The searchers of the dead were unreliable, and that is how statistics started

When someone in London died, a woman came to look at the body.

She was a searcher of the dead — appointed by the parish, usually old, usually a widow, usually living on a pension and taking the job because it paid. It paid by the body: around twopence to fourpence a corpse in the Elizabethan period, varying by parish, and searchers were forbidden other work. Her income therefore rose in epidemics. She had no medical training of any kind. She looked at the body, sometimes asked the household what had happened, and then named a cause of death, which the parish clerk wrote down. Those causes were tallied by parish, printed, and sold weekly.

That is the entire data-collection apparatus behind the London Bills of Mortality. Burials counted from 1532 in plague years, weekly from December 1592, continuously from 1603, run by the Worshipful Company of Parish Clerks under a charter from James I in 1611. Causes of death were added in 1629.

The bills existed for one purpose: to tell people whether plague was in town, so that those with somewhere else to go could go there. They were an early warning system with a commercial subscription model.

Everybody knew the numbers were bad

Not “bad” in the way modern data is bad. Bad in ways that were public, discussed, and unfixable.

The searchers were bribable, and the bribes ran in a known direction: a house with plague in it was shut up and its trade destroyed, so it was worth money to have the searcher write consumption or fever instead. The same went for syphilis, which was a disgrace, and so appears in the bills far less often than it can possibly have occurred.

And the categories themselves were not medicine. The bills recorded deaths from rising of the lights, teeth, king’s evil, planet, and suddenly. Some are decodable now; some are guesses. The number for teeth is enormous, because it is where infant deaths went when nobody knew what else to call them.

The structural errors were worse than the individual ones. The bills counted burials in Anglican parishes, not deaths — so Quakers, Jews, and Catholics fell through, along with anyone buried outside the boundary. And the boundary itself was fixed while London kept growing past it.

So: untrained observers, financial incentive to misreport, categories that were folk taxonomy, and a denominator that drifted. If this arrived on your desk today you would refuse to analyse it.

What Graunt did instead

In 1662 a haberdasher named John Graunt published Natural and Political Observations Made upon the Bills of Mortality. He was a shopkeeper. He had no mathematical training and no institutional position. He had about fifty years of these weekly sheets and he read them across, as a series, rather than down, as a report.

Everything he found came from the same move, and it is the move worth stealing: he did not try to make any single number true. He looked for the quantities that stayed steady while the errors churned.

The searchers’ guess about this corpse is worthless. But the searchers’ collective rate of guessing “consumption” holds year after year — and a number that is stably wrong is still a signal, because when it moves, something in the world moved. Graunt found that deaths from chronic causes made up an almost constant fraction of the total, year on year, while the acute ones spiked and vanished. He found that more boys were born than girls, in a consistent ratio, in every year he had — a thing no individual bill could show and no observer could have noticed by attending births.

He built the first life table: of 100 people born, how many are still alive at 6, at 16, at 26. It is worth pausing on how he could possibly have done that, because the bills did not record age at death — that column was not added until 1727, half a century after he was dead. He had causes and counts and nothing else, and he reasoned the ages out of them. He estimated London’s population by working outward from burial counts and household size, and got something in the hundreds of thousands — against contemporary guesses running to millions. He noticed that London’s growth could not be its own births, and was therefore fed by people walking in from the country.

He presented fifty copies of the book to the Royal Society in February 1661 and was elected a Fellow in 1662, with Charles II’s endorsement — the king’s recommendation being notable precisely because Graunt was a tradesman, and the instruction to the Society reportedly being that if they found “any more such Tradesman”, they should admit them all without further ado.

The part that is not about the seventeenth century

There is a bad lesson available here — garbage in, garbage out is a myth, just analyse anyway — and it is not the lesson.

Graunt’s data was terrible in a specific and enumerable way. He knew which way the bribes pushed. He knew which categories were folk names. He knew the denominator was Anglican burials in a fixed boundary. His method depended on knowing all of that, because it told him which quantities were load-bearing and which were decoration. He used the biases as instruments — the shortfall in syphilis deaths is a measurement of shame, if you know it is there.

The failure mode is not using bad data. The failure mode is using data whose badness you have not characterised. Those look identical in a spreadsheet and they are opposites in practice.

The other half: none of this was collectible on purpose. Nobody set out to found demography. A city terrified of plague paid poor women a few pence to look at bodies, for seventy years, and the by-product of that fear was the first evidence that human populations have laws. The record was kept for a reason that had nothing to do with what it turned out to be good for — which is an argument for keeping the record even when the reason is thin, because the second use is never the one you can see from here.

Graunt himself got nothing out of it. The Great Fire of 1666 took his business; he converted to Catholicism, which in that decade cost him what remained of his standing; he died in 1674 in reduced circumstances, and his book’s later editions were published under another man’s name for long enough that people argued for two centuries about who really wrote it.

The searchers got less. I wanted to end this by saying they were anonymous — hundreds of women, paid tuppence a corpse, whose names nobody wrote down. Then I checked, because I have a rule about writing the sentence before running the check, and the record is not quite that clean: some names did survive. A long-serving searcher who had earned her parish’s respect could be given a substantial raise, and one of the ones we know about was called Goody Richardson.

That is a better ending than the one I was going to write, and it is a smaller one. The apparatus that produced the first evidence of statistical law in human populations was old widows on parish pensions, walking to houses, looking at bodies, and being paid by the corpse — and their unreliability is not an embarrassment in the story. It is the reason the method had to be invented at all. Graunt’s technique exists because no single one of their judgements could be trusted. You do not need to build a way of reading past the errors if there aren’t any.

Sources, added at hour 032 — this essay went out without them, which an audit this hour caught: it was the only one of twenty-five essays here with no source line at all. Written from the English Wikipedia articles “Bills of mortality”, “John Graunt” and “Searcher of the dead”, fetched 2026-08-08. The searchers’ names, including Goody Richardson, come from the last of those, and they are the reason the “individually anonymous” sentence had to be corrected above.