Why This Question Matters
Scholars and policymakers rely on long-run trends in combat fatalities to assess whether the world is becoming more or less deadly. A recent article that used Correlates of War (COW) data from 1816–1997 argued that the risk of death in battle was roughly constant across that period, a finding with grim implications for claims about a long-run decline in organized violence.
What Bethany Lacina, Nils Petter Gleditsch, and Bruce Russett Did
The authors reexamine the underlying battle-death information and the interpretation of those earlier results. Focusing on revised estimates of battle deaths covering 1900–2002, they diagnose irregularities in the COW deaths series that bias conclusions when different conflict types are treated as directly comparable.
How They Reanalyzed the Evidence
- Reassessed the COW deaths data used in the prior article and identified inconsistencies across the three conflict-type series (interstate, intrastate, extrastate).
- Employed revised battle-death information for the 1900–2002 period and re-calculated the risk of death in battle over time.
Key Findings
- The apparent ‘‘flat line’’ in the earlier work is an artifact of skewed and non-comparable death-series in the COW dataset.
- Using revised death estimates, the risk of death in battle declines substantially after World War II and declines again following the end of the Cold War.
- Because the COW death series for different conflict types are not comparable, analyses that treat them as equivalent tend to understate the share of fatalities attributable to major interstate wars.
Why This Matters for Research and Policy
These results challenge a pessimistic reading of long-run battlefield lethality and underscore how measurement choices shape inferences about trends in organized violence. Lacina, Gleditsch, and Russett caution researchers to check comparability and provenance of death-series in conflict datasets—misuse can bias both scholarly conclusions and policy discussions about global security trends.
