These are a modified version of the inter-state war data from the Correlates of War project. Data are version 4.0. The temporal domain is 1816-2007. Data are functionally directed dyadic war-year.
Format
A data frame with 1932 observations on the following 15 variables.
warnumthe Correlates of War war number
ccode1the Correlates of War state code for side1
ccode2the Correlates of War state code for side2
yeara numeric vector for the year
cowinteronseta dummy variable for whether this is an inter-state war onset (i.e. either the year in
StartYear1orStartYear2in the raw data)cowinterongoinga numeric constant of 1
sidea1a numeric vector for the side in the war for
ccode1, either 1 or 2sidea2a numeric vector for the side in the war for
ccode2, either 1 or 2initiator1a dummy variable that equals 1 if
ccode1initiated the warinitiator2a dummy variable that equals 1 if
ccode2initiated the waroutcome1the outcome for
ccode1as numeric vector. Outcomes are 1 (winner), 2 (loser), 3 (compromise/tied), 4 (transformed into another type of war), 5 (ongoing at end of 2007, which is not observed in these data), 6 (stalemate), 7 (conflict continues below severity of war), and 8 (changed sides)outcome2the outcome for
ccode2as numeric vector. Outcomes are 1 (winner), 2 (loser), 3 (compromise/tied), 4 (transformed into another type of war), 5 (ongoing at end of 2007, which is not observed in these data), 6 (stalemate), 7 (conflict continues below severity of war), and 8 (changed sides)batdeath1the estimated deaths for
ccode1(-9 = unknown)batdeath2the estimated deaths for
ccode2(-9 = unknown)resumea dummy variable that equals 1 if this is a conflict resumption episode
Details
See data-raw directory for how these data were generated. These data are here if you want it, but I caution against using them
as gospel. There are a few problems here. One: -9s proliferate the data for battle deaths on either side, which is unhelpful. There are 10 cases where the sum
of battle deaths is exactly 1,000 or 1,001. This is suspicious. The "side" variables are not well-explained---in fact they're not explained at all in the codebook---
and this can lead a user astray if they want to interpret them analogous to the sidea variables in the Correlates of War Militarized Interstate Dispute
data. You probably want to use the initiator variables for this. Further, the war data routinely betray the MID data and the two do not speak well to each other. The language Sarkees and Wayman (2010) use in their book
talk about how MIDs "precede" a war or are "associated" with a war, which forgets the war data are supposed to be a subset of the MID data. In one case (Gulf War),
they get the associated dispute number wrong and, in one prominent case (War of Bosnian Independence), they argue no MID exists at all (it's actually MID#3557).