The data doesn't seem to indicate that judges are heavily influenced by outside forces when determining bail. This [1] is a paper from the Bureau of Justice Sta

The data doesn't seem to indicate that judges are heavily influenced by outside forces when determining bail. This [1] is a paper from the Bureau of Justice Statistics on bail stats. Those data are only for felonies at the state level, which is somewhat biased but also arguably the most important group as well -- misdemeanor offenses tend to receive $0 to negligible bail. So for instance one interesting datum for that paper is release rates for people by prior offenses. Release, in the context there, can mean they were released either for an affordable bail or for no bail at all -- 'released to personal recognizance' being the phrasing. Of those with no prior convictions, 77% end up being released. For those with misdemeanor priors, 63% end up being released. And for those with felony priors, only 46% end up being released. Actually maybe 'only' is not an appropriate term there. People charged with a felony, and who have prior felony convictions, end up being released from prison 46% of the time. Anyhow, the thing I think that arguably justifies judges' decisions is that we can have a sort of completely unplanned, but phenomenally well designed, experiment. Sometimes jails are forced to release people that they would not otherwise release because of emergency releases -- that can be triggered by things such as jail overcrowding. In these cases judges are going to release those who were seen to be less desirable candidates for release. Though of course they'll be releasing the 'least less desirable' candidates. Cutting to the chase there, these people released through emergency releases ended up engaging in some form of pretrial misconduct 52% of the time with 45% of them failing to appear for their court date. The pretrial misconduct rate for intentionally released individuals ranges from 27-36%, depending on the release type, with failure to appear rates of 14-30%. The paper includes a couple of really handy tables that show logisitic regression analysis of release decisions and reoffense rates for people broken down into various categories, alongside their pretrial misconduct rates. So for instance you can see the weighting for, or against, release for various groups contrasted against how often they reoffend. In a perfect world, these would be inverses of one another. For instance if there is a large negative bias against releasing people for [x] then we'd expect to see a large positive bias in terms of pretrial misdconduct for [x]. And in most cases this does seem to be the case. But there are some exceptions. All in all, it seems like a system that does work pretty well. Should be interesting to see how well the bail committees perform by contrast! [1] - https://www.bjs.gov/content/pub/pdf/prfdsc.pdf https://www.bjs.gov/content/pub/pdf/prfdsc.pdf