I going to guess that referring to inmates by numbers would be widely criticized as inhumane by many human rights groups. It would be an uphill battle to keep something like that going.
That would mean for 2mm prisoners there would be 1.1mm murderers, rapists, and manslaughterers. That seems like way too large a number to be realistic with the declining rate of murder, rape, and manslaughter.
Then count people who die following their sentences in the same way you count reoffenders. Problem solved. As it stands, I'd bet a big fraction of people dying within 5 years of getting out of jail are ODs, so setting up a metric that encourages prisons to work on drug issues is a good thing.
If you're worried about states executing people to keep them out of the numbers, count executions that way too.
The prisoners don't necessarily outnumber the locals outside, so I don't think it is a foregone conclusion that they will be able to.
But another way to look at it, is that by being counted as part of the population in the vicinity of the prison, it provides an incentive to have prisons and to incarcerate people, and it also reduces the representation of the places prisoners come from.
The stats talk about number of people imprisoned in a particular moment of time. It doesn't say if a part of them for example died over the next couple of months due to cold and starvation and were topped up with new inmates That would make the number look unchanged during the next count.
I don't have the statistics here but it would be interesting to see what percentage of prisoners were serving 5+ years, maybe the ratio of prisoners serving shorter sentences would make this kind of solution more plausible.
Probably a sizable number of them. Because the people in prison are still people, and there's no good reason to keep most of them locked in a cage if they're not posing a risk to other people.
The prison lobby (guard unions, etc) is actually fairly loud about this. They make a lot of noise around things that put and keep more people in jail. It's way past "probably"!
reply