From Prison Policy Initiative
Millions of people are arrested and booked into jail every year, but existing national data offer very little information about the actual criminal charges for which they are detained. In fact, the most recent national offense data published by the government about people in jail is from 2002, so we worked with the Jail Data Initiative to fill that gap using the best, up-to-date nationally representative sample available. We now have a new “snapshot” of people in jail, by offense, on a given day; new insights about how low-level offenses like misdemeanors and supervision violations impact “jail churn” over the course of a year; and we were able to analyze some variation in offenses by sex and, for the first time, by region and jail jurisdiction size.
Our unique, representative sample of jail records
The Jail Data Initiative (JDI) collects online jail rosters that are updated at least daily, and JDI excludes any jail rosters that are updated less frequently (e.g., only weekly). The dataset includes individual-level jail data for approximately 1,300 local jails across the country, representing over one-third of all jails in the country. The contents of jail rosters vary between jurisdictions: that is, some jails publish more information than others, including demographics or detailed charge data. For the purposes of our analysis, we only included jail rosters that published the relevant information on charges, reducing the sample to a subset of 865 jail rosters. While we are not able to break out data on people held pretrial versus those serving short sentences in jail using this data source, we do know that most people in jail are held pretrial: in 2023, 70% of the national jail population was unconvicted. While our analysis is based on the total number of people in the sample of jails, regardless of their conviction status, the findings are likely most reflective of the unconvicted or pretrial population, because it is so much larger than the convicted population in jails. Continue reading >>>
|
|
|
|
