Mass Incarceration Nation: How the United States Became Addicted to Prisons and Jails and How It Can Recover

Author: Jeffrey Bellin
Publisher: Cambridge University Press, 2022.  234 pages.
Reviewer: John Pfaff | September 2023
Read Author Jeffrey Bellin’s Reply to this Review

In his recent book Mass Incarceration Nation, William & Mary law professor Jeffrey Bellin seeks to explain both how the United States came to be the world’s largest jailer and how we can roll back our outsized reliance on prisons. Yet while the book builds its arguments around an insightful and underappreciated distinction—that what we call the “criminal justice system” is actually two separate systems, a criminal “justice” system that focuses on serious crimes like murder and rape, and a criminal “legal” system that targets lesser offenses like drug selling and DUIs—it ultimately falls somewhat short in both its goals. Digging into its shortcomings a bit is useful, however, because several of the missteps it makes are ones that arise frequently in the literature on mass incarceration.

Before jumping into my concerns, though, I should start by briefly summarizing the book’s main arguments. Simplifying only somewhat, the book makes two broad historical claims. The first is a fairly common “mechanistic” take on the origins of mass incarceration, that rising crime in the 1960s and 1970s pushed the criminal legal system[1] away from its long-held emphasis on rehabilitation towards a more punitive embrace of the prison. Unsurprisingly, this leads the book to caution us that rising crime in the 2020s poses an existential threat to the current push for criminal legal reform. The second historical claim is that at some point in this process we see a move away from the “criminal justice system” towards the “criminal legal system.” In other words, law enforcement shifts its resources away from harder-to-prove but more serious offenses like murder and rape to focus more on less-severe but easier-to-prove crime like drug trafficking. The book is actually quite fuzzy on the timing of this second shift, from the CJS to the CLS, but its argument seems to suggest that it took place throughout the 1990s and 2000s. Much of the solutions section then focuses on reforms that should push law enforcement to undo its (alleged) shift towards CLS offenses and back towards targeting CJS crimes.

With that summary, it may help to briefly summarize the two major oft-occurring concerns I have with the book. The first is that it does not wrestle enough with the specific politics of punishment, treating mass incarceration far more as a failure of policy than of politics. The book’s opening section on the historical causes of mass incarceration, for example, more or less ignores the large literature on the extent to which penal policy is a response not to crime but to broader social upheavals over race or class or sex.[2] And the book’s concluding section on policy recommendations lays out a long list of possible solutions, but it never addresses the political reasons why the problems it wants to fix exist in the first place. This is a significant oversight, because most of the apparent “bugs” the proposed reforms target are actually “features” for politically power actors, so it is impossible to work through how to fix things without confronting the politics that currently defend the status quo.[3]

The second concern I have is that the book takes our criminal legal system data at too much its face value. Specifically, it doesn’t confront the extent to which our outsized reliance on plea bargaining makes it very difficult to understand what, exactly, post-conviction data is really telling us: is that simple assault conviction really a conviction for simple assault? Or just an easier way to get at aggravated assault? It also doesn’t probe the extent to which even arrest data may reflect certain recording choices made by the police. Failing to think about the politics of our criminal legal data is troubling in any situation, but especially so here, where the structural shortcomings in our data are likely to over-support the book’s thesis. In fact, as I’ll show below, I’m not sure the CJS-to-CLS shift the book builds its argument around really took place at all (and if it did, it did to a much lesser degree than the book claims).

I. The Problems With Politics

Two important issues with the book’s treatment of politics deserve attention. The first is that the historical account it provides is overly simplistic; that is concerning in and of itself, but in this case it may in turn lead to ineffective policy recommendations. As noted above, the book posits a fairly mechanical take on the politics of punishment, in which rising perceptions of crime—driven by real increases in crime in the 1960s and 1970s—led to a bipartisan push for tougher sanctions.[4] The book argues that even Black political leaders largely supported the push for punitiveness, even as it impacted their own communities, although the book’s defense of this claim, a defense I have seen elsewhere as well, glosses over some very important caveats.[5] The book later argues that rising crime in the 2020s likewise threatens a decade of reform efforts.

Now, to be clear, there is certainly some truth to this account: crime (likely) rose over the 1960s and 1970s,[6] and that rising crime surely fueled at least some of the punitiveness that came in its wake. But the book’s account glosses over some critically important details that paint a far more complex story, complexity that has very real implications for how we think about our current political reality. Over the past fifteen or so years, a vibrant literature in law and sociology has probed the way in which social pressures outside of crime have driven the repeated shifts from progressive reform to punitive harshness and back over the past two centuries (a shift we’ve seen many times, contrary to the book’s assertion that we had basically embraced rehabilitation until the 1960s/70s).[7] Crime trends are not irrelevant, but broader political shifts matter too, sometimes perhaps even more than crime itself. Unfortunately, Mass Incarceration Nation not only appears to fail to cite any of the core books in this literature, but more concerningly does not even really wrestle with any of their ideas.

Bill Stuntz, for example, argued in his magisterial (and uncited) The Collapse of American Criminal Justice that the punitive turn in the 1970s, which came nearly a decade after (official UCR) crime started rising, was less a response to the crime trends themselves, and more due to suburban fears stoked by the violent social unrest in places like Newark and Watts. Politically powerful suburbanites were willing to ignore urban crime until it suddenly represented a broader social threat. We can see other examples of this as well. Contrary to an unsupported claim in Mass Incarceration Nation, as crime fell over the 1990s, people’s fear of crime fell sharply too.[8] And as people’s fear of crime declined, prison growth slowed as well—in fact, in 2001, state prison populations actually fell in absolute numbers (although Federal population growth alone more than offset that decline). All that reversed itself, however, in 2002: fear of crime spiked, and prison populations surged—even as violent victimization declined yet again. Why the spike in fear? Most likely because a population traumatized by the 9/11 attacks turned fearful and punitive. In other words, crime policy is not driven (solely, perhaps even mostly) by crime, but by broader social and political shocks.[9] Similarly, it’s important to note that prison populations didn’t start to fall in the 2000s until we were in the midst of a generational financial crisis that created new, and unexpected, political alliances that gave reformers—who were better prepared politically to capitalize on the 2008 crisis than the 2000 dot-com recession—the leverage they needed.[10]

It’s easy to see why getting the politics correct matters for our current moment. Mass Incarceration Nation argues that rising gun violence and homicide in the 2020s threatens reform efforts, and there is definitely some inescapable truth to that point. But it is worth asking if the extent and tenacity of the backlash we are witnessing driven more by fears of shootings, or by suburban reaction against the demands of (and fears of) the post-George Floyd protests? There is certainly reason to think the latter is playing a major role here. Note, for example, that even in late-2020 and into 2021, reform prosecutors continued to win the votes of those most impacted by rising gun violence, but struggled to win over less-impacted more-suburban voters.[11] And Bloomberg published a recent article showing that the media’s coverage of crime in NYC was untethered from actual crime rates in the city.[12] If this is right, then reducing crime may not rein in the backlash if the backlash is really about broader political fears.

Mass Incarceration Nation also gives too little attention to the politics of punishment in the final section of the book, which provides its “blueprint for reform.” This book, like so many other legal (and other) works, is much more comfortable addressing the “demand” side of policy change than the “supply” side. The demand side is what we want: we should have judges who are less likely to be prosecutors, counties should have to pay for costs of locking people up for lower-level crimes in state prisons, etc. The supply side, however, asks what we can actually expect to get, or what we need to do induce legislators or governors or whoever to do what reformers are proposing they do. In many ways, the supply side is more important, but it gets far too little attention (here, or pretty much anywhere).

Demand-side-only analysis can work, I suppose, in areas where policy failure comes from ignorance. Take medicine: no one is standing up “for” cancer, so it may be enough to point to a new oncology protocol and encourage others to adopt it.[13] But that is most certainly not the case when it comes to mass incarceration. It is critical to appreciate that almost every “failure” in the criminal legal system is not a failure for everyone. Almost always, someone is benefitting from the “failure,” and whoever is benefitting will almost always work to block reforms that threaten what they have.

A striking example of this centers on efforts to confront what I have called the “prosecutorial moral hazard problem.”[14] Prosecutors are county officials, but when they convict someone of a felony, that person gets sent to a state-funded prison, so the prosecutor is free to ignore the financial cost. Even worse, if the prosecutor reduces the sentence to a misdemeanor, any sentence is served in county jail or on probation, which are county costs. It’s actually cheaper for prosecutors to be harsher.

Mass Incarceration Nation proposes that one way to address this would be to make counties pay the state government the costs of locking someone up in state prison.[15] The book then notes that California adopted a policy like this for juvenile cases in the 1970s, that it seemed to work well, and that we should think about it for prisons too. And then it moves on to the next proposal.

Before getting to the demand/supply issue, there’s something quite problematic about how casually the book handles this idea. California actually adopted this idea again—in 2011, when its Realignment law imposed a policy like this for all prisons, not just juvenile ones. Moreover, the resulting decline in California’s prison population was responsible for nearly fifty percent of the overall national decline in prisons over much of the 2010s. Yet Realignment, one of the single most transformative criminal legal reforms passed in the past 30 or 40 years, is never mentioned in the book, even in a section talking about the very idea behind it.

Overlooking Realignment is concerning on its own terms. But it is also unfortunate because Realignment is a great example of just how hard the politics of reform are—of how what is demanded is often not what is supplied. First, it is important to appreciate that California adopted Realignment only under intense political pressure, after the 9th Circuit took California’s prisons into receivership and a close-call 5-4 Supreme Court opinion upheld the takeover. And at the time it was adopted, Democrats had overwhelming trifecta control of the state government, and they still struggled to pass the reform…. and a lot of its financial incentives were gutted within a year, both by statute and by referendum-led constitutional amendment.[16] Few states will face that sort of political pressure, and those that do are unlikely to have the sort of Democratic control that is generally required to pass aggressive prison-reform laws.[17] In other words, it’s a good policy to demand, but a very hard one to supply at all, and harder still to provide effectively. California’s experience demonstrates how fickle state legislators can be, and how vulnerable such reforms are.

Here’s another example. Mass Incarceration Nation proposes re-empowering “unelected” parole boards to make sentencing determinations, because they are less susceptible to political pressure.[18] It’s true that parole boards are unelected—but they are often the at-will employees of someone who is elected, namely the governor. Here is the former chairman of New York’s parole board, back in 2014: “It’s not like it’s written down anywhere, but every board member knows, if you let someone out and it’s going to draw media attention, you’re not going to be re-appointed.”[19] Parole boards may appear “unelected,” but when one drills down carefully into the actual political design, they may actually have less freedom to cut sentences than when compared to, say, prosecutors.[20]

Policy proposals that focus just on the demand side—“we should do this”—rather than the “how do we get them to adopt this” supply side are treating mass incarceration like a failure of policy, not a failure of politics. But mass incarceration is far more the latter than the former. It is not that policymakers are unaware of what the defects and problems are, it is that for many of them, they are not defects at all, but features. Which means the real issue we need blueprints for is how to navigate that sort of political quagmire.

II. The Problems With Data

The second major concern I had with the book is how it used our crime data, particularly when trying to defend the claim that mass incarceration was driven by a shift from targeting “criminal justice” cases to “criminal legal” ones. Some of my issues are specific to the book’s argument, but others are good examples of deeper, more systemic challenges that every researcher using criminal legal data has to struggle with.

But before getting to these points, I think it is important to note that multiple times the book makes empirical assertions that are simply untrue, or discusses the numbers in ways that obscure important patterns. As noted above, it claims that people’s perceptions of crime in the 1990s did not respond meaningfully to falling crime rates, even though Gallup’s measure of fear of crime was cut in half over that decade. Similarly, the book asserts that arrests did not fall over the 1990s even as crime did, a claim that is clearly incorrect (all sorts of arrests fell by a lot).[21] The book also sometimes makes empirical claims that are not technically incorrect, but which gloss over important trends in a way that makes them feel deceptive. For example, at one point the book argues that arrests and convictions for aggravated assault moved roughly in lock-step, by comparing where they were at two points in time (1986 and 2006). The claim is technically correct—both arrests and convictions were higher in 2006 than 1986—but the trends in each differ significantly within that period, in ways that have some importance for how we should think about the drivers of mass incarceration.[22] In this context, the misframing leads the book to misstate the mechanism it is considering—it leads the book to argue that police drove certain changes by glossing over the clear change in prosecutorial behavior that takes place between the end-points. The book similarly ignores everything between the end-points when talking about drug arrests and prosecutions,[23] when looking at arrest data from the LAPD,[24] or, as noted above, when talking about trends in people’s fear of rising crime. There are also times where the book misstates the findings of the papers it cites, like the time it refers to a paper about the impact of mandatory arrests in domestic violence cases to support a claim about arrests in general, even though those are two very different policies.[25] Taken together, these sorts of errors unfortunately suggest that readers should approach the book’s empirical claims with significant caution.

Now to the more structural issues about the challenges of criminal legal data. My less generalizable point is simply that the numbers themselves do not seem to support the book’s assertion that incarceration was driven in large part by a shift in focus from more-serious to less-serious offenses.[26] If nothing else, it is hard to reconcile this claim—which in many ways appears to be an effort to prop up the “standard story” that mistakenly places the war on drugs, not violent crimes, at the heart of mass incarceration—with the percent of people serving time in state prisons for violence. In 1985, ~54% of people in prison were there on a violent conviction, which falls to ~46% in 1990 and hovers around that for the next five years. At first blush, this seems consistent with the book’s story, but then a confusing wrinkle appears: by 2000, the percent serving time for violence has risen to 49%, then 52% by 2008, and finally almost 60% by the end of 2019.[27] In other words, at least when it comes to prison, the 1990s and 2000s appear to see a shift towards more criminal justice-like cases.

Mass Incarceration Nation does touch on two possible responses to this claim, but neither is empirically compelling. The first argument is that “violent crimes” as a category includes a lot of lower-level behavior that feels more “criminal legal” than “criminal justice,” and that within the category of “violent crime” we see a shift towards the less-serious.[28] It is absolutely true that a lot “violent crime” arrests are for assaults that are not particularly serious. But it is readily apparent that most of these arrests do not result in felony convictions, and far fewer still result in prison admissions. The 2006 Uniform Crime Reports, for example, state that police made ~450,000 arrests for aggravated assault and ~1.3 million for simple assault; for that same year, the National Judicial Reporting Program, a Bureau of Justice Statistics dataset on which the book often relies, reports that state courts convicted ~100,000 people of felony (aggravated) assault, and only 45% of those convictions, or ~45,000 out of nearly 2 million arrests, resulted in prison time.[29] That’s under 3% of all assault arrests (and only 10% of all aggravated assault arrests). It is almost certainly the case that almost all (although certainly not all) of those sent to prison are going to be drawn from the more serious end of the “assault” distribution—and I expect the share of minor assaults has fallen even more over the past ten years, given that reform efforts have focused primarily on keeping lower-level offenses out of prison.

The other defense Mass Incarceration Nation raises against the “prisons are full of people convicted of violence” is to argue—fairly—about “churn.”[30] People convicted of violence spend more time in prison than those convicted of other crimes; the median time to release for someone convicted of violence is ~4 years (and over 10 for murder), vs. ~1 year for all other non-violent crimes. Thus those convicted of non-violent crimes will inevitably make up a bigger share of those admitted to prison than of those currently in prison.[31] The book, however, simply asserts that this effect exist, without delving into any specifics about how significant it may be.[32] This is unfortunate, because there have been some empirical papers that have sought to estimate this effect directly. In the one that I wrote a few years back, I found that while drug offenses made up ~20% of state prison populations at the time, they made up ~24% of all admissions.[33] So I found a real “churn” effect, but one that is not particularly large. Now, that is just one paper, and the data I used stops in 2012, and it focuses more on comparing the churn of drug cases vs. non-drug cases as opposed to that of violent cases vs. non-violent cases. I’m not saying my paper provides the definitive answer. But it certainly casts serious doubt on the book’s emphasis on low-level churn, and raises the question of why the book simply makes an abstract argument about a measurable effect without engaging with any of the actual estimates that are out there.

My second major concern with the book’s use of numbers—an issue that I fear arises all too often when people use criminal legal data—is that it does not carefully confront the processes that generate the data. In many ways, our criminal legal data does not accurately reflect what is driving the highly discretionary choices that police, prosecutors, and judges are making. Here are two examples from Mass Incarceration Nation that illustrate this risk, and how it can lead to erroneous conclusions.

To start, we need to consider how plea bargaining wreaks havoc with interpreting conviction data. If someone is arrested for a serious violent crime but pleads out to a lesser violent (or even non-violent) crime, for example, they show up in the conviction and prison data under that lesser offense, even if they actually committed the more serious crime—and even if the prosecutor demanded prison as part of the plea only because of the more-serious-but-uncharged behavior. So if someone gets arrested for genuinely violent domestic assault and has heroin on him at the time of the arrest, the prosecutor may drop the assault charge and accept a plea just for the drugs. At the same time, the prosecutor may insist on prison time for the drug charge solely because of the uncharged-violence. Our data just shows the drug conviction, and so someone in prison may appear in the data as a “criminal legal” case while actually being there for something that is much more a “criminal justice” offense.

Given that over 90% of guilty verdicts come from plea bargaining, the data-distortion such deals create pose a huge empirical challenge to interpreting conviction and prison data, one that most analysts do not wrestle with enough. And, in this case, it poses a specific challenge to the book’s thesis, because it suggests our conviction and prison data will be biased in favor of the book’s claim about a shift from more-serious to less-serious offenses.

People rarely if ever plead guilty to something more serious than what they are initially charged with, so data created via plea bargaining will systematically understate the severity of the behavior that lands people in prison.[34] For an argument like the one in Mass Incarceration Nation, which argues that we’ve experienced a shift towards less serious outcomes, that’s a serious complication.

Now, to be clear, this effect poses a risk to the book’s claim only if we think that something encouraged police or prosecutors to shift their arrest and charging behavior over time—but there are several reasons that this might have very well have happened. Take prosecutors. As Mass Incarceration Nation correctly notes, legislatures kept passing increasingly tough sentencing laws from the 1970s to the 2010s. But as Bill Stuntz has argued, it’s not always clear that prosecutors want to impose the tougher sanctions legislatures provide them.[35] In fact, Mass Incarceration Nation itself cites sentencing data that shows that the median time served by those in prison barely budged from 1990 to 2009, even as legislatures kept passing tougher sentencing laws, suggesting prosecutors generally did not avail themselves of what the legislatures were providing.[36]

Some of this evasion was informal, but some of it has been much more systematic. Josh Bowers, for example, has pointed out that in the past very few of the people diverted to New York’s felony drug courts actually had a problem with drug use, even though the courts’ official function was to address addiction.[37] Prosecutors appeared to have been using the drug courts as a way to evade the tough, inflexible sentencing provisions of the notoriously harsh Rockefeller Drug Laws in effect at the time. In other words, tougher sentencing laws may encourage prosecutors to shift to lower charges than before in order to achieve the same sentencing outcome. In the data, this could look like a shift from the CJS to the CLS, even if very little was changing in terms of actual sentences imposed.

Police, too, may have had an incentive to push their statistics towards less serious offenses, even if they didn’t change how they targeted serious crime. There has long been some suspicion that computerized arrest software like CompStat—which rolled out nationwide at the same time as the alleged shift away from the CJS to the CLS—encourages police to overstate lower-level (CLS) offenses at the expense of more serious crimes (CJS). The intuition is straight-forward. CompStat both makes it easier to evaluate police commanders on their ability to reduce crime (thus the need for fewer serious offenses) while also tracking to see if their officers in the field are being “productive” (which is measured by looking at arrests, ideally lower-level ones). A recent paper by Laurel Eckhouse, in fact, claims to have detected an effect exactly like this in the NYPD statistics on rape.[38] This, again, suggests that some shifts in our criminal legal data are due to changes in the data, not in crime, nor necessarily even in punishment.

That these sorts of biases all work in favor of the book’s thesis becomes all the more concerning given that the book’s claim itself seems counter-intuitive. The book never actually provides a clear sense of when, exactly, the shift from CJS to CLS started or gained steam, although it seems to suggest that it was taking place over the 1990s and 2000s. Which is somewhat concerning, because the explanations it gives for the shift, like the fact that criminal legal cases are easier to prove than criminal justice ones,[39] seem more likely to apply in the 1970s and 1980s, when crime was high, than in the 1990s and 2000s, as crime fell. This is especially the case given that policing and prosecutorial staffing rose over the 1990s and 2000s as crime rates fell: caseload pressure, and thus the incentive to rely on the pretextual use of easier-to-prove cases, was inarguably greater in the 1970s and 1980s.

And there is evidence that points in this direction, that pretextual use of CLS-type cases to go after CJS-type crimes was greater before the 1990s than after it. In New York State, for example, the number of people incarcerated for drug offenses is almost entirely uncorrelated with the adoption of tougher drug sentencing laws or their repeal, but almost perfectly tracks rising violence in New York City, especially that tied to the crack epidemic.[40] At a more macro level, if you look at the offenses that have driven prison growth over time, it becomes immediately clear that less-serious offenses were targeted more in the 1980s than in the 1990s and 2000s, even as serious violence fell in the 1990s and 2000s—which both complicates the narrative in Mass Incarceration Nation that we are seeing a shift towards CLS offenses, and suggests that any such shift may reflect the pretextual use of CLS crime to target CJS offenses rather than a sincere effort to go after CLS crimes.[41]

III. Conclusion

The questions raised by Mass Incarceration Nation­—how did we get to be one the world’s largest jailers and how do we reverse the process—remain as critical today as they were five, ten, or twenty-five years ago. Efforts to scale back our massive, sprawling prison system continue to move ahead only slowly, and under increasingly unrelenting attack. And the book’s effort to make us think about how our approach to “criminal justice” crimes need not be the same as that towards “criminal legal” offenses is a useful way to frame some of the challenges reform faces. But the book’s failure to engage with the politics of punishment, and its struggles with our messy, contradictory, and often hard-to-understand criminal legal system data result in both a history and reform proposal that fall short.



[1] While Mass Incarceration Nation distinguishes between the “criminal justice” and “criminal legal” systems, I will generally use the term “criminal legal system” to refer to the two of them collectively, clarifying whenever necessary.

[2] As I discuss more below, the book does not cite, or even really engage with arguments of, important works on the social causes of mass incarceration, such as William Stuntz’s The Collapse of American Criminal Justice, David Garland’s Culture of Control, or Philip Goodmans, Joshua Page, and Michelle Phelps’ Breaking the Pendulum (the last of which is perhaps the least well known of these but, in my opinion, arguably the best account of the history of the politics of punishment yet written).

[3] Focusing just on what we should do without examining the politics of how we got to where we are and what that means for what we should propose is actually an acute problem in legal scholarship—this is, in fact, the second recent book review I’ve written criticizing reform proposals for either ignoring or hand-waving away the very real, very difficult politics that criminal legal reform faces. See

[4] See, e.g., Bellin, Mass Incarceration, at 37: “In the 1970s and 1980s, with crime actually rising and the public feeling its effects, the nation reached a consensus—new measures were needed to fight crime.”

[5] See Bellin, Mass Incarceration, at 41-42, citing James Forman’s Locking Up Our Own for the claim that “harsher criminal laws found support among African Americans,” and pointing to the Black Congressional Caucus’s support of Bill Clinton’s 1994 Crime Bill. It is true that Black leaders in DC in the 1960s (the subjects of Forman’s book) supported increased enforcement—which is not the same thing as increased severity, a line that this book often blurs—but Forman’s point is that they did so along with demands (ignored by Congress) for more long-run social investments as well. Their position was not either/or for severity and rehabilitation, but both-and. And while it is true the CBC eventually supported Clinton’s crime bill, it did so grudgingly, not out of real support for the bill, but because it did not want to deal a major blow to Clinton early in his first term. See @JohnFPFaff, X, (May 29, 2019 12:31 PM) There were definitely some Black leaders who came out in favor of severity—see, for example, Michael Fortner’s Black Silent Majority—but overall the issue is far more complex and nuanced than this book makes it out to be.

[6] Why the equivocation? Because one issue almost no one writing about prisons and crime has confronted is the fact that over the 1960s and the 1970s, the two primary measures of crime in the US—the FBI’s Uniform Crime Reports and the Bureau of Justice Statistics’ National Crime Victimization Survey—moved in opposite directions: reported crime rates in the UCR generally rose, while for most offenses victimization reports in the NCVS fell. It’s an issue beyond the scope of this review to tackle, but it’s a complication that at some point demands more attention than it has received so far. See, e.g., Mark T. Berg and Janet L. Lauritsen, Telling a Similar Story Twice? NCVS/UCR Convergence in Serious Violent Crime Rates in Rural, Suburban, and Urban Places, 32  J. Qualitative Criminology 61 (2016).

[7] See, e.g., Bellin, Mass Incarceration, at 49: “Prior to the 1970s, the buzzword of the American correctional system was rehabilitation, not punishment.” This both overstates the collapse of rehabilitation, which never went away even in the 1970s, and overstates the historical constancy with which rehabilitation was the animating goal of American prisons. See Goodman et al, supra, for the best history of these oscillations.

[8] The book asserts “Falling crime causes fear of crime to fall, but not by much. Crime has decreased steadily since the 1990s, but most Americans continue to perceive crime as rising.” Bellin, Mass Incarceration, at 37. This one sentence contains multiple important errors, both about crime and perceptions. First, crime has not fallen steadily since the 1990s. Crime fell from the early 1990s until the early 2000s, at which point it roughly plateaued (at which point polling questions of the sort “is crime higher this year than last year” will suddenly become noisier). Second, though, and more important, between 1993 and 2001, as crime fell sharply, the percent of Americans who answered “yes” to a Gallup poll asking if crime was higher this year compared to last year fell steadily from 87% in 1993 to 41% in 2001—a drop of more than half. Now, after 2001, even as crime stayed flat, the percent who answered “yes” to the crime-is-rising question jumped up to ~70%, but it’s problematic to compare survey answers in 1993 to, say, 2013 and argue they are roughly the same while ignoring the large decline in the middle. See, e.g., @JohnFPfaff, X (Dec. 16, 2020, 4:46 PM)

[9] @JohnFPfaff, X (Dec. 16, 2020, 4:46 PM)

[10] For a discussion of how reformers were not prepared to exploit the 2000 dot-com bubble bursting, see Marie Gottschalk, Cᴀᴜɢʜᴛ (2016).

[11] See, e.g.  John D. Pfaff, The Poor Reform Prosecutor: So Far From the State Capital, So Close to the Suburbs, Fordham Urb. L.J. (forthcoming),

[12] The Bloomberg piece suggests that crime coverage was tied less to actual crime and more to the efforts by unscrupulous politicians (obviously Republicans, but also Democrats like Eric Adams) to exploit people’s overall sense of vulnerability by invoking crime. See Fola Akinnibi and Raeedah Wahid, Fear of Rampant Crime Is Derailing New York City’s Recovery, CityLab, July 29, 2022, After all, it’s well known that at least since the time of Nixon’s Southern Strategy, if not before, “crime” has often been seen as a way to push back on civil rights without saying so explicitly.

[13] Although even here, this is surely not always true. Sam Quinones’s Dreamland, for example, notes that one reason doctors prescribed opiates to the degree that they did was because insurance companies eventually refused to cover more holistic treatment options. No one was “on the side of” chronic pain, but just saying “holistic care works better” was not enough to get insurance companies to supply that option. Even in the oncology context, established institutions may not be willing to let go of their equipment, expertise, or even just ego in the wake of some new, pathbreaking discovery.

[14] Pfaff, Locked In (2017).

[15] Bellin, Mass Incarceration, at 189. That idea, of course, raises a host of questions in turn. For example, should the counties pay the average cost, or the marginal cost—which differ substantially—and if the marginal cost (which seems like the more accurate one), how is that calculated, given that it is actually a hard number to pin down? See, e.g., @JohnFPfaff, X, (Mar 21, 2016, 9:35 AM) The book addresses none of these quite significant implementation issues; it doesn’t even note that they exist. This is a common theme in the “solutions” section, which is most just a laundry list of quick hits, most of which have been addressed in greater detail in other places.

[16] I discuss this history in some length in Pfaff, “Policy Failure,” supra, where I criticize Frank Zimring for similarly proposing that states should just pass Realignment-like laws, without confronting the specific political challenges at play.

[17] This is not a swipe at the GOP, but rather just an acknowledgement of political realities, like the “prison gerrymander”: in ~35 states, people in prison cannot vote but count as full people for the purpose of drawing legislative maps, a policy that inescapably bolsters rural GOP power at the expense of urban Democratic power. It’s telling that the states that have abolished the gerrymander have all done so only when the Democrats hold a trifecta. As a general matter, more-rural areas tend to be net beneficiaries of mass incarceration, and given the increasingly political polarization by urbanization, this means that the state Republican parties will, if just from self-preservation, resist aggressive reform.

[18] Bellin, Mass Incarceration, at 191: “That’s the genius of indeterminate sentencing. It places the real sentence-length decision in the hands of nonelected actors and allows them to make those decisions well after the offense.”

[19] Bill Hughes, Even Moden Inmates Face Steep Barriers to Parole, CityLimits (Sept. 17, 2014),

[20] The other flaw with the book’s analysis here is that it misstates the more serious political risk: a politically salient act of recidivism. In this context, the parole board is actually much closer to the risky event than the prosecutor. Parole boards decide well after the offense, but right before the possible reoffense. That’s reversed for prosecutors, and thus—contrary to what the book argues—may give prosecutors more political flexibility, not less. (Plus, as my research increasingly shows, that prosecutors are elected by counties gives them political freedom that governor-appointed parole officers may lack. See Pfaff, supra note __. Again: the weedy details really matter.)

[21] See Bellin, Mass Incarceration, at 97: “Less crime after 1990 did mean fewer arrests for crimes like murder, rape, and robbery. But it didn’t mean fewer arrests overall. That’s because police shifted their efforts to combating other crimes like drug offenses or domestic violence ….” Total crime actually rises until 1993, not 1990, and arrests push on a little more, peaking in 1995 and plateauing until 1997. From 1997 to 2014, total arrests fall from 15.3 million to 11.2 million, and then down to 10.1 million in 2019, a decline of 34%. It’s true that drug arrests did not fall—but neither did they rise, holding relatively steady at 1.6 million from 1997 to 2014 to 2019. But that ends up meaning that arrests for non-drug, non-index crimes—the quintessential “criminal legal system” offenses—fell by 37%, from 11 million in 1997 to 7.5 million in 2014, to 6.9 million in 2019. In other words, police clearly did not shift their efforts to other crimes: drug enforcement was static, and everything else declined, all by roughly the same percent shifts (arrests for index crimes fell by 41% during the time). See Aʀʀᴇsᴛ Dᴀᴛᴀ Aɴᴀʟʏsɪs Tᴏᴏʟ, for 1997 and 2014 data, and Table 29, Estimated Number of Arrests, (2019) for 2019 data.

[22] Bellin, Mass Incarceration, at 112: “Yet the general correspondence [between arrests and convictions] remains. Comparing 1986 to 2006, aggravated assault arrests went up by almost 100,000 and convictions increased by about 60,000.” I cannot reproduce the figure the book is referring to (Figure 11.3) here, but in that figure, arrests rise by about 200,000 from 1986 to 1994, then fall by about 100,000 from 1994 to 2006, while convictions steadily rise from about 40,000 in 1986 to 100,000 in 2006. This is very much not a “correspondence.” This is akin to saying that someone who walks up one flight of stairs has done the same thing of work as someone who walks up five and then down four.

[23] Bellin, Mass Incarceration, at 111: “As police arrested more people for selling drugs, prosecutors prosecuted more folks, resulting in more convictions.” Yet like with aggravated assaults, the graph (Figure 11.2) shows that drug arrests rose until about 1990 and then slowly declined, while prosecutions rose over the whole period (albeit more quickly prior to 1990 than after).

[24] Bellin, Mass Incarceration, at 105: “Overall LAPD arrest totals declined from 1980 and 2010,” which is nominally true, although Figure 10.2 on that page makes it clear that arrests rose from 1980 to 1990, and then declined through 2000 and 2010. (Perhaps: the graph and data only cover five offenses—burglary, aggravated assault, drugs, gambling, and drunkenness—and there is nothing about whether these offenses track other crimes as well.)

[25] Bellin, Mass Incarceration, at 177: “The logic of drug decriminalization applies more broadly to a host of criminal legal system offenses that are best understood as efforts to reduce harmful behaviors like gun possession, drunk driving, and parole and probation violations. … The logic arguably extends to even to unequivocally abhorrent and harmful behavior like domestic violence. … A 2020 meta-analysis of the many of studies on the effectiveness of arrests [sic] in reducing domestic violence found little evidence to support the status quo.” The title of that 2020 meta-analysis: Susan J. Hoppea et al., Mandatory Arrest for Domestic Violence and Repeat Offending: A Metanalysis, 53 Aggression and Violent Behavior 7 (2020). That “Mandatory” is key—the paper is not about arrests in general, even though that is how the book frames it, in a section arguing for the decriminalization of domestic violence. That is a significant misframing of the studies the meta-analysis considers.

[26] I italicize “incarceration” here to emphasize that the claim may have more validity for other parts of the criminal legal system, like arrests (although, as I point out in note xxix below, I’m somewhat skeptical of this claim as well).

[27] By the end of 2020, the last year with data, the percent serving time for violence had risen to over 62%, but Covid restrictions likely inflated the number of people serving time in state prisons for violence, so by stopping in 2019 I’m actually biasing my argument here in favor of the book’s assertion. Sources:,,,,

[28] Bellin, Mass Incarceration, at Ch 8. See, for example, p. 101, which states that “Crimes [referring particularly to assaults] that might have been ignored decades earlier or handled informally were becoming criminal convictions.”

[29], You might ask why I am using data that is almost old enough to vote. Fair question. It’s because 2006 is the last year for which we have NJRP data, and there is no other national-level data that provides the NJRP’s level of detail.

[30] Bellin, Mass Incarceration, at 122-123.

[31] A simple, if extreme, thought experiment can make this effect clear. Assume a state admits one person to prison in Year 1 to serve a 10 year sentence for murder, and then admits one person each year to serve a one-year sentence for drug trafficking. In any given year over the next decade, half the people in prison are serving time for violence (since there are two people in prison: the person convicted of murder serving his ten years, and the one person in on drug charges serving their one year). But over that decade, drug cases make up over 90% of all people who spend any time in prison.

[32] The book simply says “As a result [of this “churn” effect], drug offenses make up a larger share of the people ever incarcerated than they do of the people incarcerated at any particular time. And that percentage increased dramatically over the period of Mass Incarceration.” Bellin, Mass Incarceration, at 123. There is no citation for the “massively,” nor any numeric value.

[33] John F. Pfaff, The War on Drugs and Prison Growth: Limited Importance, and Limited Legislative Options, 52. Harv. J. on Legis. 1732 (2015). The data I used allowed me to see if the admissions were the same people over time, so I was able (with some limitations) to directly observe the “ever admitted” effect the book refers to.

[34] Of course, it is possible that that initial charge overstates the actual harm caused. And perhaps that overstatement itself is driven by the plea process, if the prosecutor overcharges in order to bargain down to what he wanted in the first place. So people may plead to something that overstates their behavior, but not that is greater than their initial charge. Which is to say: there are layers and layers of interpretative challenges here. Unfortunately, this book does not engage with them in any meaningful way.

[35] William Stuntz, Bordenkircher v. Hayes: The Rise of Plea Bargaining and the Decline of Law, in Criminal Procedure Stories (Carol S. Steiker, ed., 2005).

[36] Bellin, Mass Incarceration, at 124, cites a study by the PEW Research Center showing that from 1990 to 2009 the median time served for drug offenses rose by ~7 months (total, over 20 years, or by ~10 days per year), for property crimes by ~6 months (or ~9 days per year), and for violent crimes by ~15 months, or slightly more than a month per year. Mass Incarceration Nation cites these numbers as an argument that sentences have gotten longer. I actually cite this same data in Locked In, but to make the opposite point: that increases of a few days per year, especially for the more criminal-legal offenses, could not have been the primary engine of prison growth. See Pfaff, Locked In, at 58-59.

[37] Josh Bowers, Contradindicated Drug Courts, 55 UCLA L.Rev. 783 (2008)

[38] Laurel Eckhouse, Metrics Management and Bureaucratic Accountability: Evidence from Policing 66 Am. J. Pol. Sci. 385 (2022),

[39] See, e.g., Bellin, Mass Incarceration, at 97-99.

[40] Pfaff, Locked In, at 216-17. I have also heard from lawyers in New York City that the City’s Special Narcotic Prosecutor, an unelected prosecutor housed in the mayor’s office to target drug offenses, will only intervene in drug cases that come from precincts with sharply rising violence—an open acknowledgement that the drug cases are pretextual ways of targeting violence.

[41] Pfaff, Locked In, at 181-82, Tables 1A and 1B. Between 1980 and 1990, as violence rose sharply, people incarcerated for violence accounted for 36% of state prison growth, and for drugs 33%. From 1990 to 2009, as violence fell significantly but prison populations kept rising, people incarcerated for violence were responsible for 60% of prison growth, versus 14% for drug offenses. Even taking “churn” into account, it is clear that prison growth in 1990s and 2000s was driven by a shift towards crimes of violence, not lesser, non-violent crimes.


John Pfaff is a Professor of Law at Fordham University.


Start typing and press Enter to search