Start with one. A single man, in a single cell, somewhere in England or Wales. He has a wing, a landing, a cell number and, less visibly, a slot in a spreadsheet at the Ministry of Justice. That slot matters more than it sounds. His presence in the count is part of how anyone works out whether prison is becoming safer or more dangerous.

Now zoom out from that one cell. In 2024, the most recent year available, the figure recorded for males in the ministry's prison custody tables was 83,658 (Source: Ministry of Justice, Safety in Custody -- Deaths_in_prison_custody_1978_to_2024_accessible -- Table_1_1). Not a rate. Not a percentage. Men.

Here is what that means in scale. England has around 57 million people and Wales around 3.1 million, so roughly 60 million between them. Halve that for a rough count of males, and you get somewhere near 30 million men and boys. Divide 83,658 into 30 million and you land at about one in every 360. That is a crude sum, and I want to be clear it is crude: the denominator includes toddlers and pensioners, neither of whom are realistic candidates for a prison cell. Narrow it to working-age men and the share held in custody rises. The true rate for men in their twenties and thirties is higher still.

The reason this number sits inside a file about deaths is that it is the denominator. Every death rate in custody is a count of deaths measured against the number of people being held. Change the population and the rate moves even if nothing else does. That makes the population figure quietly powerful. Get it wrong and every safety statistic built on top of it is wrong too.

It also deserves reading on its own terms. A count of 83,658 men is not a statistical abstraction. It is the population of a mid-sized British town, entirely male, entirely in the state's care. The state has taken on the job of feeding them, housing them, supervising them and keeping them alive. Each one of them has someone outside who would notice if that job went badly. That is the weight this figure carries, and it is why it is worth publishing every year rather than burying it as a technical footnote.

Now the honest counterweight. This is one number from one year, and one year cannot tell you a direction. I have not been given the figures for 2023, 2020 or 1990, so I cannot tell you whether this is a high point, a low point or unremarkable. The file name suggests the series stretches back decades, but that is a label on a spreadsheet, not evidence I have checked. And the figure is from 2024, not from this morning. The custody population moves month to month.

My own view: the denominator gets far less attention than the death totals stacked on top of it, and that is the wrong way round. If you want to know whether prisons are safe, you first have to know how many people are in them. Last year, for men, the answer was 83,658.