20th Century Atlas -- Wars and Atrocities -- Margin of Error in Casualty Statistics

How accurate are casualty statistics?

The wonderful thing about computers is that you can get them do all sorts of nifty calculations even if you don't know what, exactly, they are calculating. Here, for example, are the Standard Deviations of my collection of experts.

Event	Year Began	Mean Estimated Death Toll (millions)	Standard Deviation	Confidence Level
Congo Free State	1885	7.9	88%	guess
Brazilian Indians	1900	.6	54%	uncertain
Mexican Revolution	1910	1.0	61%	guess
WW1: Civilians	1914	8.5	33%	debatable
WW1: Military	1914	9.0	14%	wide agreement
Armenian Massacres	1915	1.0	30%	debatable
Russian Civil War	1917	7.7	56%	uncertain
WW1: USA	1917	.05	6%	wide agreement
Stalin: Total	1923	30.9	47%	uncertain
Hitler: Murders	1933	14.7	27%	debatable
Holocaust	1937	5.4	9%	wide agreement
Nanjing Massacre	1937	.3	50%	uncertain
WW2: China civ.	1937	6.9	43%	uncertain
WW2: China mil.	1937	2.3	66%	guess
WW2: Japan mil.	1937	1.5	18%	wide agreement
WW2: France mil.	1939	.2	18%	wide agreement
WW2: German civ.	1939	1.9	63%	guess
WW2: German mil.	1939	3.7	16%	wide agreement
WW2: Poland total	1939	5.9	10%	wide agreement
WW2: UK mil.	1939	.4	16%	wide agreement
WW2: Japan civ.	1941	.4	38%	debatable
WW2: USA	1941	.4	0%	wide agreement
WW2: USSR civ.	1941	8.3	41%	uncertain
WW2: USSR mil.	1941	10.0	29%	debatable
WW2: Yugoslavia total	1941	1.5	14%	wide agreement
Chinese Civil War	1945	2.3	92%	guess
Fr. Indochina War: France	1945	.09	1%	wide agreement
Fr. Indochina War: Total	1945	.6	55%	uncertain
Indo-Pakistani Partition	1947	.6	55%	uncertain
Mao: Purges	1949	5.1	137%	guess
Mao: Total	1949	42.2	58%	uncertain
Korean War: China	1950	.5	61%	guess
Korean War: Civilians	1950	1.6	35%	debatable
Korean War: NKor. mil.	1950	.3	55%	uncertain
Korean War: SKor. mil.	1950	.2	63%	guess
Korean War: Total	1950	2.9	13%	wide agreement
Mao: Great Leap Forward	1959	25.6	44%	uncertain
Mao: Cultural Revolution	1965	1.1	129%	guess
Vietnam War: NVN + VC	1965	.8	22%	debatable
Vietnam War: SVN civ.	1965	.7	77%	guess
Vietnam War: SVN mil.	1965	.2	8%	wide agreement
Vietnam War: Total	1965	1.8	22%	debatable
Vietnam War: USA	1965	.06	0%	wide agreement
Bengali Massacres	1971	1.5	61%	guess
Cambodia: Khmer Rouge	1975	1.6	34%	debatable
Afghanistan	1979	1.4	26%	debatable
Iran-Iraq War	1980	.8	34%	debatable
Gulf War	1991	.08	84%	guess

Mean Estimate
- Personally, I feel that the median gives a much better average, but in order to calculate a standard deviation, my computer has to start with the mean.
Standard Deviation
- I'd say that the Line of Smugness falls at a Standard Deviation of 40%. If the experts deviate by less than that, you can support mainstream estimates with confidence, but if the experts deviate by more than 40%, then you probably shouldn't act like a know-it-all because even the experts don't know.
Confidence Level
- Wide Agreement: Standard Deviation of less than 20%
  - These numbers are as close to true as you're ever going to get in history.
- Debatable: Standard Deviation of 20% to 40%
  - There's some disagreement among experts as to the precise number, but most have agreed on the general order of magnitude.
- Uncertain: Standard Deviation 40% to 60%
  - No one has ever been able to pin down the exact number, but some estimates are more believable than others.
- Guess: Standard Deviation of more than 60%
  - Who knows?

to Table of Contents

Last updated January 2002