Counting the dead: methodology

The claim

The widely-quoted figure of roughly 100 million deaths under communist regimes comes mainly from The Black Book of Communism (1997), which totalled on the order of 94–100 million. Taking that number seriously means understanding exactly what it contains and where it is contested — because an honest case does not need a number it cannot defend.

1. The total is an aggregate of very different things. It sums executions, deaths in camps, deaths in forced deportation, and — forming most of the total — deaths by famine. Lumping deliberate shooting together with famine mortality is the first thing serious critics object to.
2. The largest single components are famines. The Chinese famine of 1959–61 (the Great Leap Forward) and the Soviet famines dominate the count. Estimates for the Great Leap alone range widely — roughly 15 to 45+ million excess deaths, with recent archive-based work (Dikötter, Yang Jisheng) clustering at the higher end.
3. Attribution is the hard part. Whether a famine death counts as a regime killing turns on intent versus policy: was it engineered, foreseen-and-accepted, or a catastrophe worsened by ideology and denial? Scholars answer differently for the Holodomor (Applebaum, Conquest) than for Maoist China.
4. The archives cut both ways. Soviet records opened after 1991 revised some components down — notably execution counts and camp populations (Getty, Rittersporn & Zemskov; Davies & Wheatcroft on famine demography) — while documenting the famines in grim new detail.
5. Even the Black Book's authors disagreed. Two of its own contributors, Nicolas Werth and Jean-Louis Margolin, publicly distanced themselves from editor Stéphane Courtois's framing and his drive to reach a round 100-million total.

The scholarly debate

Three methodological choices decide the number: whether famine deaths count as killings (the intent-versus-policy question); whether to measure excess mortality against a counterfactual or only direct killing; and how to present estimates that honestly carry error bars of tens of millions. This library's position is that the moral weight does not hinge on the last digit — a defensible range in the tens of millions is damning enough, and quoting a false-precision "100 million" as if it were counted hands the other side an easy rebuttal.