a platform for tracking granular recidivism metrics in real time

info icon We're building some exciting tools in this space; stay tuned.
Meanwhile, play around with existing datasets below or contact us.

95% of the roughly 2 million inmates in US correctional custody will one day be released.

Depending on who you ask, between 30% and 77% of those prisoners will end up back in prison. The Bureau of Justice Statistics' most recent study found a 3-year recidivism rate of 67.8% across 30 states.

The graphics below are based on BJS's data. But these numbers are likely not quite right.

Existing studies are constrained by static data collection techniques - they capture a snapshot in time, which quickly becomes stale.

They're often forced to conflate differing definitions of recidivism, lump parole infractions in with new crimes, or forgeo cross-cutting comparisons altogether.

And in most cases, the data collected isn't granular enough to be actionable.

It's really hard to find good data on recidivism, and Recidiviz wants to help.

We're building a platform that standardizes the way that recidivism data is collected.

It should be easy for states and policy makers to access granular, accurate, real time metrics, so that they can reduce crime, reduce costs, and improve outcomes for inmates and communities.

The chart below provides an introduction to the recidivism problem, with data from the United States Bureau of Justice Statistics, which determined recidivism rates among inmates released from prisons across 30 states over 5 years, starting in 2005. Ideally, we’d be able to do this calculation in real-time, in order to track progress: for example, how do recidivism rates compare to what they were, say, a decade ago? We don’t know.

Here, a recidivist is someone who has been arrested for a new crime after returning to their community. But different states, counties, and academics tend to measure recidivism differently, which makes it difficult to compare how programs are performing, what’s working, and what isn’t.

If 100 people were released from prison today...(drag the slider)

The next chart uses the same dataset, but this time dives a bit deeper, exploring how a prisoner's age, race, gender, and crime impact their likelihood of winding up back in prison. Try entering your own age, race, and gender.

Collecting this data on an ongoing basis (rather than taking a snapshot from 2005) would allow us to design more effective programs for reducing recidivism and track their effectiveness over time.

Try Your Luck

A prisoner's age, race, gender, and crime tell us something about how likely they are to end up back in prison -- and how quickly.

I am years old. I am a who has previously been arrested times. If I am released from prison after serving time for , I will probably be back in prison in...

The third chart is based on fake data because, as far as we know, this dataset doesn't exist. If we could compare prisons by recidivism rates — rather than by price per prisoner, for example — we could get a much clearer picture of what programs were working, share best practices, and reduce recidivism rates across the board.

Today, it’s virtually impossible to do this kind of comparison. Prisons themselves often don’t even know their recidivism rates, how they’re changing over time, how their programs are working for some inmates versus others, or how they might improve their programs to deliver better outcomes for all.

Compare Prisons

Show me the prisons in
ranked by their recidivism rates