Shutting the California Prison System’s Revolving Door

Asterisk: You're responsible for managing one of the most comprehensive data sets of criminal outcomes for various criminal justice systems in California. It includes 12 counties and 60% of the state population. How did you put this resource together, and what kinds of outcomes you are tracking with it?

Mia: So we need to go back a bit. For decades, California was increasing its prison population and building new prisons to accommodate this growth. In 2006, the state reached a peak prison population of 173,000 inmates, which meant prisons were operating at over 170% of their design capacity. As a result, the state began to face lawsuits focused on its inability to provide adequate health care under such crowded conditions. One of the first steps the state took was to address revocations to prison through legislation that rewarded counties for reducing the number of people who fail probation and are sent to prison. 

A: And just to clarify for readers unfamiliar with criminal justice lingo, “revocation” means that someone is incarcerated for misconduct while they are under supervision in the community. This misconduct could be a technical violation of the terms of their supervision, or it could be that they are suspected of a new offense. California had what was famously referred to as a “revolving door” because such a high percentage of people under supervision were revoked to prison. 

M: Yes. So one of the first things that California did was to pass SB 678 in 2009. When someone is sentenced for a felony, one option is that they can be sentenced to local supervision on probation. Prior to SB 678 passing, if a person violated that probation, they could be revoked to prison. What SB 678 did was create an incentive for county probation departments to use evidence-based practices to reduce revocations to prison. So if county probation departments were able to reduce revocations, they were eligible to receive some of the state’s savings from prison costs. And that money would be used to further support evidence-based programs and practices. 

A: And then this all changes again in 2011, right?

M: Yes, a lot changes in 2011. 

SB 678 was the first thing the state did to stop that revolving door, and it worked fairly well. They were able to reduce revocations to prison and start to reduce the prison population. And emphasis there was on evidence-based programming. 

Following SB 678, the state put together a package that we refer to as Public Safety Realignment. The primary legislation was AB 109, which passed in 2011. At that time, California had a state budget crisis — as many states did — and it also had a serious overcrowding problem that the Supreme Court had deemed in violation of the 8th Amendment protection against cruel and unusual punishment. The prison system was taken into federal receivership at this point in time, so there was a really strong mandate to reduce crowding. And that meant either expanding prison capacity rapidly or reducing the population. Given that the state didn’t have funding to support big expansions in capacity, it was on the state to reduce the prison population while maintaining public safety. That’s how Realignment was born.

A: And — to circle back to where we started — your project came out of wanting to track the outcomes of Realignment in California.

M: Right, exactly. What Realignment did is to relocate the responsibility for managing a large share of the felony population from the state to the counties. 

As a result, it became much more important that we understood what was happening at the county level in order to understand how state policy was affecting outcomes. And in California, the criminal legal system is very decentralized. There's not a statewide source of data for what is happening at the local level. 

This project was born out of the idea that we needed to build a data infrastructure if we wanted to be able to evaluate these policy reforms and base our political decision-making on more than anecdotal evidence. In crime policy in general in this country, a lot of policy reform ends up being driven by single events — something bad happens, and we reverse or change or initiate new policy in response. My hope was that if I could build a data infrastructure that allowed us as researchers to ask these questions, we'd be able to get a clearer picture of the effects of these changes in policy.

And that meant finding a way to bring together data not just from counties but from agencies within counties that don't typically share data, linking that together at the county level, and then linking data from as many counties as we could to statewide data on criminal history and subsequent involvement with the criminal system in the form of arrests and convictions. 

My team at the Public Policy Institute of California, including Ryken Grattet, Sonya Tafoya, and Viet Nguyen, put together a group of counties that we felt would be manageable based on our funding and team capacity. And then, given that, we asked how we could get the best representation of the state in terms of geographic, demographic, and political diversity. We came up with a list of 12 counties. And then we set on the road to talk to probation departments and sheriffs in those counties. And all of them were really interested in understanding the effects of Realignment in their counties. Every single county signed on and shared their data. 

Then we had the task of recruiting the state. We worked through the Board of State and Community Corrections to recruit the Department of Justice and the Department of Corrections at the state level to also share their data. This level of data sharing was pretty unprecedented given the number of sources and, importantly, the fact that we needed to receive an identifier from each agency that would allow us to link individuals across all of these decentralized systems. 

The end result was that we were able to put together a data infrastructure that allowed us to ask questions about the effects of Realignment. But then, following Realignment, there were other reforms that came through — importantly, Prop 47, which defelonized drug possession and some property offenses. And so that data infrastructure created a base from which we could start to research the effects of criminal legal reforms and then provide that evidence basis back to the legislature, the governor's office, and local practitioners and policymakers.

A: Let's talk about what you found. You've mentioned evidence-based practices a few times. I’m curious what those were. Because a lot of Realignment is relocating criminal justice programs in the counties, I’d expect that different counties are adopting different practices, so there's something like a natural experiment happening. What have we learned from that?

M: One example of an evidence-based practice is matching programs to meet people’s needs. This might be a very logical and obvious thing to do. But prior to this wave of reforms, much of the focus in community supervision, for example, was on enforcement rather than programming and services to address needs and risks. So the practice of conducting assessments when people start probation and matching people to interventions that would address their needs and risks and reduce their likelihood of future involvement with the system was fairly new. 

A: Replicability in criminal justice interventions is infamously low. A lot of programs either will not do well in RCTs or will do well in one RCT and then fare poorly in a different context. So I'm curious how some of these programs that had a good evidence base elsewhere performed along the outcomes they were trying to measure in California.

M: Understanding what kind of interventions work in criminal justice is deeply challenging. It’s hard to do RCTs to begin with, given that you need to randomize, for example, access to a substance use treatment program. And then the RCTs are done in different contexts, so there’s always the question of generalizability. I hate to tell you that we don’t have a perfect evidence-based literature that can tell us exactly what we should be doing and what will work where, but it is a constant pursuit. It was a big goal of this project to get there. 

One of the challenges in California, especially at the local level, is that it's very difficult to collect programming data. We may not have good data on, for example, whether people were referred to a program, whether they showed up, what kind of dosage they received, what the specific program entailed, or whether they successfully completed the program. Providing programs and other interventions is often decentralized — say we have an agency within a local government working with nonprofit groups, and then those groups are doing all different kinds of programming. Many of these nonprofits are operating on shoestring budgets, and they may not be collecting much if any data about who is participating, what they’re getting, whether they follow the program to completion. 

So even the existence of data is in question. We are working on solving the problem of how to collect good data about what people are getting. And we need to do that in order to even do an observational study, let alone an RCT. And then there are still methodological challenges once you put the data together, but I anticipate in the next decade we will have much better data to work with and improved ability to evaluate the effectiveness of these interventions. 

A: Let's zoom out, then, and talk about some of the data you were collecting in your data set and what outcomes you were tracking there.

M: I think the highest level question about Realignment is whether it was a good thing or a bad thing. Did Realignment — which fundamentally restructured California's criminal legal system and dramatically reduced the prison population — have negative public safety consequences? 

We can see that in a couple ways. We can look at crime rates overall in the state over time. We did some work at PPIC looking at crime rates both in California and in other states (to form a synthetic cohort or comparison to California). And we didn't see a lot of change in crime rates.

A: The idea here being that if crime is rising nationally or falling nationally, we wouldn't want to attribute to Realignment something that is a national trend. 

M: Exactly. We looked at crime, but because crime is high level and removed, we also wanted to look specifically at individuals moving through the criminal legal system and whether their experience had changed because of Realignment. 

We looked at different groups who experienced different treatments because of Realignment: people released from prison, people released from jail, and people who were sentenced to probation. When we compare them with very similar people that were released prior to Realignment, did their outcomes differ?

And when we looked at that, we didn’t see what was feared, which was a big increase in recidivism. Each group is different in terms of their outcomes for different measures of recidivism, but importantly, we didn’t see big increases in reconviction rates. And in some groups, we actually see reductions in reconviction. 

That helped stabilize the reform. If we saw increases in crime that deviated from other states, or we saw higher levels of recidivism for people affected by Realignment, we would have had concerns regarding the public safety impacts of the reform. 

A: Right, recidivism was mostly static. My understanding is that the way Realignment reduced prison populations is by making it harder for someone to get sent back to prison for violating the terms of their parole, which was what SB 678 did.

M: Reducing revocations was one driver of reductions in the prison population. But another really important driver was that Realignment identified a set of felony offenses that were no longer eligible for a sentence to state prison. If someone was convicted for one of these lower-level felony offenses, they were no longer eligible to be sent to prison unless they had a serious or violent criminal history. That really reduced the flow of people into the prison system. 

A: The implication, then, of all of this is that previously we had way too many people in prison from a public safety perspective. And adopting less intensive correctional sanctions did not substantially impact public safety. 

Do you think we still have too many people in prison now?

M: Great question. It depends on your criteria. One criterion the state can use is cost-effectiveness. Is our incarceration level too high, such that what we’re spending to incarcerate people is more than the value of the public safety improvement? It’s possible that at the height of mass incarceration, we were incarcerating so many people that there was no public safety improvement, and incarcerating a lot of people could have even been making outcomes worse. 

You can also think about it in terms of trade-offs in harm to people who are incarcerated, as well as their families and their communities, relative to gains in public safety for the broader community. It’s a balancing effort to figure out what we are willing to do to make society safe. What kind of society do we want to live in? How much misconduct are we willing to tolerate, as a society, in order to reduce levels of incarceration?

And then a third lens through we can look is to zoom out and ask: Is mass incarceration, the strategy that we chose for decades, really improving outcomes in society? I think what Realignment represents is a real shift toward — rather than incarcerating people who have very high needs — meeting those needs with more programming and supportive services, rather than turning to incarceration as a solution. 

A: Do you think that if we had replaced prison with supervision only — and didn’t institute these programs — we would have expected crime to rise? Or is your underlying model more like, we shouldn’t have expected Realignment to increase crime in the first place, that it would have had a null impact on public safety even without programming?

M: I would say my theoretical model is that many of the people who engage in misconduct that leads to incarceration have unaddressed needs. And if those needs were addressed, we would have improved public safety and reduced misconduct — while also generating better outcomes for those people. Reducing incarceration and generating support for, funding for, and greater connections to those programs are mechanisms that can lead to improved public safety. 

The question on whether we could reduce incarceration without doing anything else and still see no declines in public safety is an interesting one. I think that at those earlier, very high incarceration levels, we probably could have reduced incarceration by some amount — absent any other interventions — and not seen much of a change, just given the fact we were incarcerating at such a high level for such a long time, particularly people who had lower-level offenses.

But I don’t know that we’ve really proven that the increase in emphasis on programming has led to better outcomes. I certainly haven’t been able to do that study, so I think it’s an open question.