Here is a fascinating description of a UK project that makes re-offending data available to non-profits working in the justice system.

This flagship initiative allows charities to access real data on offending rates. It works like this; a charity that has worked with offenders or ex-offenders can upload to the lab the details of cohorts of people they have worked with. The lab supplies actual data on their re-conviction rates and those of a matched comparison group. The result? Charities get to find out if their approach has had an impact on re-offending, and are able to prove it. And over time the evidence on what works will be built across the sector. We know the Justice Data Lab won’t solve everything, and that measuring re-offending rates provides only a snapshot of impact. But providing charities with access to statutory data so they can prove their worth is a big deal, and a big win for NPC.

According to the online description, the next project will try to take the same approach with homelessness.  It works much more easily with binary outcomes such as re-offending or homelessness.  Note that the lab does the process of finding a matching group, based on the uploaded data, and doing the comparison — making research much cheaper and more practical.

Could we make this model work in the US?  Even in the re-offending area, we would run into obvious problems, the largest of which might come from our chaotic multiple jurisdiction system, from our relative lack of reliable identifiers, and from our privacy concerns (even is now shown to be, shall we say, inconsistently applied).  But a state with a good integrated court data system should be able to do this, and indeed, for decades researchers have been able in many states to obtain this data on an ad hoc negotiated but routine basis.

Doing it in areas like homelessness would be much harder, since so few of those in the situation receive services and are properly tracked.

One way of addressing privacy concerns would be for the central lab to only report aggregate outcomes.

A more general point is that the US here pays a heavy price for its fragmentation and lack of integrated services and data collection.

Who could take the lead in a model data lab?  Could DOJ make the establishment of such a system at the state level a condition of its massive funding of state criminal justice record automation?

Of course, the Social Security Administration could do it for employment, but the data to create a matched comparison group would be much more limited.

Thanks to Claudia Johnson for pointing this one out.