Thread: Debunking The 10:1 Ratio of Forces in COIN

Originally Posted by n21

I would be interested to know where you got that number from?

I think it was an article or post I read on BRF some years back.

Interesting study. I wouldn't argue with his data although I've always been a bit suspicious of any concrete assertions about such a fluid, fickle and frustrating type of warfare, especially if its prosecution is controlled by the type of people who need concrete assertions-- i.e. assurances, time tables, force limits, budget restrictions, poll results, exit strategies and restrictive ROE and all the rest of the CYA data-- before they can make a decision.

2. questions from an outsider:

1. Are the insurgencies from the timeframe used in the study really that simple to compare? Is there not a huge difference in the success chance of those insurgencies after automatic weapon become widely and cheaply available?

2. Would not the first answer to the 10:1 ratio be: "You might not need so much manpower, but it certainly helps?"

Finally got time to sit down and not only read but digest.

These documents never seem to relate to me the training/skill level/ability of the COIN troops, which I think is paramount in determining the levels needed as can be seen (IMO) from some of the data in the report;
e.g. Ethiopia (73-76)
and to some extent Algeria (54-62), (yes I know DeGaule made all the right noises when withdrawing, but still I cannot see it as a successful mission, even with the Foriegn Legion present.)

Never will we be 100% accurate in these estimations, but meticulous planning combined with excellent training IMO can indeed reduce the numbers required to defeat an insurgency.
That said how do we measure the big variable?
"THE WILL NOT TO BE DOMINATED AT ANY COST ......and there are a few groups out there with this will to survive.

1. Statistical analysis cannot "prove" anything. It can be used to conduct hypothesis testing and at determined thresholds, and you then either reject or fail to reject the hypothesis.

2. The author's choice to exclude 2 data points is extremely questionable. Both observed data points are within 2 standard deviations, which is typically the point at which you can start to think about classifying them as outliers, and by excluding them, you are cutting an already small sample size by 20%.

3. When the two data points are included, the "magical" ratio of 10:1 falls within the 95% confidence interval of the observed mean of the data set, meaning that you would reject the hypothesis that the ratio is < 10:1.

4. By controlling for only troops #s (with no analysis of how solid these #s are or potential biases), the study is not robust.

Bottomline, it does not prove that the 10:1 ratio - it actually rejects a hypothesis that the ratio is less than 10:1. However, these results are very fragile.

"Both observed data points are within 2 standard deviations, which is typically the point at which you can start to think about classifying them as outliers, and by excluding them, you are cutting an already small sample size by 20%."

Korea was excluded from the "Insurgent" category while El Salvador was excluded from the "Counter-Insurgent" category. With twenty-two total data points, isn't that only about 9%?

Originally Posted by S-2

"Both observed data points are within 2 standard deviations, which is typically the point at which you can start to think about classifying them as outliers, and by excluding them, you are cutting an already small sample size by 20%."

Korea was excluded from the "Insurgent" category while El Salvador was excluded from the "Counter-Insurgent" category. With twenty-two total data points, isn't that only about 9%?

Steve,
He tries to establish both a floor (the threshold at which you will fail) and a ceiling (the threshold above which you don't need more manpower to win). In evaluating the statistical validity, you have to evaluate each calculation. For the ceiling, both El Savador and Sri Lanka are dropped from the sample. To be exact, you're cutting the sample size by 18% (2/11), which is quite a large number to drop as outliers.

For any other stats buffs out there, here's some Stata output of the counterinsurgent cases.

"For the ceiling, both El Savador and Sri Lanka are dropped from the sample."

This is correct. My error. I don't know how I counted Sri Lanka as part of the sample total yet missed that Thiel had excluded it as an outlier. Selective color-blindness I guess.

Originally Posted by S-2

"For the ceiling, both El Savador and Sri Lanka are dropped from the sample."

This is correct. My error. I don't know how I counted Sri Lanka as part of the sample total yet missed that Thiel had excluded it as an outlier. Selective color-blindness I guess.

Steve,
No sweat. I like the paper's logic in trying to find a floor and ceiling, but it's statistical application and inferences is simply wrong. Also, I'm not familiar enough with the Correlates of War dataset to know why he chose a random sample instead of the full population of post-1938 insurgencies, but it appears like it doesn't contain the size of the two forces - having more observations would help make the findings more robust, although I doubt by very little because the findings wouldn't control for other variables that would impact the results. Also, what's interesting is that 2/3rds of the references he provides to the 10:1 ratio aren't specific to manpower, but to a more general "expenditure."

I'm not a statistics guy but I am a COIN guy. It seems to me that if you try to put any "reliable" data on COIN operations, opposing force levels are just one column on a chart of many. Some of you have pointed out variables. Skill, weapons, equipment, terrain, climate, logistics, money, leadership, local politics, national politics, tribal and religious influences, morale, the role of border states or supporting governments, determination and commitment, etc...Some of these are hard to measure but have impact. Pretty sure I left out a few things out.

Originally Posted by Red Seven

I'm not a statistics guy but I am a COIN guy. It seems to me that if you try to put any "reliable" data on COIN operations, opposing force levels are just one column on a chart of many. Some of you have pointed out variables. Skill, weapons, equipment, terrain, climate, logistics, money, leadership, local politics, national politics, tribal and religious influences, morale, the role of border states or supporting governments, determination and commitment, etc...Some of these are hard to measure but have impact. Pretty sure I left out a few things out.

This is what I meant by the fact that the results are fragile - there's just so many other variables to consider. I don't think that coming up with a ratio as a rule of thumb is a bad idea, but it's not a panacea that can replace good strategy that attacks the reasons for an insurgency.