Trade, Conflict, and the Gravity of Liberal Peace: Notes with Gravity-Model Interpretation
The Russia-Ukraine War and the Bargaining Model
Does the Russia-Ukraine War fit the bargaining model of war?
Yes, elements of the bargaining model are observable. Russia likely misjudged Ukrainian resolve and international response (incomplete information). Commitment problems are significant, as neither side fully trusts the other to uphold peace terms. Some issues, such as territorial sovereignty, may be framed as indivisible by one or both parties.
Is Vladimir Putin a rational actor?
From a political science perspective, a "rational actor" makes decisions by calculating costs and benefits to maximize their perceived utility, even if observers disagree with the goals or effectiveness of the actions. Putin's actions, within this framework, can be seen as rationally chosen to achieve his perceived security and geopolitical objectives, despite potential miscalculations regarding the consequences.
Is the war a result of incomplete information, commitment problems, or indivisibility?
Incomplete Information: Russia likely suffered from incomplete information regarding Ukraine's military capacity, national resolve, and the severity of international opposition and sanctions. Ukraine also likely had incomplete information about Russia's full intentions.
Commitment Problems: This is a major factor. Given deep historical mistrust and security dilemmas, both sides face challenges in credibly committing to future peace agreements, especially concerning territorial status or security guarantees. A fear that future power shifts could lead to a reneging on agreements makes current cooperation difficult.
Indivisibility: Issues like Ukraine's territorial integrity (e.g., Crimea, Donbas) or, from Russia's perspective, its perceived security sphere and historical claims, are often presented as non-negotiable or indivisible, limiting the scope for compromise.
Is the bargaining model limited to conventional military forces?
No, the bargaining model of war is a theoretical framework applicable to any costly conflict. It can encompass conventional warfare, cyber operations, economic sanctions, and even diplomatic leverage, as long as there are costs to conflict and a potential bargaining range where both sides would prefer a negotiated settlement to continued fighting.
What are Ukraine’s and Russia’s goals?
Ukraine's Goals: Primarily, to restore its territorial integrity within internationally recognized borders, safeguard its sovereignty and independence, deter future Russian aggression, and deepen integration with Western institutions (e.g., EU, potentially NATO).
Russia's Goals: Initially aimed at regime change, demilitarization, and preventing Ukraine's westward alignment. Current goals appear to include securing strategic territories (e.g., Donbas, land bridge to Crimea), weakening Ukraine's capacity for independent action, and challenging the existing European security order.
Introduction and context
Topic: Do trade and conflict influence each other reciprocally? Do they promote peace (liberal peace) or not?
Background: Prior studies (Keshk, Pollins & Reuveny 2004; Kim & Rousseau 2005) used simultaneous equations to estimate reciprocal effects of trade and conflict. They found that while conflict reduces trade, trade does not reliably reduce the risk of conflict when endogeneity is handled — challenging liberal peace.
Reassessment goal: Reanalyze the same data with a theoretically informed gravity model to account for exogenous factors (size and distance) that shape both trade and conflict. Demonstrate that once these exogenous factors are properly modeled, trade does promote peace and conflict reduces trade contemporaneously.
Core takeaway: Endogeneity is real, but when geography and size are incorporated via a gravity framework, the pacific effect of trade reappears. Trade reduces the likelihood of conflict; conflict contemporaneously reduces trade.
Key concepts and theoretical background
Liberal peace theory: Interdependence and institutions reduce interstate conflict; democracies and trade promote peace.
Reciprocal causation / endogeneity: Trade can affect conflict and conflict can affect trade; ignoring simultaneous causation biases estimates.
Gravity model of international interactions:- Trade is proportional to the sizes of two economies and inversely related to the distance between them.
Exogenous factors shaping both trade and conflict include country size and geographic proximity.
Formal intuition: larger economies have greater supply/demand and bargaining power; distance increases costs and barriers; contiguity (shared border) facilitates exchange.
Conflict and size/proximity: Militarized disputes are more likely between large, powerful states that are geographically close; proximity and relative power shape expected costs and benefits of fighting.
Measures used:- Trade: bilateral trade volumes; often logged; includes multilateral resistances and preferential trade agreements (PTAs).
Conflict: Militarized Interstate Disputes (MIDs) or fatal MIDs (KPR/KR datasets; Sherman data used in KR).
Size/power: CINC (Composite Indicator of National Capabilities) from the Correlates of War (COW); GDP and population used in trade specifications.
Geography: distance, contiguity (shared border), and regional proximity.
Political factors: democracy, alliances, rivalries, PTA membership, and system-size controls.
“System size”: accounting for the number of states in the international system over time.
Identification challenge: To estimate causal effects in a two-equation system, one must find valid instruments and satisfy exclusion restrictions so that an instrument affects the dependent variable only through the endogenous regressor.
Gravity model and its role in identifying effects
Gravity model basics:- Trade{ij} ≈ f(GDPi, GDPj, Populationi, Populationj, Distance{ij}, Contiguity
t{ij}, PTAs, Multilateral resistances, Country/Year effects)In logs:
\log Trade{ij} = \eta0 + \eta1 \,\log GDPi + \eta2 \,\log GDPj + \eta3 \,\log Populationi + \eta4 \,\log Populationj + \eta5 \,\log Distance{ij} + \eta6 \, Contiguity{ij} + \dots + \epsilon_{ij}
For conflict (MIDs): size/proximity enter via CINC, GDP, Population, Contiguity, Distance, Democracy; instruments used include relative size/balance of power and geographic variables (distance, contiguity).
Identification strategy in the article:- Trade equation instrument: factors affecting barriers to trade but not direct military risk (e.g., PTAs; multilateral resistances; country/year fixed effects to soak up resistances).
Conflict equation instrument: relative size / balance of power; and in some specifications, distance is used in generating the instrument set.
Incorporation of gravity model elements to avoid omitted-variable bias that can masquerade as a liberal peace.
Key methodological point: If geography and size are omitted or mis-specified, trade can spuriously appear to proxy for omitted variables, biasing the estimated effect of trade on conflict.
Data and methods: what was analyzed
Data sources and scope:- Conflict data: MIDs (Keshk, Pollins & Reuveny 2004) and KN KR datasets; later extensions use fatal MIDs (Oneal & Russett 2005; Long 2008) and Sherman data variants.
Trade data: bilateral trade measures; Long (2008) gravity model data; Long’s indicators augmented with contemporary controls.
Size/power: CINC scores; real GDP; population; democracy indicators; major power indicators; continuous size measures (GDPs) used in refinements.
Geography: distance between capitals; contiguity; regional proximity; PTAs; language/currency-type barriers treated as multilateral resistances.
Core estimation approach:- Two-equation system to capture reciprocal causation between trade and conflict (two-stage least squares / Maddala-type procedures for dichotomous dependent variables in conflict equations).
Use of gravity-model-informed specifications to identify exogenous drivers of both trade and conflict.
Inclusion of country and year fixed effects in some specifications to soak up unobserved resistances and time trends.
Key modeling refinements used in this study:- Replace dichotomous major-power indicator with continuous size measures (sum of the logs of the two states’ GDPs).
Include log distance in the conflict equation’s instrument set; include contiguity as a separate geographic variable.
Use a richer conflict specification that includes balance-of-power, democracy levels, alliance ties, and system size.
In trade equation, include PTAs and multilateral trade resistances; country and year fixed effects to capture time- and country-specific resistance to trade.
For alternative gravity specifications, switch between directed vs non-directed (exports vs total bilateral trade) formulations to align with the reciprocal-analysis framework.
Use spline terms to model the years of peace since the last fatal MID, capturing non-linearities in the peace–trade relationship.
Detailed findings from the replicated KPR (2004) and KR (2005) analyses
Keshk, Pollins & Reuveny (KPR, 2004) replication with gravity model in conflict equation:- Baseline finding (KPR-style): conflict reduces trade; trade’s pacific effect weak or non-existent when simultaneity is addressed.
Add distance to the conflict equation (re-estimate using distance as part of the instrument set): the bilateral trade coefficient in the conflict equation becomes negative and significant, removing the previous spurious positive correlation between trade and conflict.
Distance and contiguity become highly significant in the conflict equation; geography matters for both trade and conflict.
The pseudo-R2 for the conflict equation increases when gravity terms are included (e.g., from .419 to .429 in one replication, with a substantial likelihood-ratio statistic).
Using GDPs (continuous size) instead of a major-power dummy dramatically improves fit and shifts the conflict equation toward expected liberal-peace signs.
The major-power indicator is criticized as crude; replacing it with a continuous size measure better captures real power asymmetries.
Kim & Rousseau (KR, 2005) replication and extension:- KR’s original results suggested that bilateral trade did not reduce conflict when examining reciprocal effects; conflict reduces trade and trade’s pacific effect disappears in certain specifications.
The continuous-size approach (GDPs) in the conflict equation yields substantial improvements: coefficient on initiator’s trade-to-GDP interdependence becomes negative and highly significant; size is robustly positively associated with use of force; continuous size improves model fit (pseudo-R2 from ~0.41 to ~0.63 in some specifications).
Replacing major-power indicator with continuous GDP sums changes the sign and significance of key coefficients, supporting liberal peace under gravity-model-informed specifications.
In a non-directed (overall bilateral trade) analysis, results align with a pacific effect when size and proximity are carefully modeled; using a continuous size measure eliminates some of the prior anomalies where democracies fight or where trade appears to increase conflict.
Overall interpretation from replication efforts:- When gravity-model factors (size and distance) are properly incorporated, trade and conflict exhibit the theoretically expected liberal-peace pattern: trade reduces conflict, conflict reduces trade.
Proximity and size drive both trade and conflict; omitting them leads to spurious proxies where trade seems to spur conflict or vice versa.
Democratic ties and alliances continue to correlate with lower conflict and higher trade, but the central liberal peace effect (trade reduces conflict) depends on correct specification.
New simultaneous analyses using gravity-informed specifications
Section: “New simultaneous estimates of the reciprocal effects of trade and fatal disputes” (Table III–IV)
Approach and changes from earlier work:- Re-estimate Oneal & Russett’s conflict equation with gravity-model controls for size and distance; adopt Long’s (2008) trade equation with enhancements.
Two-stage approach: use the instruments for trade (MID with fatalities; trade during various lags) and instruments for conflict (GDP-based size, distance, contiguity, PTA indicators, etc.).
Introduce country and year fixed effects in the trade equation to capture multilateral trade resistances; preserve gravity-model exogenous controls in conflict equation.
Replace CR: log trade (instrument) in trade equation; use log GDPs for each side, contiguity, distance, democracy, alliance structure, PTA, peace-year splines, and spline terms to capture non-linearities.
Key findings across Tables III and IV:- Table III (1950–2000): Trade reduces the incidence of fatal MID; coefficient on log trade (instrument) in the conflict equation is negative and significant (e.g., roughly -0.088 with fixed effects in some specifications when using Long-like data). Distance and size controls strengthen liberal peace evidence in the conflict equation.
When adding proxies for multilateral trade resistances in the trade equation, the contemporaneous adverse effect of conflict on trade remains significant (e.g., coefficient around -0.096 in some columns).
Joint democracy, alliances, and peace-years still influence trade and conflict in expected directions; proximity and size remain central.
The results show robust liberal-peace effects: trade reduces conflict; conflict reduces trade, even with controls for political relations and for domestic/interstate conflict risks.
Table IV (1984–1997) uses Long’s gravity framework with four changes (larger state CINC, log larger GDP, naïve probability of larger state winning, and non-directed trade). Findings: trade reduces fatal disputes; fatal MID contemporaneously reduces bilateral trade even under extensive controls for domestic conflict and third-party interstate conflicts; system size and political factors continue to shape outcomes.
Overall takeaway from Table III–IV:- When the gravity model is applied and controls for multilateral resistances and politics are included, the liberal peace holds robustly across different data periods and specifications.
The contemporaneous trade impact of conflict remains negative, and the contemporaneous impact of trade on conflict remains negative (i.e., trade reduces the likelihood of conflict).
The results are consistent with liberals’ intuition that economic interdependence constrains violence, provided that exogenous factors and geostrategic context are properly modeled.
Additional robustness checks and methodological notes
Fixing potential identification issues:- The authors acknowledge that their instrumental variables are not perfectly optimal; some instruments (e.g., PTA changes) may themselves be influenced by conflict (casus belli concerns).
They argue that despite imperfect instruments, their results are informative and robust to multiple model specifications, including dropping/altering democracy indicators and alliance variables.
Replication data and resources:- Replication materials and datasets are available at http://www.prio.no/jpr/datasets for researchers to inspect and reuse.
Implications of using different datasets:- Using KPR’s or KR’s original datasets can yield different conclusions if geography and size are not correctly modeled; the gravity approach reconciles discrepancies and aligns results with liberal peace expectations.
Implications and synthesis: what this means for liberal peace theory
Main empirical implication:- The pacific impact of trade on interstate conflict is supported once fundamental exogenous determinants—size (power) and distance (geography)—are incorporated through the gravity framework, and when resistances to trade are properly modeled.
Theoretical implications:- Supports Kantian/ liberal peace hypotheses: interdependence and openness reduce conflict, while conflict imposes costs on trade.
Highlights the importance of modeling architecture in empirical tests: endogeneity alone is not enough to overturn liberal peace; specification with gravity terms and robust instruments matters.
Practical implications:- Trade agreements (e.g., PTAs) and reducing transaction costs can contribute to peaceful dyadic relations, especially when combined with other regime characteristics (democracy, alliances).
Policymakers should consider the gravity-determined exogenous factors when evaluating trade policies and their potential peace dividends.
Limitations and future directions:- Instruments are not perfect; future work should seek stronger or more exogenous instruments to further isolate causal pathways.
Ongoing research could explore network effects, third-party dispute spillovers, and the role of multilateral institutions in shaping the trade–peace relationship beyond bilateral dyads.
Takeaway for exam-ready understanding:- The core claim of this literature position is that, when you account for how big countries are and how far apart they are, trade tends to promote peace and conflict tends to reduce trade, with contemporaneous effects on both sides. This remains true even after introducing robust controls and alternative specifications.
Quick reference: key equations and specifications (LaTeX)
Gravity-driven intuition for trade:
\quad Trade{ij} \propto \frac{GDPi \, GDPj}{Distance{ij}^{\beta}} \quad (or\, \, Trade{ij} = A \cdot GDPi^{\alpha} GDPj^{\gamma} \, Distance{ij}^{-\beta} \, \, \text{times other factors})
Log-linear representation commonly used:
\quad \log Trade{ij} = \beta0 + \beta1 \log GDPi + \beta2 \log GDPj + \beta3 \log Populationi + \beta4 \log Populationj + \beta5 \log Distance{ij} + \dots + \epsilon_{ij}
Conflict model (two-equation simultaneous framework, schematic):- Conflict equation (MIDs or fatal MIDs as dependent variable):
\quad Conflict{ij,t} = f( Trade{ij,t}, \text{Size}i, \text{Size}j, Distance{ij}, Contiguity{ij}, Democracyi, Democracyj, PTA{ij,t}, SystemSizet, PeaceYear{t}, \text{Splines}, \text{Other controls}) + u{ij,t}
Trade equation (bilateral trade as dependent variable):
\quad Trade{ij,t} = g( Trade{ij,t-1}, Sizei, Sizej, Distance{ij}, Contiguity{ij}, PTA{ij,t}, Democracyi, Democracyj, CountryYearFE, \text{Multilateral resistances}, \text{KPR/KR instruments}) + v{ij,t}
Instruments (representative components):- For trade equation: Preferential Trade Agreements (PTAs), multilateral trade resistances, country and year fixed effects to proxy resistances.
For conflict equation: relative size / balance of power (e.g., CINC-based measures), distance, contiguity, and democracy indicators; in some specifications, the logarithm of distance is included among the instruments for trade while distance is excluded from the conflict equation to avoid endogeneity in instruments.
System-size and splines (example placeholders):
\quad SystemSizet \text{ (number of states in system in year } t) \quad Spline1t, \, Spline2t, \, Spline3t \text{ represent the peace-year splines used to capture non-linear effects of years since last MID}
Replication and data notes
Replication materials available at: http://www.prio.no/jpr/datasets
The study emphasizes that replication with alternative data and model specifications is crucial for assessing the robustness of the liberal peace claim.
Hegre, Oneal, & Russett (2010) Study Questions
What is the research question?
The research question is: "Do trade and conflict influence each other reciprocally? Do they promote peace (liberal peace) or not?"
What are the key independent and dependent variables in the study?
Independent Variables: Bilateral trade volume, country size/power (e.g., CINC, GDP, population), geographic factors (distance, contiguity), political factors (democracy levels, alliance ties, PTA membership), multilateral resistances, system size, and peace-year splines.
Dependent Variables: Militarized Interstate Disputes (MIDs) and fatal MIDs.
What do the authors mean by “liberal (or commercial) peace”?
By "liberal (or commercial) peace," the authors refer to the theory that increased interdependence (especially through trade), strong international institutions, and democratic governance significantly reduce the likelihood of interstate conflict.
What is the difference between the “liberal peace” and “democratic peace” theories?
Liberal Peace (Commercial Peace): A broader concept emphasizing how interdependent economic relations (such as trade) and international institutions create mutual interests and raise the costs of conflict, thus promoting peaceful interactions.
Democratic Peace: A more specific hypothesis, often considered a subset of liberal peace, which posits that democracies are less likely to engage in conflict with other democracies due to shared norms of peaceful dispute resolution and institutional constraints on warmaking.
Why is the inclusion of size and geography in the analysis relevant when exploring the relationship between trade and peace?
The inclusion of size and geography (e.g., distance, contiguity) is crucial because these are exogenous factors that inherently shape both trade volumes and the likelihood of conflict. Omitting or mis-specifying these variables can lead to omitted-variable bias, where trade might spuriously appear to cause conflict (or vice-versa) simply by proxying for the unmodeled effects of country size and geographic proximity. Properly accounting for them clarifies the true causal relationship between trade and peace.
What is the main finding in the analysis?
The main finding is that when fundamental exogenous determinants like country size and geographic distance are properly incorporated into the analysis using a gravity model, and when resistances to trade are adequately modeled, the pacific effect of trade on interstate conflict is strongly supported. Essentially, trade reduces conflict, and conflict contemporaneously reduces trade.
What do the authors mean by the following excerpt? o “Whether paid prospectively or contemporaneously, the economic cost of conflict should reduce the likelihood of military conflict, ceteris paribus, if national leaders are rational.”
This excerpt means that if national leaders are rational—meaning they calculate costs and benefits to make decisions—they will anticipate the significant economic costs associated with military conflict. These costs, whether they are immediate and incurred during the conflict ("contemporaneously") or are expected in the future once the conflict starts ("prospectively"), should act as a deterrent. "Ceteris paribus" signifies "all other things being equal"; thus, assuming all other factors remain constant, the economic burden of war should make leaders less inclined to