A recent article using the new Correlates of War (COW) data on the distribution of interstate, intrastate, and extrastate wars from 1816 to 1997 claims there was a relatively constant risk of death in battle during that time. We show that the authors' information is skewed by irregularities in the COW deaths data, and contest their pessimistic interpretation. Using revised information on battle deaths from 1900 to 2002 we demonstrate that the risk of death in battle by no means followed a flat line, but rather declined significantly after World War II and again after the end of the Cold War. Future users should note that the deaths data collected for the three conflict types by COW are not comparable, and using them as such tends to underestimate the share of fatalities due to major interstate conflicts.
Note to the Replication Data
The absolute numbers of battle-deaths in Figure 4 (The Risk of Death in Battle Worldwide, 1900-1997) are also found here, along with documentation of the coding decisions. Go to 'Dataset for use with the Correlates of War Dataset' and the corresponding 'Documentation of Coding Decisions: Correlates of War Data'. An update of the battle deaths data for 1946-2008 was completed in October 2009 and is found here.