Compassionate Release and Decarceration in the States


The California Medical Facility (“CMF”) is a medium-security prison that houses some of the oldest and sickest people incarcerated in the California state prison system.1 Some of the people incarcerated at CMF have metastatic cancer or end-stage liver or lung disease.2 When they have to be admitted to a community hospital for treatment, they are shackled to their hospital beds and monitored 24 hours a day by prison guards.3 Some people at CMF are “bedridden, incontinent, and lost in the fog of dementia”;4 others are physically frail and unable to get dressed, use the bathroom, or bathe without assistance. In CMF’s seventeen-bed hospice unit, other incarcerated people care for patients in their final days of life, brushing their teeth and changing their soiled bedding.5

CMF is not unique. Prisons across the United States increasingly resemble “nursing homes behind bars.”6 The number and proportion of people age 55 and older in prison has skyrocketed over the last 30 years, a phenomenon sometimes described as the “graying” of the prison population.7 Today, 13.1 percent of people in prison are age 55 or older8 and, as is true of the prison population as a whole, they are disproportionately Black or Latinx.9 As the population of older people in prison has increased, so too has the prevalence of diseases closely associated with aging, such as cancer, heart disease, and dementia.10 Some states have built specialized prison units that provide nursing home level care.11 Others have built designated “prison wards” within community hospitals to accommodate the growing number of people in prison who require medical care beyond what a prison infirmary can provide.12

More than a decade into an ostensible consensus—however thin—on the need for some degree of decarceration,13 progress toward reducing the prison population has been glacially slow. At the current pace of decarceration, it will take 60 years just “to cut the U.S. prison population in half.”14 Releasing people with serious or terminal medical conditions is often described as a commonsense way to accelerate the pace of decarceration.15 And indeed, a legal mechanism for releasing people with serious or terminal medical conditions already exists in nearly every state. Today, 49 states and Washington, D.C. provide for some form of compassionate release: early release from prison based on a serious or terminal medical condition.16

Despite the ubiquity of compassionate release provisions, very few people are granted compassionate release from state prisons. On average, each state grants compassionate release to between four and seven people per year.17 Even as the COVID-19 pandemic has torn through prisons across the country, putting older and medically vulnerable people at severe risk, compassionate release grants from state prisons have remained rare.18 One advocacy group has described compassionate release in the states as simultaneously “everywhere and nowhere.”19

Legal scholars have largely ignored compassionate release in the states.20 This is not surprising. State laws are one “of criminal justice scholarship’s most familiar blind spots[.]”21 Though nearly 90 percent of people in prison in the United States are held in state rather than federal prison, criminal law scholarship focuses disproportionately on the federal legal system, which is distinct and in many ways unrepresentative.22

The dearth of scholarship on compassionate release in the states is a missed opportunity, in two respects. First, examining compassionate release in the states sheds light on the decarceral possibilities and limitations of other early release mechanisms, such as geriatric release and second look sentencing.23 Second, compassionate release in the states is a case study in one of the thorniest questions about achieving decarceration: what about people incarcerated for violent convictions?

In referring to “violent convictions,” I am mindful of the many compelling arguments challenging the coherence and utility of this category. Scholars have rightly argued that the category of “violent” crime is frequently overbroad and criticized the use of the violent/nonviolent binary to structure criminal law and policy.24 At the same time, I argue, it is important to acknowledge the unique and daunting obstacles to the decarceration of people serving prison sentences for convictions broadly understood to be violent.25 The category of “violent convictions” is harmful when it is used to reinforce a rigid distinction between people incarcerated for nonviolent convictions, who deserve leniency and second chances, and people incarcerated for violent convictions, who deserve fear and loathing. But the category is useful when it highlights the particular challenges facing people incarcerated for violent convictions, in order to address them.

Addressing those challenges is key to ending mass incarceration. People incarcerated for violent convictions serve the longest sentences and account for more than half of the people in prison.26 The outsize attention to the federal legal system—where people incarcerated for violent convictions account for a far smaller share of the prison population than in the states27—has often obscured the challenge of violent convictions for decarceration.28 But there is wide agreement among scholars that meaningful reductions in the prison population will never be possible unless reforms reach people incarcerated for violent convictions.29 So far, however, decarceral measures have largely been limited to the group of people Marie Gottschalk calls the “non, non, nons”: people accused or convicted of nonviolent, nonsexual, nonserious offenses.30

Leading scholars and advocates have emphasized the need to make people accused or convicted of violent crimes eligible for decarceral reforms.31 This is a critically important step. But eligibility is not the end of the story. Even if people incarcerated for violent convictions are eligible for a decarceral measure, they may not actually benefit.

Compassionate release in the states illustrates the challenges beyond eligibility. People age 55 and older are, by far, the age group most likely to meet the medical criteria for compassionate release.32 They are also the age group most likely to be incarcerated for a violent conviction,33 under a relatively narrow definition of that term.34 Two in three people age 55 and older in state prisons are incarcerated for a violent conviction.35 Perhaps surprisingly, given their broad exclusion from other decarceral measures, many people incarcerated for violent convictions are eligible to seek compassionate release.36 In practice, however, people incarcerated for violent convictions face what I call an anti-release default, due to the twin obstacles of extreme risk-aversion and retributivism. Extreme risk-aversion leads to compassionate release denials based on speculative, far-fetched fears that the applicant will go on to commit a crime if released; retributivism leads to denials based on the belief that the applicant deserves harsh punishment, even if they do not present a public safety risk. These obstacles are not unique to compassionate release. As is true for compassionate release, most people who qualify for geriatric release or second look sentencing will also be incarcerated for violent convictions.37

Obscuring or de-emphasizing the prevalence of violent convictions among potential beneficiaries of compassionate release and other back-end release mechanisms may be a useful political strategy for getting laws enacted in the first place. But ultimately, the success of these laws as tools for decarceration must be judged by how many people they benefit. While people incarcerated for violent convictions will be the hardest cases for any early release measure, those hard cases will be the rule, not the exception.   

This Article proposes a new approach to the challenge of violent convictions for decarceration. If they are to realize their potential as tools for decarceration, I argue, compassionate release and other early release measures must reduce the obstacles for people incarcerated for violent convictions. In other words, they should be designed for the hardest cases. This Article models this approach using compassionate release as a case study.

The approach I propose is intentionally provocative. Designing early release measures for the hardest cases puts people incarcerated for violent convictions—the “third rail” of criminal legal reform38—at the center of decarceral efforts. To be clear, implementing this proposed approach will not be politically easy. But the political calculus around questions of crime, punishment, reform, and abolition is evolving rapidly.39 Leading advocates and scholars have emphasized the need for the deep, difficult, and long-term work of challenging the reductive ways many members of the public and decisionmakers within the criminal legal system think about people accused or convicted of violent crimes. Violence is not an identity, as the pernicious label of “violent offender” suggests, but rather a discrete act.40 Acts of violence that lead to conviction and incarceration are frequently committed in conditions of profound deprivation, where the line between perpetrator and victim is blurry and in flux.41 The visionary work of restorative and transformative justice practitioners, prison abolition advocates, and people directly impacted by both violence and mass incarceration (sometimes referred to as “dual victims”42) is critically important to realizing lasting changes in the criminal legal system’s treatment of people accused or convicted of violent crimes.43 The approach I propose complements this deeper, longer-term work.

The Article proceeds in six parts. Part II situates compassionate release in the broader context of back-end release measures. Part III argues that to be effective tools for decarceration, back-end release measures must work effectively for people incarcerated for violent convictions. Part IV describes the ways in which compassionate release laws and policies disadvantage people incarcerated for violent convictions, even when they are technically eligible to apply. Part V proposes changes to compassionate release laws and policies that would reduce the obstacles to compassionate release for people incarcerated for violent convictions. Part VI concludes.

Finally, two brief notes about terminology. This Article describes people currently serving a prison sentence for a violent conviction as “people incarcerated for violent convictions.”44 This Article also uses the phrase “older people” to refer to people in prison who are age 55 or older. Age 55 or older is the standard definition for who counts as an older person in prison, in order to account for the phenomenon of accelerated aging among people in prison.45

Back-End Tools for Decarceration in the States

After decades of steady increases in the imprisonment rate and the total prison population, there are some signs the United States has entered a period of slight decarceration.46 Since 2009, both the imprisonment rate and the total number of people in prison have slowly but consistently decreased.47 The questions of how far decarceration should go,48 and whether significant decarceration should be a goal in itself or merely a step toward prison abolition,49 are highly contested. But by nearly any metric, progress toward decarceration has been exceedingly modest.50 Both the total U.S. prison population and the U.S. imprisonment rate remain remarkably high compared to either historical or international baselines,51 and the racial disparities in the prison population remain staggering.52 In sum, mass incarceration is alive and well.53

There is no silver bullet reform that will achieve significant decarceration overnight. Mass incarceration was built piece by piece and must be dismantled the same way.54 Achieving significant decarceration will require an “all-of-the-above” approach that incorporates many different tools and strategies,55 including changes at both the front- and back-end of the criminal legal system. At the front-end, decarceral strategies include changes to policies and practices in areas such as policing, charging, bail, plea bargaining, and sentencing.56 Back-end release measures create or expand the mechanisms for early release from prison.57 Though often overlooked in the scholarship about decarceration,58 back-end release measures are an important tool for decarceration.59 While some back-end measures, such as expanded good time credit, operate largely by default,60 I focus on discretionary early release measures: those where a decisionmaker evaluates an individual’s case and decides whether to grant early release.

I also focus on back-end release measures in the states. Despite the outsize attention paid to the federal system in media coverage and criminal legal scholarship,61 the federal criminal legal system accounts for just 12 percent of the prison population.62 Nearly 90 percent of people incarcerated in U.S. prisons are serving sentences for state convictions.63 Due to the significant differences between the federal criminal legal system and the states, effective strategies for decarceration in the states will look quite different from those that work in the federal system.64

In general, expanding or creating back-end release measures “has proven far less feasible politically than shortening sentences going forward.”65 Where a person has already been sentenced to prison, early release means “giving up something society thinks it owns – the longer sentence.”66 Back-end release mechanisms that target particular groups, rather than sweeping across the board, are one way of making these measures more palatable.67 Each of three discretionary back-end release measures I describe targets a particular group of people and has attracted significant attention from state policymakers and advocates: second look sentencing measures (people who have served a lengthy period of time in prison), geriatric release (older people who have served a lengthy period of time in prison), and compassionate release (people with a serious or terminal medical condition).

Second Look Sentencing

Second look sentencing measures would empower judges to review and reduce the sentences of people who have served many years in prison. Proposals for second look sentencing measures typically provide for automatic eligibility for review by a sentencing court after someone has served a lengthy period in prison, such as ten, fifteen, or twenty years.68 After reaching the time-served threshold, second look review would occur either automatically or upon application of the incarcerated person.69 The sentencing court’s second look review would be similar to a resentencing hearing. The judge would consider whether the original sentence was still justified in light of the purposes of sentencing.70 Notably, the leading proposal for second look sentencing measures does not exclude anyone based on their crime of conviction.71

Among legal scholars, “there has been a remarkable consensus that a second-look approach . . . is necessary in our criminal justice system.”72 Advocates for second look sentencing laws argue, persuasively, that second looks would provide some counterbalance to the U.S. criminal legal system’s excessively punitive sentencing practices. The United States is unique among democracies in relying so heavily on the most severe sentences—decades-long prison terms and life sentences, including life without parole.73 The reliance on such severe sentences has increased dramatically over the last half century.74 Douglas Berman has argued that a second look provision is important “because we can expect, we should expect, first looks to be dysfunctionally harsh.”75

While no state has yet enacted a general second look sentencing measure of the sort I describe above,76 several states have enacted second-look provisions narrowly targeted at particular groups, such as people who were under age 25 at the time of their crime of conviction.77 Legislation that would create such provisions is pending in several states78 and calls for the enactment of second look sentencing legislation continue to gain steam.79

Geriatric Release

Older people who have served a lengthy period of time in prison are often described as particularly attractive targets for decarceration.80 Geriatric release benefits people in this group. Geriatric release provides for early release from prison based on advanced age and time served, with no medical requirements. It is similar to second look sentencing measures in that there is a time served requirement, but unlike second look sentencing measures, geriatric release is only available to people above a certain age.

The most common age/time served requirement for geriatric release is at least 60 years old and at least ten years of time served.81 Once a person reaches a certain age and satisfies the time served requirement, they can petition for early release; the most common decisionmaker is the parole board.82 Unlike many proposed second look sentencing measures, geriatric release laws exclude some people who meet the age and time served criteria based on their conviction or sentence.83

Today, 11 states and D.C. provide for geriatric release.84 Geriatric release laws have been a legislative priority of the Justice Reinvestment Initiative (“JRI”), an influential partnership between the Department of Justice Bureau of Justice Assistance and the Pew Charitable Trusts that has played a significant role in state efforts to reform the criminal legal system and reduce spending on prisons.85 JRI championed geriatric release measures as a means of accomplishing “two common goals of the JRI process: reserving prison for people who pose a high risk for re-offending, and reducing costs.”86

While the cost saving arguments for geriatric release may be overstated,87 the arguments for geriatric release draw support from a large body of social science research on the age-crime curve, which finds that advanced age is one of the strongest predictors of non-offending.88 Recidivism rates for older people released from prison are much lower than for younger age groups, including for people who served prison sentences for the most serious violent crimes.89

Compassionate Release

Compassionate release, which I define as early release for people who have a serious or terminal medical condition, is the only back-end release measure available in nearly every state.90 Today, every state except Iowa provides for compassionate release.91

Compassionate release in the states is always a multi-step process. To apply for compassionate release, someone must meet the medical eligibility criteria; usually, this determination is made by a prison doctor. The two dominant models of medical eligibility are the prognosis model and the physical incapacitation model.92 The prognosis model defines medical eligibility by life expectancy (e.g., death expected within six to twelve months).93 The physical incapacitation model defines medical eligibility by some measure of physical ability, ranging from an inability to perform “activities of daily living”94 to complete physical incapacitation, as where someone is in a coma or a permanent vegetative state.95 Some people who meet the medical eligibility criteria for compassionate release are not eligible to apply because they are excluded based on their conviction or sentence.96

If someone meets all eligibility criteria, their application is reviewed by the decisionmaker: most commonly the state parole board, but sometimes by the head of the state Department of Corrections, the governor, or the sentencing court judge.97 The compassionate release decision is always discretionary. Compassionate release is not the end of punishment: typically, the person is released onto community supervision, a significant form of punishment in itself.98

Compassionate release statutes first emerged in the 1980s, but the practice of compassionate release is not new. In the nineteenth and most of the twentieth century, release from prison based on a serious or terminal medical condition was accomplished through executive clemency99 or parole100 rather than through statutory compassionate release provisions. The 1980s and 1990s saw the first sizeable wave of formal compassionate release laws, sometimes enacted in the wake of a remarkable constriction in parole boards’ release authority.101 By 1994, 27 states and the federal system had formal compassionate release statutes or policies.102 In the summer of 2021, Illinois became the 49th state to enact a compassionate release provision.103

The ubiquity of compassionate release laws reflects the compelling justifications for these provisions. First, given how strictly medical eligibility criteria are defined, people with a serious or terminal illness will rarely present a significant public safety risk if released.104 Because people age 55 and older are the most likely to be medically eligible for compassionate release,105 the public safety arguments in favor of geriatric release also apply here.106 Second, compassionate release alleviates some of the suffering, indignity, and health-harming effects of incarceration for people with a serious or terminal medical condition.107

Like geriatric release, compassionate release is frequently described as a cost-saving measure. Again, many of the boldest cost-saving claims are overstated.108 Nonetheless, while the cost savings of compassionate release are difficult to estimate precisely,109 increasing the number of compassionate release grants will reduce spending on prescription medications,110 outside medical care not covered by Medicaid or Medicare, and staff overtime associated with transporting and guarding people who require hospitalization, visits to specialists, or other outside medical care.111

The three back-end release measures I have described—second look sentencing, geriatric release, and compassionate release—overlap to some degree. For example, someone who is in their late sixties, has served over twenty years in prison, and has a serious or terminal medical condition could conceivably qualify for all three types of back-end release. Together, these three measures could reach a significant portion of the state prison population and are potentially powerful tools for decarceration.

The Challenge of Violent Convictions

The biggest challenge to realizing the decarceral potential of the back-end release measures I described in Part II is the prevalence of violent convictions among potential beneficiaries. This is a variation on a familiar theme. One of the biggest challenges to achieving significant decarceration through any means is the reluctance, so far, to make meaningful changes to the treatment of people convicted of violent crimes, who account for over half of the people in state prisons and serve the longest sentences, increasing their impact on the total prison population.112

Despite a consensus among scholars that decarceration is impossible if it fails to reach people incarcerated for violent convictions,113 people in this group have been largely excluded from decarceral reforms over the last decade.114 For example, decarceral measures focused on people accused or convicted of nonviolent drug offenses have been a mainstay of criminal justice reform legislation since 2008.115 But the United States could release every person serving a prison sentence for a drug offense—of any type—tomorrow and the U.S. prison population would still be the world’s largest and the U.S. incarceration rate far higher than historical averages.116

This Part defines “violent convictions” and describes their prevalence among potential beneficiaries of the early release measures described in Part II. I argue that for compassionate release and other back-end release measures to realize their potential as decarceral tools, they must benefit significant numbers of people incarcerated for violent convictions. In light of this reality, this Part proposes a new lens for evaluating back-end release measures: how well do they work for people incarcerated for violent convictions? In other words, are they designed for the hardest cases?


Though the concept of violence feels familiar and intuitive (“I know it when I see it”), scholars have persuasively argued that the category of “violence” is difficult, if not impossible, to define coherently.117 Criminal laws frequently define violent crimes expansively, to include offenses that result in no physical harm to a person, such as burglary, unlawful weapon possession, or drunk driving without an accident.118 At the same time, definitions of violence frequently exclude “violence that doesn’t fit our preconceptions.”119

In describing the prevalence of violent convictions, I rely on the narrow Bureau of Justice Statistics (“BJS”) definition of violent offenses.120 But no matter how narrow the definition, any discussion of “violent crime” “encompasses widely varying conduct performed by widely varying individuals in widely varying circumstances.”121 Similarly, any definition of “violent” convictions will include some convictions that do not satisfy any commonsense definition of violence.122 Given the incentives to plea bargain rather than take a chance at trial, and the profound resource shortages that plague indigent defense services throughout much of the country,123 we should also be wary of assuming that “conviction and crime commission are the same thing.”124 Finally, focusing on the violent nature of the underlying convictions should not obscure or distract from the violence inherent in incarceration and the prior violent victimization and trauma frequently seen in the life histories of people convicted of violent crimes.125

David Sklansky has observed that “[n]o distinction plays a larger role in contemporary American criminal law than the line between violent and nonviolent offenses.”126 The critiques of the violent/nonviolent binary are persuasive. At the same time, however, we must acknowledge the challenge of violent convictions for decarceration in order to address it squarely. In a legal proceeding under one of the back-end early release measures described in Part II, a conviction for homicide, sexual assault, felony assault, or robbery is a highly salient fact. The conviction—and, likely, the person convicted—will be perceived as violent. That perception has real consequences. Articulating the shortcomings of relying on “violent crime” as a coherent category is important. But so too is understanding—and redressing—the unique obstacles to early release facing people incarcerated for violent convictions.


People serving sentences for violent convictions account for over half of all people incarcerated in state prisons.127 The prevalence of violent convictions is likely even higher among people in state prison who may qualify for second look sentencing, geriatric release, or compassionate release based on their time served in prison, age, or medical condition.128

People age 55 and older in state prisons are by far the age group most likely to meet the medical eligibility criteria for compassionate release,129 and the only age group possibly eligible for geriatric release. They are also the age group most likely to be incarcerated for a violent conviction.130 Two thirds of older people in state prisons are serving a prison sentence for a violent crime under the narrow Bureau of Justice Statistics definition,131 compared to 55 percent of the state prison population as a whole.132 In other words, about one in two people of any age in state prison is incarcerated for a violent conviction; among older people in state prison, that number is two in three. The category of older people in prison incarcerated for violent convictions includes both people serving long sentences imposed when they were young, as well as a significant number of people admitted to prison after they were already age 55 or older.133

The prevalence of violent convictions among people who would qualify for proposed second look sentencing is likely similar.134 People incarcerated for violent convictions serve, on average, more than twice as long in prison as those incarcerated for drug, property, or public order convictions.135 People incarcerated for violent convictions—especially people convicted of homicide—also account for a strikingly large proportion of those who have served ten years or more.136

The numbers in this Section reflect only the conviction for which someone is currently serving a prison sentence, not their prior convictions.137 For this reason they almost certainly understate the true prevalence of violent convictions among potential beneficiaries of compassionate release, geriatric release, and second look sentencing. But a prior violent conviction is likely to be a salient fact in early release proceedings, even where the person is currently incarcerated for a nonviolent conviction.

Designing for the Hardest Cases

For the early release measures described in Part II, the challenge of reaching people incarcerated for violent convictions is twofold. First, at the policy level, people incarcerated for violent convictions may not be eligible for release. The remedy for this challenge is straightforward legally (though complicated politically): Statutes enacting back-end release mechanisms should not exclude people from eligibility based on their crime of conviction. The second challenge, however, is more complex. Even if they are technically eligible, people incarcerated for violent convictions will face significant obstacles when they seek early release.

People incarcerated for violent convictions will be the hardest cases for any early release measure. But the hardest cases will also be the most common. To maximize their potential as tools for decarceration, early release measures should seek to reduce the obstacles to release for people incarcerated for violent convictions. In other words, early release measures should be designed for the hardest cases. In evaluating early release measures, we should ask: how well does this legal mechanism work for people incarcerated for violent convictions? What changes are necessary to ensure that an early release measure reaches people incarcerated for violent convictions? This approach would align policy with demographic reality.138

For people with deep retributive commitments, this proposed approach may trigger strong objections.139 The question of whether and when early release for people incarcerated for violent convictions is compatible with retributivism is a complicated one.140 So, too, is the question of whether retributive objections should trump efforts at decarceration for people serving sentences for violent convictions.141 I do not attempt to resolve these questions here. My point is an instrumental one. In order for the back-end release measures I have described to realize their decarceral potential, they must reach significant numbers of people incarcerated for violent convictions. There will often be a tension between the need for significant decarceration and fidelity to retributive values.142 I identify a path forward if the choice is to prioritize achieving decarceration over fidelity to retributivism.

The first step in designing early release measures for the hardest cases is to understand the obstacles people incarcerated for violent convictions face when seeking early release. I argue that the dynamics of extreme risk-aversion and retributivism combine to produce an “anti-release default” for people incarcerated for violent convictions who seek early release. Extreme risk-aversion is the desire to keep someone incarcerated just in case they might go on to commit a serious crime if released, no matter how unlikely that may be. Extreme risk-aversion leads to release denials based on speculative, far-fetched fears that the applicant will go on to commit a crime if released. Retributivism is the desire to keep someone incarcerated because they deserve harsh punishment. Even where it is highly unlikely that the person would commit another crime if released, retributivism may provide independent grounds for continued incarceration.143

Extreme Risk-Aversion

Throughout the criminal legal system, from initial bail decisions to parole, the default position in release decision-making is “to keep someone locked up, ‘just in case.’”144 Typically, the decisionmaker’s risk assessment inquiry is framed as: Could this person, upon release, do something that I, the decisionmaker, might regret?145 Given this framing, the answer will almost always be yes. There is almost always some risk—no matter how small—that any person released from prison could commit a crime (just as there is almost always a risk that any person, regardless of past involvement with the criminal legal system, could commit a crime).

Extreme risk-aversion finds release permissible only where risk is eliminated, not just significantly reduced. The salient risks to be avoided are two-fold. First, and most obviously, there is the public safety risk: that the person will commit some crime (though what type of crime is often unspecified) after release. Second, there is professional risk for the decisionmaker. If the person granted release does commit a serious crime, the decisionmaker could lose their job or face political punishment. This is not a speculative fear. Frequently, in the wake of a violent offense committed by someone granted early release, the individual decisionmaker pays the price with their job.146 As a result, decisionmakers have profoundly asymmetric incentives: there is little personal cost to denying release, but the possibility of professional ruin if they grant early release to someone who goes on to commit a serious violent crime. A release decisionmaker “never assumes a personal risk by voting to deny release—and always takes a chance when letting someone out.”147

Extreme risk-aversion is likely to be a particularly strong obstacle to early release for Black people and other people of color. Pernicious and enduring racist stereotypes that associate Blackness with dangerousness and criminality have shaped individual and systemic outcomes throughout the criminal legal system.148 It would be shocking if they did not exacerbate obstacles to release in the early release context as well.

Compared to people incarcerated for drug, property, or public order offenses, people incarcerated for violent convictions are perceived as a higher risk of serious violent recidivism if released.149 Research does not bear this out—especially for older people incarcerated for violent convictions. Recent research has found a remarkably low reincarceration rate (a proxy for recidivism) for older people released from prison after serving at least five years for a homicide offense.150 For example, a recent study of 3,000 older people released from prison in New York and California after serving at least five years for a homicide offense found that only nine people were reincarcerated for a new crime within three years of release.151

A robust body of empirical research demonstrates that older people incarcerated for violent convictions are a very low risk of recidivism if released.152 This body of research extends even to the most feared category of people in prison: people convicted of sex offenses against children.153 But social science research findings, no matter how robust, are no match for the entrenched beliefs about the high risk of serious violent recidivism among people serving sentences for violent convictions.154 In individual cases where the early release applicant is serving a prison sentence for a violent crime, decisionmakers’ risk aversion may lead them to insist on a near guarantee that the person will not commit future crimes,155 which is only possible in the most extreme circumstances, such as where the person is in a permanent vegetative state.


Extreme risk-aversion is rooted in a utilitarian view of incarceration, as primarily aimed at preventing future crimes.156 Retributivism is a different type of obstacle to discretionary early release. Retributivism leads to compassionate release denials based on the belief that the person should be incarcerated on retributive grounds—not because they present a risk to public safety if released, but simply because they deserve more punishment. Retributive justifications for punishment are the most forceful where someone has been convicted of a violent crime (rather than a drug, property, or public order crime). The desire for retribution in such cases is often bound up with fear and loathing. The American criminal legal system treats people convicted of serious violent crimes “as morally deformed people rather than ordinary people who have committed crimes.”157 Unsurprisingly, judgments about desert—like judgments about dangerousness and risk158—are warped by racism, particularly anti-Black racism.159

Contemporary American punishment practices embrace both retributive and utilitarian theories.160 Under a retributive theory of punishment, the purpose of punishment is punishment itself (limited, in theory, by proportionality principles).161 Whereas utilitarian punishment purposes aim to prevent future crime (through deterrence, incapacitation, or rehabilitation), retribution’s goal is to punish crime for punishment’s sake.162 There is robust debate about the specifics of retribution as a punishment theory but the basic idea of all retributive theories of punishment is that the “wrongdoer” is punished because they deserve to be punished, and punished to a degree commensurate with their wrongdoing—even if there is no real concern that they will go on to commit a crime in the future.163

While extreme risk-aversion may be refuted—at least somewhat—with data, claims about the retributive need for continued incarceration are more difficult to rebut. Retribution is a squishy concept; there is no template for determining how much prison time serves the retributive purpose of a sentence. Nor do sentencing judges divide their sentences into utilitarian chunks and retributive chunks.164 Claims about how much punishment is necessary to serve a retributive purpose in a particular case are difficult to refute—not because they are right, but because they are unfalsifiable.165 Even where decisionmakers are persuaded that early release would not present too great a risk to public safety, retributivism will remain a most powerful, all-purpose objection to early release in many cases.166

The Anti-Release Default in Compassionate Release

Compassionate release is the ideal case study for illustrating the dynamics of extreme risk-aversion and retributivism described in Part III. Perhaps surprisingly, no state compassionate release provision categorically excludes everyone serving a sentence for a violent conviction.167 From this perspective, then, compassionate release in the states is a success story about extending decarceral tools to people convicted of violent crimes. But despite the technical eligibility of many people incarcerated for violent convictions for compassionate release, people in this group still face significant obstacles to compassionate release.168

In keeping with the approach I proposed in Part III—designing compassionate release and other early release measures for the hardest cases—this Part identifies specific aspects of compassionate release laws and policies in the states that create an anti-release default for people incarcerated for violent convictions.169 This default operates against people incarcerated for violent convictions at various stages of the compassionate release process—even stages where the underlying conviction is ostensibly irrelevant.

Legal Eligibility

Every compassionate release provision defines medical eligibility, as discussed in the next Section. But not everyone who meets the medical eligibility criteria for compassionate release can apply. Some people who meet the medical eligibility criteria are excluded based on their sentence (e.g., a sentence of life without parole), offense classification (e.g., Class A felony), or offense type (e.g., homicide or a sex offense).170 Frequently, other early release mechanisms affirmatively limit eligibility to people in a narrow, specified category (such as people serving a sentence for a nonviolent drug offense) or categorically exclude anyone serving a sentence for what that state defines as a “violent” conviction.171 No state limits eligibility for compassionate release so significantly. Nonetheless, some people incarcerated for violent convictions are excluded from eligibility for compassionate release.

Surprisingly, nearly a third of state compassionate release programs do not categorically exclude anyone.172 Among programs that do impose legal eligibility exclusions, the most common exclusion is to exclude people sentenced to die in prison: those serving death or life in prison without the possibility of parole (“LWOP”) sentences. Eighteen state compassionate release programs exclude only people serving death and/or LWOP sentences.173 In all, over half of state compassionate release programs have no categorical exclusions or exclude only people serving death/LWOP sentences. State programs that exclude more broadly than death/LWOP sentences typically specify excluded offenses rather than broad categorical exclusions based on offense category (e.g., “violent offenses”),174 though five state compassionate release programs categorically exclude anyone serving a sentence for any sex offense.175 Common excluded offenses are first-degree murder176 or the highest level of felony convictions.177 Some state programs impose time-served requirements only for people serving a sentence for a specific category of conviction.178 Because states do not make available data that cross references age, crime of conviction, and sentence, it is difficult to know how many older people in prison are excluded from applying for their state’s compassionate release programs.179

Advocates and commentators seeking to expand the use of compassionate release provisions have called for the elimination of all eligibility exclusions.180 This is an important and worthy reform. But on its own, it is insufficient to address the obstacles to compassionate release for people serving sentences for violent convictions. The technical eligibility of people serving sentences for violent convictions does not mean they have a real shot at achieving compassionate release. Even when people serving sentences for violent convictions are legally eligible for compassionate release, multiple, interlocking aspects of compassionate release laws and policies disadvantage them.

Medical Eligibility

To be eligible to apply for compassionate release, someone must meet specific medical eligibility criteria. While the medical eligibility determination should be independent of the potential applicant’s crime of conviction, many states define medical eligibility partly in terms of public safety risk. This invites consideration of the underlying conviction, disadvantaging people incarcerated for violent convictions.

Medical eligibility criteria should be value-neutral medical determinations wholly independent of someone’s underlying conviction. The most common definitions of medical eligibility are based primarily on prognosis or physical incapacitation.181 But frequently, medical eligibility criteria also define eligibility partly in terms of public safety risk.182 Medical eligibility may require a finding that the person is “physically incapable of presenting a danger to society,”183 that the person “poses an extremely low risk of physical threat to others or to the community,”184 or that as a result of the medical condition the person is “no longer an unreasonable threat to public safety.”185

Medical eligibility criteria that include public safety invite extreme risk-aversion and retributivism into the medical eligibility determination. Defining medical eligibility in terms of public safety shifts the political and professional risk of supporting compassionate release from the ultimate decisionmaker onto doctors.

Doctors may fear liability or professional consequences if they certify that someone is not a public safety risk and that person goes on to commit a crime.186 Like other decisionmakers in the compassionate release process, doctors are also influenced by retributive factors. One of the few studies on doctors’ decision-making about compassionate release medical eligibility found that prison doctors “considered older and ailing prisoners with convictions for nonviolent drug offenses as appropriate candidates for medical parole” and considered “the nature of the crime in determining [medical] eligibility.”187 Requiring doctors to evaluate a person’s public safety risk if released only exacerbates the tendency to consider retributive factors when determining medical eligibility.

Doctors are ill-equipped to evaluate public safety risk, which is a legal judgment rather than a medical fact. Asking doctors to evaluate public safety risk also blurs their role of providing care for patients.188 Additionally, defining medical eligibility in terms of public safety risk rather than observable medical facts means there are not systematized triggers to initiate consideration of whether a person meets the medical eligibility requirements for compassionate release (e.g., a terminal cancer diagnosis automatically triggering compassionate release consideration). The lack of application triggers and the enormous discretion of prison doctors in deciding which patients they will support for compassionate release makes it more likely that racial bias will influence doctors’ decision-making.189

Release Planning

If someone meets the medical eligibility criteria for compassionate release and is not otherwise barred from applying, they must then navigate the application process, which is typically complex and burdensome.190 A key part of a strong compassionate release application is a viable and verified release plan.191 For people serving sentences for violent convictions, however, developing such a release plan is often challenging and sometimes nearly impossible.

For many people seeking compassionate release, a nursing home, hospice, or assisted living facility is the only viable option given their serious medical needs. Yet nursing homes are frequently hesitant to accept any person coming out of prison, particularly people with convictions for sex crimes, violent crimes, or arson.192 For people with sex convictions, even if accepted into a nursing home, the nursing home may not be a permitted residence because of sex offender residency restrictions.193 Even where the person plans to live with family or friends, there may be legal obstacles to that release plan, such as supervision prohibitions against residing with another person who has a felony conviction or is on supervision, or public housing policies barring residency by people with certain convictions.194

The release plan obstacle is compounded by the failure of most compassionate release laws to require release planning assistance for compassionate release applicants. Despite the difficulties in release planning for compassionate release applicants serving sentences for violent convictions, compassionate release laws and policies typically leave release planning to the individual applicant, without assistance.195 Decisionmakers also unrealistically expect or require a fully formed release plan at the time of application, which creates a chicken-and-the-egg dynamic.196 Compassionate release decisionmakers want a verified release plan before making a decision, but nursing homes cannot promise a bed without a definite release date.197 The end result is that some people who meet the medical eligibility criteria for compassionate release are thwarted by challenges in developing a release plan.

The Release Decision

After medical and legal eligibility are confirmed and the application (including release planning) is complete, the case then reaches the decisionmaker.198 In over half of state compassionate release provisions, the state’s parole board is the final compassionate release decisionmaker.199 The next most common decisionmakers are the head of the Department of Corrections or the Governor.200 The sentencing court is the least common compassionate release decisionmaker in the states.201 The power of parole boards and the Department of Corrections over compassionate release in the states—and the rarity of sentencing courts as decisionmaker—is a significant difference from the federal process, where the sentencing court is the ultimate decisionmaker.202

The review and decision-making process can be lengthy, with multiple stages of review even in cases where the applicant is facing imminent death and no deadlines for review. Review is typically on the papers; hearings are rare, and few states allow lawyers to represent people seeking compassionate release.203 With no required timelines for making a decision on a compassionate release application—even in cases of imminent death—some people die in prison waiting for a decision or while waiting for release after a grant.204

Compassionate release decision-making standards vary widely from state to state, but public safety risk is the dominant consideration. In some state programs, public safety is the sole criteria for deciding whether to grant compassionate release to someone who meets the medical eligibility criteria.205 More commonly, public safety is one factor for the decisionmaker to consider in evaluating a compassionate release application.206 The meaning of public safety risk, however, is typically left vague and undefined, which encourages extreme risk-aversion.207

Like many bail statutes, state compassionate release statutes typically “invoke . . . risk without defining it.”208 Defining risk requires specifying the type of behavior or outcome being assessed and deciding whether the likelihood or probability of that behavior or outcome justifies incarceration rather than release. The risk that someone will commit a homicide is different than the risk that someone will commit a simple assault, which is different still from the risk of a possessory offense or petty theft or the risk that a person will violate the rules of their supervision by, for example, consuming alcohol, staying out past curfew, or associating with another person who has a felony conviction.209 And to meaningfully evaluate risk, there must be a probability threshold for the risk assessment (high risk, moderate risk, etc.) to justify incarceration rather than release. Where the risk is undefined, extreme risk-aversion will fill in the gap. Any level of risk that the person will commit any crime after release will be seen as intolerable. This commonplace, gut-level risk assessment severely disadvantages people incarcerated for violent convictions, who are (wrongly) perceived as a high risk of serious violent recidivism if released.210

The fact of terminal illness or physical incapacitation may not be enough to assuage the decisionmaker’s fear that the compassionate release applicant will commit a crime after release.211 As an example of how extreme risk-aversion shapes compassionate release outcomes, consider Steven Martinez, who was the first applicant for California’s medical parole, a compassionate release provision enacted in 2011. Mr. Martinez, who was serving a prison sentence for violent convictions, became a quadriplegic while incarcerated, due to a knife attack that severed his spinal cord.212 Though he met the medical eligibility criteria of “permanent incapacitation,” he was denied medical parole because he could still speak.213 The parole board reasoned that “he could possibly use his vocal cords, which are not paralyzed, to order crimes, maybe attacks on state employees.”214 Similar examples are available from other states.215

Retributive considerations present another ground for denying compassionate release. Few compassionate release statutes explicitly require consideration of retributive factors, such as whether the person has served sufficient time for punishment and whether release would unduly depreciate the seriousness of the crime. But these retributive factors will be a familiar touchstone for decisionmakers. Parole boards—the most common compassionate release decisionmaker—are heavily influenced by “backward looking retributive”216 concerns, separate and apart from forward-looking fears of a public safety risk if the person is released. In a national survey from 2015, for example, parole board chairs ranked the nature and severity of the current offense as the most important factors in the release decision—above empirically based assessments of the likelihood the person would reoffend.217

Proposals for Reform

In Part IV, I argued that people incarcerated for violent convictions face an implicit anti-release default, at multiple stages of the compassionate release process. In this Part, I propose reforms designed to reduce the obstacles to compassionate release for people incarcerated for violent convictions. My proposed reforms fall into two categories: those designed to cabin the influence of extreme risk-aversion and retributivism, and those designed to counterbalance the anti-release default.

The proposed reforms I discuss below are far from an exhaustive list of how state compassionate release laws should be reformed. I focus on reforms that would reduce specific obstacles to compassionate release that arise based on a person’s violent conviction; this focus illustrates the broader approach I propose of designing early release measures for the hardest cases.218 This focus, however, means I do not discuss other important reforms proposed by commentators, such as requirements that prisons publicize the existence of compassionate release219 or create systems for fast-tracking compassionate release applications by people facing imminent death.220

Cabin the Influence of Extreme Risk-Aversion
& Retributivism

The first step in designing compassionate release laws for the hardest cases is to cabin the influence of extreme risk-aversion and retributivism on the compassionate release process. This requires reforming medical eligibility criteria, the release planning process, and public safety considerations.

Meaningful Medical Eligibility Criteria

Extreme risk-aversion and retributivism impact medical eligibility determinations where eligibility criteria are vague, extreme, or both.221 States should enact clear, operationalizable medical eligibility criteria developed by medical experts and disentangle medical eligibility from public safety risk assessment. Doctors can describe a person’s physical condition, which may inform public safety risk assessment. But doctors cannot and should not make judgments about public safety risk or the level of punishment someone deserves. Those judgments should be reserved for the ultimate compassionate release decisionmakers.

Brie A. Williams, the leading medical expert on compassionate release, has proposed eligibility criteria that identify three distinct groups of people who should be eligible for medical compassionate release: (1) people “who have a terminal illness with a predictably poor prognosis”; (2) people who have “profound cognitive impairment or dementia”; (3) people who have a “serious, progressive, irreversible illness with profound functional or cognitive impairment.”222 These highly specific criteria are far superior to current medical eligibility criteria, especially for people incarcerated for violent convictions, because they prevent doctors from needing to assess public safety risk, and thus limit the influence of risk-aversion. Moreover, these criteria do not place outsize weight on medically unreliable standards such as prognosis223 or permit compassionate release only in the most extreme situations (such as where the person is in a permanent vegetative state).224

Further, clear medical criteria would enable systematized compassionate release consideration, limiting the warping influences of doctors’ risk aversion and moral judgments about who deserves release. Dr. Williams has identified, for each of the three medical eligibility categories described above, a specific medical trigger for an assessment of compassionate release eligibility.225 For people who have a terminal illness with a predictably poor prognosis, for example, the triggering event is the diagnosis (e.g., of new cancer or rapidly progressive terminal illness).226 Such systematized processes for identifying people who meet the medical eligibility criteria—disentangled from public safety considerations—would benefit people serving sentences for violent convictions because they are the group most likely to be negatively impacted by medical staff’s discretionary judgments about who deserves compassionate release.

Creative Approaches to Release Planning

As described in Part IV, nursing homes and hospice facilities are often reluctant to accept people coming out of prison who have a serious violent conviction. While compassionate release laws cannot mandate that nursing homes accept people, they can ease some of the obstacles to release planning. Namely, state compassionate release statutes and policies should require (as some already do) that the Department of Corrections provide assistance with release planning.227 But it is also necessary to expand the universe of housing options for people granted prison who have serious health problems.

Beyond providing assistance with release planning, states should also consider creative approaches to the release planning obstacle. For example, in order to expand post-release housing options for people required to register as sex offenders, states may consider exemptions from residency restrictions where someone has a serious or terminal medical condition.228 Because of the difficulties surrounding release planning for elderly people leaving prison (whether at the end of their sentence or because they are granted compassionate release), some states have considered establishing a nursing home facility specifically for people leaving prison.229 Connecticut has already done so and other states have considered similar proposals.230 The Connecticut nursing home model raises concerns about what Chaz Arnett has described as “the outsourcing of aspects of prison into communities under the guise of carceral humanism: the repackaging or rebranding of corrections and correctional programming as caring and supportive, while still clinging to punitive culture.”231 Yet the problem to which the Connecticut nursing home model responds is very real. For people leaving prison who require nursing home level care—especially those with violent convictions—placement options are few and of low quality.

Addressing release planning challenges is an area that requires creative thinking and program experimentation. Innovation by community organizations—and funding by philanthropies and government to support such innovation—is critical. Nursing homes with a commitment to challenging mass incarceration and promoting racial justice should consider ways to support people leaving prison. Similarly, there is an immense need to expand supportive or transitional housing programs for people released from prison—programs that are already stretched thin—to accommodate those with serious health problems.232

Clear and Rational Public Safety Standards

There are no easy answers when crafting public safety risk standards for judges, parole boards, or other decisionmakers to apply when making release decisions.233 What we do know, however, is that where public safety standards are vague and undefined, extreme risk-aversion and “just in case” thinking will fill the void.234 A clear and rational public safety standard must define risk in two respects: the type of risk that is being assessed (i.e., “risk of what event?”) and the degree of risk that justifies continued incarceration rather than release (i.e., “how likely is this event to occur within a particular period of time?”).235 California’s Committee on the Revision of the Penal Code has recently proposed defining public safety risk in the parole context as “imminent risk that the parole candidate will commit a serious or violent felony if released.”236 That standard would also work well in the compassionate release context.

Additionally, the public safety risk evaluation in compassionate release cases must consider how the medical condition (and, where applicable, advanced age) reduces risk. The role of risk assessment instruments in release decision making is complex and controversial.237 But if risk assessment instruments are used,238 they must account for the applicant’s serious or terminal medical condition. A risk assessment instrument conducted at the time of someone’s admission to prison or years before the diagnosis has no place in compassionate release decision-making. While age is a standard factor in risk assessments, medical condition and physical limitations are not.239 Risk assessment instruments that cannot account for medical condition should be discounted heavily where they indicate high risk levels.

Counterbalance the Anti-Release Default

Cabining the influence of extreme risk-aversion and retributivism is important, but insufficient. To increase compassionate release grants, states must also counterbalance these anti-release dynamics. To do so, state compassionate release laws should broaden the grounds for release and create a presumption of release. Compassionate release laws and policies should also permit the involvement of advocates, who serve a counterbalancing function by marshaling the complex facts of a medical condition into a coherent and compelling narrative.

Broaden the Grounds for Release

Where someone is serving a prison sentence for a serious violent crime, retribution will almost always be a significant obstacle to compassionate release. Statutes should establish that compassionate release is justified where the applicant’s medical condition (alone or in tandem with other factors) has significantly reduced any one of the four conventionally recognized purposes of incarceration as punishment: incapacitation, rehabilitation, deterrence, or retribution. The focus here is on reduced justifications for imprisonment, not eliminated justifications. This recognizes that the justifications for imprisonment will rarely be eliminated, except in the most extreme circumstances where, for example, someone is in a permanent vegetative state.

In their analysis of mitigation in sentencing proceedings, Carissa Byrne Hessick and Douglas Berman have argued that a factor that reduces the justification for any one of the purposes of punishment is legitimate grounds for imposing a lesser punishment.240 States universally recognize both utilitarian and retributive purpose of punishment and neither judges nor legislatures divide sentences into retributive versus utilitarian parts.241 “If legislatures and the general public assume they are accomplishing both retributivist and utilitarian goals with criminal sentences,” they argue, “then it is logical to reduce a defendant’s sentence whenever additional punishment does not further one of those goals.”242 While Hessick and Berman make their argument about front-end sentencing decisions, the same logic should apply to compassionate release.

Compassionate release statutes and policies should also empower decisionmakers to grant release based on a range of other factors that are not grounded in punishment theory. The project of punishment theory is to “justify the imposition of punishment,” not its remission,243 and punishment theory (at least in its conventional form) does not account for the systemic issues of mass incarceration and racial injustice in the criminal legal system.244 Broadened grounds for compassionate release may include, for example, that compassionate release of a particular applicant would protect the health and safety of other people in prison by reducing overcrowding and reducing the demands on limited prison healthcare resources, such as hospice or infirmary beds.245 The COVID-19 pandemic has highlighted the serious safety risks of continued incarceration in overcrowded facilities.246 But compassionate release standards rarely require decisionmakers to consider the health-harming consequences of incarceration for the compassionate release applicant and the dangers of overcrowding for other incarcerated people.247

Any definition of “public safety” should include incarcerated as well as non-incarcerated people. Advocates challenging the use of money bail and the pretrial detention system have redefined and reimagined public safety in that context, challenging the prevailing “defendant-community dichotomy, [which] pits the defendant against the community.”248 A similar reimagining of public safety is important in early release decisions, including compassionate release.

Create a Presumption of Release

The most forceful way to counterbalance the anti-release default would be to create a presumption of release where someone meets the legal and medical eligibility criteria. The presumption of release should be rebuttable only where the state can establish a high risk of present dangerousness by clear and convincing evidence, in spite of the applicant’s health problems (and advanced age, if applicable). The questions of how to best evaluate present dangerousness (and risk, more generally) are deeply complicated, and I will not attempt to answer them here. What is important is that the present dangerousness standard shifts the focus away from the crime of conviction, the typical focus in discretionary back-end release decisions.

Though uncommon, a presumption of release from prison if the applicant meets certain criteria is not without precedent. D.C.'s compassionate release statute provides that the court "shall" modify the sentence if the applicant meets the eligibility criteria.249 In the parole context, at least four states (Maryland, Mississippi, Oklahoma, and South Dakota) have recently adopted administrative parole provisions which create a presumption of release on parole if a person meets certain standardized criteria and is serving a sentence for a narrow category of convictions (all nonviolent offenses).250 Colorado has adopted a rebuttable presumption of release from prison for people convicted of serious crimes as children who have met a time served requirement and other criteria.251 A presumption of release should apply in compassionate release cases.

Involve Advocates

The involvement of an attorney or other advocate is also a means of counterbalancing the anti-release default. Practitioners and academics have long recognized the persuasive power of narrative-driven advocacy in combating punitiveness.252 Without an advocate’s involvement, however, compelling narratives supporting compassionate release are unlikely to emerge.

The apparent assumption of compassionate release laws and policies is that a serious or terminal medical condition speaks for itself. But this is rarely true. The facts of someone’s daily life in prison as they live with a serious or terminal medical condition—the shackles they wear during chemotherapy or their struggles to stand for count three times a day—are not found in a medical record or a doctor’s brief statement of diagnosis and prognosis. Even critical facts about someone’s medical diagnosis or prognosis may not be obvious to compassionate release decisionmakers from the medical documentation they review.253 Decisionmakers lack medical training and are ill-equipped to decipher medical records never intended to be read by laypeople; where doctors provide a statement of an applicant’s medical condition, their statements are frequently cursory and opaque.254 An advocate can translate medical facts into a cognizable story and expand the universe of facts considered by the decisionmaker beyond the narrow confines of the medical condition and prison record.255

Creating a right to state-provided counsel for compassionate release applicants is appealing, but without additional resources could exacerbate the strain on indigent defense systems already operating under severe resource constraints.256 Pro bono attorneys, law school clinics, and volunteer organizations should step into the gap and represent people seeking compassionate release in the states.257

The involvement of an advocate is one way to counterbalance the strong incentives to deny release. The potential harms of granting release are clear for decisionmakers: the nightmare scenario of a horrific crime and the resulting headline holding them responsible.258 An advocate’s involvement can help make the harms of continued incarceration for a compassionate release applicant equally vivid.


Compassionate release is a case study in the possibilities and limits of early release mechanisms as tools for decarceration in the states. There is widespread agreement that meaningful decarceration will not be possible if it excludes people accused or convicted of violent crimes. Many commentators have called for an end to the outright exclusion of people convicted of violent crimes from reform measures meant to reduce the prison population and lower the incarceration rate. Their critique is right: the exclusion of people serving sentences for violent convictions from decarceral reform measures is troubling. But examining compassionate release in the states reveals that simply ending outright exclusions is insufficient to ensure that the benefits of decarceral measures will reach people serving sentences for violent convictions.

If compassionate release and other early release measures are to realize their potential as tools for decarceration, they must reduce the obstacles to release for people incarcerated for violent convictions. Early release measures should be designed for the hardest cases. The approaches I propose to cabin the influence of extreme risk-aversion and retributivism and counterbalance the anti-release default will allow compassionate release and other early release measure to benefit these hardest cases.   




  1. [1]. See Chris Feliciano Arnold, The Dying American Prisoner, Atlantic (Dec. 23, 2019), https://
    603988 [].

  2. [2]. Id.

  3. [3]. This is not unusual. Michele DiTomas, Joseph Bick & Brie Williams, Shackled at the End of Life: We Can Do Better, Am. J. Bioethics, June 2019, at 61, 61–62 (2019) (“Most custodial jurisdictions in the United States require shackling of the hands and feet when inmates are transported outside of the prison setting. Generally, prisoners who are hospitalized in the outside community are chained to their hospital beds, even in situations with minimal or no risk to public safety.”); see Pew Charitable Trs. & the John D. and Catherine T. MacArthur Found., State Prison Healthcare Spending: An Examination 9 (2014),
    2014/07/stateprisonhealthcarespendingreport.pdf [].

  4. [4]. Suleika Jaouad, The Prisoners Who Care for the Dying and Get Another Chance at Life, N.Y. Times Mag.: The Health Issue (May 16, 2018),
    16/magazine/health-issue-convicted-prisoners-becoming-caregivers.html [

  5. [5]. Id.

  6. [6]. Rachel E. López, The Unusual Cruelty of Nursing Homes Behind Bars, 32 Fed. Sent’g. Rep. 264, 264 (2020); see also Nat’l Rsch. Council, The Growth of Incarceration in the United States: Exploring Causes and Consequences 212 (Jeremy Travis, Bruce Western & Steve Redburn eds., 2014) (“[P]risons increasingly are becoming a critical delivery site for nursing home-level care ....”).

  7. [7]. Pew Charitable Trs., supra note 3, at 9.

  8. [8]. See E. Ann Carson, U.S. Dep’t Just., Prisoners in 2019, at 15 tbl.9 (2020), https:// [].

  9. [9]. Id.

  10. [10]. See Pew Charitable Trs., supra note 3, at 9; Pew Charitable Trs., State Prisons and the Delivery of Hospital Care 2 (2018),
    prisons-and-hospital-care_report.pdf [] [hereinafter Delivery of Hospital Care].

  11. [11]. Maura Ewing, When Prisons Need to Be More Like Nursing Homes, Marshall Project (Aug. 27, 2015, 7:15 AM), [].

  12. [12]. Delivery of Hospital Care, supra note 10, at 2.

  13. [13]. See, e.g., John F. Pfaff, Locked In: The True Causes of Mass Incarceration—and How to Achieve Real Reform 186 (2017) (describing “the rhetoric of transformative change coming from both the Left and the Right” and noting that decarceration “has been taken seriously since about 2008”); Todd R. Clear, Decarceration Problems and Prospects, 4 Ann. Rev. Criminology 239, 256 (2021) (“There is a lot of talk about undoing mass incarceration, and an uneasy political consensus has developed that this should happen.”). This ostensible consensus masks deep disagreement about mass incarceration’s causes and consequences. See Benjamin Levin, The Consensus Myth in Criminal Justice Reform, 117 Mich. L. Rev. 259, 276–83 (2018).

  14. [14]. Nazgol Ghandnoosh, Can We Wait 60 Years to Cut the Prison Population in Half?, Sent’g Project (Jan. 22, 2021), [].

  15. [15]. See, e.g., Rebecca Silber, Alison Shames & Kelsey Reid, Vera Inst. Just., Aging Out: Using Compassionate Release to Address the Growth of Aging and Infirm Prison Population 2 (2017),
    -to-Address-the-Growth-of-Aging-and-Infirm-Prison-Populations—Full-Report.pdf [
    ZFB4-EL6G] (“A ripe opportunity for strengthening the justice system examined by a number of states is the prospect of reducing the population of elderly or severely infirm people in prison.”).

  16. [16]. See infra Appendix. Compassionate release is the only early release mechanism currently available in nearly every state. See infra Section II.C.

  17. [17]. Gina Barton, Release Programs for Sick and Elderly Prisoners Could Save Millions. But States Rarely Use Them., Milwaukee J. Sentinel (Apr. 18, 2018),
    2018/4/18/release-programs-for-sick-elderly-prisoners-could-save-millions.html [
    H8HM-ZK6N] (finding an average of less than four compassionate release grants per state in 2015); Margaret Holland, Stephanie Grace Prost, Heath Hoffmann, & George Dickinson, Access and Utilization of Compassionate Release in State Departments of Corrections, 26 Mortality 49, 56 (2021) (finding an average of seven compassionate release grants per state from 2013 to 2015).

  18. [18]. See, e.g., Mark Osler, Criminal Justice Amid the Pandemic of 2020 4 (U. St. Thomas Sch. L., Legal Stud. Rsch. Paper No. 20-23, 2020), (describing the low numbers of incarcerated people granted compassionate release in Minnesota during the COVID-19 pandemic); Andrea Armstrong, Hardwired Against Change: Race, Incarceration, and COVID-19, Just Security (Aug. 26, 2020), [] (describing “[t]he lack of significant releases” from prisons during the COVID-19 pandemic); Emily Widra, With Over 2,700 Deaths Behind Bars and Slow Vaccine Acceptance, Prisons and Jails Must Continue to Decarcerate, Prison Pol’y Initiative (June 23, 2021),
    06/23/june2021_population [] (“Many states’ prison populations are the lowest they’ve been in decades,butthis is not because more people are being released from prisons. The limited data available from six states shows that the number of prison releases did not change significantly between 2019 and 2020, suggesting thatmost of the population drops that we’ve seen over the past year are due to reduced prison admissions.”) (emphasis omitted).

  19. [19]. See Mary Price, Fams. Against Mandatory Minimums, Everywhere and Nowhere: Compassionate Release in the States (2018), [] (reviewing state compassionate release policies).

  20. [20]. Most legal scholarship on compassionate release has focused on the federal system. See, e.g., William W. Berry III, Extraordinary and Compelling: A Re-Examination of the Justifications for Compassionate Release, 68 Md. L. Rev. 850 (2009); Jalila Jefferson-Bullock, Quelling the Silver Tsunami: Compassionate Release of Elderly Offenders, 79 Ohio St. L.J. 937 (2018) [hereinafter Quelling the Silver Tsunami]; Shon Hopwood, Second Looks & Second Chances, 41 Cardozo L. Rev. 83 (2019); Stephen R. Sady & Lynn Deffebach, Second Look Resentencing Under 18 U.S.C. § 3582(c) as an Example of Bureau of Prisons Policies That Result in Overincarceration, 21 Fed. Sent’g. Rep. 167 (2009); Mary Price, A Case for Compassion, 21 Fed. Sent’g. Rep. 170 (2009); Jalila Jefferson-Bullock, Consensus, Compassion, and Compromise? The First Step Act and Aging Out of Crime, 32 Fed. Sent’g Rep. 70 (2019) [hereinafter Consensus, Compassion, and Compromise?]. Legal scholarship on compassionate release in individual states or the states as a whole includes: Marjorie P. Russell,Too Little, Too Late, Too Slow: Compassionate Release of Terminally Ill Prisoners—Is the Cure Worse than the Disease?, 3WidenerJ.Pub.L. 799 (1994); John A. Beck,Compassionate Release from New York State Prisons: Why Are So Few Getting Out?, 27J.L.Med.&Ethics 216 (1999); Nicole M. Murphy,Dying to Be Free: An Analysis of Wisconsin’s Restructured Compassionate Release Statute, 95Marq.L.Rev. 1679 (2012); Gregory J. O’Meara,
    Compassion and the Public Interest: Wisconsin’s New Compassionate Release Legislation, 23Fed. Sent’gRep. 33 (2010); E. Lea Johnston, Smoke and Mirrors: Model Penal Code § 305.7 and Compassionate Release, 4 Wake Forest J. L. & Pol’y 49, 60–64 (2014). The most comprehensive overview of compassionate release in the states is a 2020 report from Families Against Mandatory Minimums. Price, supra note 19.

  21. [21]. Andrew Manuel Crespo, The Hidden Law of Plea Bargaining, 118 Colum. L. Rev. 1303, 1305 (2018).

  22. [22]. Pfaff, supra note 13, at 189 (“Even though the federal government holds only about 12 percent of the nation’s prisoners, its criminal justice system receives almost all of the national media and scholarly attention. Problematically, federal criminal justice outcomes look much different from those in the states. While people with drug convictions make up about sixteen percent of state prisoners, they make up approximately 49 percent of federal prisoners. The federal system is also distinctly more punitive in general, and especially so when it comes to drugs. This focus on federal policy leads people to overestimate how many people are in prison for drugs and how long they spend in prison.”).

  23. [23]. Geriatric release, which provides for early release based on advanced age and time served, and second look sentencing, which would create a mechanism for sentence reduction and release for people who have served significant time in prison (such as ten, fifteen, or twenty years). See infra Section II.A–B.

  24. [24]. See infra Section III.A.

  25. [25]. This Article uses the Bureau of Justice Statistics’ definition of violent convictions. U.S. Dep’t Just., Violent Offenses, Bureau of Just. Stats.,
    #glossary-terms-block-1-irrqipyxfvlnp-ak []. The BJS definition includes five categories of crimes: murder, rape, robbery, assault, and “other violent offenses” (including kidnapping, vehicular manslaughter, and child abuse). Id. The BJS definition excludes weapons offenses, driving while intoxicated, and unarmed burglary, which are sometimes categorized as violent under state or federal law. See Michael O’Hear, Managing the Risk of Violent Recidivism: Lessons from Legal Responses to Sexual Offenses, 100 B.U. L. Rev. 133, 138–39 n.23 (2020) (noting that “the BJS uses a relatively narrow definition of what counts as ‘violent[]’” and excludes burglary, driving under the influence, and weapons offenses from its definition of violent crimes); accord J.J. Prescott, Benjamin Pyle & Sonja B. Starr, Understanding Violent-Crime Recidivism, 95 Notre Dame L. Rev. 1643, 1649 (2020) (describing the BJS definition as “relatively restrictive”).

  26. [26]. Carson, supra note 8, at 20.

  27. [27]. In federal prisons, nearly half of all incarcerated people are serving a sentence for a drug conviction, while eight percent of people are serving sentences for violent crimes. In state prisons, however, people serving sentences for drug convictions account for 14 percent of people in state prisons; people serving sentences for violent convictions account for over half. Carson, supra note 8, at 20. In state prisons, people serving sentences for homicide convictions account for 14 percent of the total population - nearly the same share of as people serving sentences for drug convictions. Id. at 1; see also Wendy Sawyer & Peter Wagner, Mass Incarceration: The Whole Pie 2019, Prison Pol’y Initiative (Mar. 24, 2020),
    2020.html [] (providing statistics on mass incarceration and convictions in 2020).

  28. [28]. See Pfaff, supra note 13, at 189; James Forman, Jr., Racial Critiques of Mass Incarceration: Beyond the New Jim Crow, 87 N.Y.U. L. Rev. 21, 47 (2012) (“[C]ontrary to the impression left by many of mass incarceration’s critics, the majority of America’s prisoners are not locked up for drug offenses.... Federal prisons are the only type of facility in which drug offenders constitute a majority (52%) of prisoners, but federal prisons hold many fewer people overall.”).

  29. [29]. See infra Part III.

  30. [30]. Marie Gottschalk, Caught: The Prison State and the Lockdown of American Politics 165 (2015); accord Rachel Elise Barkow, Prisoners of Politics: Breaking the Cycle of Mass Incarceration 12 (2019) (reviewing state reforms and concluding that reforms thus far have been “only modest changes, focused predominantly on the harshest punishments for nonviolent drug and property offenders who do not have much in the way of a criminal record.”).

  31. [31]. See, e.g., Pfaff, supra note 13, at 230–31 (arguing against the exclusion of people incarcerated for violent convictions from back-end release reforms); James Forman, Jr., Locking Up Our Own: Crime and Punishment in Black America 228–31 (2017) (calling for the inclusion of people accused or convicted of violent crimes in reform measures).

  32. [32]. See infra note 129.

  33. [33]. See E. Ann Carson & William J. Sabol, U.S. Dep’t of Just., Aging of the State Prison Population, 1993-2013, at 9 (2016), [https://].

  34. [34]. See infra Section III.A.

  35. [35]. Carson & Sabol, supra note 33 at 1 (“Between 1993 and 2013, more than 65% of prisoners age 55 or older were serving time in state prison for violent offenses ....”).

  36. [36]. See infra Section II.C.

  37. [37]. See infra Section III.B.

  38. [38]. Pfaff, supra note 13, at 185; accord David Alan Sklansky, A Pattern of Violence: How the Law Classifies Crimes and What It Means for Justice 42 (2021) (“Even advocates of reducing prison sentences, and improving the treatment of prisoners after their release, often treat violence as something of a third rail.”).

  39. [39]. See Amna A. Akbar, An Abolitionist Horizon for (Police) Reform, 108 Calif. L. Rev. 1781, 1783 (2020) (“The nationwide protests [in the summer of 2020, after George Floyd’s murder] catapulted prison and police abolition into the mainstream.”).

  40. [40]. See, e.g., Forman, supra note 31, at 230–31 (2017) (“[T]he label ‘violent offender,’ tossed out to describe a shadowy group for whom we are supposed to have no sympathy, encourages us to overlook their individual stories. It causes us to separate those other people—the ones who did something violent, the ones who belong in cages—from the rest of us. It leads us, as Bryan Stevenson has written, to define people by the worst thing they have ever done. And it ensures that we will never get close to resolving the human rights crisis that is 2.2 million Americans behind bars.”); Pfaff, supra note 13, at 190–91 (“For almost all people who commit violent crimes, however, violence is not a defining trait but a transitory state that they age out of.”).

  41. [41]. Bruce Western, Homeward: Life in the Year After Prison 63–82 (2018) (describing the histories of violence exposure among a cohort of adults recently released from Massachusetts’ state prisons); James Austin, Vincent Schiraldi, Bruce Western & Anamika Dwivedi, Reconsidering the “Violent Offender” 9–14 (2019),
    loads/2019/09/executive-session-pdf-Reconsidering-the-violent-offender-report-ONLINE_FINAL.pdf [] (“[P]rovid[ing] empirical evidence on high rates of victimization among incarcerated youth[] [and] adults in Arkansas state prison[.]”). The sources here focus on violent crime that is prosecuted and punished; much interpersonal violence is never reported or prosecuted. Cecelia Klingele,Labeling Violence, 103Marq.L.Rev. 847, 857–58 (2020).

  42. [42]. Vaidya Gullapalli, Recognizing the ‘Dual Victim’, Appeal (Aug. 30, 2019), https://the [] (describing organizing by people who identify as “dual victims” of mass incarceration and violent crime); John L. Micek, Commentary, ‘No One Can Ever Tell Me We Don’t Exist’: Dual Victims Caught in the Cracks of Pa.’s Criminal Justice System Deserve Recognition, Pa. Cap.-Star (Oct. 23, 2019, 5:53 PM), https://www.penn [].

  43. [43]. See, e.g., I. India Thusi, Book Review, Feminist Scripts for Punishment, 134 Harv. L. Rev. 2449, 2479–82 (2021) (describing the work of radical feminists of color to develop noncarceral responses to gender-based violence); Danielle Sered, Until We Reckon: Violence, Mass Incarceration, and a Road to Repair 14–15 (2019) (calling for a new, more complex understanding of violent acts “that include the context in which violence takes place not as an excuse, but as a piece of the puzzle of ending it ... [and] hold[s] everyone accountable for harm, both the individuals who commit it and the societies that allow it”); Our Mission, UnCommon Law (2020), [] (describing their mission of “ensur[ing] that all people incarcerated for violent crime have access to healing, justice, and effective legal representation”). See generally Mariame Kaba & Rachel Herzing, Transforming Punishment: What is Accountability Without Punishment?, in Mariame Kaba, We Do This ‘Til We Free Us: Abolitionist Organizing and Transforming Justice 148­–56 (Tamara K. Nopper ed., 2021) (criticizing the criminal legal system’s approach to addressing interpersonal harm and violence and discussing abolitionist perspectives on accountability).

  44. [44]. People who are currently serving a prison sentence for a nonviolent conviction, but have previously been convicted of violent crimes, are not included in this group. Nonetheless, they likely experience many of the same dynamics I describe for people incarcerated for violent convictions. This language is an attempt to avoid the reductive, stigmatizing, and dehumanizing labels frequently used to describe incarcerated people. See, e.g., Eddie Ellis, An Open Letter to Our Friends on the Question of Language, Ctr. for NuLeadership on Urb. Sols., [] (“When we are not called mad dogs, animals, predators, offenders and other derogatory terms, we are referred to as inmates, convicts, prisoners and felons—all terms devoid of humanness which identify us as ‘things’ rather than as people.... [W]e are asking everyone to stop using these negative terms and to simply refer to us as PEOPLE. People currently or formerly incarcerated, PEOPLE on parole, PEOPLE recently released from prison, PEOPLE in prison, PEOPLE with criminal convictions, but PEOPLE.”). See generally Alexandra Cox, The Language of Incarceration, 1 Incarceration 1 (2020) (describing justifications for, and critiques of, person-first language in the criminal legal system).

  45. [45]. Accelerated aging refers to the fact that people in prison, on average, experience chronic illness and geriatric health conditions at younger ages than their non-incarcerated peers. The reasons for accelerated aging are a complicated combination of pre-prison risk factors (such as a history of substance abuse and poor health care) and the negative health impacts of incarceration itself. See generally Brie A. Williams, James S. Goodwin, Jacques Baillargeon, Cyrus Ahalt & Louise C. Walter, Addressing the Aging Crisis in U.S. Criminal Justice Health Care, 60 J. Am. Geriatrics Soc’y 1150 (2012) [hereinafter Addressing the Aging Crisis] (discussing the causes and consequences of accelerated aging in prison).

  46. [46]. Avlana K. Eisenberg, Incarceration Incentives in the Decarceration Era, 69 Vand. L. Rev. 71, 71 (2016) (“After forty years of skyrocketing incarceration rates, there are signs that a new ‘decarceration era’ may be dawning; the prison population has leveled off and even slightly declined.”).

  47. [47]. The total number of people in prison has fallen 11 percent since 2009, when it peaked. Carson, supra note 8, at 1. The imprisonment rate (the number of people in prison per 100,000 U.S. residents) has declined 17 percent since 2009. Id. In 2019, the imprisonment rate was the lowest since 1995; the total number of people in prison was the lowest since 2002. Id. at 1–2. California’s “realignment”—prompted by the Supreme Court’s holding that the state’s prisons were unconstitutionally overcrowded—has been an outsize contributor to national decarceration, accounting for nearly half of the total national decrease between 2010 and 2014. Pfaff, supra note 13, at 14; Brown v. Plata, 563 U.S. 493, 502 (2011). Twenty-five other states also reduced their prison populations over the same period (though the other half of the states saw the number of people in their state prisons rise over the same period). Pfaff, supra note 13, at 14.

  48. [48]. Pfaff, supra note 13, at 8 (“[A]lthough pretty much everyone agrees that we need to move away from today’s ‘mass’ incarceration to something less, what that number should be is unclear.”).

  49. [49]. Decarceration is one of the three phases of prison abolition, as conceived by the Prison Research Education Action Project in 1976. See John Washington, What Is Prison Abolition?, Nation (July 31, 2018), [https://perma.
    cc/6T3G-TBF3]. The three phases “are; moratorium, decarceration, and excarceration.” Id.; see also Allegra M. McLeod, Prison Abolition and Grounded Justice, 62 UCLA L. Rev. 1156, 1156 (2015) (arguing for a conception of abolition as “an aspirational ethic and a framework of gradual decarceration”).

  50. [50]. See Ghandnoosh, supra note 14. See generally Andrew D. Leipold, Is Mass Incarceration Inevitable?, 56 Am. Crim. L. Rev. 1579 (2019) (noting that efforts to reduce incarceration have made modest gains).

  51. [51]. Nat’l Rsch. Council, supra note 6, at 33–42; Inst. for Crime & Just. Pol’y Rsch., Highest to Lowest – Prison Population Rate, World Prison Brief, [].

  52. [52]. Despite a 29 percent decrease in the imprisonment rate for Black adults between 2009 and 2019, in 2019 the imprisonment rate for Black adults was still five times higher than for white adults. Carson, supra note 8, at 9–10 tbl.5. In some states the Black/white disparity in imprisonment rate is far greater. Ashley Nellis, The Color of Justice: Racial and Ethnic Disparity in State Prisons, Sent’g Project, (June 14, 2016),
    color-of-justice-racial-and-ethnic-disparity-in-state-prisons [].

  53. [53]. David Garland, Introduction: The Meaning of Mass Imprisonment to Mass Imprisonment: Social Causes and Consequences 1–2 (David Garland ed., 2001).

    What are the defining features of mass imprisonment? There are, I think, two that are essential. One is sheer numbers. Mass imprisonment implies a rate of imprisonment and a size of prison population that is markedly above the historical and comparative norm for societies of this type. The US prison system clearly meets these criteria. The other feature is the social concentration of imprisonment’s effects. Imprisonment becomes mass imprisonment when it ceases to be the incarceration of individual offenders and becomes the systematic imprisonment of whole groups of the population. In the case of the USA, the group concerned is, of course, young black males in large urban centres. For these sections of the population, imprisonment has become normalized. It has come to be a regular, predictable part of experience, rather than a rare and infrequent event.

    Id.; see also Levin, supra note 13, at 276 (describing Garland’s definition as “the clearest and most cited definition of the phenomenon” of mass incarceration).

  54. [54]. Forman, Jr., supra note 31, at 229 (“I have described mass incarceration as the result of a series of small decisions, made over time, by a disparate group of actors. If that is correct, mass incarceration will likely have to be undone in the same way.”).

  55. [55]. Id. at 12.

  56. [56]. See John G. Malcolm, Criminal Justice Reform at the Crossroads, 20 Tex. Rev. L. & Pol. 249, 259–67 (2016) (describing front-end reforms).

  57. [57]. See Jonathan J. Wroblewski & Sean M. Douglass, Two Targeted, Back-End Reforms for the Federal Criminal Justice System, 28 Fed. Sent’g Rep. 264, 264 (2016) (defining “back-end” reforms as those that provide incarcerated people with “a chance to reduce their time in prison”); Cecelia Klingele, Changing the Sentence Without Hiding the Truth: Judicial Sentence Modification as a Promising Method of Early Release, 52 Wm. & Mary L. Rev 465, 487 (2010) (identifying the three major “back-end” release mechanisms as “increased parole eligibility, reinstatement or expansion of sentence credit, and creation of infirmity-based release for geriatric and seriously ill prisoners”).

  58. [58]. See Kevin R. Reitz, Prison-Release Reform and American Decarceration, 104 Minn. L. Rev. 2741, 2742 (2020) (noting that prisons release mechanisms “ha[ve] received little attention in contemporary analyses of mass incarceration”).

  59. [59]. Back-end release measure affect one of the two variables that determines the number of people in prison. See Todd R. Clear & James Austin, Reducing Mass Incarceration: Implications of the Iron Law of Prison Populations, 3 Harv. L. & Pol’y Rev. 307, 312 (2009) (“[T]he size of a prison population is completely determined by two factors: how many people go to prison and how long they stay.”).

  60. [60]. See Cecelia Klingele, The Early Demise of Early Release, 114 W. Va. L. Rev. 415, 442 (2012)

  61. [61]. Pfaff, supra note 13, at 189 (“[The federal] criminal justice system receives almost all of the national media and scholarly attention.”).

  62. [62]. Carson, supra note 8, at 3.

  63. [63]. Id.

  64. [64]. Pfaff, supra note 13, at 13 (“[I]t’s likely that the two states that differ the most from each other when it comes to criminal justice policy have more in common with each other than either does with the federal system.”). Most notably, federal and state prisons incarcerate people convicted of very different crimes. In federal prisons, nearly half of all incarcerated people are serving a sentence for a drug conviction, while just eight percent of people are serving sentences for violent crimes. In state prisons, however, people serving sentences for drug offenses account for just 14 percent of people in state prisons; people serving sentences for violent convictions account for over half. Carson, supra note 8, at 1, 20.

  65. [65]. Sarah Lustbader, Potential Second Chances for Prisoners in D.C., Appeal (Sept. 26, 2019), https:// [].

  66. [66]. Barkow, supra note 30, at 75 (arguing that the “endowment effect”—people react worse to losing something they already have than never receiving something in the first place
    —explains why early release measures have been less durable than reforms aimed at reducing sentences prospectively).

  67. [67]. Leipold, supra note 50, at 1599 (“No jurisdiction has yet proposed that the prison population as a whole be reduced in a sweeping, across-the-board manner. For example, no state has proposed that all current sentences be shortened by 10%, or that all sentences authorized by statute be reduced by 5%, regardless of the crime of conviction.”).

  68. [68]. The leading proposal for second look sentencing is section 305.6 of the Model Penal Code: Sentencing, which authorizes second look review after 15 years of imprisonment for adults. Model Penal Code § 305.6 (Am. L. Inst., Proposed Final Draft 2017).

  69. [69]. Id.

  70. [70]. Id.

  71. [71]. Id.

  72. [72]. Meghan J. Ryan, Taking Another Look at Second-Look Sentencing, 81 Brook. L. Rev. 149, 155 (2015); accord Margaret Colgate Love & Cecelia Klingele, First Thoughts about “Second Look” and Other Sentence Reduction Provisions of the Model Penal Code: Sentencing Revision, 42 U. Tol. L. Rev. 859, 873–74 (2011) (“[T]here now appears to be consensus that courts must have some power to reexamine a lengthy sentence after a period of years, particularly if the public mood that produced a particularly harsh sentence has mellowed or the overall legal environment has changed.”).

  73. [73]. Ashley Nellis, Still Life: America’s Increasing Use of Life and Long-Term Sentences, Sent’g Project (May 3, 2017),
    -use-life-long-term-sentences [].

  74. [74]. See id.

  75. [75]. Margaret Colgate Love, Sentence Reduction Mechanisms in a Determinate Sentencing System: Report of the Second Look Roundtable, 21 Fed. Sent’g Rep. 211, 217 (2009) (quoting Professor Douglas Berman). Similarly, the drafters of the Model Penal Code: Sentencing have argued that the second look measure “reflects a profound sense of humility that ought to operate when punishments are imposed that will reach nearly a generation into the future[.]” Model Penal Code § 305.6, cmt. 568 (Am. L. Inst., Proposed Final Draft 2017).

  76. [76]. In 2018, California enacted what some commentators have described as a second look provision. Cal. Penal Code § 1170(d)–(e) (West 2020). I do not consider it a general second look sentencing provision for the following reasons. It does not include any time-served requirement and does not provide for any sort of automatic review. Nor does it empower the person serving the prison sentence to seek second look review. Rather, second look review by the court requires a recommendation by the Department of Corrections, parole board, or the district attorney in the county where the person was sentenced. The court’s power to reduce the sentence of someone serving a prison sentence is conditioned upon a recommendation by the Secretary of the Department of Corrections, the Board of Parole, or the district attorney. See Rory Fleming, Prosecutor-Driven “Second Look” Policies Are Encouraging, but Not a Panacea, 32 Fed. Sent’g Rep. 205, 205 (2020) (discussing the California provision).

  77. [77]. See, e.g. 68 D.C. Reg. 1034 (Jan. 22, 2021). See generally Nazgol Ghandnoosh, A Second Look at Injustice, Snt’g Project (May 12, 2021),
    lications/a-second-look-at-injustice [] (providing an excellent overview of second look legislation).

  78. [78]. Families Against Mandatory Minimums updates weekly their list of pending second look legislation in the states. Second Chances Agenda, FAMM, [https://].

  79. [79]. Leading advocacy organizations like the National Association of Criminal Defense Lawyers (“NACDL”) have made second look legislation a top policy priority. See generally Jane Anne Murray, Sean Hecker, Michael Skocpol, & Marissa Elkins, Nat’l Assoc. of Crim. Def. Laws., Second Look = Second Chance: Turning the Tide through NACDL’s Model Second Look Legislation (2020),
    aa83589/second-look-second-chance-the-nacdl-model-second-look-legislation.pdf [https://perma.
    cc/ZRN5-86ZC] (advocating for adoption of second look legislation).

  80. [80]. See, e.g., Frank O. Bowman, III, Freeing Morgan Freeman: Expanding Back-End Release Authority in American Prisons, 4 Wake Forest J. L. & Pol’y 9, 38 (2014) (arguing that “‘old [people] who have been [incarcerated] a long time’” should be considered for “back-end discretionary release”); Quelling the Silver Tsunami, supra note 20, at 971 (“[I]ncarceration of the elderly fails to fulfill any theory of criminal punishment.”); Steven L. Chanenson, Guidance from Above and Beyond, 58 Stan. L. Rev. 175, 191 (arguing for sentence review of “long-serving, older inmates”).

  81. [81]. Alaska Stat. Ann. § 33.16.090(a)(2) (West 2021) (60+, served at least ten years); Cal. Penal Code § 3055(a) (West 2021) (50+, served at least 20 years); D.C. Code Ann. § 24-403.04(a)(2) (West 2021) (60+, served at least 20 years); Ga. Const. art. IV, § II, para. II(e) (62+, no time served requirement); La. Stat. Ann. § 15:574.4(2) (2021) (45+, served at least 20 years or 60+, served at least ten years); Md. Code Ann., Crim. Law § 14-401(g) (West 2021) (60+, served at least 15 years); Okla. Stat. Ann. tit. 57, § 332.21 (West 2021) (60+, served at least ten years or 1/3 of sentence); Miss. Code Ann. §§ 47-7-3, 47-7-4 (West 2021) (60+, served at least ten years); S.D. Codified Laws § 24-15A-55(4)–(5) (2021) (70+, served at least 30 years or 65+, served at least 10 years); Tex. Gov’t Code Ann. § 508.146 (65+, no time served requirements); Va. Code Ann. § 53.1-40.01 (West 2021) (60+, served at least ten years or 65+, served at least five years); Wis. Stat. Ann. § 302.113(9g)(b)(1)–(2) (2021) (West 2021) (60+, served at least ten years or 65+, served at least five years). This list of geriatric release laws does not include those that impose additional medical eligibility requirements. See, e.g., D.C. Code Ann. § 24-403.04(a)(3)(B) (West 2021) (60+, served the lesser of 15 years or 75 percent of sentence, and meets medical eligibility criteria).

  82. [82]. In ten states, the parole board decides whether to grant geriatric release. Alaska Stat. § 33.16.090(a)(2); Cal. Penal Code § 3055(a); Ga. Code Ann. § 42-9-42(c); La. Stat. Ann. § 15:574.4.(A)(2); Md. Code Ann., Crim. Law § 14-101(f); Miss. Code Ann. §§ 47-7-3, 47-7-4; Okla. Stat. Ann. tit. 57, § 332.21; S.D. Codified Laws § 24-15A-55(4)–(5); Tex. Gov’t Code Ann. § 508.146; Va. Code Ann. § 53.1-40.01. In Wisconsin and D.C., the sentencing court makes the release decision. D.C. Code Ann. § 24-403.04(a)(2); Wis. Stat. Ann. § 302.113(9g)(b)(1)
    –(2). In Wisconsin the sentencing court may only grant release if the Department of Corrections has recommended doing so. Wis. Stat. Ann. § 302.113(9g)(b)(1)–(2).

  83. [83]. See e.g., Wis. Stat. Ann. § 302.113(9g)(b)(1)–(2) (excluding people incarcerated for a Class B felony).

  84. [84]. See supra note 81 and accompanying text.

  85. [85]. JRI provides guidance and technical assistance to states seeking to enact bipartisan criminal justice reform legislation that will reduce spending on corrections and criminal justice. Pew Charitable Tr., 35 States Reform Criminal Justice Policies Through Justice Reinvestment 1 (July 2018),‑/media/assets/2018/07/pspp_reform_
    matrix.pdf [].

  86. [86]. Silber et al, supra note 15, at 2.

  87. [87]. Inflated claims about the cost-savings of decarceration measures frequently confuse average and marginal costs. Pfaff, supra note 13, at 99 (discussing the conflation of average and marginal costs in criminal justice reform rhetoric). See generally Chris Mai & Ram Subramanian, Vera Inst. Just., The Price of Prisons: Examining State Spending Trends, 2010-2015 (2017), [] (same).

  88. [88]. Nev. Advisory Comm’n on Admin. Just., Justice Reinvestment Initiative 29 (2019), [
    28BW-Y9A7] (“Researchers have consistently found that age is one of the most significant predictors of criminality, with criminal activity decreasing as a person ages. Studies on parolee recidivism found that the probability of a parole violation also decreases with age, with older parolees the least likely to be re-incarcerated.”) (footnote omitted). In Virginia, the state commission that proposed the geriatric compassionate release provision enacted in 1994 argued, “[S]ome of the worst criminals with the longest sentences may remain incarcerated past the point at which, by virtue of their age and physical condition, they have ceased to pose a threat to the community.” Governor’s Comm’n on Parole Abolition and Sent’g Reform, Final Report 47 (1994), []; see also Patrick Lussier & Jay Healey, Rediscovering Quetelet, Again: The “Aging” Offender and the Prediction of Reoffending in a Sample of Adult Sex Offenders, 26 Just. Q. 827, 828 (2009) (noting that the age-crime relationship is “one of the most robust and stable empirical findings of criminological research.”); Bianca E. Bersani & Elaine Eggleston Doherty, Desistance from Offending in the Twenty-First Century, 1 Ann. Rev. Criminology 311, 313 (2018) (“[E]ventual desistance from crime is the norm, even among those characterized as high-rate, chronic offenders.”).

  89. [89]. Prescott et al., supra note 25, at 1671–81 (reviewing the literature).

  90. [90]. States use different language to describe their compassionate release provisions, such as “medical parole” or “medical release.” Some commentators define “compassionate release” more broadly. 18 U.S.C. § 3582(c)(1)(A) (2018), the federal provision often referred to as “compassionate release,” authorizes release based on any “extraordinary and compelling reasons.” See Hopwood, supra note 20, at 103; Consensus, Compassion, and Compromise?, supra note 20, at 71.

  91. [91]. See infra Appendix; see also Price, supra note 19, at 12.

  92. [92]. Beck, supra note 20, at 225. Some states also recognize a third category of medical eligibility: cognitive incapacitation because of dementia, Alzheimer’s, and related conditions. See, e.g., Alaska Stat. Ann. § 33.16.085(a)(3)(B) (West 2021) (“[S]evere medical or cognitive disablity ... and the medical or cognitive disability has progressed so that the likelihood of the prisoner’s [sic] committing the same or a similar offense is low ....” (emphasis added)); N.Y. Exec. Law § 259-s(1)(a) (McKinney 2021) (“[A] significant and permanent non-terminal condition, disease or syndrome that has rendered the incarcerated individual so physically or cognitively debilitated or incapacitated as to create a reasonable probability that he or she does not present any danger to society ....” (emphasis added)).

  93. [93]. Prognosis is an inexact science. Brie A. Williams, Rebecca L. Sudore, Robert Greifinger & R. Sean Morrison, Balancing Punishment and Compassion for Seriously Ill Prisoners, 155 Annals Internal Med. 122, 124 (2011) [hereinafter Balancing Punishment and Compassion].

  94. [94]. The standard definition of activities of daily living is “bathing, eating, toileting, dressing, and transferring (getting in and out of bed).” Brie A. Williams, Karla Lindquist, Rebecca L. Sudore, Heidi M. Strupp, Donna J. Willmott & Louise C. Walter, Being Old and Doing Time: Functional Impairment and Adverse Experiences of Geriatric Female Prisoners, 54 J. Am. Geriatrics Soc’y 702, 703 (2006) [hereinafter Being Old and Doing Time].

  95. [95]. See, e.g., California’s Medical Parole statute, which defines medical eligibility to include circumstances where the person “is permanently medically incapacitated with a medical condition that renders them permanently unable to perform activities of basic daily living, and results in the prisoner requiring 24-hour total care, including, but not limited to, coma, persistent vegetative state, brain death, ventilator-dependency, loss of control of muscular or neurological function.” Cal. Penal Code § 1170(e)(2)(C) (West 2020). Similarly, a person who is required to register as a sex offender in Texas is only eligible for compassionate release if they are “in a persistent vegetative state or” have “an organic brain syndrome with significant to total mobility impairment.” Tex. Gov’t Code Ann. § 508.146(a)(1)(B) (West 2017).

  96. [96]. See infra Section IV.A.

  97. [97]. See infra Section IV.D.

  98. [98]. See generally Fiona Doherty, Obey All Laws and Be Good: Probation and the Meaning of Recidivism, 104 Geo. L.J. 291 (2016) (describing the onerous standard conditions of probation in 15 states and probation officers’ expansive powers to enforce them); Kate Weisburd, Sentenced to Surveillance: Fourth Amendment Limits on Electronic Monitoring, 98 N.C. L. Rev. 717 (2020) (describing the electronic surveillance of people on community supervision).

  99. [99]. Some governors’ official pardon criteria specifically listed life-threatening illness as grounds for pardon. Carolyn Strange, Discretionary Justice: Pardon and Parole in New York from the Revolution to the Depression 102 (2016) (“The capacity of governors to feel pity was the last hope of the condemned, and most officeholders were known to favor particular categories of offenders ... [including] prisoners suffering from serious illnesses ....”). Similarly, in the federal system between 1870 (when the Department of Justice was formed) and 1900, the rules governing Presidential clemency made specific provisions for applicants with serious health problems. W. H. Humbert, The Pardoning Power of the President 126–28 (1941) (describing Presidential pardons based on “ill health” and “old age” between 1870 and 1930); see also Margaret Colgate Love, The Twilight of the Pardon Power, 100 J. Crim. L. & Criminology 1169, 1183 (2010) (same).

  100. [100]. See Don M. Gottfredson, Leslie T. Wilkins, & Peter B. Hoffman, Guidelines for Parole and Sentencing: A Policy Control Method 35 (1978) (describing “serious medical problems” as a factor in parole decision-making).

  101. [101]. Kevin R. Reitz & Edward E. Rhine, Parole Release and Supervision: Critical Drivers of American Prison Policy, 3 Ann. Rev. Criminology 281, 283 (2020) (“From 1976 through 2000, sixteen states and the federal system abolished parole-release discretion for most or all cases ....”); Joan Petersilia, When Prisoners Come Home: Parole and Prisoner Reentry 66–67 (2003) (listing the 16 states that eliminated parole during this period).

  102. [102]. See Russell, supra note 20 at 818–27.

  103. [103]. Joe Coleman Medical Release Act, Pub. Act 102-0494 (2021) (creating a new statutory provision at 730 Ill. Comp. Stat. Ann. 5/3-3-14).

  104. [104]. See Balancing Punishment and Compassion, supra note 93, at 122.

  105. [105]. See infra note 129.

  106. [106]. See supra notes 88–89.

  107. [107]. For a description of the experience of incarceration for people with a serious or terminal medical condition, see Brie A. Williams et al., Caregiving Behind Bars: Correctional Officer Reports of Disability in Geriatric Prisoners, 57 J. Am. Geriatrics Soc’y 1286, 1290–91 (2009) [hereinafter Caregiving Behind Bars]; Being Old and Doing Time, supra note 94; DiTomas et al., supra note 3.

  108. [108]. See supra note 87.

  109. [109]. See Silber et al., supra note 15, at 3. See generally Cyrus Ahalt, Robert L. Trestman, Josiah D. Rich, Robert B. Greifinger & Brie A. Williams, Paying the Price: The Pressing Need for Quality, Cost, and Outcomes Data to Improve Correctional Health Care for Older Prisoners, 61 J. Am. Geriatric Soc’y 2013 (2013) (describing the lack of reliable data about prison healthcare expenditures or the drivers of prison healthcare spending, both of which vary dramatically from state-to-state).

  110. [110]. It is cheaper to provide the same drug to someone who is not incarcerated and is covered by Medicaid, than to someone in prison. See generally Pew Charitable Trs., Pharmaceuticals in State Prisons (2017), [] (describing cost and varying methods of purchasing medications for state prison systems).

  111. [111]. See Off. Inspector Gen., U.S. Dep’t Just., The Impact of an Aging Inmate Population on the Federal Bureau of Prisons, 14–15 (2016),
    2015/e1505.pdf [] (“Institution staff also told us that aging inmates with chronic conditions require treatment from specialists outside the institution and that overtime paid to Correctional Officers who escort inmates to such appointments is a significant budget item.”).

  112. [112]. See Pfaff, supra note 13, at 185–96; Gottschalk, supra note 30, at 165–95.

  113. [113]. Pfaff, supra note 13, at 185 (“The emphasis current reform efforts place on reducing punishments for people convicted of low-level, nonviolent crimes is understandable, but it should be clear by now that the impact will be limited. Any significant reduction in the US prison population is going to require states and counties to rethink how they punish people convicted of violent crimes, where ‘rethink’ means ‘think about how to punish less.’”).

  114. [114]. Barkow, supra note 30, at 12–13.

  115. [115]. Id. at 12.

  116. [116]. See Forman, supra note 28, at 47–48.

  117. [117]. See, e.g., Sklansky, supra note 38, at 43 (“[A]lthough the definition of ‘violence’ is often taken to be self-evident, the category can be surprisingly difficult to delineate.”); Alice Ristroph,Criminal Law in the Shadow of Violence, 62Ala. L.Rev. 571, 581 (2011) (“It is tempting to think that the right account of violence will lie beyond moral disagreement.... But [in attempting to define violence], we soon find ourselves back in the contested territory of the normative.”).

  118. [118]. Ristroph, supra note 117, at 574 (“For reporting purposes, violent crime includes only a few enumerated offenses that involve injury to human bodies: murder, manslaughter, rape, assault, and robbery. These offenses are a relatively small portion of criminal activity. For sentencing purposes, violent crime is defined much more broadly. A number of sentencing laws impose enhanced penalties on offenders with prior convictions for ‘violent crime.’ These laws typically define violent crime in terms of risk of physical injury—or even more remotely, ‘potential risk.’ These broad definitions have led courts to consider new candidates—such as burglary of an unoccupied home, drunk driving, or obstruction of justice—to bear the mantle of violent crime.”) (footnote omitted).

  119. [119]. Sklansky, supra note 38, at 12.

  120. [120]. The full BJS definition of violent offenses is:

    Murder—Includes homicide, nonnegligent manslaughter, and voluntary homicide. It excludes attempted murder (classified as felony assault), negligent homicide, involuntary homicide, or vehicular manslaughter, which are classified as other violent offenses. Rape—Includes forcible intercourse, sodomy, or penetration with a foreign object. It does not include statutory rape or nonforcible acts with a minor or someone unable to give legal consent, nonviolent sexual offenses, or commercialized sex offenses. Robbery—Includes unlawful taking of anything of value by force or threat of force. It includes armed, unarmed, and aggravated robbery, carjacking, armed burglary, and armed mugging. Assault—Includes aggravated assault, aggravated battery, attempted murder, assault with a deadly weapon, felony assault or battery on a law enforcement officer, and other felony assaults. It does not include extortion, coercion, or intimidation. Other violent offenses—Includes vehicular manslaughter, involuntary manslaughter, negligent or reckless homicide, nonviolent or nonforcible sexual assault, kidnapping, unlawful imprisonment, child or spouse abuse, cruelty to a child, reckless endangerment, hit-and-run with bodily injury, intimidation, and extortion.

    U.S. Dep’t of Just., supra note 25.

  121. [121]. O’Hear, supra note 25, at 163. O’Hear argues that this variability reflects the expansive nature of American substantive criminal law, which imposes broad accomplice liability (felony murder is the most notorious example) and makes little allowance for diminished capacity or situationally dependent conduct. Id.

  122. [122]. Consider, for example, someone convicted of homicide under the felony murder rule for playing a nonviolent role in the underlying felony. See id.; Abbie VanSickle, If He Didn’t Kill Anyone, Why Is It Murder?, N.Y. Times (June 27, 2018),
    27/us/california-felony-murder.html [].

  123. [123]. See, e.g., Irene Oritseweyinmi Joe, Systematizing Public Defender Rationing, 93 Denv. L. Rev. 389, 391 (2016) (“[P]ublic defenders are constantly tasked with representing more individuals than their limited resources support.... Insufficient resourcing ... has created a public defender system that is commonly described as unfair, struggling, and even broken.”); Andrew Manuel Crespo,Regaining Perspective: Constitutional Criminal Adjudication in the U.S. Supreme Court, 100Minn. L. Rev. 1985, 2018–20 (2016) (describing the underfunding of indigent defense services nationally).

  124. [124]. Anna Roberts, Convictions as Guilt, 88 Fordham L. Rev. 2501, 2507 (2020).

  125. [125]. See Austin et al., supra note 41, at 9–14.

  126. [126]. Sklansky, supra note 38, at 41. Sklansky also points out that the sharp line between violent and nonviolent crimes did not emerge until the 1970s. Id. at 45.

  127. [127]. Carson, supra note 8, at 1.

  128. [128]. I focus on the prevalence of violent convictions among people who meet the substantive eligibility criteria for these early release measures, based on their time served in prison, age, or medical condition. As discussed in Part II, some early release measures exclude people based on their crime of conviction or sentence. See supra notes 83-96 and accompanying text.

  129. [129]. Compared to younger people in prison, older people are much more likely to have a chronic disease, comorbid conditions (meaning multiple co-occurring medical problems) and mobility challenges. Kimberly A. Skarupski, Alden Gross, Jennifer A. Schrack & Gabriel B. Eber, The Health of America’s Aging Prison Population, 40 Epidemiologic Revs. 157, 162 (2018) (metareview of the literature finding “that [the] population [of older people in prison] reports more chronic diseases, comorbid conditions, mental health issues, and mobility challenges than their younger counterparts”); Addressing the Aging Crisis, supra note 45, at 1151 (“[O]lder adults have more medical needs than younger adults ....”). Older people in prison have “a substantially higher burden of chronic conditions such as hypertension, diabetes mellitus, and pulmonary disease than younger prisoners and older nonprisoners.” Id. Older prisoners also have high rates of geriatric syndromes: common conditions associated with aging such as visual or hearing impairment, incontinence, and falls. Caregiving Behind Bars, supra note 107, at 1288–91; Being Old and Doing Time, supra note 94, at 704–06.

  130. [130]. Carson & Sabol, supra note 33, at 9.

  131. [131]. Id.

  132. [132]. Carson, supra note 8, at 20. The percentage of older people in prison serving a sentence for a violent crime has stayed steady for 20 years (it was 65 percent in 1993, 68 percent in 2003). Carson & Sabol, supra note 33, at 9. This two thirds figure is—and has consistently been—the highest of any age group in state prison. Id.

  133. [133]. Forty percent of older people currently incarcerated in state prison were admitted to prison after they were 55 or older. This finding is consistent with the increasing arrest rates for people 55 or older since the early 1990s, though arrest rates for this age group remain the lowest of any age group. Overall arrest numbers have declined since 1993, but the number and proportion of arrests of people 55 or older has increased. Carson & Sabol, supra note 33, at 9. The reasons for the increase in arrests and prosecutions of older people are not well understood. See generally Jeremy Luallen & Ryan Kling, A Method for Analyzing Changing Prison Populations: Explaining the Growth of the Elderly in Prison, 38 Evaluation Rev. 459 (2014) (evaluating possible reasons for the shift in prison population age distribution). See generally Lauren C. Porter, Shawn D. Bushway, Hui-Shien Tsao & Herbert L. Smith, How the U.S. Prison Boom Has Changed the Age Distribution of the Prison Population, 54 Criminology 30 (2016) (describing the recent change in prison population age distribution).

  134. [134]. Proposals for second look measure vary in how much time served is required to trigger eligibility. See supra Section II.A.

  135. [135]. Danielle Kaeble, U.S. Dep’t Just., Time Served in State Prison, 2018, at 4 (2021), []. “The average sentence length for persons released after serving time for violent offenses (10.8 years) in 2018 was more than twice the average sentence length for those released after serving time for property (4.9 years), drug (5.2 years), or public-order (4.4 years) offenses.” Id.

  136. [136]. Pfaff, supra note 13, at 66 (reporting the finding that in a study of people who had served at least 11 years in state prison by the end of 2013, across 17 states, “fully one-fourth were in for murder or manslaughter, and 65 percent had been convicted of an index violent crime; all told, 83 percent of these long-serving [incarcerated people] had been convicted of some sort of violent offense.”); Kaeble, supra note 135, at 1 (“Among persons released from state prison in 2018 after serving 20 years or more, 70% had been imprisoned for murder or rape.”); Ashley Nellis, No End in Sight: America’s Enduring Reliance on Life Imprisonment 22 (2021), [] (reporting the finding that 91 percent of people serving life sentences are incarcerated for a violent conviction); John Pfaff, Decarceration’s Blindspots, 16 Ohio St. J. Crim. L. 253, 258 (2018) (“Looking just at California, I found that over 57% of those in [prison] for at least fifteen years had been convicted of homicide, and over 85% of those in for at least 25.”).

  137. [137]. Carson, supra note 8, at 20 (the data on prevalence of violent convictions report most serious offense); Kaeble, supra note 135, at 4 (same); Nellis, supra note 136, at 22 (the data on prevalence of violent convictions is based on the crime of conviction).

  138. [138]. See supra notes 127–36 and accompanying text.

  139. [139]. Leipold, supra note 50, at 1586–87.

    Protecting public safety and predicting future criminality are important, but they are not the only goals of the justice system. One of the critical weaknesses in our analysis of crime is the lack of agreement on why we punish, either in general or in a particular case. Focusing exclusively on deterring future crime and incapacitating those who are not deterred misses the vital role that retribution plays in our sentencing policy and decisions. No matter how confident the prediction that an inmate can be returned safely to society, release will not (and should not) happen if the inmate has not been adequately punished for his behavior. (Think, for example, of the child-murderer who for health reasons is no longer at risk for reoffending).


  140. [140]. Several scholars have considered the tension between retributivism and forms of legal mercy, such as clemency or compassionate release. See, e.g., Dan Markel,Against Mercy, 88Minn. L. Rev.1421, 1425–28 (2004) (critiquing mercy as a basis for decision-making in the criminal legal system). E. Lea Johnston and William Berry have considered this tension specifically in the context of compassionate release. Johnston, supra note 20, at 64 (“In essence, compassionate release laws appear to reflect the determination that cost and incapacitation considerations trump retributive concerns of just deserts.”); Berry III, supra note 20, at 874–76 (describing the tension between compassionate release and retributive purposes of punishment).

  141. [141]. See Russell L. Christopher, Deterring Retributivism: The Injustice of “Just” Punishment, 96 Nw. U. L. Rev. 843, 852–69 (2002) (summarizing the age-old debate over retributive versus utilitarian theories of punishment).

  142. [142]. Robert Weisberg, Reality-Challenged Philosophies of Punishment, 95 Marq. L. Rev. 1203, 1208 (2012) (arguing that retributive theories of punishment “either ignore the empirical realities of mass incarceration or address them only in evasive and feckless ways.”); Ekow N. Yankah, Punishing Them All: How Criminal Justice Should Account for Mass Incarceration, 97 Res Philosophica 185, 187 (2020) (“[O]ur criminal law is hostage to a kind of retributivist hunger that premises punishment solely on how punishment reflects and affects individual wrongdoers. The wider effects of punishment on one’s family, one’s neighborhood, and the wider community are rendered invisible.”).

  143. [143]. The lines between extreme risk-aversion and retributivism are not always clear. Sometimes, what sounds like extreme risk-aversion (“this person is too great of a threat to be released”) is retributivism in disguise, dressed up and dignified as a sober assessment of public safety risk (“this person deserves to stay in prison, and I’ll justify denying release using the language of public safety risk”).

  144. [144]. W. David Ball, The Peter Parker Problem, 95 N.Y.U. L. Rev. 879, 879 (2020).

  145. [145]. See id. at 884–86.

  146. [146]. Beth Schwartzapfel, Parole Boards: Problems and Promise, 28 Fed. Sent’g Rep. 79, 80–81 (2015); Kevin R. Reitz & Edward E. Rhine, Parole Release and Supervision: Critical Drivers of American Prison Policy, 3 Ann. Rev. Criminology 281, 286 (2020) (“Members or entire boards have been forced to resign after a single high-profile crime committed by a released prisoner. In the wake of episodes like these, firings or no, release rates plummet, and the cautionary stories quickly spread to parole boards nationwide.” (citation omitted)).

  147. [147]. Reitz & Rhine, supra note 146, at 286.

  148. [148]. See generally Elizabeth Hinton & DeAnza Cook, The Mass Criminalization of Black Americans: A Historical Overview, 4 Ann. Rev. Criminology 261, 269 (2021) (arguing that stereotypes of “[B]lack people as a uniquely dangerous and lawbreaking group [have] justified the perpetual expansion of the American prison system, sustained harsh sentencing practices, informed decisions surrounding capital punishment, and sanctioned racial profiling in general”); Katheryn Russell-Brown, The Color of Crime 23 (2d ed. 2009) (describing “the link between the deviance attached to Blackness and racial disproportionality in the justice system”); Khalil Gibran Muhammad, The Condemnation of Blackness: Race, Crime, and the Making of Modern Urban America 7 (2010) (describing the historical origins of “the emergence of an enduring statistical discourse of black dysfunctionality” and criminality).

  149. [149]. See, e.g., Megan Denver, Justin T. Pickett & Shawn D. Bushway, The Language of Stigmatization and the Mark of Violence: Experimental Evidence on the Social Construction and Use of Criminal Record Stigma, 55 Criminology 664, 680 (2017) (“Among members of the public, a violent conviction signals a uniquely high level of recidivism risk ....”); Prescott et al., supra note 25, at 1644 (noting the “popular belief” that “those who have killed before will eventually kill again”).

  150. [150]. Prescott et al., supra note 25, at 1647.

  151. [151]. Id.

  152. [152]. See supra Section II.B.

  153. [153]. See Barkow, supra note 30, at 45.

  154. [154]. See, e.g., Denver et al., supra note 149, at 680.

  155. [155]. This is a version of what Jonathan Simon terms “total incapacitation”: “[T]he idea that imprisonment is appropriate whenever an offender poses any degree of risk to the community.” Jonathan Simon, Dignity and Risk: The Long Road from Graham v. Florida to Abolition of Life without Parole, in Life Without Parole: America’s New Death Penalty? 282, 293 (Charles J. Ogletree, Jr. & Austin Sarat eds., 2012).

  156. [156]. Three of the four classic purposes of punishment are considered utilitarian: deterrence, incapacitation, and rehabilitation.

  157. [157]. Joshua Kleinfeld, Two Cultures of Punishment, 68 Stan. L. Rev. 933, 941 (2016).

  158. [158]. See supra Section III.C.1.

  159. [159]. See Justin D. Levinson, Robert J. Smith & Koichi Hioki, Race and Retribution: An Empirical Study of Implicit Bias and Punishment in America, 53 U.C. Davis L. Rev. 839, 883 (2019) (describing the “deep and inextricable connection between race and retribution,” and the history of harsher punishments for Black Americans as compared to white Americans); Alice Ristroph, How (Not) to Think Like a Punisher, 61 Fla. L. Rev. 727, 746 (2009) (“[J]udgments of ‘desert’ may serve as an opportunity for racial bias to enter the criminal justice system. Research on capital sentencing, a context in which jurors are frequently urged to make a direct assessment of desert, reveals an unsettling tendency to find [B]lack defendants who kill white victims more deserving of death than those who commit similar crimes but with a different defendant-victim racial matchup.”).

  160. [160]. Carissa Byrne Hessick & Douglas A. Berman, Towards a Theory of Mitigation, 96 B.U. L. Rev. 161, 165, 210 n.193 (2016).

  161. [161]. See Ristroph, supra note 159, at 729–30.

  162. [162]. Hessick & Berman, supra note 160, at 179 (“Under the theory of retributi[on] (also sometimes called the theory of ‘just deserts’), a defendant is punished because she deserves it. Put another way, the goal of punishment for a retributivist is the punishment itself.”) (footnote omitted).

  163. [163]. Id. at 179–82.

  164. [164]. Ryan, supra note 72, at 168 (“Moreover, sentencing decisions usually do not explain what part of a sentence is due to desert considerations and what part is due to other concerns such as deterrence.”).

  165. [165]. Ristroph, supra note 159, at 746 (“Claims of desert are not falsifiable ....”).

  166. [166]. Johnston, supra note 20, at 64. To be sure, there are some compassionate release cases where retributive arguments have little force. Imagine a person in their early twenties or thirties sentenced to a relatively short prison term, such as three years; they are diagnosed with a terminal illness within their first year of incarceration and doctors predict they have 18 months or less to live. The terminal illness has turned their three-year prison term into a life sentence, a wildly disproportionate sanction. My clinic, Legal Assistance to Incarcerated People, has represented people in their twenties and thirties seeking compassionate release based on a terminal illness. Karen Koethe, The Case for Compassionate Release, Gargoyle (Sept. 17, 2020), https://gargoyle. [].

  167. [167]. See supra Section III.A.

  168. [168]. In California, for example, release on medical parole is available—at least on paper
    —to anyone who has not been sentenced to death or life without the possibility of parole. See Cal. Penal Code § 1170(e)(2) (West 2021). Yet one of the few empirical studies on compassionate release decision-making found that people serving sentences for violent convictions were significantly less likely to be granted compassionate release than people serving a sentence for a nonviolent conviction (drug, public order, or property offense). See generally Ashley L. Demyan, What’s Compassion Got To Do With It? An Empirical Examination of Medical Release in California Prisons (2013) (Ph.D. dissertation, University of California-Irvine) (on file with author). This study—one of the only empirical studies on compassionate release decision-making in the states—was conducted in California using a comprehensive dataset of compassionate release applications and decisions from 1995-2011. Demyan found that a conviction for a violent or sexual crime was negatively associated with release, while a conviction for a property or drug crime was positively associated with release. Other factors strongly associated with a violent conviction—longer sentence length and longer time served—were also negatively associated with release. Among compassionate release applicants who were serving a sentence of ten years or less, more applicants were granted compassionate release (around 60 percent) than denied (around 40 percent). The percentage of grant rates dropped significantly as sentence length increased. This strongly suggests that decisionmakers see compassionate release as a remedy for unforeseen changes in circumstance that render a relatively short (i.e., a single digit term of years) prison sentence disproportionate. Longer time spent in prison was also negatively associated with compassionate release. The majority of successful compassionate release applicants had served a relatively short time in prison: 69.8 percent of successful applicants had served three years or less at time of release, and 85.5 percent had served six years or less.

  169. [169]. The information in this Part is based on a fifty-state survey of compassionate release laws and policies. The extraordinary work of Mary Price and Julie Clark at Families Against Mandatory Minimums (“FAMM”) was the foundation for the fifty-state survey. In 2018, FAMM published both a report on compassionate release in the states and detailed state memos describing the compassionate release policies for each state. Price, supra note 19; Everywhere and Nowhere: Compassionate Release in the States, FAMM, []. The Appendix lists the 78 different state legal regimes reviewed. There are more policies than states because many states have multiple compassionate release laws. The multiple compassionate release provisions in a state may reflect different sentencing regimes (e.g., one process for people serving a parole-eligible sentence, one for people sentenced after the state abolished discretionary parole). They may empower different decisionmakers (e.g., one executive clemency process in which the governor makes the decision; one parole process in which the parole board makes the decision), or recognize different grounds for medical eligibility (e.g., one process for people with a terminal illness who are near death; one process for people who are physically incapacitated).

  170. [170]. These categories reflect the statutory language; in many cases, their practical effects are intertwined. For example, where a life sentence is required for a first-degree murder conviction, everyone serving a sentence for a first-degree murder conviction will be excluded regardless of whether the exclusion specifies a life sentence or a first-degree murder conviction.

  171. [171]. Michael O’Hear, Third-Class Citizenship: The Escalating Legal Consequences of Committing a “Violent” Crime, 109 J. Crim. L. & Criminology 165, 168, 172 (2019) (cataloging the exclusion of people incarcerated for violent convictions from various forms of early release); Klingele, supra note 60, at 429–31 (discussing Washington’s early release program limited to people “convicted of most drug and property offenses whose risk assessments suggested they were at a low risk of recidivism” and Kansas’s early release program limited to people convicted of low-level crimes).

  172. [172]. See infra Appendix.

  173. [173]. Id. This is a significant exclusion given the increasing number of people serving LWOP sentences. See Ashley Nellis, supra note 136, at 5 (reporting that more than 55,000 people are serving LWOP sentences).

  174. [174]. To be sure, some states significantly limit eligibility for compassionate release. Virginia’s new compassionate release legislation, enacted in 2020, provides for compassionate release for someone with a terminal illness who has a life expectancy of less than 12 months but includes a lengthy laundry list of 17 categories of excluded convictions, ranging from first- and second-degree homicide, kidnapping, and rape to burglary and second convictions for animal fighting. Va. Code Ann. § 53.1-40.02 (West 2021); see also Ned Oliver, Thousands of Virginia Prisoners Could Be Released Early Under New Earned Sentence Credit Program, VA. Mercury (Oct. 26, 2020, 12:03 AM), [] (describing exclusions). Such sweeping exclusions, however, are the exception rather than the rule.

  175. [175]. See infra Appendix.

  176. [176]. See, e.g., D.C. Code Ann. § 24-464 (West 2021); D.C. Code Ann. § 24-467; N.M. Stat. Ann. § 31-21-25.1 (West 2021).

  177. [177]. See, e.g., N.C. Gen. Stat. Ann. § 15A-1369.2(b) (West 2021) (excluding capital felonies and class A, B1, or B2 felonies); Wis. Stat. Ann. § 302.113(9g)(b) (West 2021) (excluding Class B felonies).

  178. [178]. See, e.g., N.Y. Exec. Law § 259-s(1)(a) (McKinney 2021) (time served requirements where the applicant is serving a sentence for 2nd degree murder, 1st degree manslaughter, or a sex conviction).

  179. [179]. The Bureau of Justice Statistics data on age and crime of conviction do not disaggregate by specific offense. For example, first-degree intentional homicide and non-negligent manslaughter are both classified under the umbrella of “homicide” in the BJS data, while state compassionate release policies may distinguish between the two for eligibility purposes. See Carson & Sabol, supra note 33, at 31.

  180. [180]. See, e.g., Price, supra note 19, at 21; Jorge Renaud, Eight Keys to Mercy: How to Shorten Excessive Prison Sentences, Prison Pol’y Initiative (Nov. 2018),
    reports/longsentences.html [].

  181. [181]. See supra notes 92–95 and accompanying text.

  182. [182]. See, e.g., Ala. Code § 15-22-42 (2021) (“poses an extremely low risk of physical threat to others or to the community”); Ala. Code § 14-14-2(4) (2021) (medical “condition that prevents him or her from being able to perpetrate a violent physical action upon another person or self or initiate or participate in a criminal act”); Alaska Stat. Ann. § 33.16.085(a)(5) (West 2021) (“incapacitated to an extent that incarceration does not impose significant additional restrictions on the prisoner”); Colo. Rev. Stat. Ann. § 17-22.5-403.5(1)(b) (West 2021) (“the special needs offender is not likely to pose a risk to public safety”); Conn. Gen. Stat. Ann. § 54-131k(a) (West 2021) (“physically incapable of presenting a danger to society”); Kan. Admin. Regs. § 45-700-1(b) (2021) (“to permanently render the inmate physically or mentally incapacitated to the extent that the inmate lacks effective capacity to cause physical harm”); La. Dep’t of Pub. Safety & Corr., Health Care Pol’y No. HC-06, § 5(E) (2010), https:// [] (“Any offender who, by reason of an existing physical or medical condition, is so permanently and irreversibly physically incapacitated (including, but not limited to, a prolonged coma or mechanical ventilation) that he constitutes only minimal danger to himself or to society.”); Md. Code Ann., Corr. Servs. § 7-309(b) (West 2021) (“so chronically debilitated or incapacitated by a medical or mental health condition, disease, or syndrome as to be physically incapable of presenting a danger to society”); Nev. Rev. Stat. Ann. § 209.3925(1)(a)(1) (West 2021) (“[p]hysically incapacitated or in ill health to such a degree that the offender does not presently, and likely will not in the future, pose a threat to the safety of the public”); N.M. Stat. Ann. § 31-21-25.1(F)(2)(b)-(c) (West 2021) (“by reason of an existing medical condition, is permanently and irreversibly physically incapacitated; and [] does not constitute a danger to himself or to society”); N.Y. Exec. Law § 259-s(1)(a) (McKinney 2021) (“so physically or cognitively debilitated or incapacitated as to create a reasonable probability that he or she does not present any danger to society”); N.C. Gen. Stat. Ann. § 148-4(8) (West 2021) (“so incapacitating that it is highly unlikely that the inmate poses a significant public safety risk”); N.C. Gen. Stat. Ann. § 15A-1369.2(a)(2) (West 2021) (“Incapacitated to the extent that the inmate does not pose a public safety risk.”); Okla. Stat. tit. 57, § 332.18(B) (2021) (“medical condition has rendered the inmate no longer an unreasonable threat to public safety”); S.C. Code Ann. § 24-21-715(A)(3) (2021) (“no longer poses a public safety risk”); Tenn. Code Ann. § 41-21-227(h)(1) (West 2021) (“no such furlough shall be granted where there is good reason to believe that the release of a designated inmate will present a serious threat to the safety of the public or of escape by the inmate”); Tenn. Dep’t Corr., Admin. Pol’ys & Procs., Medical Furloughs No. 511.01.1, § (IV)(A)(2) (2019),
    51101-1.pdf [] (physically incapacitated “to the point where the inmate does not pose a threat to the public”); Utah Admin. Code R671-314-1(4)(a) (West 2021) (“public safety and recidivism risk is significantly reduced due to the effects or symptoms of advancing age, medical infirmity, disease, or disability, or mental health disease or disability”); Vt. Stat. Ann. tit. 28, § 502a(d) (West 2021) (“unlikely to be physically capable of presenting a danger to society”); Vt. Stat. Ann. tit. 28, § 808(e) (West 2021) (“unlikely to be physically capable of presenting a danger to society”); Wash. Rev. Code Ann. § 9.94A.728(1)(c)(i)(B) (West 2021) (“low risk to the community because he or she is currently physically incapacitated due to age or the medical condition”).

  183. [183]. Conn. Gen. Stat. Ann. § 54-131k(a) (West 2021); Md. Code Ann. Corr. Servs. § 7-309(b) (West 2021).

  184. [184]. Ala. Code § 15-22-42 (2021).

  185. [185]. Okla. Stat. tit. 57, § 332.18(B) (West 2021).

  186. [186]. Under previous versions of New York’s compassionate release law, prison doctors were hesitant to provide the required dangerousness evaluations, for two reasons: evaluating dangerousness was outside their area of expertise and some physicians were concerned about liability if they certified that someone was not dangerous, and that person then committed another crime. See Beck, supra note 20, at 227–28. Prison doctors who did make the dangerousness evaluations began to rely on an informal standard: was the person able to walk? See Ian Fisher, Weighing Caution Against Compassion, N.Y. Times (Mar. 7, 1994), https://www.ny [

  187. [187]. George Pro & Miesha Marzell, Medical Parole and Aging Prisoners: A Qualitative Study, 23 J. Corr. Health Care 162, 170 (2017) (emphasis added).

  188. [188]. See generally Andreas Mitchell & Brie Williams, Compassionate Release Policy Reform: Physicians as Advocates for Human Dignity, 19 AMA J. Ethics 854 (2017) (discussing how physicians should be advocates for patients in compassionate release programs).

  189. [189]. While there is not research on how race influences medical eligibility determinations for compassionate release, these determinations occur at the intersection of the criminal legal system and the healthcare system, both areas where racial bias and racial disparities in treatment and outcomes are well documented. Research on perceptions of patients’ pain by physicians has demonstrated that physicians are more likely to underestimate pain in Black patients than in white patients. See, e.g., Lisa J. Staton, et al., When Race Matters: Disagreement in Pain Perception Between Patients and Their Physicians in Primary Care, 99 J. Nat. Med. Assoc. 532, 532 (2007) (finding that “black race was a significant variable associated with underestimation of pain by physicians”). Perception then shapes action. Compared to white patients, Black patients are systematically undertreated for pain—meaning that they are less likely to be given pain medication and, if given pain medication, receive lower quantities. This disparity in pain treatment is found among both adult patients and child patients. The disparities in perception and treatment of pain may reflect medical providers’ fantastical beliefs about biological differences between blacks and whites (e.g., that black skin is thicker than white skin—a belief held by around 40 percent of first- and second-year medical school students, and 25 percent of medical residents in a 2016 study). See generally Kelly M. Hoffman, Sophie Trawalter, Jordan R. Axt & M. Norman Oliver, Racial Bias in Pain Assessment and Treatment Recommendations, and False Beliefs About Biological Differences Between Blacks and Whites, 113 Proc. Nat’l Acad. Scis. 4296 (2016) (discussing the results of two studies conducted on racial bias present in pain assessment and treatment).

  190. [190]. There are many practical barriers to people applying for compassionate release. Particularly for people with serious or terminal medical conditions, coordinating medical documentation and navigating the bureaucratic complexities of the application process will prove daunting. Knowledge about the availability of compassionate release among incarcerated people is limited and prisons typically do little to publicize it. See Alexa Kanbergs, Cyrus Ahalt, Irena Stijacic Cenzer, R. Sean Morrison & Brie A. Williams, “No One Wants to Die Alone”: Incarcerated Patients’ Knowledge and Attitudes About Early Medical Release, 57 J. Pain & Symptom Mgmt. 809, 813 (2019).

  191. [191]. Some states explicitly require the decisionmaker to consider the strength of a person’s release plan. See, e.g., Ala. Code § 14-14-3(c) (2021) (“No inmate shall be considered for medical furlough unless he or she would be Medicaid or Medicare eligible at the time of release or a member of the inmate’s family agrees in writing to assume financial responsibility for the inmate, including, but not limited to, the medical needs of the inmate.”).

  192. [192]. Donna Cohen, Teresa Hays & Victor Molinari, State Policies for the Residency of Offenders in Long-Term Care Facilities: Balancing Right to Care with Safety, 12 J. Am. Med. Dirs. Ass’n 481, 482–85 (2011) (describing state legislation regulating nursing home placements for people with criminal records); Rebecca Silber, Léon Digard, Tina Maschi, Brie Williams & Jessi LaChance, Vera Inst. Just., A Question of Compassion: Medical Parole in New York State 30 (2018),

    501_154111.pdf [] (reporting that in a survey of community care providers, including skilled nursing homes: “40 percent indicated that they had policies prohibiting them from accepting people convicted of sex offenses; 24 percent were prohibited from accepting people with a history of arson; and 20 percent said they could not accept people convicted of violent offenses”); Allison Frankel, Pushed Out and Locked In: The Catch-22 for New York’s Disabled, Homeless Sex-Offender Registrants, 129 Yale L.J.F. 279, 282–89 (2019) (describing the difficulties in finding community placements for people with sex convictions who meet the medical eligibility criteria for compassionate release).

  193. [193]. See Frankel, supra note 192, at 279–80.

  194. [194]. Id.

  195. [195]. Price, supra note 19, at 18–19 (“[O]nly a handful of states provide ... assistance [to compassionate release applicants] in developing [release] plans.”).

  196. [196]. Silber et al., supra note 192, at 31–32 (“Waiting until parole is granted is likely to be too late for a thorough discharge-planning process, especially for people in the late stages of a terminal illness. Beginning the process before the parole hearing, however, may require that discharge planners find a bed in a residential care facility that the provider can hold for the patient without knowing if or when the individual will be released from prison. This further narrows an already limited range of options.”).

  197. [197]. Id.

  198. [198]. I focus here on the ultimate decisonmaker. But some state policies provide for initial review (typically by a staff member with the Department of Corrections) before the application gets to the ultimate decisionmaker. Sometimes, the initial reviewer has veto power: if the initial reviewer does not recommend release, the application never gets to the final decisionmaker. See, e.g., Wis. Stat. Ann. § 302.113(9g) (West 2021) (requiring initial review by a Department of Corrections committee; if the committee does not recommend release, the petition is never reviewed by the ultimate decisionmaker).

  199. [199]. Of the 78 state compassionate release policies I reviewed, the parole board makes the ultimate decision about release under 44 state policies (56 percent of state policies). See infra Appendix.

  200. [200]. A Department of Corrections leader is the ultimate decisionmaker under 14 state policies (18 percent). The Governor makes the release decision under 12 state policies (15 percent). See infra Appendix.

  201. [201]. The sentencing court makes the release decision under eight state policies (ten percent). See infra Appendix.

  202. [202]. 18 U.S.C. § 3582(c) (2018).

  203. [203]. See Price, supra note 19, at 18.

  204. [204]. Silber et al., supra note 192, at 24 (“In Vera’s sample, six applicants who were granted medical parole died before they could be released from custody. These people died within one month of their parole interview; this dramatically highlights the tight time constraints that discharge planners face, and the need for early identification of cases and speedy case processing.”). In New York between 1992 and 2014, 20 percent of certified applicants for medical parole (meaning people who met the medical eligibility criteria) died during the review process. N.Y. State Corr. & Cmty. Supervision, Medical Parole 2014, at 3 (2015), https://doccs. [

  205. [205]. See, e,g., Ark. Code Ann. § 12-29-404(c)(2) (West 2021); Cal. Penal Code § 3550(a) (West 2021); Conn. Gen. Stat. Ann. § 54-131b (West 2021); Del. Code Ann. tit. 11, § 4217(b) (West 2021); D.C. Code Ann. § 24-468(b)(1) (West 2021); Fla. Stat. Ann. § 947.149(1)(a)–(b) (West 2021).

  206. [206]. See, e.g., Ala. Code §14-14-5(e) (2021); Cal. Penal Code§1170(e)(2)(B) (West 2021).

  207. [207]. See, e.g., Ark. Code Ann. § 12-29-404(c)(2) (“If the facts warrant and the board is satisfied that the inmate’s physical condition makes the inmate no longer a threat to public safety, the board may approve the inmate for immediate transfer to parole supervision.”).

  208. [208]. Ball, supra note 144, at 889.

  209. [209]. See Cecelia Klingele, The Promises and Perils of Evidence-Based Corrections, 91 Notre Dame L. Rev. 537, 576 n.178 (2016) (describing how the risk of homicide differs from the risk of a bar fight).

  210. [210]. See supra notes 149–54 and accompanying text.

  211. [211]. The Illinois State Commission on Criminal Justice and Sentencing Reform expressed these concerns when proposing a compassionate release program. Ill. State Comm’n on Crim. Just. & Sent’g Reform, Final Report (Parts I & II), at 66 (2016),
    cjreform2015/pdf/CJSR_Final_Report_Dec_2016.pdf [] (“A physically incapacitated inmate might still be dangerous if he or she retains the ability to direct a criminal enterprise, and terminally ill individuals can still pose a risk if the illness is not debilitating.”).

  212. [212]. Tony Perry, California Authorities Deny State’s First Medical Parole Case, L.A. Times (May 30, 2011, 12:00 AM),
    0530-story.html [].

  213. [213]. Id.

  214. [214]. Id.

  215. [215]. In Massachusetts, one of the last states to enact a compassionate release provision, the decisionmaker (the head of the DOC) rejected a compassionate release application from Alexander Phillips, who had a diagnosis of terminal cancer, on public safety grounds. Arjun Singh, Despite Passage of Law, Advocates Say State is Still Slow to Allow Compassionate Release, GBH News (June 18, 2019), []. The head of the DOC issued a statement: “While a clinician employed by the Department of Corrections (DOC) health care provider has determined that Mr. Phillips’ life expectancy is not expected to exceed eighteen months, I have determined that if released, he would not be able to live and remain at liberty without violating the law.” Id.

  216. [216]. Edward E. Rhine, Kelly Lyn Mitchell & Kevin R. Reitz, Robina Inst. Crim. L. & Crim. Just., Levers of Change in Parole Release and Revocation 14 (2018), https://robina [https://perma
    .cc/CS6H-X877]. Two of the most common reasons for denying parole are that the person has not served sufficient time for punishment or that granting parole “would diminish the seriousness of the offense.” Id. (quoting Mont. Code Ann. § 46-23-208 (2018)).

  217. [217]. Kaleena J. Burkes, Edward E. Rhine, Jason P. Robey & Ebony L. Ruhland, Robina Inst. Crim. L. & Crim. Just., Releasing Authority Chairs: A Comparative Snapshot Across Three Decades 24 (2017),
    parole_chairs_report_final.pdf [].

  218. [218]. See supra Section III.C.

  219. [219]. Price, supra note 19, at 21.

  220. [220]. See Balancing Punishment and Compassion, supra note 93, at 125 (proposing “a fast-track option for evaluation of rapidly dying prisoners”).

  221. [221]. See supra Section IV.B.

  222. [222]. See Balancing Punishment and Compassion, supra note 93, at 124–25.

  223. [223]. See id. at 123–25.

  224. [224]. Texas imposes this eligibility requirement in some compassionate release cases. Tex. Gov’t Code Ann. § 508.146(a)(1)(B) (West 2021) (a person required to register as a sex offender is medically eligible for compassionate release only where they are “in a persistent vegetative state or” have “an organic brain syndrome with significant to total mobility impairment”).

  225. [225]. See Balancing Punishment and Compassion, supra note 93, at 124–25.

  226. [226]. Id. at 124.

  227. [227]. See, e.g., Minn. Dep’t Corr., Pol’y No. 203.200, Conditional Medical Release (2020), [https://].
  228. [228]. Meryl Kornfield & Sky Lebron, Florida’s Elderly Sex Offenders Shut Out of Housing, WUFT (May 2, 2019), [] (reporting that “Gail Colleta,” president of Florida Action Committee, “has asked lawmakers to consider lifting the state’s residency restrictions if an offender is a certain age or has ailments”).

  229. [229]. Christine Vestal, For Aging Inmates, Care Outside Prison Walls, Pew Charitable Tr. (Aug. 12, 2014),
    for-aging-inmates-care-outside-prison-walls [].

  230. [230]. Id.

  231. [231]. See Chaz Arnett, From Decarceration to E-carceration, 41 Cardozo L. Rev. 641, 645 (2019). Arnett focuses on electronic surveillance technologies, but his critique may apply with equal force to state-run nursing homes that serve only formerly incarcerated people. See id. Further research on the experiences and opinions of formerly incarcerated people at the Connecticut nursing home facility, as well as the facility’s administration and staff, may illuminate the degree to which the nursing home perpetuates a punitive or quasi-carceral environment. See Vestal, supra note 229.

  232. [232]. For example, the Missionaries of Charity, in California, is a hospice facility that in recent years has increasingly served people coming out of prison. Arnold, supra note 1.

  233. [233]. In a variety of contexts (both pre-trial and postconviction), scholars have described the knotty questions of line-drawing in public safety risk evaluations. See, e.g., Sonja B. Starr, Evidence-Based Sentencing and the Scientific Rationalization of Discrimination, 66 Stan. L. Rev. 803, 839 (2014); Sandra G. Mayson, Dangerous Defendants, 127 Yale L.J. 490, 494–95 (2018).

  234. [234]. Ball, supra note 144, at 884–85.

  235. [235]. Id. at 895 (“One model suggested that we are, at most, capable of understanding only five levels of probability: ‘surely true,’ ‘more probable than not,’ ‘as probable as not,’ ‘less probable than not,’ and ‘surely false.’” (citing David Salsburg, The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century 307 (W.H. Freeman & Co. 2001))).

  236. [236]. Cal. Comm. on Revision Penal Code, Annual Report and Recommendations 57 (2020), [

  237. [237]. See, e.g., Megan Stevenson,Assessing Risk Assessment in Action, 103Minn.L.Rev.303 (2018); Starr, supra note 233; Mayson, supra note 233; Jessica M. Eaglin, Constructing Recidivism Risk, 67 Emory L.J. 59 (2017); Erin Collins, Punishing Risk, 107 Geo. L.J. 57 (2018). One area of particular concern is the heavy reliance of risk assessment instruments on criminal history—a variable highly influenced by race, as mediated through racialized policing, charging, and punishment practices. Bernard E. Harcourt, Risk as a Proxy for Race: The Dangers of Risk Assessment, 27 Fed. Sent’g Rep. 237, 237 (2015) (“The fact is, risk today has collapsed into prior criminal history, and prior criminal history has become a proxy for race. The combination of these two trends means that using risk-assessment tools is going to significantly exacerbate the unacceptable racial disparities in our criminal justice system.”).

  238. [238]. Rhine et al., supra note 216, at 10 (“The use of actuarial tools to assist in parole release decision-making is well-established in most, but not all, states.”).

  239. [239]. The 137-item CORE questionnaire used for COMPAS risk assessments does not address medical needs (other than substance use) or physical condition. Cynthia Rudin, Caroline Wang & Beau Coker, The Age of Secrecy and Unfairness in Recidivism Prediction, Harv. Data Sci. Rev., March 31, 2020, at 1, 3, [https://].

  240. [240]. Hessick & Berman, supra note 160, at 188.

  241. [241]. Id. at 209.

  242. [242]. Id. at 212.

  243. [243]. Id. at 177.

  244. [244]. See Ristroph, supra note 159, at 746–49; Weisberg, supra note 142, at 1220–24; Yankah, supra note 142, at 193–201.

  245. [245]. Jaouad, supra note 4 (noting that the hospice unit at the California Medical Facility has only 17 beds).

  246. [246]. See, e.g., Beth Schwartzapfel, Katie Park & Andrew DeMillo, 1 in 5 Prisoners in the U.S. Has Had COVID-19, Marshall Project (Dec. 18, 2020, 6:00 AM), https://www.themarshall [].

  247. [247]. One notable exception to this general trend is Illinois’ new compassionate release law, enacted in 2021. H.B. 3665, 102d Gen. Assemb., Reg. Sess. (Ill. 2021) (creating a new statutory provision at 730 Ill. Comp. Stat. Ann. 5/3-3-14 (2021)). In considering a compassionate release application, one factor the decisionmaker must consider is “the impact that the inmate’s continued incarceration may have on the provision of medical care within the Department [of Corrections].” Id.

  248. [248]. Jocelyn Simonson, Bail Nullification, 115 Mich. L. Rev. 585, 612 (2017) (emphasis omitted).

  249. [249]. D.C. Code Ann. § 24-403.04(a) (West 2021).

  250. [250]. Criteria generally include serving a minimum amount of time in prison, no serious misconduct while in prison, and compliance with the parole release plan. Rhine et al., supra note 216, at 17–18. The state administrative parole provisions are: Md. Code Ann., Corr. Servs. § 7-301.1 (2019); Miss. Code Ann. § 47-7-18 (2019); Okla. Stat. Ann. tit. 57, § 337.2 (2019); S.D. Codified Laws § 24-15A-38 (1996).

  251. [251]. In 2016, Colorado enacted legislation creating a presumption of parole release for some people who received long prison sentences for crimes committed as children. See Colo. Rev. Stat. § 17-34-102 (2021). If they complete a specific three-year DOC program designed for juveniles with long sentences and have met the time served requirement (25 or 30 years depending on the crime of conviction), they are presumed to meet the criteria for early parole. See id. § (8)(a)–(b). This is a rebuttable presumption. Id. § (8). The ultimate decision is made by the governor. See Colo. Rev. Stat. § 17-22.5-403(4.5) (2021).

  252. [252]. See, e.g., Lindsey Webb, Slave Narratives and the Sentencing Court, 42 N.Y.U. Rev. L. & Soc. Change 125, 141–42 (2018) (“At sentencing, a powerful narrative on the client’s behalf can impact the court’s decision-making in significant ways. By drawing on the powerful storytelling traditions common to all societies, defense lawyers have an opportunity to craft an emotionally and factually persuasive narrative at sentencing.”); Thomas P. Gressette Jr., A Practical Guide to Storytelling at Sentencing, 36 Champion 16, 16 (2009) (advising defense lawyers that “[y]our clients stand to gain or lose freedom in direct proportion to the story you tell for them”).

  253. [253]. Silber et al., supra note 192, at 37 (urging the New York Department of Corrections and Community Supervision to require, in their compassionate release policy, “a narrative explanation of people’s conditions using descriptive language about their functionality, their degree of impairment, their symptoms, what kinds of care they require, an assessment of their prognosis and likely trajectory, and recommendations about the kind of care and placement they will need in the community[,]” and noting that this information is not consistently shared with the parole board, the compassionate release decisionmaker).

  254. [254]. Id. (noting that in New York, compassionate release “medical evaluation forms are designed for the review and comprehension of health care professionals, not the members of the parole board, who make release decisions”); see also Sample: COVID-19 Accelerated Release Letter, Amend (June 14, 2020), [] (modeling how doctors can effectively describe a patient’s medical condition in support of a compassionate release application).

  255. [255]. Renagh O’Leary,Early Release Advocacy in the Age of Mass Incarceration, 2021 Wis.L.Rev. 447, 456 (2021) (“[T]he story of the client’s day-to-day life in prison is often the most persuasive element of a compassionate release petition.... The harshness and cruelty of incarceration for someone with a serious or terminal medical condition is made most vivid by details that only the client can share.”).

  256. [256]. See Richard S. Frase, Second Look Provisions in the Proposed Model Penal Code Revisions, 21 Fed. Sent’g Rep. 194, 200 (2009) (“[H]ow can we ensure that the injustices remedied by effective second look procedures do not further impoverish already-strained appointed counsel resources, thus creating new injustices? At least in jurisdictions where appointed second look counsel are funded from the same budget as direct-appeal and/or trial counsel, it will be necessary to increase those budgets.”).

  257. [257]. There are encouraging signs that this has begun happening during the COVID-19 pandemic. Amanda Robert, Advocacy Organizations Call for Compassionate Release of Elderly, Sick Prisoners, ABA J. (Apr. 21, 2020, 8:30 AM), [] (describing the creation of the Compassionate Release Clearinghouse and its “emergency release effortto recruit, train and support pro bono lawyers who agree to represent federal prisoners who are eligible for compassionate release, as well as those especially at risk due to COVID-19.”); Ann E. Marimow, Sick, Elderly Prisoners Are at Risk for COVID-19. A New D.C. Law Makes It Easier for Them to Seek Early Release, Wash. Post (Dec. 30, 2020, 6:00 AM), https://
    29/5342816c-3fcd-11eb-8db8-395dedaaa036_story.html [] (“A coalition of lawyers led by the Washington Lawyers’ Committee and the National Association of Criminal Defense Lawyers, with help from the Public Defender Service and local law school clinics, has filed hundreds of motions ... “ seeking compassionate release under a 2020 D.C. law expanding eligibility for compassionate release.).

  258. [258]. Pfaff, supra note 13, at 224.


Clinical Assistant Professor of Law, University of Wisconsin Law School.

I am grateful to Kristen Bell, Mira Edmonds, M. Eve Hanan, Amy Kimpel, Rachel Lopez, Kathryn Miller, Jyoti Nanda, India Thusi, Margo Schlanger, Nicole Summers, and Susannah Tahk for thoughtful feedback. Joshua Engelkemier and Olivia Schwartz provided excellent research assistance.