The nine million witches figure came from one man's bad arithmetic

Most explanations of the number of witch trial executions stop at the point where it starts to matter. This one carries on.

The number in circulation

A figure of nine million executed witches has appeared in pamphlets, films and political writing for well over a century. It is usually presented as a suppressed fact, which gives it a resistance to correction that ordinary statistics do not have.

The number is enormous, implying a scale of killing comparable to the largest atrocities in European history. An event of that size would have left demographic traces visible in parish records across the continent. No such traces exist, which is the first indication that something has gone badly wrong upstream.

How the figure was produced

A German writer in the late eighteenth century took the recorded executions from the archives of one town over a period of years. He then calculated a rate per head of population and applied it to the population of Europe over several centuries of trials.

The arithmetic was performed correctly and the inputs were not, which is the classic failure mode of extrapolation. His chosen town had experienced an unusually severe episode, so it was among the worst possible samples to generalise from. The method also assumed a constant rate across regions and centuries, when the actual pattern was intensely clustered in time and place.

Why the sample was so misleading

Witch trials were not evenly distributed but occurred in bursts, with large parts of Europe experiencing few or none. Some jurisdictions prosecuted rarely and acquitted often, while a small number of intense local panics produced most of the deaths.

Legal systems differed enormously in whether they permitted torture, allowed appeals or required corroborating evidence. Applying the rate from a panic region to a jurisdiction that barely prosecuted at all inflates the total by orders of magnitude. The distribution matters more than the average, and the eighteenth-century calculation had no way of seeing the distribution.

What the archives actually support

Historians have spent decades working through surviving court records across many European jurisdictions. The resulting estimates for executions across the whole early modern period are generally in the tens of thousands. That is a substantial number of judicial killings and does not need inflating to be a serious historical subject.

Uncertainty remains, because records are incomplete and some regions are far better documented than others.

Responsible accounts give a range and explain why it is a range, which is exactly what the nine million figure never does.

Why the inflated number spread

It was adopted in the nineteenth century by writers arguing about the historical role of organised religion, where scale served the argument. It was later taken up by twentieth-century movements that used it as a foundational account of persecution.

Tested properly, in both cases the number was doing rhetorical work, and rhetorical numbers are rarely audited by the people they persuade. It also has the appeal of a round figure, which reads as authoritative while actually indicating that nobody counted. Correcting it is often heard as minimising the persecution, which makes the correction socially awkward as well as unwelcome.

Printing a correction has a poor record of taking a myth out of circulation.

Getting the scale right without minimising

The accurate account includes that a majority of those executed were women, that accusations often followed local disputes, and that the poor were disproportionately vulnerable. It also includes that trials ended largely because legal elites lost confidence in the evidence, not because popular belief changed.

Somewhere in the retelling, none of that requires an inflated total, and the real figures are quite sufficient to make the point. Insisting on accuracy here protects the subject, because an argument built on a discredited number can be dismissed along with it. A discipline that says tens of thousands and shows its working is stronger than one that says nine million and cannot.

The takeaway

One town, one bad decade, multiplied across a continent and four centuries, and the result outlived the arithmetic by two hundred years.

The satisfying version of a story is the one that travels, which is the whole problem.