Read on Twitter
Quantitative data on military alliances in IR can be really bad. Just eyeballing the data finds major problems.
A few fun examples from the Correlates of War Formal Alliances dataset below ...
https://correlatesofwar.org/data-sets/formal-alliances
A few fun examples from the Correlates of War Formal Alliances dataset below ...
https://correlatesofwar.org/data-sets/formal-alliances
Some definitions first:
"A defense pact (Type I) is the highest level of military commitment, requiring alliance members to come to each other’s aid militarily if attacked by a third party."
OK. Cool.
"A defense pact (Type I) is the highest level of military commitment, requiring alliance members to come to each other’s aid militarily if attacked by a third party."
OK. Cool.
Apparently, Russia and Germany signed a mutual defence pact in 1905. Wait, what? Oh, you mean the Treaty of Björkö--a secret agreement signed by the Kaiser and the Tsar that their countries ended up rejecting and never ratifying. That should not belong here.
At first I thought this agreement was the Salonika Agreement (not a defence pact) but it turns out to be the Nyon Conference (not a defence pact). This agreement tried to address attacks on int'l shipping in the Mediterranean in the context of the Spanish Civil War, but whatever.
Evidently, the Soviet Union and the Baltic states concluded a set of defence pacts in 1939, which is a funny way of describing military occupation. Nope, these "agreements" should not belong here either.
I can't figure out what defence pact East Germany and Romania agreed to establish on 12 May 1972 and what relationship this agreement had with the Warsaw Pact. In fact, East Germany signed other bilateral defence pacts in 1967 with Poland, Czechoslovakia, and Hungary. OK ...
A discussed merger between Syria and Libya is also coded as a defence pact, one that lasted three months in 1980 according to the data. Obviously, this proposed union state that never was is comparable to NATO.
I can go on, but what does this all mean?
- The historical knowledge here is highly suspect.
- Lots of apples and oranges in the COW alliance data.
- Definitely do not take these data off-the-shelf and unquestioningly run regressions. Very little to learn here in so doing.
- The historical knowledge here is highly suspect.
- Lots of apples and oranges in the COW alliance data.
- Definitely do not take these data off-the-shelf and unquestioningly run regressions. Very little to learn here in so doing.
I am sympathetic to the enterprise and do admire the effort at transparency here. Sincerely. However, these data have been around for years and I am not sure why these mistakes are still in the books. I am very concerned about research that have drawn on these data. END.
