Recorded on March 18, 2025, this video features a talk by Alexander F. Roehrkasse, Assistant Professor of Sociology and Criminology at Butler University. Roehrkasse’s research focuses on inequality, victimization, punishment, families and children, and quantitative and historical methods. His work has been published in the American Sociological Review, Demography, Proceedings of the National Academy of Sciences, Science Advances, Social Forces, and other leading journals. He received his Ph.D. in Sociology from UC Berkeley.
This talk is part of a symposium series presented by the UC Berkeley Computational Research for Equity in the Legal System Training Program (CRELS), which trains doctoral students representing a variety of degree programs and expertise areas in the social sciences, computer science and statistics. The talk was co-sponsored by the Berkeley Institute of Data Sciences (BIDS).
Abstract
The dynamics of inequality in mass incarceration are rapidly changing and poorly understood. In this talk, I present new evidence of declining Black–White inequality and skyrocketing educational inequality in U.S. prison admissions. I qualify these findings by documenting vast racial disparities in indirect contact with the carceral system through families and neighborhoods. I conclude by discussing possible causes of recent inequality trends and potential research strategies for identifying them.
Podcast and Transcript
Listen to the podcast version of this talk below or on Apple Podcasts.
[MUSIC PLAYING]
DAVID HARDING: Welcome, everyone. My name is Dave Harding. I’m a professor in sociology and faculty director of the CRELS program, which is an acronym for Computational Research for Equity in the Legal System, which is a National Science Foundation-funded training grant for PhD students here at Berkeley. And today is the first talk in our spring speaker series for the CRELS program. And our talk today is co-sponsored by the Berkeley Institute for Data Science, or BIDS.
And just a quick plug for our next talk, which is April 1, and AJ Alvaro from Cornell is coming to talk then. We hope to see you then right back here on April 1. Today we’ll hear from Alex Roehrkasse, who is assistant professor of sociology and criminology at Butler University and a Berkeley PhD graduate. We’re very excited to welcome him back.
His talk is titled “The New Contours of Mass Incarceration,” and his research focuses on inequality, victimization, punishment, families and children, and quantitative and historical methods. His work has been published in the American Sociological Review, Demography, Proceedings of the National Academy of Sciences, Science Advances, Social Forces, and other leading journals.
So please welcome– please join me in welcoming [Alex Roehrkasse].
[APPLAUSE]
ALEX ROEHRKASSE: Thanks. Thank you. OK. Thank you so much for being here today. I’m really excited to be here. As Dave said, I was trained downstairs on the fourth floor, and so it’s fun to be back in the Social Sciences Building. I want to thank Dave and Harpreet for running this very cool program. I want to thank Sarah for making my visit possible.
So I don’t know your preferred rules of engagement here, but I’m very happy to take questions on the fly. So if you have a question as I’m talking, particularly if anything is unclear, please just flag me down. I’m happy to handle questions on the fly. My talk today is about what I understand to be a pretty big change in the contours of inequality in imprisonment in the United States.
Most of what I’m going to be talking about today are a variety of findings that come from two papers I’ve published recently with a co-author, Chris Muller, who’s formerly of the Sociology Department here, now at Harvard. Chris and I are still working in this area. We’re following up on some of these results. I’ll talk a little bit about some work in progress, really more about future directions we’re hoping to take this work.
And my goals for the talk today are really threefold. First, I want to describe some of these inequality trends in prison admissions in the United States, particularly regarding racial inequality, class inequality. I’ll talk about class inequality. Really, mostly, I’m going to be talking about educational inequality today. We can talk about. Whether or not that’s a good proxy for class. But empirically speaking, I’m mostly going to be talking about educational inequality.
And then I’ll talk through some candidate causes of these inequality trends, some things that might be driving some of these trends. I’ll talk a little bit about some of our early efforts to understand these causes. Though, as you’ll see, what I mean by a cause in each case is a little bit different. And then I want to honestly offer a pretty big caveat for a lot of the initial evidence that I’ll present to you.
And I’ll talk about how actually, a lot of these evidence– pieces of evidence about prison admissions are meaningfully complicated by evidence about inequality in vicarious contact with the prison system through family members, through neighborhoods. And so I’ll talk through some seeming paradoxes in inequality, in the lived experience of mass incarceration. I’ll introduce some ideas that I think are helpful for reasoning through some of these paradoxes, thinking about the lived experience of incarceration more holistically.
So regarding these changing inequality patterns, how do social and legal scholars currently think about these inequalities? What are Chris and I in these papers doing to intervene in this conversation? What do we learn from these interventions? At the risk of gross oversimplification, I do think it’s still helpful to describe, on the one hand, a fairly predominant narrative about mass incarceration over the last half century or so that was very much popularized by the 2010 publication of Michelle Alexander’s The New Jim Crow.
And Alexander draws our attention in particular to the war on drugs as a campaign of racialized social control, one that’s functionally consistent with prior regimes of racial domination. But Alexandra’s account really draws our attention to disparities in the prison system. The disparities that most warrant our attention and warrant action are racial disparities. And the primary drivers of these disparities are things like drug enforcement and drug sentencing. So just a matter of emphasis, but a clear emphasis.
This narrative is not without its critics, as I’m sure many of. And so on the other hand, a variety of scholars who have pointed to a number of facts that are somewhat inconvenient for this narrative. So for example, the role of the Black political leadership class in driving or shaping a lot of America’s punitive turn, or the fact, as John Pfaff has pointed out, that fewer people in American prisons are there for only nonviolent drug offenses than we sometimes assume.
So these critics are by no means in consensus about everything, but they tend to believe that Alexander’s account undersells or downplays the role of violence and violent crime in explaining inequality trends in American imprisonment, and that they also– and that the account also sort of downplays the extent to which poor people, lower class Americans, of all racial identities have been targets of the prison system.
I think this debate has been helpful in many respects. But to be honest, I think there are two really important limitations with this debate, at least as I’ve construed it. One is that the evidentiary basis for some of the claims making about these inequalities is sometimes a little bit thin, or at least unsystematic. And so one of the things we’re trying to do here is just bring more systematic data to this debate about racial and class inequality in imprisonment.
Another problem though is that it’s maybe a little bit out of date. So Alexander and many of her critics were writing from what we now know was a high water mark of mass incarceration, or at least I hope it is. And so now, looking back from where we stand now, this begs some questions about inequality patterns in this– what appears to be somewhat of a new era of modest decarceration. So we’d like to know how things look more recently.
And so just on a very basic level, we’re going to start from a set of questions about how it is that racial and class inequality in prison admission rates in the United States have changed in the 21st century. How does this compare to trends that we know more about in the late 20th century? Credit where credit is due, our approach is very much informed by the earlier work of Bruce Western, who, in his 2006 book, documented pretty high levels of racial inequality in prison admissions. So these are rate ratios, say, comparing the Black prison admission ratio to the White prison admission ratio.
So high levels of racial inequality, but levels that were relatively stable over the period that he was examining. While on the other hand, even higher and indeed increasing levels of educational inequality measured in terms of the rate ratios of people who had not attended college compared to people who had attended college.
Bruce’s work is, I think, grossly understudied in some of these debates or undecided in some of these debates. And it’s really helpful, I think, in adjudicating claims about the late 20th century. But to a very large extent, we’re extending Bruce’s methods, extending the data that he used into the 21st century and seeing how it changes some of our stories about mass incarceration. So how did Bruce create this figure? What did he do to actually get here, and what do we do to update this figure?
So like him, we’ll use– our primary source of data is going to be the National Corrections Reporting Program. The NCRP is really the primary administrative data resource for sociodemographic information on people entering prisons in the United States. So information about people’s race, ethnicity, educational attainment, things like this. This is really the best place to go for information like this. Yeah, clarifying question.
AUDIENCE: Yeah. I’m wondering if you break it down at all by state?
ALEX ROEHRKASSE: We’ll talk about geographic inequality– I’m happy to talk about rural urban divides a little bit more. Hold on for just one second. I’ll talk a little bit why that’s difficult. Yeah. So the NCRP though is– only tracks admissions to state prisons. So I’m not going to talk about federal prisons today. It’s important to acknowledge dynamics that are actually pretty different, so other people should do that work.
The main problem with the NCRP, and I just want to be transparent that this is probably the biggest limitation in our study, is that it’s– the reporting has actually gotten quite good in recent years. The NCRP captures almost every state every year these days, not so much true in the 80s and 90s. So there are a lot of states that are not regularly reporting to the NCRP in earlier years, OK?
In a variety of sensitivity analyzes, what we do do is try to impute missing values of variables in observed or reported prison admission records. So if we have a record and there’s missing information in it, we’re often imputing that information, missing information about people’s ethnicity, more often, their educational attainment. We find that of strategies for dealing with missing values, our results are not really sensitive to different reasonable approaches.
But this is different than missing records altogether. So there are very many records that just don’t show up in the NCRP. And I want to be clear that we are not trying to impute the counts of records that don’t get reported to the NCRP. If we were to use the NCRP just to tabulate admissions to state prisons, we would have wildly erratic figures just as a function of irregular reporting to the NCRP.
So what we do instead is supplement our approach with data from the National Prisoner Statistics Program, and this is a much more reliable count of all people admitted to prison in any given year. The problem is, we don’t have the same level of sociodemographic detail in those data as we do in the NCRP. So trying to leverage the demographic detail of the NCRP and the reliability of the NPS and combine them to get some sort of count of admissions by race and educational attainment.
So we’ll use these sources to count admissions, and then we’ll calculate population rates of admission using population data from the merged outgoing rotation groups of the current population survey. OK. So what does this actually look like? How do we marshal these data? So our analysis is going to start in 1984 when the NCRP gets decent, and then it’ll go through 2019, which is the last year for which we have data from the National Prisoner Statistics. So I want to be clear that we don’t know as much about what’s happened in the last five years. Things could be quite a bit different, honestly, over that period. So I want to be clear that we’re– I might talk about today. But really, by today, I mean 2019. We only have data up through 2019.
We’re going to focus on Black and white, non-Hispanic people. We’re working on incorporating more Hispanic people into our analyzes, but I won’t be showing any results for Hispanic folks today. And then we’re going to subdivide the population into two education groups. And we’re going to slice the educational attainment distribution at the margin of whether or not you’ve attended college. So if you attended a college and not necessarily gotten a degree, have you ever attended college? Have you spent any time in college or have you spent no time in college?
This isn’t necessarily the education margin we would have chosen if we had our choice. Although it ends up being a pretty important one, a pretty interesting one. We’re working on analyzing other margins, and this does actually seem to be a pretty big one. This is a margin at which a lot of inequality is actually arising. The main reason we do it is that this is the most– the margin for which we have the most consistent measures over the whole period of our analysis. It’s also what Western did, so it makes our results comparable with his. That’s always nice.
And so the way we’re going to combine these two pieces of– two sources of data to get account of prison emissions is that in any given year, we’ll look at, what proportion of all admissions observed in the NCRP, whether it’s from 20 states or 45 states or 50 states, what proportion of prison admissions observed in the NCRP correspond to each racial and educational group? So what percentage of observed records belong to, say, Black Americans who had attended college or white Americans who had not attended college?
So we’ll calculate a proportion corresponding to each year racial and educational group, and then we’ll multiply that proportion by the much more reliable count of all prison admissions in any given year. And so by multiplying the NCRP proportion by the NPS count, we’ll get an estimate of the number of admissions for each year, each racial group, each educational group. And we’ll just use our population data to calculate a prison admission rate. So nothing too fancy.
What do we find? So here are our prison admission rates annually from 1984 to 2019, stratified by racial group and educational group. I think most striking is obviously the very, very high prison admission rate for Black Americans without a college education over the whole period of our analysis. Obviously, this meteoric rise in the late 20th century. But it’s fallen by about half gradually over the last 20 years or so.
Also striking is this initially low but steadily increasing rate of prison admissions among white Americans without a college education. This is absolutely a large increase, but because of the low initial levels, proportionately, it’s a huge increase from about 200 per 100,000 to about 1,200 per 100,000. So it’s about a six-fold increase for that group. It’s a pretty striking increase. Obviously, much lower admissions among people who have been to college, so it’s a little bit hard to interpret these trends. I promise, I’ll get you there.
But you can see that for no college or, sorry, any college Black Americans, the rate has been decreasing since about 1990 or so. It’s hard to even see any change for white college-educated Americans, but it’s actually steadily increasing over the whole period. OK. This is talk about inequality though. So how are we going to measure inequality? How do we think about inequality in prison admission rates?
We think it’s most helpful to think in terms of rate ratios. So I’ll be showing you evidence where we divide the Black prison admission rate by the White prison admission rate in any given year for any given educational group. And then conversely, we’ll divide the no-college rate by any college rate for each racial group. What are these rate ratios look like?
Obviously, in the early part of our analysis, racial inequality, at least between Black and white Americans, was pretty extreme. So the pattern here is the same. Obviously, inequality is a little bit higher for people with any college. But Black Americans were roughly eight times more likely to enter prison in any given year than their white counterparts in the early part of our analysis.
Obviously, and quite happily, this has fallen considerably. So in 2019, the numbers are much lower. But I really want to emphasize in this talk that these numbers are still very high. So among people without a college education, Black Americans are still about twice as likely as their white counterparts to go to prison in any given year. And for people with a college education, Black Americans are about 2.6 times more likely than their white counterparts to be admitted to prison in any given year, OK? So it’s just, we need to be careful when we talk about these things. Huge decrease, still really large disparities.
The story for educational inequality is quite a bit different. So in the early period, disparities between people without a college education and with a college education, both among Black Americans and white Americans, were roughly comparable in scale to racial disparities. So in the early part of our analysis, I think we can say that, at least as we measure it– it’s a little bit measure dependent. But at least as we measure it, racial inequality, educational inequality, we’re roughly comparable in scale. But of course, they have very different trends.
This is a logarithmic scale, so that’s important to note. By 2019, among Black Americans, people without a college education were 26 times more likely to be admitted to prison in any given year than their counterparts who had attended college. And among white people, people who had not attended college were 32 times more likely to enter prison than people who had not gone to college. So really, just enormous disparities.
I’ll give you a minute to take pictures. [LAUGHS] I’m happy to send you any of these slides though. OK. So these are just the preliminary descriptive findings about overall inequality. But what’s driving these trends? I think obviously, the causes are many. It’s probably going to take a generation of research to really figure out what’s going on here, but I do want to talk through for candidate causes here that I’m thinking about that Chris and I have published a little bit about that’s guiding some of my research going forward. So four things.
The first is the role of specific offenses in contributing to prison admissions in different ways. So what kinds of offenses are driving these inequality trends? Second, how might changes in the educational attainment of Americans over this period, which have been pretty substantial, drive or maybe just confound some of our analyzes? How would we want to think about that fact?
Third, Americans are not sorted into different communities randomly, different kinds of people live in different kinds of places systematically. And we know from a research including people in this room that subnational criminal legal system reform is uneven across the country. And so maybe it is that some people in our analysis are experiencing criminal justice reforms that other people aren’t.
And then lastly, I want to speculate about some technological economic changes over this period that might be driving differential change in people’s risk of coming into contact with the criminal legal system. OK. Regarding offenses and how different offenses might be contributing to these inequality trends. Recall that part of what was at stake in this debate, at least as I described it, is not only the relative salience of racial and educational inequality, but also the kinds of offenses that were driving it. How important were drug offenses? How important is violent crime to explaining some of these stories? So we’ll just ask straightforwardly, how did specific offenses contribute to some of these inequality patterns?
To do this, what we’ll do is group prison admissions according to the offense for which people are admitted to prison, and we’ll use the Bureau of Justice Statistics classification scheme for doing this. BJS tends to classify offenses into five different categories– drug offenses, violent offenses, property offenses, public order offenses, and then a catch all other category. It’s important to note that admissions get classified according to the most serious offense. So that’s the offense with the longest sentence.
Many people are admitted though for multiple sentences, and so this does generate some complications. I’m happy to talk about them in the Q&A. If we don’t think carefully about how admissions are classified according to offense, we might lead us to some erroneous inferences. But it’s important to understand that the estimated prison admission rate for each year, racial group, and educational group is just the sum of all of the offense specific rates for that group. So we just add up the offense-specific rates to get to the overall admission rate. And then the rest of our calculations are just the same as before, except that we’re indexing our count of admissions by the offense category.
AUDIENCE: [INAUDIBLE]
ALEX ROEHRKASSE: Yeah. Like disturbing the peace. I think that resisting arrest goes in there, not in the violent category. Things where causing trouble but not necessarily harming someone or don’t intend to harm someone. Drunk driving might be in there. You’ll see it’s– I’m really going to talk about the first three. The last two just actually don’t matter all that much, so I haven’t thought all that much about it. Yeah, I don’t actually know too much what goes into that category. Yeah.
AUDIENCE: Small stuff.
ALEX ROEHRKASSE: Small stuff, yeah. Well, but these are people who are going to prison, not jail. So it’s not, strictly speaking, really small stuff. Yeah.
AUDIENCE: Just not violent.
ALEX ROEHRKASSE: Yeah. OK. So what do we start to see when we stratify our previous analysis by the offenses that lead people to be admitted to prison? So here you see, for each of our four groups, each ribbon represents an offense-specific admission rate. And these ribbons are stacked. So you can think of the top of the top ribbon as being the total admission rate that I showed you in the previous slides, but now you just see it broken down by the different offenses that are contributing to it.
So note that each panel has a different scale here, so we shouldn’t be comparing the panels. Really, what we’re interested in here is like, how do these ribbons sort of compose the total admission rate? And there was an important thing to take away from this slide, is that it really is these three offense categories, namely drug, violence, and property offenses that are contributing to the bulk of admissions, but also the majority of change in admissions. And so it’s really those three offense categories that we’re going to focus on, just for simplicity in the rest of our analyzes.
What do we see when we start to compare the scale of admissions though across these different offense categories? Well, in some cases, say for white Americans without a college degree, the offense-specific patterns look pretty similar to one another, and therefore also pretty similar to the overall admission rate. So not a lot of interesting things happening for white no-college-educated people, for example.
In other cases though, there are some interesting qualitative divergences. So if we look at the no-college Black American group, the highest lines here, we see that for drug offenses as well as property offenses, admission rates have declined pretty substantially, although, obviously, at different rates. This is not true for violent offenses. Admissions for violent offenses among Black Americans with no college education have kind of leveled out over the last 20 years, not showing too many signs of change.
One consequence of that leveling out is that racial inequality in admissions for violent offenses has decreased much less than it has for drug offenses and property offenses. Less interesting to say about inequality in educational or educational inequality and admissions stratified by offenses. But honestly, this is not why we did this stratification. What we really want to do is decompose the overall admission rate and ask, to what extent did the offense-specific rates contribute to overall inequalities? How much can we attribute racial inequality, how much can we attribute educational inequality to drug admissions, to violent admissions, to property admissions?
So how do we do this? Define some set of offenses O prime. That’s essentially those five offenses that I said minus any particular offense. So O prime is just all the offenses less offense O. And then we’ll define a counterfactual rate for Black Americans where what we do is substitute– let’s say we’re talking about drug offenses. We’ll substitute the white admission rate for drug offenses but retain the Black rate for all the other offenses.
So essentially, what we’re doing here is asking how much the total racial or total educational inequality in admissions would be different if admissions for one specific cause were equalized? We do the same for no college people where we substitute any college rate for one specific offense, retain the no college rate for all the other offenses. And then we take essentially a ratio of the rate ratios where we divide the observed rate ratio, so the evidence that I showed you before, by the counterfactual rate ratio.
And then this estimated phi here essentially tells us the factor by which overall inequality, as we actually observe it, increases as a result of inequality in admissions for a particular offense. So, say, for drug offenses, phi hat would tell us here, how much does racial inequality increase as a result of inequality in drug offenses compared to a scenario where drug offenses were actually equal across these two racial groups? So what do we see when we do this decomposition?
So racial inequality is on the top. Educational inequality is on the bottom. I’ve drawn a line here at one– if the line were at one, it would essentially say that offense contributes nothing to the overall inequality. And so the trends here– the lines indicate the factor by which, how many times by which the disparity increases as a result of inequality in that specific offense. So it’s a little counterintuitive, I grant.
Essentially though, higher values mean that that offense is contributing more to overall inequality, OK? What do we see? It’s actually quite interesting, quite striking how similar three of these panels are. The patterns for racial inequality among people with and without college, and the patterns for educational inequality among Black Americans are strikingly similar. What do we see in these three panels?
Property crimes are the primary driver of inequality in the earliest parts of our analysis, so this dotted green line. Property crimes are here driving most of the inequality, the greatest share of the inequality. But then in the decades surrounding the turn of the century, we see this huge increase in the significance of admissions for drug offenses. And that’s sustained for a couple of decades, but then it starts to drop off pretty meaningfully over the last 10 or 15 years.
And then Meanwhile, the significance of violence or violent offenses either remain stable or starts to somewhat increase such that toward the end of our analysis, for each of these three dimensions of inequality, violence starts to emerge as the primary driver of disparities of these three types. OK. Patterns are a bit different though for educational inequality among white Americans. So some things are similar. We see this steadily declining significance of property crimes for all kinds of inequality here. Although for educational inequality among white Americans, it just remains absolutely more significant. Property crime is the main driver over the whole period of analysis for this group.
The main difference is that the significance of drug offenses to educational inequality among white Americans starts out very small, as it does for all the other groups, but rises steadily over the whole period of analysis such that it’s now one of the main drivers of inequality for that group. OK. So this is of what we learned from decomposing our findings in this way.
Another consideration though is honestly, as I said, more of a confounder than a cause. We would want to make sure we understood or were addressing this factor if we were to give confident interpretation to our results. And that’s the fact that during the period we study, there’s a pretty significant change in the educational attainment of Americans, particularly more people are going to college.
This has the consequence that both our lower education group, people without college, and our higher education group, people who attend college, are increasingly negatively selected over time. This can bias our result of educational inequality. And it essentially means that without some sort of adjustment or correction, if we only rely on nominal education groups, we’re not comparing people of similar, relative social advantage or disadvantage over time. We’re comparing apples and oranges if we compare people who have and haven’t been to college in 1984 and 2019.
And so we do a adjustment to correct for this. I should note that this has been a point of some methodological contention in the research program around educational inequality in mortality, particularly due to deaths of despair. And so there’s actually been a lot of methodological innovation in this space as well. I think what we do is probably the most rigorous thing you could do with our data. But if you’re interested in more computationally intensive approaches to dealing with educational selection, I’m happy to talk about them.
So essentially, we’re asking here, to what extent are these changing levels of educational attainment contributing to our results? Might they be biasing our results? What do we see if we were to control for this changing selection? I’m going to spare you the math here and just reason through some illustrations. So what do we do here?
Recall that in our main analysis, we’re comparing people who have and have not been to college. So just two groups. In our adjustment, what we do is split that lower group into two more detailed groups– people who completed four years of high school and people who didn’t complete four years of high school. The completers, we don’t actually know if they have their diploma or not. But we’ll just call them high school completers, and we’ll call the other folks high school dropouts, OK?
So this figure is for, Black Americans and white Americans, the proportion of the population– I should have said earlier, this whole analysis focuses on people aged 20 to 39. So this is the proportion of the population who fall into these three, more-detailed education groups, OK? So first what we do is calculate a prison admission rate for each of these three more-detailed groups. It’ll obviously be the same for this any college group. That doesn’t change. But we’re calculating two different admission rates for these two different detailed educational groups.
Then what we do is calculate the proportion of people falling into the no-college or any-college group in 2019. And 2019 will be our reference year. It will the year to which we index everything. We’re trying to replicate conditions in 2019. So that’s this dotted line here. So in 2019, 60% of Black Americans in this age group had attended college, 40% had not. Slightly different numbers for white Americans. So you want to think about that dotted line as being like a threshold that we’re trying to reproduce in any given year. We’re trying to redistribute people so that they fall either above or below that threshold.
Then what we do is essentially ask, what proportion of the high school completers, this middle group, fell above or below that line in any given year? So say in 2000, this proportion of high school completers would have attended college if we put them in a time machine and moved them up to 2019 given their position in the educational distribution in that year. And then the last move we make is to essentially calculate a weighted average of the detailed admission rates in any given year.
So for our no college group, what we’re going to do is essentially add up this chunk of folks and this chunk of folks and calculate a weighted admission rate for them. And then in– for the no-college or any-college group rather, we’ll do a weighted average of these folks and the rest of the folks. So it’s essentially just a weighting exercise where we’re using a more-detailed set of prison admission rates and then reassigning people according to whether they would or would not have attended college in 2019 given their position in the educational distribution in any given year. OK.
So this is essentially analyzing fixed proportions of the educational attainment distribution. So it’s more or less achieves our goal of analyzing people with similar levels of social and economic advantage. They’re analyzing their position in the distribution. This, of course, is only valid on the assumption that people within any given group have a uniform risk of imprisonment. Maybe that’s a strong assumption, but it’s the best we can do.
What changes when we do these adjustments? Well, as regards racial inequality, basically nothing. It doesn’t really change our analysis. So that’s good to know. As regards educational inequality though, things do change a bit. So the red line is our adjusted results. Black line is without the adjustment. So you can think of these as analyzing nominal educational categories, these as analyzing the kind of fixed proportions of the educational attainment distribution.
By construction, they’re going to be identical in 2019. But as you can see in earlier years, they’re quite a bit different. More specifically, we see lower levels of initial educational inequality in the adjusted results. And that’s more or less to be expected, given the way we do the adjustment, because we’re essentially reassigning some of these high school completers to any college group. They’re going to have higher risk of imprisonment. And so to an extent, it’s an artifact of our method.
But this, I don’t think, really changes qualitatively, our conclusions, our findings. Our story doesn’t change. If anything, it really just shows that if we were to think about fixed levels of advantage, this story about educational inequality, it’s even more dramatic. The absolute scale of the change is even bigger, and the pace of that change is actually accelerating even faster.
I want to briefly talk through a couple more candidate causes that Chris and I are now just kind of starting to work on. So I won’t show you any evidence here, but I’ll reason through how we plan to go about studying some of these things. So a third consideration is that any explanation for these inequality trends would ideally account for the fact that the people we study are nonrandomly sorted into different places.
And we know, from a variety of research, that the punitiveness of the US prison system is not uniform, and it’s not stagnant. It’s changing. Katherine Beckett and her colleagues have shown recently that the ratio of prison admissions to crime rates or to arrest rates is actually decreasing in urban areas, but increasing in rural areas. That is to say, we can think about the punitiveness of rural areas as increasing over the last 15, 20 years and the punitiveness of urban areas as decreasing somewhat, OK?
Of course, people of different racial and educational groups are more or less likely to live in rural or urban areas, and so some of our findings might be driven, at least in part, by the fact that Black Americans and college-educated Americans might be disproportionately likely to enjoy some of the liberal or progressive criminal legal system reforms that are concentrated in urban areas. Whereas white Americans, less educated Americans who disproportionately live in rural areas, might be more likely to disproportionately experience more punishing, harsher criminal legal system regimes.
And so we’re working on this. We’re trying to further stratify and decompose our findings according to whether or not people are admitted to prison from a rural or an urban county. We only have county-level data, so this complicates things. We only have good crosswalks since 2006, so we’re actually not able to say so much about the late 20th century. We can say much more about the recent past. I’m happy to talk more about this in the Q&A, but we really only have very, very preliminary findings here.
Another striking feature of our results is that the inequality patterns we document correspond very closely to some of the inequality patterns documented by Anne Case and Angus Deaton in their research on changing life expectancy and changing mortality. They also correspond very closely to some recent research by Raj Chetty and his colleagues on changing patterns of socioeconomic mobility.
This points us to think about factors, honestly, beyond the criminal legal system, factors that shape Americans life chances more broadly. That might explain not only mortality, not only socioeconomic mobility, but also imprisonment risk. So one of the things we’re doing now is to, I think, think more carefully about the degree to which this is consonant more generally with arguably with Bill Wilson’s thesis about when work disappears. To what extent are some of the changes we’re currently seeing among low-college or no-college white Americans the result of decreases in demand for low skilled labor that earlier hit black Americans in the late 20th century, black Americans without a college education.
And so one of the things we’re doing to try to understand the changes in labor markets and their effect on imprisonment rates is to combine data– our imprisonment data with data on local labor market exposure to automation. So these data come from Daron Acemoglu’s recent paper on robots and jobs, and we’re trying to ask the degree to which exposure to automation, which we now know had a really significant effect on both employment and wages, might also explain some of the changes we see in imprisonment trends.
I said I was going to give you a caveat at the end, and it’s time for the caveat. Mostly, I’ve been documenting or describing our results regarding declining racial inequality and skyrocketing educational inequality. I said that these were roughly comparable in the early part of our analysis, although they’re obviously measure-dependent. But I think, honestly, by any reasonable measure it’s, I think, fair to say that educational inequality now dwarfs racial inequality in prison admissions, OK?
But to state the obvious, going to prison oneself is not the only way that one comes into contact with the carceral system. Our friends, our family members, our neighbors might go to prison. And we know, from a lot of research, that these vicarious contacts with the prison system are pretty consequential for our health, for our income, for our civic engagement, for our trust in the law.
And so I want to highlight a seeming paradox that Chris and I have documented recently. Most of these results come from our social forces paper from a few years ago. But I take pains to emphasize this because I think it really does complicate some of the prior trends that I showed. And that’s that despite declining racial inequality in prison admissions and skyrocketing educational inequality in prison admissions, it actually remains the case that as regards vicarious contact with the prison system, that is, the imprisonment of people’s family members or their exposure to high-imprisonment neighborhoods, racial inequality remains at least as large and maybe even larger than educational inequality, OK?
So we have a divergence in inequality patterns between direct and indirect contact with the prison system. And this seems a little bit paradoxical. It wasn’t immediately clear to us how this could be the case. So let me show you first some evidence about this, and then I’ll try to reason through this seeming paradox. So I’ll spare you the detailed data and methods, but these results come from an analysis of the Fam HIS Survey.
Fam HIS is a survey that was fielded in 2018, and it was designed specifically to measure family member incarceration. So how likely people are to have a family member incarcerated. So we look more specifically at the imprisonment– not just the incarceration, but the imprisonment of a close family member. So a spouse, a child, a parent.
And what we see is that using the same divisions here, that the Black-white disparities in the likelihood that someone has a close family member imprisoned are actually a bit larger than the educational disparities in family member imprisonment. This is somewhat at odds with the evidence I showed you about prison admissions. Similar difference in results when we look at neighborhood imprisonment. So here what we do is rely on a resource called the Justice Atlas of Sentencing and Corrections. And this is a little known data set that covers, I think, only 13 states, but it has the unique value of measuring imprisonment rates at the census tract level. So this allows us to get a better sense of neighborhood world imprisonment.
So we use this resource to ask then, what neighborhoods in these places have very low levels of imprisonment , so the lowest half of the neighborhoods in terms of their imprisonment rate, or, say, very high levels of imprisonment, so the top 5% of neighborhoods with respect to their imprisonment rate? So we first measure neighborhoods in terms of their imprisonment rate. And then we ask, how likely are people of different racial groups or different educational groups to live in a very low-imprisonment neighborhood or a very high-imprisonment neighborhood? OK? What do we see here?
Well, regarding very high-imprisonment neighborhoods, the Black-white ratio in the likelihood that one lives in a very high-imprisonment neighborhood is very large. And indeed, considerably larger than the educational disparities in the likelihood that someone lives in a very high-imprisonment neighborhood. The inverse is true for low-imprisonment neighborhoods. Black Americans are about half as likely as white Americans to live in a low-imprisonment neighborhood, whereas the corresponding educational disparities aren’t quite as large.
Essentially, what we’re seeing here is that whereas educational inequality in prison admissions is now much larger than racial inequality in prison admissions, the same is not true for indirect contact or vicarious contact with the prison system. There, racial inequality seems to continue to predominate. How? This seems like a paradox. We think we can make sense of it, though, using this idea of class permeability, which is an idea that comes from Erik Olin Wright’s work.
Classical permeability is essentially the degree to which we are socially connected to people from other classes. It’s a version of social capital, you might think. And so if class glass shapes inequality in imprisonment, then one’s total exposure to the prison system is a function of one’s class position, which shapes the likelihood that you go to prison, but also your class permeability, which shapes the imprisonment risk for the people to whom you’re connected. OK?
And so a large literature documents higher rates of downward social mobility among Black Americans compared to their white counterparts. And this means that Black Americans of any given class position are more likely to have poorer family members than their white counterparts. Again, countless studies show that as a result of segregation, ghettoization, residential discrimination, middle class Americans, even upper class– sorry, middle class Black Americans, even upper class Black Americans, are much more likely to live in poor neighborhoods than their white counterparts.
So what this means essentially is that Black Americans have much more downward class permeability than their white counterparts, given any class position. And so it’s actually racial inequality in class permeability that helps explain this paradox in results. It’s racial inequality and class permeability that really helps us explain how persistent racial inequality in vicarious contact with the prison system can continue even in an era where racial inequality in imprisonment is declining and class inequality in imprisonment is skyrocketing.
I think it’s time to wrap things up. So what have we learned by way of conclusion? Black-white inequality in US prison admissions is declining, but it remains really high. So I just want to restate that among people without a college education, Black Americans are still twice as likely as white Americans to go to prison. Among people with a college education, Black Americans are 2.6 times more likely to go to prison than their white counterparts.
Meanwhile, educational inequality in prison admission is just skyrocketing. So depending on whether we’re talking about Black or white Americans, people without a college education are somewhere between 24 and 32 times more likely to be admitted to prison in any given year than people who have attended college. Drug offenses really did drive high black inequality in– high Black-white inequality in prison admissions during the war on drugs. But in more recent years, it’s been violent offenses that are now the primary driver of both Black-white inequality among people with and without college, and also educational inequality among Black Americans.
And so to return to this debate, I think our findings do largely corroborate Alexander’s account about what happened in the United States during the war on drugs. But I think some of her critics are vindicated in their description of more recent years. And so maybe it’s a cop out. I think the diplomatic answer here is that actually both sides are right, but they’re actually right about successive, somewhat distinct historical periods. Alexander, looking backward, really was describing accurately what was going on in the United States at the time. In more recent years, I think we need to update some of these accounts about both the primary contours and also the primary drivers of inequality in mass incarceration.
I think it’s important to state– to clarify that educational selection does not account for our findings. They’re not an artifact of the changing educational distribution in the United States. And I want to restate this caveat, that patterns of vicarious exposure significantly qualify any claims that someone might make about the declining significance of race to mass incarceration. Across generations, racial differences in residential attainment and social mobility have led to really large racial differences in what we call class permeability, and this makes racial inequality in vicarious exposure to the prison system at least as large as educational inequalities.
And so I think this helps us see that we would do well to think not only intersectionally, but also holistically about the lived experience of mass incarceration. Not only the likelihood that we go to prison, but the likelihood that our loved ones, our community members go to prison. I think this helps us understand much better. The new contours of mass incarceration in the United States today and beyond. That’s my talk. Thank you so much for your time and attention. If you have questions, or you can email me or find out more about my research on the website. Thanks.
MODERATOR: Thank you so much, Alex.
ALEX ROEHRKASSE: Yeah.
MODERATOR: If you have questions, fell free to send [INAUDIBLE]. We have approximately 10 ish minutes. Questions, observation.
AUDIENCE: Hi.
ALEX ROEHRKASSE: Hi.
AUDIENCE: Are you familiar with Paul Butler’s paper on his belief that incarceration is not a solution for various offenses, but rather that Black men should be returned to the community where their community would take care of them rather than going into prison? And he puts forward that African-American jurors should not convict African-Americans for low-level offenses.
ALEX ROEHRKASSE: I’m not familiar with the paper you describe, but I think I’m broadly sympathetic to the argument. Yeah.
AUDIENCE: Yeah. It was very controversial, and it was in a lot of places, just the idea that rather than incarceration, that sending black men with low-violent– nonviolent offenses back into the community because that reduces just what you were talking about, the exposure to the prison system to the community.
ALEX ROEHRKASSE: Yeah. I think the preponderance of evidence indicates that we could significantly decarcerate the population with probably net positive effects on– net negative effects, I guess, on crime rates in the United States. And it’s needless to say that this would have a host of positive social effects as well. Yeah.
MODERATOR: One more thing. Can you also introduce yourself?
AUDIENCE: Sure. Neil Fligstein, sociology. So at the end of this– so we talked earlier about policies. And so if I was to walk away from this and say, well, what we should be doing is working on sentencing policies or giving people more education or worrying about drugs in rural areas of America, where do you come down on that?
ALEX ROEHRKASSE: Yeah.
AUDIENCE: That makes sense?
ALEX ROEHRKASSE: It does make sense. Yeah. So I think part of my answer dovetails with my previous one. I think that we’re not looking here at the consequences of incarceration. But if we take incarceration, all else equal, to be something that we would prefer not to occur in the United States, I think you’re asking, how might we achieve both lower levels of incarceration, but also lower disparities in incarceration? So those are the ends you’re–
AUDIENCE: Yes.
ALEX ROEHRKASSE: –trying to– OK. So if we wanted to decrease incarceration in the United States and we also wanted to decrease disparities in incarceration, what would we do? I think obviously, many of the solutions lie in the criminal legal system themselves. I think that– I didn’t show you my evidence from urban-rural divides. But I think there’s mixed evidence that some of the criminal legal system policies that have been concentrated in urban areas are having decarceration effects, but also effects that decrease disparities, particularly racial disparities.
But I think the fact that our inequality results are so consistent with a variety of other outcomes like mortality and– especially mortality seem to indicate that what we’re documenting here in terms of incarceration is like an epiphenomenon of deeper inequalities in life chances in the United States. So when we say that imprisonment is increasing among this population, essentially what we’re saying is, the opportunities that this group faces for gainful employment, meaningful engagement in a community are decreasing.
And so I’m quite sympathetic to arguments that many of the solutions here actually lie outside of the criminal legal system. I’m a big supporter of a full employment industrial policy. I think that would have probably the single largest effect in decreasing some of the disparities that we see here. But obviously, that’s expensive and fairly radical. So yeah.
AUDIENCE: Hello. Thank you. William Welsh–
ALEX ROEHRKASSE: Hello, William.
AUDIENCE: –PhD student, and Berkeley sociology. Yeah, I was interested in two things. One was the declining contribution of property crime to inequality, I think, among all four groups.
ALEX ROEHRKASSE: Yeah.
AUDIENCE: And I was wondering if your interpretation of that is that property crime is becoming a more general phenomenon, not limited to one or the other subgroup. And then the other question I had is that while among white voters, no-college voters voted– in 2024 voted for Trump at about half, again, as high a rate as the college-educated counterparts, among Black voters, no college voters voted for Trump three times more likely than their college-educated counterparts. So I’m wondering what connection you might see, if any, between the results that you found and these election– voting results.
ALEX ROEHRKASSE: OK. Yeah. Your first question was about property crimes and whether we’re seeing, if I understood you correctly, some convergence in property offending across the different groups. I don’t think we’re seeing that. If you look at the absolute rates, I think those property offenses largely track admissions– offense-specific admission trends for the groups more broadly. So I don’t think I would explain that declining significance as a general convergence in offending.
I have to be careful with my words here, because there’s a big difference between offending and being admitted to a prison for that offense. Yeah, regarding elections, I don’t know. What would you say? What’s the connection you see? I can’t say I’ve honestly thought about it all that carefully. It’s a quite nuanced question. You’re saying essentially that partisan divide– educational partisan divides are actually larger in the Black community than in the White community?
AUDIENCE: Yeah.
ALEX ROEHRKASSE: Yeah. And so you’re asking how that would explain our results or maybe be an effect of our results.
AUDIENCE: I guess I’m asking, how might it be an effect of the–
ALEX ROEHRKASSE: An effect of the results?
AUDIENCE: Voting effect?
ALEX ROEHRKASSE: Yeah.
AUDIENCE: Basically vote against Democrats [INAUDIBLE].
ALEX ROEHRKASSE: Yeah. So yeah. So well, it’s important to acknowledge actually that– if I’m remembering correctly, that generally speaking educational inequality in imprisonment rates is larger among white Americans than among Black Americans. But we also know that criminal legal system contact has much greater effects on things like trust in the law for Black Americans than for white Americans.
So I think that’s part of what’s part of what might be going on or a connection point between what we’re talking about and what you’re describing, where the class disparities in incarceration among Black Americans have this outsized effect on the politics of low-college or no-college Black Americans that might contribute to the fact that you see these actually outsized partisan divides among the Black community that aren’t necessarily reflected in the prison admissions statistics.
AUDIENCE: Thank you.
ALEX ROEHRKASSE: Yeah
AUDIENCE: So you mentioned full employment. I wanted to ask you to flesh out your thoughts on the interaction between your story and labor markets. I mean, that could be anything. It could be over aggregate unemployment or sectoral or regional. But I guess specifically, I wanted to ask about the last few years when we’ve moved towards full employment and we’ve actually had a reduction in economic inequality, how do you think that would affect your story? And I mean, is it too soon to tell or do you already have some sense of what impact that’s having on this kind of inequality?
ALEX ROEHRKASSE: Yeah. So you’ve got me with my back against the wall as regards, say, racial and educational disparities in employment rates over the last 18 months or something. I don’t know those. But look, I think that we need to think very seriously about, I guess, offending, but especially prison admission for different offenses as a function of labor market opportunity.
I think there’s decent evidence– a growing body of evidence that shows that at different levels of geographic specificity, at different levels of demographic specificity, that when folks have more meaningful opportunities for gainful employment, meaningful employment, they’re less likely to end up in circumstances, no worries at all, that might expose them to the carceral system. So I can’t give a responsible answer as regards recent trends, particularly past the scope of our data here. But I do think that for me, it’s a policy area where we should be thinking more and more seriously, more aggressively. Sorry. It’s a bit of a cop out, but–
AUDIENCE: Hi. My name is Taylor. I’m a PhD student over at the law school.
ALEX ROEHRKASSE: Oh, cool,
AUDIENCE: Would thinking about juvenile justice and maybe changes or rates that’s going on across the story over time and thinking about how that may be impacts educational attainment, does that change the story at all or does it fit into the model?
ALEX ROEHRKASSE: Yeah, I think it could change the story. A lot of government statisticians, when they measure educational attainment, often start at age 25. It’s reasonable assumption that our educational attainment process is more or less complete by age 25. We’re analyzing 20 to 39-year-olds here, so there’s a decent number of people whose educational attainment process may not have completed by the time they come into contact with the system. And of course, we know that system contact affects your educational trajectory. You’re, of course, pointing to the juvenile justice system, which we would expect to have an even more significant effect on interrupting educational attainment.
We have stratified our analysis by five-year age band, and we don’t– it doesn’t change our story very meaningfully. So just in terms of what we can show with the prison system data, we’re not terribly concerned about reverse causality here, where prison– our story doesn’t meaningfully change when we think more carefully about the possibility that prison admission is affecting educational attainment and not the other way around. But look, that’s obviously not the case when it comes to juvenile justice. And I’m less well versed in the juvenile justice research than I should be. But I think it’s something we should probably think more carefully about. What is your thinking about how juvenile justice might be shaping this inequality story?
AUDIENCE: I wonder if it would tell a parallel story. Maybe it would, but I don’t know if it might parallel [INAUDIBLE].
ALEX ROEHRKASSE: Right. Where we couldn’t, say, use educational attainment as a measure of class because we’re talking about even younger people. But if we were to, say, look at parents’ education and see how things were changing in the juvenile justice system, would we see similar patterns or not? Yeah, that’s a very good question. I’d be willing to bet that we are, but– or we would, but I don’t yet know.
AUDIENCE: Yeah, thanks.
MODERATOR: Alex, this would be the last question.
AUDIENCE: I guess your study didn’t really address social causes to these declines, but I wonder if– it seems to me some social movements like Black Lives Matter or the George Floyd protest might have contributed to some of the declines in drug-related prison admissions.
ALEX ROEHRKASSE: Yeah. Those mostly would have come after the period we’re studying, but not entirely. And I think you see the declining significance of admissions for drug offenses much earlier than some of those events. They’re starting in the early 2000. Why are they declining? Yeah. So OK. So why are so many fewer people going to prison for drug offenses, why this massive drop off?
It’s a good question. I think part of it– well, so it’s a massive drop off for Black Americans, but not for white Americans. So I was talking briefly about successive waves of economic, social-technological change that affected less-advantaged Black Americans in the late 20th century and are now arguably coming to effect less advantaged white Americans to greater degrees.
So part of it might be that the kind of drug epidemics that affected the Black population in the United States have waned, to a degree. And now we see the opioid epidemic, which affects, obviously, all Americans, but particularly lower-education white Americans these days. So part of it might be in sequential drug epidemics that are differentially affecting these different populations.
Part of it may be drug sentencing and enforcement reforms. So we’ve seen big– just look around, big changes in the enforcement of marijuana law, for example, that’s really affected less educated Black Americans who used to go to prison for marijuana use and no longer do. I cite those as the two major contributing factors, but they’re probably not the only ones. Thank you so much for your time and attention. I really appreciate it.
MODERATOR: Thank you so much–
[MUSIC PLAYING]
