rogercooper.com

The cliometrician, Peter Turchin, recently posted on his substack, https://peterturchin.substack.com/p/is-ukraine-russia-war-entering-the, an analysis of wars of attrition that is seriously flawed. You can find more detail in his preprint https://osf.io/preprints/socarxiv/k5b42_v1

Turchin has used the Correlates of War datasets (https://correlatesofwar.org/) to find 7 modern wars of attrition. He concluded that once a critical point is passed at battle deaths between 2 and 3% of the total population, the nation will be defeated, unless their foe is in equally bad shape, which results in stalemate.

The most obvious flaw in the analysis is that it excludes the winners of wars. You can’t understand wars without including the winners. Here is the data, using the same datasets that Turchin used, with the winners included.

Attrition

War	Start Year	End Year	State	Outcome	BatDeath	Pop	C%
American Civil War	1861	1865	CSA	Loss	258	8.7	2.97%
American Civil War	1861	1865	USA	Win	360	23.6	1.53%
World War One	1914	1918	France	Win	1385	33.2	4.17%
World War One	1914	1918	Germany	Loss	1773	67	2.65%
WW2-Europe	1939	1945	Germany	Loss	3500	79	4.43%
WW2-Europe	1939	1945	USSR	Win	7500	197	3.81%
WW2-Pacific	1941	1945	Japan	Loss	1740	73	2.38%
WW2-Pacific	1941	1945	USA	Win	405	133	0.30%
Korean War	1950	1953	North Korea	Stalemate	316	10.5	3.01%
Korean War	1950	1953	South Korea	Stalemate	113	19	0.59%
Vietnam War	1965	1975	North Vietnam	Win	700	19	3.68%
Vietnam War	1965	1975	South Vietnam	Loss	254	37.5	0.68%
Iran-Iraq War	1980	1988	Iran	Stalemate	750	38.5	1.95%
Iran-Iraq War	1980	1988	Iraq	Stalemate	500	13.6	3.68%

Battle deaths are in thousands, population in millions.

So how does Turchin’s hypothesis hold up. Statistically, it does not do well. The correlation between casualty percent and loss/stalemate is actually -.08, not statistically significant and actually indicates the lower casualties are slightly more likely to be associated with loss/stalemate. Counting stalemate as 0.5 or using relative casualties does not give better results.

To see what happened, let’s look at the individual wars.

American Civil War actually follows his hypothesis. The Confederates break at just under 3% with Union casualty rate at half the Confederate. There is disagreement among leading military historians of how important attrition was, but all say it was a factor.

World War One violates his hypothesis. The French had higher casualty rates and did not break. Of the seven wars listed, the French had higher casualty rate than any losed except Germany in WW2. Turchin in a post suggested that the French had higher asabiya (social cohesion) than the Germans but that was not part of his hypothesis. I also doubt that anyone in 1914 would have thought the Third Republic (with divisions revealed by the Dreyfus affair) was more cohesive then the Second Reich. Most historians cite the blockade and the demoralizing impact of American entry for Germany’s collapse.

World War Two in Europe does not fit his hypothesis. Both Russian and German losses exceeded 3% and German losses were as a percent, not much higher than Russia’s (4.43% to 3.81%). Under his hypothesis, stalemate was to be expected. If the war had just between Russia and Germany, that might have happened. It should be noted that the totalitarian regimes on both sides reduced the sensitivity to losses.

World War Two in the Pacific fits quite well (even though it ignores China). However, the Japanese did not surrender until the use of the Atomic Bomb rendered resistance pointless.

Korean War violates his hypothesis. The losses of South Korea were quite low. If the United States is considered the antagonist the percentage is even lower. Turchin listed North Korea, if China was seen as the main participant its losses were 422,000 compared to a population of 572,000,000 (0.07%). However, the reluctance of China to continue the war is understandable if you include the casualties of the immediately proceeding Chinese Civil War and the Second Sindo-Japanese War.

Vietnam War is another war the victor took much higher casualties than the loser. The asabiya argument looks much better here as South Vietnam was ruled by its small Catholic minority, but it still not part of the hypothesis. The Vietnam War is an outlier in the data but we can’t just arbitrarily exclude outliers (https://en.wikipedia.org/wiki/No_true_Scotsman)

Iran-Iraq War fits the hypothesis as the Iranians had taken substantial casualties, although lower than the Iraqis. Both states had low asabiya.

So of the seven wars he covered, 3 fit the hypothesis and 4 did not. It is no surprise that the statistical correlation was non-existent.

The real surprise is that Turchin, who has done some excellent work would make such an error. Part of the problem is analysis by eyeballing, even the values he used did not really “cluster tightly”.

The more substantial problem is what I call “FakePolitik”. It sounds smart to say that wars are always won by the bigger army/economy/population and dismiss those who disagree with you as idealists/romantics/fools. You can even create a simulation to “prove” it. However, examining the Correlates of War database or the Wikipedia war list will show you that many wars are lost by the “bigger” side for a variety of reasons.