Thread by @ProfPaulPoast

1/ In my "Quantitative Security" grad class, students "deconstruct" an observation from a widely used dataset: they pick an observation and see if its coding makes sense.

👇I walk through an observation that I deconstructed as a class demo.

2/ The observation I deconstructed was the Correlates of War (correlatesofwar.org) interstate war dataset coding of the 1999 Kosovo war (BTW: the 20th anniversary of that war is next month)

en.wikipedia.org/wiki/Kosovo_War

3/ I don't ask the students to explore everything that is coded about a observation. Instead, they should pick one part of the coding to deconstruct.

In this case, I focus on COW's coding of the Kosovo war's participants. Here is the coding:

4/ Looking at the "Side" column, notice that the war has 7 countries on one side v Serbia (labeled as "Yugoslavia") on the other.

5/ On the one hand, it makes sense that the one side had multiple countries. After all, it was a @NATO campaign (Operation Allied Force).

6/ On the other hand, things aren't quite so simple.

This is evident by turning to this war's entry in the excellent CQ Press (@SAGECQPolitics) book by Meredith Sarkees and Frank Wayman "Resort to War", which chronicles the COW narratives for each war:

amazon.com/Resort-War-181…

7/ The book's Kosovo war entry includes a "Coding" note. The note acknowledges how the COW team held lengthy discussions about who to include as participants. They opted to code the individual @NATO members, rather than @NATO itself, as the participants.

8/ But notice a key phrase in their description: "the overall dominance of the United States". Others picked up on this idea, namely Dani Reiter, Allan Stam, and @mchorowitz.

9/ They wrote this JCR piece in 2016 detailing their effort to recode all of the COW inter-state wars.

journals.sagepub.com/doi/abs/10.117…

10/ A war that received a lot of attention in their paper was the 1999 Kosovo war, namely the identification of participants.

They write that identifying the participants is tricky because COW is inconsistent in the handling of naval crews and aerial personnel

11/ Reiter, Stam, and @mchorowitz make the judgement call to exclude support personnel, naval personnel, and aerial personnel.

12/ The end result is that they code the 1999 Kosovo as a bilateral war between the United States and Serbia (which, again, is labeled "Yugoslavia")

13/ In a supplemental coding appendix, they offer more details on the decision

14/ Was that the right decision? It's not unreasonable, but it would be news to @NATO to find out that no other members actually "participated" in the war!

15/ More generally, this case highlights how our existing procedures for coding "war participation" are probably inadequate.

What if an actor supplies ALL of the weapons, but doesn't actively fight?

What if an actor pays for the entire war, but doesn't fight?

16/ These are questions I'll address in a future thread (though I've addressed them a bit in my #ModernMajorPowerWar threads).

But for now, this case demonstrates well why my students should not take conflict data at face value!

(end)