Militarized Interstate Dispute Analysis
- Introduced by
- Daniel Jones, Stuart Bremer & J. David Singer (Correlates of War project)
- Year
- 1996
- Type
- Coding and statistical analysis of interstate militarized confrontations
- Family
- Observational quantitative IR / conflict studies
What does it do?
Militarized interstate dispute (MID) analysis is the coding and quantitative study of confrontations in which one state threatens, displays, or uses military force against another.
Built on the Correlates of War project's MID dataset and the coding rules codified by Jones, Bremer, and Singer (1996), it provides the standard observational measure of interstate conflict short of and including war, structured as dyad-years so that the onset, escalation, and outcomes of disputes can be modeled statistically across two centuries of the international system.
The intuition behind the method
War is rare, but the threat and use of force short of war are common, and studying only full wars would discard most of the relevant variation. MID analysis widens the lens to every militarized confrontation between states — a mobilization, a border clash, a naval blockade — and records its severity on a graded ladder from threat to use of force to war. By organizing these into pairs of states observed year by year, it lets researchers ask which conditions make a dispute more likely to start and more likely to escalate, turning the messy history of interstate hostility into a dataset on which hypotheses about the causes of war can be tested.
History
The MID dataset is the conflict-measurement backbone of the Correlates of War (COW) project, founded by J. David Singer and Melvin Small in the 1960s to study war scientifically. The 1996 article by Daniel Jones, Stuart Bremer, and J. David Singer in Conflict Management and Peace Science laid out the rationale, coding rules, and empirical patterns of disputes from 1816 to 1992, defining the modern MID standard. Successive versions extended and refined the data, and MIDs became the dependent variable in a vast literature on the democratic peace, deterrence, and the causes of war.
When to use it
Use MID analysis when the research question concerns the causes, escalation, or resolution of interstate conflict short of and including war, and you want a carefully coded, theory-driven measure spanning the modern state system. It is the field standard for testing dyadic theories of conflict — the democratic peace, power preponderance, alliance reliability. It is less suited to intrastate conflict (which has its own datasets), to high-frequency or real-time monitoring (where automated event data are better), or to questions requiring the contextual nuance that the coding scheme deliberately abstracts away.
Limitations and considerations
- Hand coding is labor-intensive and updated only periodically, so the data lag current events and cannot support real-time monitoring.
- Restricted to interstate disputes, omitting civil wars, terrorism, and non-state violence that other datasets cover.
- Coding choices (what counts as a display of force, how incidents aggregate) embed judgments that affect counts and have been revised across dataset versions.
- The all-dyad-years design generates enormous numbers of non-conflict observations, raising rare-events and temporal-dependence problems that require statistical correction.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Provides a carefully validated, theoretically grounded measure of interstate conflict across nearly two centuries.
- The graded hostility scale supports analysis of escalation, not merely the binary onset of war.
- The dyad-year structure cleanly aligns conflict with dyadic and monadic covariates for hypothesis testing.
- Widely adopted and documented, enabling cumulative, replicable research and direct comparison across studies.
Applications
- Testing the democratic peace by relating regime type to the probability of dispute onset.
- Evaluating whether capability preponderance or parity is more conflict-prone between states.
- Studying whether and how alliances deter or provoke militarized confrontations.
- Examining the role of territorial contiguity and disputed borders in conflict initiation.
- Serving as the canonical dependent variable for dyadic models of interstate conflict and escalation.
How it works
How it works
A militarized interstate dispute is defined as a set of incidents in which the threat, display, or use of force by one state is explicitly directed at the government, official representatives, forces, property, or territory of another state. The coding rules specify what qualifies, how incidents combine into a dispute, and the temporal and actor boundaries — the conceptual core that makes counts comparable across cases and time.
Each dispute and each participant is coded for a hostility level on an ordinal scale — from no militarized action, through threat to use force, display of force, and use of force, up to interstate war — along with start and end dates, initiator and target roles, fatalities, and outcome. This graded coding is what allows escalation, not just onset, to be studied.
Formula
\text{hostility level} \in \{1{:}\text{none},\,2{:}\text{threat},\,3{:}\text{display},\,4{:}\text{use},\,5{:}\text{war}\}The coded disputes are reorganized into the dyad-year, the standard unit of analysis: every pair of states is observed in every year, with an indicator of whether a MID began (onset) and at what hostility level. This structure aligns the conflict data with dyadic covariates such as capabilities, alliances, contiguity, and regime type.
Formula
D_{ijt} = \big(\text{dyad}(i,j),\, t,\, \text{onset}_{ijt},\, \text{hostility}_{ijt}\big)The probability that a dispute begins in a given dyad-year is modeled, classically with logit or probit regression (and corrections for the rarity of conflict and for temporal dependence), as a function of dyadic predictors. The estimated coefficients test theories — democratic peace, capability balance, alliance ties — about what drives interstate conflict.
Formula
\Pr(\text{MID}_{ijt}=1) = \Lambda\!\big(\mathbf{x}_{ijt}'\boldsymbol{\beta}\big)Common pitfalls
- Analyzing dyad-years with ordinary logit while ignoring rare-events bias and temporal dependence, producing misleading inferences.
- Using the politically relevant dyad subset versus all dyads without justifying the choice, which can change results substantially.
- Mixing dataset versions (MID 2.1, 3.0, 4.0, 5.0) within a study, since coding rules and the universe of cases differ across releases.
- Treating a dispute's hostility level as a clean measure of intensity when it reflects the single highest action, not the overall severity.
Frequently asked questions
Why is MID analysis classified as an observational design rather than a statistical model?
Because its defining contribution is a measurement and research-design strategy: rigorously defining and coding militarized confrontations and organizing them into dyad-years so that observational comparisons across pairs of states and time become possible. The regression models applied afterward are tools borrowed from elsewhere; the distinctive method is the observational design — the coding rules and the dyad-year framework — that makes those models interpretable.
What is a 'dyad-year' and why is it the unit of analysis?
A dyad-year is a single pair of states observed in a single year. It is the standard unit because most theories of interstate conflict are about relationships between two states (their relative power, alliance ties, joint regime type), and observing every pair every year lets researchers model the yearly probability of a dispute as a function of those dyadic conditions. It also creates the methodological challenges of rare events and temporal dependence.
How does MID data relate to automated conflict event data?
They are complementary. MID data are hand-coded, low-frequency, carefully validated records of interstate confrontations involving force, optimized for theoretical precision. Automated event data (KEDS/TABARI, ICEWS, GDELT) are machine-coded, high-frequency, broad-spectrum records from news, optimized for scale and timeliness. Researchers often use event data for monitoring and forecasting and MID data for cumulative theory testing. See the related Event Data Analysis of Conflict entry.
Related methods
Sources
Jones, D. M., Bremer, S. A., & Singer, J. D. (1996). Militarized interstate disputes, 1816–1992: Rationale, coding rules, and empirical patterns. Conflict Management and Peace Science, 15(2), 163–213.
Open DOI referenceCopied.