Probable Causation

View Original

Episode 11: Steven Raphael

Steven Raphael

Steven Raphael is Professor and James D. Marver Chair in Public Policy at UC Berkeley's Goldman School.

Date: September 3, 2019

A transcript of this episode is available here.


See this content in the original post

Episode Details:

In this episode, we discuss Professor Raphael's work on the effects of sentencing reform in California: 

"The effect of sentencing reform on crime rates: Evidence from California’s Proposition 47" by Magnus Lofstrom and Steven Raphael.


OTHER RESEARCH WE DISCUSS IN THIS EPISODE:


Transcript of this episode:

 

Jennifer [00:00:07] Hello and welcome to Probable Causation, a show about law, economics and crime. I'm your host, Jennifer Doleac of Texas A&M University, where I'm an Economics Professor and the Director of the Justice Tech Lab.

 

Jennifer [00:00:17] My guest this week is Steve Raphael. Steve is Professor and James D. Marver Chair in Public Policy at the Goldman School at UC Berkeley. Steve, welcome to the show.

 

Steve [00:00:27] Thank you for having me.

 

Jennifer [00:00:28] So today we're going to talk about your recent paper, coauthored with Magnus Lofstrom, about sentencing reform and particularly the effects of California's Prop 47. But to start out, could you tell us about your research expertise and how you came to study this topic?

 

Steve [00:00:44] Well, sure. I'm an economist. I've been a professional researcher probably for about a quarter century at the moment. I started out doing work on low wage labor markets and racial disparities in pay and employment in particular. And very early in my career, I started studying the effects of having a criminal record on unemployment outcomes. And from there, my work just sort of naturally bridged into policy issues that are sort of more general to the criminal justice system, issues involving incarceration, issues involving the determinants of crime, and more generally, how policy responses impact social inequality and criminal justice outcomes.

 

Jennifer [00:01:35] It's been an interesting few years in terms of criminal justice policy in this country with lots of conversation about ending mass incarceration. But California has been making some especially large changes on this front, or at least that's how it seems from the outside. So give us a bit of background on what's been going on in California over the past decade in terms of efforts to reduce incarceration rates.

 

Steve [00:01:57] Sure. So California historically has had an incarceration rate that has pretty much charted the national average. It increased with the increases in incarceration in the 70s, 80s, and 90s. And and it sort of paralleled the U.S. Up until about 2010, 2011. And what what was really unique about California was the extent to which our prison population, relative to the capacity of the system was, was very, very large. So as of, say, 2008, 2009, 2010, the number of people in prison was around 200 percent the rated capacity of the system. And it prompted a number of lawsuits brought by brought by incarcerated plaintiffs, basically, that were alleging that they were not receiving adequate health care and adequate adequate mental health care services. And the overcrowding of the system was the primary culpable factor. These two lawsuits were joined and heard by a federal three-judge panel, which ultimately ruled that California had to address the system by either increasing the capacity of the prison system or reducing the population. There was an ensuing legal battle that went all the way to the Supreme Court. And then in 2011, the Supreme Court basically ruled in favor of the plaintiffs. So starting in 2011, there have been a series of reforms that have been aimed at reducing prison overcrowding and more generally sort of reducing the punitiveness of the criminal justice system.

 

Jennifer [00:03:49] And before you wrote the study that we're going to talk about today, what had we known about the effects of these policy changes in California, and perhaps similar policy changes in other places around the world, in terms of their effects on crime rates and recidivism?

 

Steve [00:04:02] Well, in terms of California, there are two big reforms- there are actually a number of reforms, but but there there are two pretty large reforms that that stand out. The first was in 2011, the state passed the corrections realignment, AB 109. And basically what the legislation did is it transferred responsibility for for many people who commit felonies from the state to local probation departments and made it much more difficult to send somebody back to prison for a technical parole violation. And essentially what we saw quite quickly was a drop in the prison population of about 26, 27 percent, and a partially offsetting increase in the jail population associated with people serving out violations or short periods of time for for technical violations in county jails.

 

Steve [00:05:00] And then the second major reform in California, Prop 47, which passed a few years later, was actually voted in by by the voters, and it redefined a series of of wobbler offenses, which are offenses that can be charged either as a felony or a misdemeanor to straight misdemeanors. And that led to an almost immediate decline of 8- or 9,000 people in the jail population on average, and then also subsequent declines in prison population as fewer people were being sentenced for felonies.

 

Steve [00:05:31] In terms of what we know about the effects on crime of those- of the earlier reform in California, Magnus Lofstrom and I have written several papers that that tried to test to see whether there was an impact of that reform on on crime rates as measured by the FBI. And we found basically no evidence of any impact of that decline on violent crime, but we did find a relatively small impact on motor vehicle theft. And oftentimes the way the the these sort of effects are summarized is how many crimes are prevented per year of prison served. And we were finding on average, about one crime per year of prison served. Now, there's lots of other research that looks at the United States at different points in time, looks at large prison releases in other parts of the world, such as in Italy and France. And generally speaking, what that research seems to suggest is that in the lower incarceration rate settings, the incapacitation effect associated with putting away someone for a year tends to be higher, suggesting that that the more intensively you use prison, the less effective it is on the margin in terms of crime prevention.

 

Jennifer [00:06:51] So policymakers and researchers alike, I think, wish we knew what optimal sentencing policy looks like, so exactly who we should lock up and for how long in order to maximize some measure of social well-being. But that's difficult to figure out even with- you know, this is a place where I think we have a relatively large literature relative to a lot of other areas in criminal justice space. So what are the main challenges in studying the effects of sentencing on crime and social welfare more broadly? Is it mainly data challenges, identification challenges? What do you see as the hurdles here?

 

Steve [00:07:22] Well, I think it's basically all of the above. Identification challenges are always central in the sense that we know that crime rates and prison populations and jail populations for that matter, tend to move in tandem and the direction of causation can go in either way, right. So it could be that there's some sort of shock to society, whether it's drugs or some change in violence or some change in health that might increase crime and it'll increase crime in prison simultaneously to the extent that the criminal justice system responds by incarcerating people. But at the same time, we know if we are incarcerating people, we have theoretical reasons to believe that that might impact crime, right. That we may be incapacitating people who are criminally active or we may be deterring people or in the steady state, we may be turning people into more criminally active people by by exposing them to to to the experience of being incarcerated. So it's very difficult to break that that sort of causal- or to kind of make that causal claim, given that that we have this this sort of simultaneous soup of things being being determined at the same time.

 

Steve [00:08:35] But I think researchers in recent years have have made some pretty good progress in the sense that there are many instances where there are factors that are kind of third factors outside the system, whether it's, you know, a sort of response to a series of inmate lawsuits or a mass release associated with advocacy by the pope or what have you, that that creates, you know, exogenous sort of independent shocks to incarceration that can then be used to study the effect.

 

Steve [00:09:12] I think, one, aside from the identification issue, another another issue that I think is quite important in this particular field is that by design, this effect should be heterogeneous, right. So you could imagine if you visit any any prison in the United States, you have very young people. You have very old people. You have people with very deep rap sheets. You have people with not very deep rap sheets. You have people that have committed property crime only. You have people who have committed violent crime and so on and so forth. And so there's a lot of heterogeneity in the population. And so, to say what the effect is of of incarcerating someone is obviously going to be heterogeneous and then trying to evaluate a policy change or an incarceration change more generally to some degree is always going to be context specific in terms of how that change has affected who is impacted and whose sentence is being altered by by the factor that is being used to kind of study the question.

 

Jennifer [00:10:18] Yeah, it's a great point about the heterogeneity. And that's an issue even if we're just thinking about what's the impact on future crime rates. I talked with Mike Mueller-Smith recently about his CJARS project and, you know, thinking there about all the other stuff that we wish we knew about this population to try to get at these broader effects of locking someone up and what the impacts on their family are and what the impacts on employment are and all of that, which most of these studies don't even attempt to get at because it's just impossible given the data that we have. OK, so let's dive into this paper. So this most recent paper that you have with Magnus is titled "The Effect of Sentencing Reform on Crime Rates: Evidence from California's Proposition 47." You already mentioned this briefly, but tell us more about Prop 47. What led to that policy change and what did it do?

 

Steve [00:11:03] Sure. So Proposition 47 was a voter initiative that was on the California ballot. There have been a number of popularly passed initiatives that have been directed at criminal justice reform in the last few years. And this is perhaps one of the most prominent. What the proposition basically did was it it redefined a series of drug possession felony offenses, and then also property crimes that could be charged as a felony, down to straight misdemeanors. In terms of drugs, most of them were were possession, and there was a weight amount involved. And then in terms of property, what what they had basically done was raise the limit in terms of the dollar value, the amount stolen, that defines the difference between a misdemeanor offense and and a felony offense. So that's the sort of main thrust of the proposition.

 

Steve [00:12:02] But aside from that, there there are also provisions to resentence people that were currently serving a sentence, whether they were in jail or prison or under a community corrections system, they could petition to have their their case resentenced as a misdemeanor if it was a Prop 47 case. And for people who had prior convictions, they could also file petitions to have their their prior convictions redefined down from misdemeanor to felony convictions. So that that's that's roughly speaking what the proposition did.

 

Steve [00:12:39] What we saw in terms of practice was- there was an almost immed- it passed in November 2014 and it went into effect basically the next day. And what we saw was sharp declines in felony drug arrests, sharp declines in property crime drug arrests, and then and then to some degree, offsetting increases in misdemeanor drug arrests, but for for reasons that are a little unclear, no real offsetting increase in misdemeanor property crime arrest. In terms of the jail population from from one month to the next, we saw a pretty sizable decline in the population of county jails on the order of, say, 13 percent of of the average daily population of county jails. And then and then, of course, we saw- and that was driven almost entirely by a decline in county jail admissions.

 

Steve [00:13:40] The other thing that that that you see, which is quite interesting, is if we move back to the realignment reforms that happened in 2011, I think I had mentioned that one of the main aspects of that reform was that people who who violated the terms of their community corrections supervision who were in prison, would be violated back to jail and not go back into the state prison system. And so it- that reform created a bit of a crisis in that many of the jails in California are overcrowded and are under their own court ordered population caps. And and essentially we we had this shock that moved some population from the state prison system into these county institutions. And and and of course, what happens when that happens is the jails use capacity releases where they just let people out early to to sort of meet those caps. And one of the things that Prop 47 did by by relieving the pressure associated with arrest for these relatively less serious offenses was you also saw a reduction in capacity releases associated with the passage of this proposition.

 

Jennifer [00:14:54] And how often did people take advantage of the option to petition to have their sentence reduced? So it sounds like this policy was effectively retroactive, but people had to initiate that process. Did we see that happen much in practice?

 

Steve [00:15:07] So that's still evolving and to some degree, there seems to be quite a bit of active effort on the part of certain organizations to inform people that they can- that they can petition and to actually help them with the process of of petitioning for for these reclassifications for their prior offenses. So there's a group called Code for America who has developed basically an algorithm for automatically reading people's criminal histories and then identifying people who have penal codes, convictions for charges where the penal code was impacted by the proposition and is making an effort to contact them and so on and so forth. So that's something that is still playing out. And I don't- there is some- you know, there are definitely reclassification petition totals that have been tabulated by an organization that oversees the California courts. But there's not a whole lot of information on volume as of yet.

 

Jennifer [00:16:10] OK, so we should think about kind of the main impacts, at least in these early years after after the reform, as being a reduction in people being arrested and booked for these kind of low-level offenses. And it sounds like — was it for property crimes? — that sounds like there was no displacement. So the police, it sounds like, just aren't enforcing some of these laws anymore? Is that the way to interpret the reduction in arrests there?

 

Steve [00:16:34] Well, that that, I guess, is what what many initially thought. I believe what I have heard from from journalists, from officers, from other people who studied the effects of Proposition 47 is true is that for for somebody who commits a misdemeanor property crime, an arrest can't be made unless the person is caught- is observed by the officer committing the offense. And so there is a procedural issue that for one reason or another was essentially — I don't know if there's a higher probable cause threshold or exactly what the what the reasoning is behind it — but there is a procedural obstacle, apparently, that makes it difficult to to make arrests when it when it isn't viewed by the officer. And for the most part, that seems to be part of what's going on.

 

Jennifer [00:17:28] Interesting. OK, so given all of that, when we think about what the likely effects of this policy reform are on crime rates and criminal behavior, what are the mechanisms we should have in mind here for why this reform might matter?

 

Steve [00:17:41] Well, I mean, there are a few different possibilities. So one clearly is, you know, if we have fewer people in jail and we have fewer people in prison as a result of this reform, and if those folks are criminally active on average, then we're basically would have less incapacitation than we than we might otherwise have. And that could lead to an increase in crime. Of course, to sort of hearken back to our conversation of heterogeneity, that depends on on who the people are and how criminally active they are, right. So, in essence, it's kind of an empirical question as to whether that matters. You know, another mechanism that that that comes up in economics, literature, and criminology literature more generally, about about the incarceration of crime link has to do with deterrent effects associated with with the expected likelihood of punishment and the severity of punishment. And in this instance, there's definitely a reduction in the severity associated with certain offenses. And there also appears to be a reduction in the likelihood that you're going to be arrested at all. So, one might suggest that that could have an impact on crime via general deterrence.

 

Steve [00:18:57] And then there's also the issue of the sort of long term effects of cycling in and out of jail and prison has on the person in question. And there are some people who believe that doing a spell in prison might actually deter someone specifically because they don't want to repeat that experience. And then there are other people who argue that actually doing time might make it harder for you to find a job, might increase your connections to other people who are prone to committing criminal offenses, might cause people to become acclimated towards social norms among incarcerated populations that might make them more likely to commit crimes when they go out and is what a criminologist would call a criminogenic effect, right. The idea that somehow spending time behind bars actually makes one more criminally prone. And if we have people cycling in and out, it could be that on average there's a higher crime rate. So those are all the mechanisms that could occur. In this particular study that that we're looking at, we can't really get at exactly which one would be in operation. But what we try to do is we try to say something about what the average effect is or the net effect of all of those factors.

 

Jennifer [00:20:03] Yeah. So you're going to look at the net effect of this policy change, and in some ways, this might sound like a very straightforward policy evaluation, but it turned out to be not so straightforward because of other stuff going on in California at the same time. As I was reading the paper, I counted three potential challenges you guys had to deal with. The first is that property crime rates were unusually low in 2014, leading up to the policy change. The second was that Los Angeles changed its data system at the same time this policy was passed, which was very inconsiderate of them, I must say. And three, the FBI changed how it defined rape during this period and that change was implemented at different times throughout the state. So first, did I miss anything on that list? And second, could you talk a bit about these issues and why they made your analysis more difficult?

 

Steve [00:20:50] Sure. Sure. So the latter two are are measurement issues, right. So the what happened in L.A. was there was some press reports that had suggested that the LAPD was undercounting the amount of aggregated assault, basically by defining when someone brandishes a weapon and many domestic abuse cases as simple assault rather than aggravated assault. And the official crime rate, the Part 1 Crimes, you know, the aggravated assault — which is the major contributor to violent crime — does not include simple assault. So that that that sort of practice would understate aggravated assault in L.A. And basically the same month that Prop 47 went into effect, a data integrity unit was introduced in the LAPD with an eye on accurately measuring aggravated assault and taking care of this problem. There was a report by the inspector general for LAPD that suggested that in the years previous, they were underestimating aggravated assault by anywhere from 30 to 40 percent. And once they introduced this this data integrity unit, you basically saw aggravated assault in the state- or in L.A. — in the jurisdiction covered by LAPD — jump 40 percent. L.A. is 10 million people in a state of 40 million people, and so you can imagine that's going to impact the overall crime rate. So- and we, you know, we try to address that by estimating models with and without Los Angeles included. So we have like a California minus L.A. and try to address it that way.

 

Steve [00:22:24] So the rape question is another interesting question in the sense that the definition of rape was was formally expanded by the Department of Justice in 2014, or 2013 I believe. And California adopted, or at least the deptart- the California Department of Justice had advised everybody to adopt, the new definition in 2014. And then different agencies sort of implemented it either in 2014 or 2015. The more expansive definition mechanically increased the number of incidents, and that also coincided with the implementation of this particular law. But thankfully, the FBI, in their tabulations of state level crime rate, also provide a legacy estimate where they use the old definition as well as the new definition that allows you to to sort of make comparisons across time.

 

Steve [00:23:18] The first issue that you mentioned about 2014, so the the the proposition passed at the end of 2014. So most of 2014 is a pre-proposition year. And one of the interesting things we saw is that that property crime in particular in 2014 was at the lowest level recorded since the early 1960s for for California. And it was actually discretely lower than what we were observing in in 2013 and 2012. And so there was an issue that, you know, if you're trying to fit a trend line and then project forward to say what would have happened in the absence of this proposition, if 2014 was an aberration, then fitting that trend line to that aberration and projecting forward is essentially going to overstate what would have happened in the absence of Proposition 47. On the other hand, if that really was the trend — that crime was just falling through the floor and that would have continued happening in the absence of Prop 47 — that's exactly what you should do. So that that really presented a problem for us when we were trying to figure out, you know, what do we say? We can look at what happens to crime before and after. But but how do we get some sense of what would have happened in the Parallel Universe California, in the absence of this intervention?

 

Jennifer [00:24:39] Yeah, so I'd love for you to talk a little bit more about how the sausage got made here in terms of the research process. So a lot of our listeners, you know, don't do research themselves or are students just embarking on research for the first time. I know you pay close attention to criminal justice policy in California. So in this case, I mean, did you know about all these issues going into this study or did you have to do some digging when you saw your results in Los Angeles, for instance?

 

Steve [00:25:02] Well, we sort of discovered them along the way. So so they're- the aggravated assault result, other researchers had noted in looking at data that there appeared to be a big jump in ag assault around, you know, between 2014, 2015. And and it's very visible in the data where you see it just go up from one period to the next. And it did cause us, given that that we didn't really see the same jumps in anything else, it sort of gave us pause. And so what we started to do is we, you know, we just got the the the county level data for every county in the country. And then we even got data for every single law enforcement agency in the state and just looked agency by agency to say, OK, where is this happening? Is this something that is happening all across the state, reflecting maybe some deeper factors at one agency driving the trend? And that's when we discovered it was L.A. And so when we saw L.A., then we called, you know, LAPD and and asked the people who are basically in their crime analysis unit, what- if there's if there's some reason why this might have occurred? And they said, oh, yeah, we had a big issue here with aggravated assault. And there was an L.A. Times report and there was a- there was an inspector general report. And then we made this effort to to sort of change it and make sure that we got it right. And and so we you know, just just by virtue of just kind of pushing the data and trying to make sure that we weren't seeing something that was sort of an artifact of the way data was collected, we uncovered the the L.A. issue. So that that that's basically how that that came to be.

 

Steve [00:26:40] In terms of the unusually low crime — for the low property crime for 2014 — that was another issue where it just looked out of the ordinary. However, that was a pattern that we were seeing that was pretty much broad based across the state. So it wasn't the case that there was one agency driving it, or there was at least to the best of our knowledge after after consulting people who curate that data and collect from the localities, there was no change in the way property crime was measured from one period to the next. And so it just looks like crime was just lower in 2014. So these are just- I think when you're doing empirical work, sometimes when you see an abrupt change, it could be an effect, but it's sort of incumbent upon you as the researcher to make sure that it's not not not something mechanical happening. And then when when you start digging, you know, I mean, of course, it's to some degree a little bit of sausage making where- when do you stop digging? But, you know, when you've exhausted all your leads and nothing nothing smells like fish, then you can... *laughter*.

 

Jennifer [00:27:47] Right, yeah, I mean, in some ways, this is the fun part of research, right, like learning along the way about all the other stuff going on. OK, so let's dig into the analysis now. So because of all these issues we just talked about, you guys used three different empirical strategies to paint a picture of what the effects of Prop 47 were. So could you give us a brief overview of these three strategies and how they measure the causal effects of the policy change? And then as you do that, I think it'd be really helpful if you could kind of step through what you see as the strengths and weaknesses of each strategy.

 

Steve [00:28:17] Sure, sure. So the first thing we did is we we you know, we wanted to say, OK, well, we could observe what's happening in California. And then one might ask, well, what happens to the states next door? What happens to states in the rest of the country? And is the trend for California somehow differing from the trend for for other places? Now, of course, that that kind of raises the issue of, well, what other states do you use? Would you use a neighboring state? Would you use some average of other states? How do you pick the average? And it turns out that there is a method called Synthetic Comparison Estimator, which basically allows you to select a group of states that best matches the sort of crime trends of California before an intervention, right. Where it's essential you allow the data to tell you which group of states best matches. And then you can compare what happens in California to what happens in what people refer to as Synthetic California, right — this average of states that match California pretty well in the pre-period — and what happens in the post-period, right. And so if it's the case, for example, that 2014 was, even though crime was really low, if we had a bunch of other states where they also had a low crime rate in 2014, we can then see what happened to those states in 2015 to see whether it was just a continuation of trend or whether that was the trend or whether that would have ebbed in the coming year, right. But what it basically does is it allows you to generate something to compare California to. Now now, of course, there are there are- so that that's a strength in the sense that, you know, you have your Parallel Universe California, right. You have your California without Prop 47. But I think that there are potential pitfalls in the sense that, let's say, for example, there are similar policy changes in those other states and you don't observe it or you as the researcher are ignorant of them. You know, so I know a whole lot about what's going on in California, but I have no idea what's happening in Arizona, right. And or I have a passing familiarity with things that are happening in other states where I don't do as much research. So to some degree, you're assuming that California is unique and they don't have a realignment Prop 47 there as well, right. And so that that's something that has to be looked into.

 

Steve [00:30:46] Another issue is that it's- even though there are methods to sort of look at whether you see any departure pre-/post- California and say whether it's statistically significant, it's a little bit more challenging to draw inference in this setting. And what people tend to do is they'll they'll take the same estimate or apply it to the all the other 49 states as if they had the same policy reforms as California and then estimate basically a placebo estimate for each one of those states. And then ask, you know, does California stand out relative to the other 49 estimates where we know no policy change happened? If it does, then you can assume it's it's statistically significant. The problem there is that tends to be kind of a low-powered test, right. So, you know, it might be the case that there are significant effects that that are there, but they're of a size that they're below the minimum that you can detect, given the fact that you just have this weird placebo distribution of 49 other estimates. So those are those are some of the sort of strengths and weaknesses there.

 

Steve [00:31:57] The other things we did is we simply used- we sort of assembled monthly data for California for the months previous to the proposition's implementation and then the months after the proposition's implementation. And we basically use the pre-intervention trend to project out what would have happened, right. And that that essentially assumes that that 2014 low crime rate is essentially an indicator of what would have happened in the future. And then we we sort of estimated how much crime was incapacitated by looking at just the discrete change in crime around November 2014. And then also the difference in crime between this projected path and what actually occurred over the whole subsequent year. The benefits of that to some degree are, at least with this high frequency data, there was a very sharp change in the jail population. There was a very sharp change in arrests and and changes in incentives to some degree. And so at that that break in time, you might think, OK, there's a large shock here. It's salient. Perhaps we can identify something here that would be you know, this is this is the immediate effect of of of of the proposition on crime rates that are occurring in the neighborhood. Now, of course, it might be the case that over time the police may adjust their practice or probation, or whoever are the criminal justice actors, may essentially incorporate these new constraints into their policies and practice and they can undo any negative impact. So if you project that forward, it could be incorrect, right. You're making an extrapolation from a small area of the small time interval around the intervention and you're pushing it forward. And the further and further out we project from using the pre-intervention trends, the less confidence one should have in the quality of that projection. Right. So the degree of uncertainty just grows and that that's just another issue.

 

Steve [00:34:02] And then I think that the most important issue with, or limitation here, is simply that we have this odd year in 2014 and using that to project out what would have happened may be the right thing to do or may be the wrong thing to do. And there's really no way to assess that. Although we do try to say, well, what happened in other states and in other states, we saw a decline, but no further declines. And so it sort of casts a little bit of aspersions on our second method.

 

Steve [00:34:31] And then the third thing that that we did is a method that I've used in several projects in the past and one one with Magnus, and that's namely to exploit the fact that different counties were impacted differently by this proposition. So, for example, in California, Central Valley, Fresno, Kern County, so on and so forth, the incarceration rates are very high and the criminal justice system seems to be more punitive than in the coastal areas, in particular in northern California. So, for example, before real- on the eve of realignment, San Francisco had had an incarceration rate in terms of the people from San Francisco in state prison. That was about 200 per 100,000. And Kings County had it over 1,000 per 100,000. Right. So you have everything from what appears to be a European country to what appears to be the highest incarceration rates in the United States, all in the same state in these counties. And what you would expect is in that setting, when you introduce a reform, the places like Alameda County and San Francisco County that that don't use prison and jail systems as intensively as other areas are not really that impacted, while the areas that use them more intensively are really impacted. And you can measure that by changes in in jail commitments, changes in jail population in those areas, change of prison population in those areas and so on and so forth. And so what we do is we we basically identify counties that experience big decreases in their jail populations per per capita and then compare those counties that have smaller decreases and see whether or not crime increased by more in the counties experiencing larger, larger increases. And that's pretty much our method.

 

Jennifer [00:36:21] OK, so what data do you have at your disposal for all of this?

 

Steve [00:36:26] Almost everything that- I think actually everything that we use in in this particular dataset was publicly available. So the crime data are Part 1 Incident Defense data that's reported by every law enforcement agency in the United States to the FBI. It's usually first sent to the State Department of Justice. The State Department of Justice compiles it and sends it to the FBI. And so there's a dataset one can download known as Offenses Known and Cleared by Arrests — I can't remember what the exact title is — and basically what the data will have is a total for every single month for for many years, for every law enforcement in the agency of Part 1 Offenses which are murder, rape, robbery, aggravated assault, burglary, motor vehicle theft, and larceny and then- and arson. And then they also have a series of of Part 2 Offenses as well. And then also some data on arrests. And you- basically we use that. That's that's the data we use in various forms, sometimes aggregated at the county level, sometimes aggregated at the state level, in some instances aggravated state level, ditching L.A. for the reasons that we already had mentioned. And and then also, you know, data in comparison states as well. So the same data for other states to to generate counterfactuals for California.

 

Jennifer [00:37:53] I thought you were going to say you had all the data that California had. I know you have access to some great California data, so it's very cool that you didn't need to use it here. OK, so so what are the main results for your your three different strategies?

 

Steve [00:38:08] Well, there's a little bit of heterogeneity across strategy, though not much. Pretty much they they all failed to find evidence of any impact of these changes on violent crime. So so in essence, it seems to be very similar to what we found with realignment that we couldn't really find any impact on on violence. And I think the other thing that we see is there does appear to be a somewhat small effect, anywhere from a 3 to 7 percent increase in larceny theft, which is the one Part 1 Offense that is perhaps the the least serious, but it's also the one where we- one would expect there to perhaps be an impact, given the nature of the of the crimes that were targeted by the by the by the proposition. They all- you know, there's some heterogeneity across them. So, for example, in the you know, when we looked at the synthetic comparison estimates, we actually find some some results suggest murder declined relative to comparison states, but we don't see it in the other two. When we were projecting using pre-intervention trends, we found some evidence that robbery might have increased. But we didn't see that with the other two estimates or the other two methods. But pretty much in all of the methods we found we found evidence that that largely theft had gone up.

 

Steve [00:39:37] Oftentimes when I talk to audiences in California about what the research seems to suggest, we're very interested in trying to precisely measure an impact associated with a given shock. And so we use language like robust and statistically significant, things like that. And I think in this instance, one could say, yeah, we- you know, that the effect seems to be robust on property crime in the sense that it is not sensitive to these alternative specification choices and methods being used — that it peers out of the cloud of numbers that were thrown at you. But that doesn't necessarily mean that it's large. So what what I often- you know, a set of figures that I like to show people when they ask what the net effect of all of this has been is you could look at a long term time series of the incarceration rate in California. And basically what you see is, you know, the incarceration rate rising rising rising rising til about 2006 and plateaus, hits 2010, and then it drops. And we we probably have- the state has erased 20 years of incarceration growth basically, in terms of the rate. So we're back down at around 1990, 1991 levels. And then you can juxtapose that against crime trends where what we see for violent and property crime rates again parallels the country, that there are increases through the 60s, 70s, and 80s, peaks that are around 1990 and then starts to decline. And when we look at the last 6 or 7 years, what's happened since 2011 or so, basically it just looks like tiny little wiggles at the bottom of a of a of a very, very, very long down-sloped trend. So we've reduced our incarceration trends back to what it was in 1990, which was much lower than what it is today. But our crime rates are are at early 1960s levels. Right. They really haven't budged all that much. And the bottom line seems to be we probably have more larceny theft than we otherwise would have had, but our crime rates are still at historical lows.

 

Jennifer [00:41:50] Yeah, and you guys you don't have a formal cost benefit analysis in the paper, but people surely do point to this tradeoff. Have you crunched the numbers at any point? Do you have any sense of kind of what the cost savings is from that big drop in incarceration? I mean, incarceration is expensive, so dropping incarceration at all saves a whole lot of money. But then you do have this, you know, 5 percent-ish increase in larceny, which is not nothing. So so how does- where does that leave us in terms of- I mean, it sounds like your general view is that this increase in crime was worth it, but do you have a sense of how worth it it was?

 

Steve [00:42:23] Well, it's a little bit more complex. Well, I guess anything's complex. So we have done at least in our realignment work, we did make some effort to do cost benefit calculations. And, you know, one couldn't find numbers on the average expenditures per year per inmate. And then, you know, where we were actually finding, at least for realignment, that there was an impact of motor vehicle theft, you can come up with an estimate of the cost of that crime. You could, you know, say, well, this much is prevented by incarcerating the average person for one year and this is how much it costs then you can come up with a benefit cost ratio. And if you go through that exercise, at least for that period, it was on the order of 20 cent return on the dollar. So it didn't appear to be a cost effective.

 

Steve [00:43:14] But but that being said, I think, you know, I mean, I love to give people a straight number. My dog's barking at me because it doesn't like my- she-. But but I think there are other complications in the sense that that, you know, the marginal cost of incarcerating somebody is not the same as the average cost of incarcerating somebody. And essentially, we could have a situation where, sorry, my dog is howling. I don't know if you can hear that.

 

Jennifer [00:43:50] Very, very faintly.

 

Steve [00:43:52] OK. All right. But at least in California, you know, you don't really save any money unless you shut a prison. Right. And and I think that that sort of marginal reductions in population, given the large fixed costs associated with incarcerating somebody, it's not clear that you save on average the $80,000 a year. I do think the the the one thing that you can say, however, is that relative to the counterfactual of what the state would have had to have done to come into compliance with the court order. Where, you know, to get from 200 percent rated capacity to 135 percent rated capacity, which is what the court was demanding, you could either reduce the population or you could build a whole series of new prisons. Right. And and so it's just a it's just a fundamentally difficult enterprise to do. And we have made some attempts at it, but I don't know how much confidence I have in those numbers.

 

Jennifer [00:44:56] That's very responsible of you. OK, so this paper came out sometime last year, I think. Have there been any other studies since that time about the impacts of either California's decarceration efforts or sentencing reform more broadly that you think are relevant to this conversation?

 

Steve [00:45:12] There certainly are a few studies that are relevant. There's another paper by Charis Kubrin and colleagues at at UC Irvine, that also looks at Prop 47, finds things that are pretty similar to what it is we're finding. Right now, we're working on on some work and it's still preliminary, but we're trying to assess whether or not statewide these reforms have have led to narrowing of racial disparities in criminal justice outcomes. And in particular, what we're looking at are differences in arrest, differences in incarceration, differences in bookings and so on and so forth. And hopefully we'll have that that done soon. Where we're- it's not so much a criminal justice outcome, but it's it's trying to understand what the the sort of social equity issues and and some of the the kind of inequality issues that are, of course, paramount concern in U.S. corrections if it's impacted by this reform.

 

Steve [00:46:15] And then along those lines, John MacDonald and I did a study for the San Francisco district attorney where they were principally interested in racial disparities in how their office was processing cases. But the data they gave us actually sort of spanned the time period where Proposition 47 was implemented. And we saw quite interesting patterns, if you just looked at the pre-period and the post-period. That that- to summarize, basically what you see is prior to Prop 47, in terms of cases presented to the San Francisco District Attorney's Office, the African American defendants were more likely to be detained pretrial and for longer that that to some degree narrows quite a bit relative to what you see for white defendants in the post Prop 47 period. And you also see that when you look at case outcomes, the sentencing disparities narrow pre-/post- Prop 47, and you can almost attribute that entirely to less weight being placed on on having a prior criminal history or not- or being detained pretrial in the process and it having a disparate impact by race. So we're we're kind of hoping to sort of build in that direction a little bit more to try to understand some of the other dimensions of how this reform are impacting society and impacting Californians more generally.

 

Jennifer [00:47:51] Yeah, that's really interesting. So based on all of your research in this area and the other work in this area, what's the main takeaway for policymakers thinking about both how to reform sentencing in California and policymakers in other states that also want to reduce their incarceration rates?

 

Steve [00:48:08] Well, the big picture seems to be that that it is certainly possible to reduce the use of incarceration in this country without having a huge crime wave- I mean that's essentially what happened in California, so we have have proof by concept. And you don't even need a lot of fancy research to show that because it's just evident right there in trends that one could plot out and look at. You know, so in terms of policy, it does suggest that there are many ways to approach public safety and the incarceration route that many localities and jurisdictions have pursued over the last four decades is probably, you know, gotten beyond the point where it's as effective as other ways of doing things and is likely doing damage in the process. So I think that that's one thing that comes from California, right. So California- I mean, right now the state has reduced its incarcerated population by roughly a quarter, right, in a very short period of time. And there doesn't appear to be much by way of additional crime as a result.

 

Steve [00:49:11] I do think, you know, in terms of a research frontier, there's just so many things that people could be studying in the state because you have, you know, several large pronounced exogenous changes that have occurred both in terms of institutional populations as well as arrest probabilities and booking probabilities and all these sorts of things. So I always whenever I whenever I talk to grad students, I just try to encourage people to pay attention to what's happening here, because there's there's lots that that can be done and a lot a lot to be learned from from these experiments. And I think the more the merrier. Right. It would be a very good place for for graduate students to look for for PhD dissertations and for people just interested in just the interplay between the criminal justice system and society more generally to find testing grounds for various hypotheses. So I would encourage people to to jump in the pool.

 

Jennifer [00:50:10] Yeah, I totally agree. Are there other states in the U.S. that you think of as being kind of leaders in terms of experimenting like this, or is California way out ahead?

 

Steve [00:50:19] I think California is way out ahead. There there certainly are- I mean, reform is is sort of in the air in the United States. And I think in the last the last decade or so, we've seen a number of things, everything ranging from community corrections reform to redefining, you know, the dollar value for for felony larceny and so on and so forth that are all kind of tending in the same direction. Right. Trying to figure out how we could do less of this and not impact public safety. But that being said, I do think nowhere has experienced the kinds of declines that we've seen in California. So we even I think I've done some back of the envelope calculations that if you look at the declines in the incarceration rate that have happened, you know, say from from 2010 to 2017, I think California drives about 40 or 50 percent of the decline. So it's very much a unique situation, in part driven by legal battles and- but also in part driven by our voter initiative. Right. And the fact that the public mood has shifted and seems to be predisposed towards passing reforms that that are scaling things back somewhat.

 

Jennifer [00:51:40] My guest today has been Steve Raphael from UC Berkeley. Steve, thanks so much for doing this.

 

Steve [00:51:45] Thank you for having me.

 

Jennifer [00:51:51] You can find links to all the research we discussed today on our website, probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts to make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show. Our sound engineer is Caroline Hockenbury. Our music is by Werner, and our logo is designed by Carrie Throckmorton. Thanks for listening and I'll talk to you in two weeks.