carries an exclusionary logic. Doyle’s reply to counter-examples, that aggressive states were simply “not truly liberal”, makes the theory self-sealing, immune to falsification by re-coding offenders out of the category. That move continues a classical-liberal property logic and a modern civilizational hierarchy: societies organized around the commons rather than private property (recall the Manifest-Destiny portrayal of Native societies as “backward” for lacking private property) are excluded from “democracy” by definition. So “democratic peace” is really “liberal peace,” bound up with the global expansion of capitalist order, which is exactly why critics like Rosato, Barkawi, Laffey, and Jahn see it as masking the exclusionary, colonial side of the liberal order. We must therefore separate two layers: DPT as a conditional empirical proposition whose causal mechanism needs to be shown independent of structure, and DPT as a political discourse that has been used to legitimize intervention and “democracy export”, an ideological device that both describes the world and reshapes its legitimacy.
Rosato (2003): The Flawed Logic of Democratic Peace Theory
American Political Science Review · grant the correlation, attack the causal mechanisms
Rosato’s paper is a sharp stress-test: he grants the correlation (“democracies rarely fight each other”) but attacks whether democracy causes the peace through any standard mechanism. I found it probing and often fair on internal-logic and measurement grounds, especially where each mechanism implies extra, testable predictions that don’t systematically hold; but several criticisms rest on selective historical illustrations and contestable assumptions (e.g., the equivalence of “accountability” across regime types) that defenders can challenge. His method: evaluate each causal pathway against process-tracing expectations, derive auxiliary predictions (about imperialism, surprise attack, accountability penalties), and check them against history. He targets not the fact of democratic–democratic peace but whether it flows from democratic institutions: since all the mechanisms, norms, accountability, signaling, run through domestic politics, we should unpack “democracy” into specific mechanisms (which structures channel public constraint? which party competition, media, collective-action costs shape war decisions? which preferences like nationalism or alliance dependence drive external policy?). He operationalizes “liberal democracy” as the intersection of Przeworski et al.’s procedural definition (competitive elections, alternation of power) and Doyle’s substantive liberal criteria, so even Britain, France, and the U.S., coded liberal, have waged imperial wars, contradicting the normative logic; the cost is that his coding ignores hybrid regimes, and several “liberal” cases he cites read better as semi-liberal empires. Ultimately he argues the democratic peace is a by-product of American hegemony: if democratic norms/institutions caused peace, they should restrain all uses of force, which they don’t, pushing us from binary regime-type logic toward domestic preferences and institutional structure (the later selectorate and audience-cost path). (1) Normative externalization & mutual trust/respect: DPT says democracies socialize elites into “live-and-let-live” norms exported abroad, so liberal dyads trust and accommodate one another. Flaw: democracies often fail to externalize liberal norms, imperial wars and coercive interventions belie it, and even among democracies, serious clashes don’t reliably produce trust (Layne’s four Anglo-American crises and Rock’s nineteenth-century cases find realism, not liberal forbearance, drove peace). Fair as a scope critique, though defenders restrict it to dyads where both meet liberal criteria. (1b) “Repaired” logic, perceptions of who is democratic: trust operates only if each side perceives the other as democratic, and misperception explains exceptions. Flaw: perceptions are contested, unstable, and interest-laden, so the repair lacks predictive bite (Owen’s data show frequent elite disagreement inside the U.S. about others’ liberal status). I find this strong, if who “counts” shifts with partisanship, the theory risks post-hoc coding. (2) Public constraint: an accountable, war-averse public restrains leaders. Flaw: publics often rally rather than restrain, and leaders shape opinion (Mueller’s “rally round the flag,” Baum & Groeling on media filters). Fair. (3) Group constraint: anti-war groups in a broad coalition block escalation. Flaw: leaders outmaneuver groups, and sometimes act despite opposition (Madison 1812). Persuasive caution. (4) Slow mobilization: open debate slows democracies, aiding bargaining. Flaw: leaders bypass deliberation when they deem speed essential (Korea, Cambodia, the Gulf War). Strong, at least for the U.S. presidency. (5) No surprise attack: public mobilization precludes operational surprise. Flaw: surprise depends on defender misperception, not attacker secrecy, democracies have surprised (Suez 1956, Israel 1967) and suppress transparency at will; as Finel & Lord note, more transparency can mean more noise. Compelling. (6) Information/credible signaling: accountability and opposition scrutiny make democratic threats more informative. Flaw: transparency ≠ clarity, open systems spew too much information and observers cherry-pick (Jervis; Finel & Lord), and opposition “blessing” often follows a rally logic, so it isn’t a clean signal. Strong. And the cornerstone, that democracies are more accountable than autocracies, is itself shaky: revising Goemans’s data, Rosato’s Table 4 shows losing autocrats are more often removed or severely punished than losing democrats, so autocrats may face an equal or larger downside and be just as constrained ex ante.
Three stand out. First, information ≠ clarity: “the mistake has been to equate plentiful information with perfect information” is devastating to a naïve transparency-leads-to-peace claim, and his use of Finel & Lord plus the fragility of opposition signals is convincing, it echoes Schelling/Jervis on noise and the selection-effects debates (Reiter & Stam vs. Downes). Second, comparative accountability (Table 4): if autocrats’ expected punishment can be harsher, DPT’s unique-constraint premise wobbles. Third, the perception-repair critique: if whether another state “counts” as a democracy is politicized and unstable (Owen), the repaired normative logic risks circularity. On fairness, the internal-logic checks are fair, he makes each mechanism generate extra predictions (imperial restraint, inability to surprise, slower mobilization, clean signals, greater accountability), and their absence weakens the story; the evidence is often persuasive but U.S.-centric on the institutional side, and Rosato himself acknowledges data gaps (e.g., covert interventions). Class examples reinforce it: 1967 and the Cuban Missile Crisis fit Jervisian misperception; personalist leaders who fall lose everything (strengthening the penalty logic) while many democratic leaders survive costly wars; and Cold-War-era allies were strategically deemed “democratizing” while hostile elected governments were called “illiberal”, exactly the perception endogeneity he flags. Three lines. First, clarify scope conditions and selection: DPT predicts peace between recognized liberal democracies, not between democracies and colonies/illiberal states, nor during incomplete transitions; imperial wars are out-of-scope dyads. (Rosato flags this as ad hoc unless backed by stable recognition rules.) Second, emphasize probabilistic, not deterministic, mechanisms: DPT is about tendencies, so a few democratic surprise attacks or rapid mobilizations don’t overturn the average, and Rosato himself grants the correlation. Third, add interaction effects: even if any single channel is noisy, shared liberal identity plus institutional checks plus repeated interaction can jointly reduce misperception and bargaining failure, so single-case vignettes understate the complements.
Weeks (2014): Dictators at War and Peace
Ch. 1 · variation within authoritarianism, and elite audiences
Weeks wants to explain why authoritarian regimes vary so much in their tendency to initiate conflict, their success in war, and the survival of their leaders after defeat. Treating all autocracies as one aggressive category, she argues, is a fundamental mistake; instead she offers a typology built on how authoritarian leaders are constrained by domestic actors. Her novelty: most IR treats regime type as a binary (democracy vs. autocracy) and assumes autocrats are unconstrained and thus more war-prone; Weeks shows some autocrats face strong constraints, not from voters but from elites. She builds a two-dimensional framework, personalist vs. nonpersonalist, and civilian vs. military leadership/audience, yielding four ideal types, and she ports domestic-audience theory (previously used mainly for democracies) into authoritarian contexts, showing elite audiences can impose ex post accountability just as voters do. Her contribution is theorizing variation within authoritarianism through institutional logic and elite incentives. Ex ante constraints are institutional barriers that prevent a leader from choosing or implementing a policy at all (prevention, can’t act); ex post accountability punishes leaders after the fact, especially after failure (the leader can act but may face removal, exile, or execution). Weeks argues ex ante constraints are rare in crises because leaders need flexibility, so ex post accountability does the work: voters punish in democracies; elites, generals, party bosses, punish through coups or purges in autocracies, so even without elections, fear of elite backlash can make a dictator cautious. On preferences, she focuses on four factors that shape leaders’ and audiences’ choices: the benefits of winning relative to the peaceful status quo (greedy or fearful leaders may prefer war), the costs of fighting (material and moral, she denies autocrats are insulated from them), the costs of defeat (lost territory, resources, even regime survival; defeat reveals weakness and invites future attack), and the likelihood of winning (optimism vs. pessimism). I read costs of defeat and likelihood of winning as especially important for elite audiences, since defeat threatens their survival. What’s missing is non-rational drivers: ideology (revolutionary regimes like Maoist China or Khomeinist Iran), nationalism and identity (irredentism, ethnic conflict), external alliances (superpower backing lowering perceived costs), and the use of war to suppress domestic dissent, any of which can override her cost-benefit calculus. Two dimensions, personalist vs. nonpersonalist (power centered on one individual through kinship/patronage, vs. shared among an elite group with merit/seniority promotion) and civilian vs. military leadership/audience, give four ideal types: machine (nonpersonalist civilian), boss (personalist civilian), junta (nonpersonalist military), strongman (personalist military). Examples: machine, post-Mao China (Deng onward) and the post-Stalin USSR (collective, Politburo-based); boss, Saddam’s Iraq (family/tribal loyalists) and early Kim Il-sung; junta, Argentina 1976–83 (Galtieri ousted after the Falklands) and post-2014 Thailand; strongman, Gaddafi’s Libya and Stalin’s USSR (party-based but personalized through purges and fear). Strengths: simple, clear, theoretically meaningful, tied directly to audience strength and preferences, and good at explaining war behavior. Weaknesses: it over-simplifies hybrids (Putin’s Russia blends personalism with institutional elites, siloviki, United Russia, and fits none cleanly; Iran is a theocracy with competing centers; Singapore is a dominant-party machine yet with strong personal leadership under Lee Kuan Yew), it’s static (it can’t capture a junta drifting toward personalism), and personalism is really a spectrum forced into binaries. Versus other typologies: Geddes (1999) classifies by dominant institution (military/party/personalist), similar, but Weeks adds the civilian/military audience distinction; Linz (2000) focuses on ideology and pluralism, less useful for foreign policy; Brownlee (2007) distinguishes sultanistic/bureaucratic/competitive authoritarianism, overlapping but less war-focused. Weeks’s typology is superior for foreign policy because it links institutions directly to decision-making incentives.
Weeks’s key assumption is that regime insiders in nonpersonalist autocracies are not necessarily dependent on the leader for their own survival. In practice this means machine elites can oust or punish the leader and expect to keep their positions afterward, in places like China, party officials rise through seniority and merit, not personal loyalty, so a new leader won’t purge everyone. That career continuity gives them incentives to monitor decisions, including war: if a war goes badly, they replace the leader to protect their own interests, which makes leaders consult and avoid high-risk wars. I agree, with caveats: in institutionalized autocracies like the CCP, norms of collective leadership and retirement hold in stable periods, but Xi Jinping has weakened them by consolidating power, showing personalism can re-emerge during transitions or crises. Weeks also thinks the literature underestimates nonpersonalist autocrats’ vulnerability for two reasons: machine leaders can’t easily buy off or intimidate elite opponents (they lack full control over resources and repression), and they find it much harder to massage domestic opinion (informed, autonomous elite audiences see through propaganda). So selectorate-style claims that a small winning coalition means less accountability miss the point, audience strength matters more than coalition size: if a machine leader starts a costly, losing war, informed elites can and will remove him, just as voters would. On juntas, Weeks doesn’t directly rank hawkishness in this excerpt but implies military audiences see force as more routine and low-cost, yet they are less tolerant of losing, because defeat damages the military’s reputation and power (the Argentine junta removed Galtieri after the Falklands). So juntas may be more likely to start wars but quick to punish failure. I agree: officers often treat force as a professional tool, not a last resort, but they protect the institution’s prestige, so losing is unacceptable, hence adventurism followed by coups (Argentina, Nigeria). On strongmen (personalist + military), they are most likely to initiate because they face no effective audience: unconstrained, deciding alone, and prone to overconfidence with no dissent. And they are most likely to be defeated because, lacking institutional decision-making, they misjudge costs, enemy strength, and odds, no one challenges their optimism, and defeat is catastrophic since they have no elite support network and their cronies fear removal by any successor rather than acting as a stabilizing audience. No checks plus poor information equals high risk of failure. Reverse causation/endogeneity: does personalism cause war, or does war (by consolidating power) cause personalism? If the security environment shapes regime type, threats producing personalism or military rule, then regime-type/conflict correlations could reflect reverse causation: a leader who wins becomes more personalist; a losing war triggers elite rebellion that makes a regime look “less constrained” after the fact. Weeks relies on theoretical logic and process tracing in this chapter, assuming regime type shapes behavior; she also runs supplementary analyses showing states facing past security threats are not more likely to transition to personalism or military rule, and uses cases to trace how, given a fixed external environment, domestic politics affected decisions. For qualitative theory-building this multi-pronged strategy (especially the process-tracing) is reasonable; remaining worries are measurement error in regime coding and slow-moving unobserved confounders (organizational culture, ideology) correlated with both regime type and threat exposure. On Table 1.3, which predicts initiation and defeat rising machines < juntas < bosses < strongmen, high punishment for machines and low for bosses/strongmen: I’d flag within-type variation and transitions (Chile junta → personalist), that machines do sometimes fight (North Vietnam, post-Stalin USSR, the claim is relative frequency), that some juntas fight competently before eventual defeat (early-1930s Japan, so separate operational effectiveness from risk propensity), and that elite coordination problems may sometimes shield junta leaders, softening the “medium” punishment rate. Applied today: Weeks lists contemporary China as a machine, predicting caution, broad consultation, and high punishment risk for defeat, low initiation, with diplomacy/coercive leverage short of major war; if Xi has increased personalism since 2014, he shifts toward boss dynamics (higher initiation, lower punishment risk). Putin is a personalist civilian without a professional military background, closest to a boss, predicted high on initiation, high on defeat risk (riskier wars), low on punishment probability, so conflict-prone and resilient even after setbacks. These are model-based inferences; outcomes depend on shifting elite coalitions, the economy, and the specifics of ongoing wars beyond the book’s scope.
Bell & Quek (2018): Authoritarian Public Opinion and the Democratic Peace
International Organization · a survey experiment that takes the public-opinion mechanism to China
They test an important but unexamined assumption in recent DPT work: that publics in democracies are uniquely reluctant to use force against other democracies, while publics in nondemocracies are not. Recent survey experiments (Tomz & Weeks 2013; Johns & Davies 2012) show democratic citizens are less supportive of attacking democracies than autocracies, and read this as a public-opinion microfoundation for the democratic peace. Bell & Quek’s critique is that this only works if nondemocratic publics do not share the preference: if citizens in authoritarian states also oppose attacking democracies, then public opinion alone cannot explain why democracies avoid war specifically with other democracies. If both publics dislike fighting democracies, the observed pattern (peace among democracies but not across mixed dyads) can’t come from differences in opinion content, the difference must lie in how institutions respond to opinion, shifting the mechanism from what people think to how governments react. So their claim isn’t that prior work is wrong, but that its theoretical implications are overstated unless tested beyond Western democracies. They run a large national online survey experiment in China (2,920 adults, Q2 2015), using quota sampling on gender, ethnicity, geography, and income from the 2010 census, representative, though not a random probability sample. Following Tomz & Weeks closely, respondents read a vignette about a country six months from its first nuclear weapon, with conventional military half China’s size, unclear motives, and refused diplomacy; the randomized treatment tells them the country either “is a democracy and shows every sign it will remain one” or “is not a democracy and shows no sign of becoming one.” Alliance and trade status are also randomized as controls. The dependent variable asks whether they’d favor or oppose the Chinese military joining an attack on the nuclear sites (five-point scale), a between-subjects factorial design estimating the average treatment effect of regime type. Finding: Chinese citizens are significantly less likely to support force when the target is described as a democracy, about a six-point gap, p < 0.001, similar in magnitude to U.S. and U.K. effects, robust across education, age, and gender and to alliance/trade controls. Figure 1 plots support for strikes in China, the U.S., and the U.K. by whether the target is democratic; in all three, support is lower against democracies. The gap is ~6 points in China, vs. ~5–6 in Johns & Davies and 11–13 in Tomz & Weeks, and China’s absolute support level sits between the U.S. (highest) and U.K. (lowest). Visually it shows the reluctance to attack democracies is not unique to democracies, it appears in China too, undermining the idea that the preference is a product of democratic culture alone. Presentation could improve with error bars/confidence intervals, numeric effect-size labels, and a footnote explaining cross-country baseline differences (higher U.S. militarism, more U.K. restraint). They identify a pattern but don’t test why Chinese citizens are less supportive of attacking democracies. They note (citing Tomz & Weeks) several mechanisms that might apply universally rather than only in democracies: threat perception (democracies seen as rule-following, dispute-resolving, domestically constrained, hence less threatening, Doyle 1986; Russett 1993), anticipated costs and risks (wars against democracies seen as costlier and harder to win given resilient institutions and mobilization, Reiter & Stam 1998), moral legitimacy (even authoritarian publics may grant elected governments greater legitimacy, raising the bar to attack, a global norm), fear of escalation (democracies have strong allies like NATO, so attacking one risks a wider war), and cultural/identity distance (respondents may associate democracies with powerful Western states, making aggression seem futile). To find the real driver, add mediator questions after the dependent variable, “How threatening is this country to China?” (threat perception), “How likely is China to win?” (expected success), “How morally justified would an attack be?” (norms), “How concerned are you that others would get involved?” (alliance fears), and use causal mediation analysis to estimate how much of the democracy effect runs through each path.
First objection, Chinese public opinion doesn’t influence policy, so the finding is irrelevant. They reply with a growing literature showing authoritarian leaders do respond to public opinion, especially on foreign policy: research challenges the idea that Chinese elites are insulated, citing Zhao (2013) on a nationalist “strident turn” and Chen, Pan & Xu (2016) on elite responsiveness. I find this strong: even without electoral accountability, leaders are sensitive to mass protests, social-media backlash, and patriotic fervor, China’s South China Sea assertiveness often tracks nationalist sentiment, so dismissing Chinese opinion as irrelevant is outdated. Second objection, Chinese citizens may imagine different countries when they hear “democracy” (the U.S., Japan, Germany) vs. “nondemocracy” (North Korea, Iran), so the effect could be country associations rather than regime type. They address it by showing the findings hold after dropping respondents who thought the scenario was North Korea or Iran, and that the democracy effect doesn’t vary by alliance or trade condition (so “democracy” isn’t a proxy for alliance). They argue minzhu is the standard Chinese term, well known from media and textbooks; and that more globally exposed respondents (foreign language, friends abroad, interest in international news, above-high-school education) show little difference in the treatment effect, which they’d expect to diverge if understandings of “democracy” varied widely. I find this partially convincing: excluding NK/Iran guessers reduces but doesn’t eliminate the risk that “democracy” evokes powerful, threatening Western states. A stronger test would ask open-ended “which country do you think this is?” and analyze free text, or use implicit-association measures. Still, given practical constraints, the consistency across subgroups and conditions suggests the finding isn’t simply driven by country stereotypes. Not especially, given our theoretical priors: if wars against democracies are expected to be costlier, riskier, or less legitimate, all publics, Chinese included, should show lower support for attacking them. Realpolitik (democracies are often powerful, wealthy, and allied, so attacking them is risky), norm diffusion (the legitimacy of democratic government has globalized, even if not fully accepted), socialization (education, media, globalization spread the idea that democracies behave differently), and strategic culture (China emphasizes avoiding unnecessary conflicts with major powers) all point the same way. In conversation with the rest of the week, the result is that the democratic peace likely results from a combination of factors rather than a single democracy-specific mechanism: if the same public reluctance shows up in an autocracy, then content of opinion can’t be the whole story, and the action moves to institutions, structure, and hegemony, exactly where Rosato and the imperial-peace critics locate it.
Application · Theory in current events
Colby and Kaplan. 2020. “The Ideology Delusion: America’s Competition with China Is Not About Doctrine.” Foreign Affairs.
The standard Washington view is that US-China competition is increasingly ideological. The evidence: a rare bipartisanship, “the two parties do share something: a deep concern about China”, framed in moral terms (Pelosi’s “we have agreement in that regard” at Munich), legislation on Hong Kong, Taiwan, and sanctions that is symbolic and normative as much as strategic, and Democratic thinkers like Campbell and Sullivan warning that China’s fusion of authoritarian capitalism and digital surveillance may be a stronger, more durable ideological challenge than the Soviet Union. The worry is that ideological framing makes the rivalry “an existential cage match”: the goal becomes defeating and transforming the other’s political system, room for coexistence shrinks because the adversary is morally illegitimate, every move is read as part of a grand battle, and such conflicts, like the Crusades or the Cold War at its peak, lack clear off-ramps, where realist conflicts over power can at least be managed by deterrence and balance. Colby and Kaplan, however, argue the competition is not about doctrine: “the very scale of China’s economy, population, and landmass… would cause profound concern… even if the country were a democracy.” Beijing seeks regional hegemony over Asia, to shape a vast market to its standards, and to make neighbors take security cues from it, largely non-ideological aims; many ideologically diverse great powers (Nazi Germany, imperial Japan, post-revolutionary France, the nineteenth-century U.S., liberal Britain) pursued regional hegemony and preferential trade empires, so the behavior tracks power position, not doctrine. An over-ideological view is counterproductive: it raises escalation risk, pressures Washington toward regime change (hardening Beijing’s resolve “at potentially catastrophic lengths”), magnifies peripheral crises into system tests (as Vietnam became), alienates central illiberal partners (Vietnam, India, Malaysia, much of Southeast Asia) essential to denying Chinese hegemony, and feeds a millenarian myth that spreading democracy ends competition. Applied to DPT: since the U.S.-China dyad is not a democracy-democracy pair, baseline DPT predicts no special “separate peace” here and its mechanisms (mutual respect, audience costs, transparency) don’t bind when one side isn’t liberal; DPT is useful only indirectly, warning us not to overgeneralize liberal peace to non-liberal dyads, and reminding us that ideological regime-change projects undercut stability, exactly Colby and Kaplan’s caution. I largely agree: for grand strategy the material stakes (market access, coercion risks, denial of regional hegemony) should organize the framework while values shape the how. But their argument has real limits. First, ideology and geopolitics can’t be cleanly separated: ideology is central to domestic legitimacy and mobilization, China ties authority to national rejuvenation, U.S. consensus is framed as “freedom versus autocracy”, so as constructivists (Wendt) note, identity shapes interests, and a fully de-ideologized strategy may lack domestic feasibility. Second, they invoke “hegemony” without defining it measurably (military denial? economic dominance? institutional power?); a composite index, U.S. access to the first island chain, the RMB’s share of regional settlement, shifts in allied alignment, would set a clearer threshold for action. Third, they underestimate the governance costs of a coalition of hedging illiberal partners: Vietnam stays cautious about deep military ties, India insists on strategic autonomy (it would not sanction Russia over Ukraine), and intelligence/technology sharing carries leakage and reputational risks, a broad but loose “anti-hegemony coalition” may fail in a crisis, a “reverse enclave.” Fourth, reducing Vietnam to an “ideological trap” is too simple: structural realist domino fears, bureaucratic competition, and domestic fear of “losing Vietnam” all drove overcommitment, so even realist logic can lead to overreach. The most sustainable approach is layered: at the strategic level prioritize interests and power balances, at the operational level use values-based rhetoric to bind coalitions, and at the domestic level manage public expectations so idealism doesn’t override prudence.
← Back to the genealogy tree · ← Prev: Week 6 · Next: Week 8, Causes of War →