Obviously higher than 0, all else being equal. I have priorities other than keeping people locked up on the off chance that a tiny fraction of them might commit

Obviously higher than 0, all else being equal. I have priorities other than keeping people locked up on the off chance that a tiny fraction of them might commit a crime- given that most people would at least claim to support fair trials with evidence and such, I suspect that 0 is actually a pretty rare response. edit: I think your edit made things a little clearer. I doubt we agree, but we've talked past each other here. The relevant statistic is [number of people who commit crimes on bail]/[number of people released on bail]. If we had Minority-Report-style precogs, we could have had 0 such cases in the last 6 months while denying only 3 more people bail. But we don't have that option- there's not much to be done about the sensitivity, so all we can do is crank the cutoff of doubt higher or lower. With tens of thousands of people released on bail, and assuming that identifying the nth potential murderer roughly follows a power law distribution (a fairly reasonable assumption in a sensitivity/specificity situation like this) we're talking about jailing thousands of innocents to prevent one guilty man from going free. If look at violent crime in general the numbers are less skewed (around 2%), but I'd hope that most people balk at the idea of preemptively jailing ten people without trial to prevent just one from committing a crime. I mean, recidivism rates after prison are way higher than that- perhaps "the rest of us" would favor life sentences for every crime, but I doubt it.