💥

Lecture: A Quantitative Approach to Conflict (11:03)

Nov 19, 2025

Overview

The speaker introduces a large-N dataset in international relations to examine how conflict type relates to death estimates using basic R linear modeling.

Large-N Data and Research Framing

  • Large-N data: 1,848 observations; contrasts with single case studies.
  • Substantive focus: US foreign policy, national security, international relations.
  • Research question: Which conflict type yields more deaths on average?
  • Variables:
    • Independent variable: Conflict type (three categories).
    • Dependent variable: Death estimate (numerical).

Conflict Types (Concepts and Examples)

  • Internal: Government vs. rebel group (e.g., Colombia).
  • Internationalized: Government vs. rebel group supported by an outside power (e.g., Sierra Leone; Nicaragua Contra War).
  • Interstate: Conventional conflict between states (e.g., Ukraine and Russia).

Modeling Approach (R / RStudio)

  • Environment: R (statistical language) with RStudio interface.
  • Model: Linear model (lm) with dependent ~ independent.
  • Specification: death_estimate ~ conflict_type, data = “National Security” dataset.
  • Factor handling: Internal used as baseline because of alphabetical order.

Findings and Interpretation

  • Overall significance: Conflict type significantly predicts death estimates.
  • Coefficients (relative to internal baseline):
    • Interstate: Large positive estimate; highest t-value; indicates largest increase.
    • Internationalized: Positive estimate; increase relative to internal.
  • Practical interpretation:
    • Interstate conflicts have the highest average deaths.
    • Internationalized conflicts also exceed internal in average deaths.
    • All pairwise differences are statistically significant (p < .05).

Pairwise Comparisons (AMeans Function)

  • All conflict-type pairs differ significantly (p < .05).
  • Sign directions:
    • Internal vs. Internationalized: Negative for internal baseline; fewer deaths than internationalized.
    • Internal vs. Interstate: Negative for internal baseline; fewer deaths than interstate.
    • Internationalized vs. Interstate: Negative for internationalized baseline; interstate has more deaths.
  • Policy relevance: Interstate conflicts tend to entail markedly higher death tolls.

Model Output Highlights (Quantitative Details)

  • Reported estimates (relative to internal):
    • Interstate ≈ +4,346 average deaths.
    • Internationalized ≈ +1,430 average deaths.
  • T-values: Very high for interstate; both positive coefficients significant.

Structured Summary

Conflict Type ComparisonDirection vs. First ListedRelative DeathsStatistical Significance
Internal vs. InternationalizedInternal lowerInternationalized > Internalp < .05
Internal vs. InterstateInternal lowerInterstate > Internalp < .05
Internationalized vs. InterstateInternationalized lowerInterstate > Internationalizedp < .05
Internal (baseline) → Internationalized (coef)Positive≈ +1,430Significant
Internal (baseline) → Interstate (coef)Positive≈ +4,346Significant

Key Terms & Definitions

  • Large-N dataset: Data with many observations enabling statistical inference.
  • Independent variable: Predictor; here, conflict type.
  • Dependent variable: Outcome; here, death estimate.
  • Linear model (lm): Regression estimating outcome as a function of predictors.
  • Baseline (reference) category: Factor level against which others are compared.
  • T-value: Test statistic assessing coefficient significance.
  • Standard error: Measure of estimate precision.
  • p-value: Probability indicating statistical significance (commonly p < .05).

Action Items / Next Steps

  • Replicate lm model with death_estimate ~ conflict_type using the provided dataset.
  • Review factor levels to confirm baseline and interpret coefficients correctly.
  • Examine model diagnostics for assumptions and robustness.
  • Extend analysis: Control for region, time, or conflict duration if available.
  • Translate findings to policy memos on expected lethality by conflict type.