China's Deadliest Rebellions and the Problem With Their Death Tolls

The figures you see for China's largest rebellions — tens of millions dead, sometimes more than the First World War — are not counts. They are almost always derived from the difference between two imperial censuses, and an imperial census recorded registered households for taxation and labour service rather than living human beings. When a rebellion destroys the administration that does the registering, the register falls. Separating the part of that fall caused by death from the part caused by flight, displacement and administrative collapse is the central problem in this subject, and it is why the same event can be quoted at wildly different scales by sources that are all acting in good faith.

What an imperial census actually was

A modern national census tries to enumerate a population. A Tang or Ming register was a fiscal instrument: it existed so the state knew which households owed grain, cloth, cash or corvée labour, and it was compiled by officials whose reach ended where their control ended.

That difference does most of the work in this article. Consider what happens when a rebellion takes a third of the country. Households in rebel-held territory stop appearing in the imperial register. Refugees who move are not necessarily re-registered where they arrive. Local officials who have fled or died stop filing returns. Households with an interest in not being taxed take the opportunity to disappear. When the next register is compiled, the recorded total has collapsed — and every one of those mechanisms has contributed to the collapse alongside actual mortality.

A death toll built by subtracting the second register from the first therefore measures several things at once and cannot tell you their proportions. It is a real measurement of something. It is not a measurement of deaths.

The An Lushan case, and why the biggest number is the weakest one

The eighth-century An Lushan Rebellion is regularly listed among the deadliest conflicts in human history, and the figure used to support that claim is a very large drop in the Tang registered population across the rebellion years.

Take the mechanism above and apply it here. The rebellion broke the Tang state's administrative grip over large regions for years. Fiscal registration in those regions stopped or became nominal. Post-rebellion Tang governance was substantially more decentralised, with regional commands retaining revenues that had previously been registered centrally. Every one of those changes reduces the registered figure without requiring a single additional death.

That does not mean the rebellion was small — it was a catastrophe by any reading, with mass killing, famine and long-term dislocation. It means the specific figure most often attached to it is measuring the visibility of the population to the state, and should not be quoted as a body count. Specialists in Tang demography have made this argument for decades in the scholarly literature, including in venues such as the Journal of Chinese History and in the institutional research published by Academia Sinica's Institute of History and Philology.

The same caution applies to the Three Kingdoms period, where a comparable fall in Han-era registration figures is often converted into a casualty total. The registers moved; the reason they moved is exactly what is in dispute.

The Taiping Rebellion, where the evidence is much stronger

The nineteenth-century Taiping Rebellion is the case with the best documentary footing, and it is worth separating from the earlier ones for that reason.

Columbia University's Asia for Educators dates it to 1850-1864 and describes a movement led by Hong Xiuquan, who had failed the imperial examinations for the third time in 1847 and came to believe that he and his followers had been chosen to conquer China. Its programme combined Confucian and Christian-derived elements and proposed common property, land reform, equal position of women, and abstinence from opium, tobacco and alcohol. By the late 1850s the Taiping "controlled over a third of the country", and the Qing government required fifteen years and enormous expenditure to defeat them. The resource puts the cost at "more than twenty million lives", and notes that "even by the 1950s, some parts of central China had not yet fully recovered".

Here the evidence base includes foreign consular and missionary reporting, treaty-port commercial records, Qing military and fiscal documents, and later fieldwork on regional recovery. That is not the same evidentiary situation as a Tang register, and treating the two as equivalent data points on one list is the error that produces those tidy "deadliest wars in history" tables.

There is a smaller illustration in the sources themselves. Columbia's page gives 1850-1864; Fordham University's Internet East Asian History Sourcebook indexes the same event as "The Taiping Rebellion, 1851-1864". Neither is wrong. They have chosen different starting points for the same process, which is precisely what how historians date events describes happening across the whole discipline.