It's all how you count settlements. From data used by RAND in this monograph, which I used for my graduate thesis:
RAND considered 89 post-1945 insurgencies. I removed the 16 ongoing insurgencies in the dataset and came up with:
# Pct Avg Length
All N 73 12.25
Govt Win 28 38% 14.64
Political Settlement 19 26% 9.89
Govt Loss 26 36% 11.38
Be aware the standard deviation of the length is very, very high, 10.46.
So wins and losses in this dataset are about even, with political settlements making up the rest.
"Insurgent Win" is all about how you define the middle group - settlements. Lyall and Wilson count settlements as insurgent victories (using a different dataset, 1800-2005), while RAND counts them separately. (Full Disclosure: I have co-authored a response essay to Lyall and Wilson's very flawed study and submitted it to International Organizations)
It should be noted once you actually try and code "win" and "loss" with insurgencies, you dive into a subjective morass. For example, RAND cites El Salvador as a "settlement", while some COIN literature calls it a win. The areas of grey are what kill you, because some settlements favor the govt more and some favor the insurgent, but neither are clear victories. A lot of Solomonesque baby splitting in the coding. Hence, the writer can make the data say what he wants based on the use of the "middling" outcomes. Also when you introduce the 19th century, governments tend to win more. My hypothesis for that is that in the pre-global media age, they could be more genocidal and ruthless than today with less impact. (ex. Native Americans, Boer war, etc.) Another observation is that no democratic regime in the dataset has lost an insurgency. (note - regime experiencing the insurgency, not the host nation, see this thread)
Niel
Bookmarks