March 19, 2025: Join Prison Policy Initiative’s Advocacy Department at 1 PM EST on Wednesday, March 19, 2025, for a webinar about understanding recidivism. We will cover strategies from our recidivism toolkit, including an overview of recidivism statistics and what they really mean, how to combat the “Willie Horton” effect, and ways that advocates can transform the narrative around reentry to make it less focused on negative outcomes and more focused on the ways that returning citizens contribute to their communities.
Local jails, which hold one out of every three people behind bars, have become America’s misguided answer to problems faced by the most vulnerable people, like poverty and homelessness. Despite jails’ central role in mass incarceration, comprehensive national data about the 5.6 million people who cycle through them each year is collected infrequently, leaving even basic questions about jails unanswerable.1 Fortunately for researchers, advocates, journalists and many others, the Jail Data Initiative is collecting present-day data from roughly 900 jails to provide a better understanding of those who are criminalized and locked up, including the approximately 205,000 unhoused people who are booked into jails each year.
In this briefing, we present what we know about unhoused people who are booked into jails, using the best available dataset, collected from jail rosters2 by the Jail Data Initiative (JDI). (Last year we published our first analysis of JDI data, focused on repeat bookings; we intend to publish additional analyses this year.) We find that people booked into jail who were marked as unhoused at intake are held for longer than average, while being handed some of the lowest-level charges like trespassing or petty theft.
As we’ll explain, the data have limitations, and some jails are simply not collecting important demographic data such as housing status. But we know that jurisdictions have grownincreasinglyhostile toward people with nowhere to call home: Instead of extending a helping hand to people simply trying to rest, eat, or otherwise survive, local law enforcement is handing them a criminal record and further destabilizing their lives.
Key findings
Only 20% of the jail rosters in the full dataset (175 of 889) contained one or more entries indicating an unhoused person, but the data from those jails suggest that cities and counties are turning to their jails to address behaviors that unhoused people often engage in because they are unhoused and/or poor.
Note: When referring to “unhoused people,” we mean those who are known to us to be unhoused, based on the jail roster data; everyone else may or may not have housing, but it’s unknown. As such, we also don’t know about people entering jails who are facing housing insecurity. For more information on our process, see our methodology section.
About 4.5% of jail bookings in our sample are of unhoused people: Across the 175 jail rosters in our dataset, there were 22,839 bookings of people known to be unhoused, out of 503,571 total bookings over the course of one year. These bookings represent over 15,000 unique unhoused individuals (and about 406,000 people whose housing status was unknown, or who were housed, before their admission to jail.) This translates to about 205,000 different unhoused people going to jails each year nationwide — nearly one-third of the number of people experiencing homelessness on a single night in 2023.3
Unhoused people are more likely to be booked multiple times: More than one out of every five jailed people are booked again within a year. Unhoused people made up a disproportionate share of those rebooked, representing 4% of all unique jail bookings but 8% of those rebooked. Said another way, over 40% of unhoused people booked into jail were booked multiple times, while only 20% of people who were housed or had an unknown housing status were booked multiple times. This finding affirms many observations of ineffective targeting and sweeps of homeless populations.
Unhoused people are held in jails for longer than average: The overall average stay in jails whose rosters included at least one unhoused person was 21 days, but for unhoused individuals was 32 days — almost 50% longer. We also looked at median length of stay, because the average could be skewed by very long or very short jail stays. The median length of stay for all bookings in this sample was 4 days, but for unhoused individuals was 14 days — which is 2.5 times longer. Our analysis didn’t include bond (bail) amounts, but it’s safe to assume that unaffordable cash bail is keeping many unhoused people in jails longer than those who can afford it.
People aged 55 and older make up a disproportionate share of bookings of unhoused people: About 10% of all jail bookings in our sample were older people (those age 55 or older), but 15% of bookings of unhoused people were older people. It’s important to note that older adults are more likely to spend 50% or more of their income on rent compared to people in other age groups, making them severely housing-cost-burdened and closer to housing insecurity or homelessness.
The mass jailing of unhoused people overburdens Black people: While Black people accounted for 31% of all bookings, they accounted for 40% of all bookings of unhoused people. Meanwhile, white people made up 62% of all bookings but only 55% of all bookings of unhoused people. Similarly, people of color are overrepresented in the unhoused and severely housing-cost-burdened populations.
Unhoused people face a litany of unfair criminal charges simply because they’re unhoused: In general, most people are jailed on public order, property, or drug charges, but bookings of unhoused people made up a disproportionately large share (8%) of bookings where the most serious charge was a property charge, and a slightly greater-than-expected portion (4.8%) of bookings for which the most serious charge was a drug charge. Unhoused people were most commonly booked for a top charge of trespassing — a charge frequently used to criminalize people for having nowhere else to go. They were also more commonly booked for possession of amphetamines, disorderly conduct/drunkenness, and petty theft (of less than $500) compared to all jail bookings in our sample. In contrast, bookings of unhoused people made up disproportionately small shares of bookings where the top charge was “violent” (3%), or related to DUI (<1%) or criminal traffic (2.1%) offenses.
Inconsistent data collection in jails leaves gaps in understanding
Although the data suggest that U.S. cities and counties are unnecessarily and excessively jailing unhoused people, 4% of jail bookings is a significant underestimate of unhoused people in local jails. Our methodology relies on positively identifying people as unhoused, but many more unhoused people may have chosen to list a shelter address, a family member’s address, or another location as their address when they were booked into jail.
And clearly, housing and other demographic data are not consistently collected by jail jurisdictions. While jail rosters are by no means a traditional source of data, it’s telling that only 20% of the nearly 900 rosters in the Jail Data Initiative’s sample recorded even a single person as unhoused during the yearlong study period. Smaller jails, in particular, were underrepresented in our dataset and may be less likely to record housing status information: Jails with an average daily population of less than 100 people made up 26% of our dataset, but make up 54% of jails across the U.S.
Criminalization will never solve homelessness
While not the complete picture of jails that we all wish for, the Jail Data Initiative data provide the best and most recent look at our national reliance on jails for addressing the ongoing crisis of homelessness. Our analysis reveals that unhoused people in jails are kept there for longer, are more likely to be booked multiple times, and are disproportionately Black. In total, over 200,000 unhoused people are coming in contact each year with law enforcement agents who are supposed to be keeping them safe, but the thinly veiled case for bringing them to jail only exacerbates their homelessness4 and despair.
There may be reports of unhoused people “choosing” to go to jail over sleeping in the streets, suggesting that jail is an acceptable solution. But it’s been shown time and time again that providing housing, services, and treatment instead of jail incarceration is more sustainable, a huge relief for taxpayers, and much less harmful to individuals. This is where diversion programs and permanent supportive housing can be utilized, before someone is arrested — better yet, before any police encounter.
Methodology
The Jail Data Initiative (JDI) collects, standardizes, and aggregates individual-level jail records from more than 1,000 jails in the U.S. every day. These records are publicly available online in jail rosters — the online logs of people detained in jail facilities that often include some personal information like name, date of birth, county, charge type, bail bond amounts, and more. JDI uses web scraping — the process of automating data collection from webpages — to update their database of jail records daily. The more than 1,000 jails included in the JDI database represent more than one-third of the 2,850 jails identified by the Bureau of Justice Statistics’ Census of Jails, 2019 and are nationally representative.
For the purposes of our analysis of unhoused populations, we limited the JDI sample to bookings with admission dates between July 1, 2023 and June 30, 2024 from 175 jail rosters for which at least one person was categorized as “unhoused” upon intake (except when looking at rebookings; see below). To do so, our partners at JDI searched parts of the “Address” field in jail rosters for words like “homeless,” “unhoused,” “transient,” and similar keywords, then dropped rosters and/or bookings that did not meet our standards for robustness.
In all, there were 503,571 jail bookings captured across these 175 rosters, representing over 420,000 unique individuals booked into jail. Because our sample of jails for this analysis is so much smaller than the full JDI sample, ours may not be nationally representative.
Of course, not all of the jails included in the Jail Data Initiative database provide the same information. For the more detailed analyses of jail bookings by demographic and other characteristics, we used subsets of this sample due to inconsistencies in data collection:
Length of stay: This sample included 440,120 bookings from 175 jail rosters.
From the original sample described above, bookings were removed due to potential issues with date range overlap or missingness. Next, any active bookings as of the end of the date range (2024-06-30) were excluded.
Rebookings: This sample included 599,423 bookings from 140 jail rosters. To look at unhoused people booked into jail multiple times, we began with data from 648 jail rosters for which there was available data for a two-year window (July 1, 2021 to June 30, 2023), plus an additional 365 days for a look-forward review of rebookings (to June 30, 2024). We looked at people who were both booked into jail and released within the two-year study period, and counted people as “rebooked” or “booked two or more times” if they were booked into the same jail system within 365 days of their first jail admission in the study time period. We elected to use a two-year time frame to capture a larger sample of bookings than we could in a single calendar year. From this sample of 648 jail rosters, there were 140 rosters that included at least one individual categorized as unhoused upon admission for any of their bookings.
Gender: While we did look at the gender distribution of unhoused bookings, we did not find a notable difference between bookings of unhoused people and those whose housing status was unknown. The sample here included 477,363 bookings from 167 jail rosters, where at least one person was categorized as unhoused and where sex and/or gender was also reported.
Race and ethnicity: This sample included 401,544 bookings from 138 jail rosters, where at least one person was categorized as unhoused and where race and/or ethnicity was also reported.
Age: This sample included 448,273 bookings from 155 jail rosters, where at least one person was categorized as unhoused and where age was also reported.
Charge type: This sample included 437,241 bookings from 173 jail rosters where at least one person was categorized as unhoused and where the charge type was also reported. Charges were standardized by JDI into both broad categories (Violent, Public Order, Property, Drug, DUI Offense, and Criminal Traffic) and more specific offense types, as described by the Uniform Crime Classification Standard (UCCS) schema. An overall “top charge” category per person was determined by selecting the most severe charge from among all bookings for an individual.
Unbelievably, the 2002 Survey of Inmates in Local Jails (SILJ) is the most recent, nationally representative data available on people held in local jails (whereas an annual Survey of Jails and occasional Census of Jails capture some demographic information, but only from administrative records). The SILJ collects information like who is held pretrial due to inability to afford bail; the racial distribution of people detained pretrial; and what offenses people are locked up for before trial. The next edition of the SILJ was sent out to respondents in late 2024, a full 17 years off-schedule. ↩
A jail roster is a publicly available, online log of all individuals detained in a jail facility (or in some cases, multiple facilities or counties) on a given date. Jail rosters are typically updated daily, hourly or even in real-time, and contain information obtained at booking, like someone’s basic identifying information, where they were arrested, and the dollar amount of their bond. ↩
To estimate the number of different (unique) unhoused people booked into jail in a year, we started with the share of unique individuals in our sample, 3.63% (or 15,297) who were known to be unhoused. We applied this percentage to the number of unique individuals booked over the year from July 1, 2022 to June 30, 2023, which was 5,658,992, yielding our estimate of 205,268. ↩
Yesterday, Prison Policy Initiative’s Advocacy Director, Sarah Staudt, testified in the House Judiciary Committee in the Colorado Legislature to support a bill that would guarantee visitation rights to people in Colorado prisons.
HB25-1013, sponsored by Representatives Regina English and Jennifer Bacon, would make visitation a right, not a privilege, for people in Colorado prisons. While the Department of Corrections would maintain the ability to restrict visitation for safety reasons, visitation would no longer be able to be taken away for disciplinary reasons, like as punishment for refusing work within the prison.
At the request of Together Colorado, we provided testimony about the wealth of research that shows that visitation is vital for helping incarcerated people maintain their mental health, reintegrate into society, and avoid returning to the criminal legal system.
Visitation has numerous benefits for incarcerated people, including decreasing recidivism; people who experience visitation were 25% less likely to be rearrested within two years of release from prison. Visitation also improves the likelihood of employment after release. Incarcerated people who had family visitation had odds of finding employment almost 2 times higher than those who were not visited by family. Lastly, visitation helps incarcerated people cope with the inherent stressors of incarceration, decreasing mental health issues.
Visitation is also a key tool in decreasing the harm that incarceration causes to children and families beyond prison walls. Nationwide, half of people in prison are parents to minor children, including 80% of all incarcerated women. Incarceration of a parent is a stressor for children that affects overall wellbeing, family dynamics, poor school performance, and a heightened risk of eventual involvement in the criminal legal system. Visitation with family has been shown to lessen these negative outcomes.
Prison Policy Initiative is proud to support HB25-1013, and hopes that it can serve as a model for legislatures around the country to improve access to visitation for incarcerated people.
A little-known 1988 law called the Thurmond Amendment stripped people with drug distribution convictions of federal protections under the Fair Housing Act, making it even more difficult for many people with criminal records to secure housing - even when they are qualified in every other way, and even when the conviction is decades old. By our count, this law makes it more difficult for as many as 3 million people with these kinds of convictions to secure housing.
You’ve probably never heard of the Thurmond Amendment, but for almost 40 years, it has been quietly enabling landlords to deny housing to people based solely on a past conviction for selling drugs — no matter how qualified they are as tenants, no matter what drugs or quantities were involved, and no matter how long ago they were convicted. This lifelong collateral consequence, tied to this one specific type of conviction, makes it even more difficult for people with criminal records to find housing at a time when it’s already a near-impossible task given the highly competitive rental markets in the U.S. As we and many other researchers haveexplainedbefore, safe and stable housing is a critical factor in reentry success, while homelessness puts people directly in the crosshairs of law enforcement. These facts put the Thurmond Amendment directly at odds with correctional and public safety interests.
A coalition of advocates is working to repeal the law, led by a real estate investor who was directly impacted by the Thurmond Amendment. Their first step: gathering the available data about its impact and raising awareness of this overlooked relic of the failed “war on drugs.” That’s where we at the Prison Policy Initiative come in. One form of support we offer advocates — particularly through our Policy & Advocacy team — is to bring to light the relevant data that exists about an issue, and to find ways to fill in the gaps where the necessary information simply isn’t available. When the coalition came to us with the question of how many people are potentially impacted by the Thurmond Amendment, we combed through the available historical data to come up with an estimate, ultimately finding that over 3 million people in the U.S. have received drug distribution convictions since 1986 (which is as far back as the available data go).
How we developed our estimate
Our approach, sources, and assumptions
You might think it would be easy to answer the question, “How many people in the U.S. have a drug distribution conviction and could therefore be excluded from Fair Housing Act protections?” but the unfortunate fact is that criminal legal system data are rarely collected or reported in ways that provide simple answers to simple questions. In this case, we were lucky to find that the annual number of felony convictions in state courts for “drug trafficking” (manufacturing or distribution) was published every other year from 1986 to 2006; for federal court convictions, this number was published every year from 1990 to 2014. We found no data for years prior to 1986, and had to use less-relevant but related data to produce estimates for the missing years. Then, to avoid double-counting people who had more than one such conviction (because they would be excluded under the Thurmond Amendment for their first conviction), we had to estimate how many of these convictions were “firsts” for people as opposed to subsequent convictions of the same individuals. Missing from our analysis are people with only a misdemeanor drug distribution/manufacturing conviction (who are also excluded by the Thurmond Amendment), because we couldn’t find a data source on which to base an estimate of their number.1 Inevitably, because of the gaps in the available data, our estimates are conservative and inexact, but we believe they are the only ones produced to date. This is how we did it.
For convictions in state courts:
In even-numbered years from 1986 to 2006, the National Judicial Reporting Program reported estimated numbers of felony convictions based on a nationally representative sample of counties. These were published in a Bureau of Justice Statistics report series, Felony Sentences in State Courts.
For odd-numbered years from 1987 to 2005, we used the reported even-numbered year estimates to calculate the least squares regression line and interpolated values for the missing years along that line.
For the years 2007-2023, we calculated estimates based on (a) annual arrests of adults for drug trafficking and (b) the most recent known ratio of felony convictions for drug sale or manufacturing to arrests for drug trafficking, which was 62 per 100 arrests in 2009. However, the number of annual arrests specifically for drug trafficking are not reported; instead, the FBI reports the number of all drug arrests — combining both sale/manufacturing and possession — and the percentage of drug arrests that were for sale or manufacturing versus possession. Once we had estimates of how many people aged 18 years and older were arrested for drug-related offenses each year, we multiplied those estimates by the annual percentages attributed to sale/manufacturing to produce estimates of adult arrests specifically for drug distribution or manufacturing. 2
We wanted to exclude people under the age of 18 from our estimates because minors are typically (though not always) referred to juvenile courts, where they would not receive a felony conviction that would exclude them from fair housing protections. But this, too, required a few steps. Because not every jurisdiction reports its arrest data to the FBI every year, the FBI reports arrest data by most serious charge in two ways: the reported number of arrests and the estimated number of arrests. The estimated number adjusts for jurisdictions that did not report their data, providing the FBI’s best estimate of how many arrests were made nationwide in a given year. Unfortunately, these estimates can’t be broken down in as much detail as the reported numbers are, by characteristics such as age or sex. However, relying only on the reported numbers would lead to a serious undercount of arrests: for almost every year from 2007 to 2023, for example, the reported total was roughly 20% to 30% lower than the estimated total for drug arrests. For this reason, we calculated our own estimate of adult drug arrests based on the percentage of reported drug arrests that involved adults for each year, multiplying the FBI’s estimated total drug arrests by the adult percentage, which was between 89% and 96% for all years.
Once we had annual estimates of adult arrests for drug sale or manufacturing for each year, we multiplied those estimates by 62% (the most recently reported ratio of drug sale/manufacturing arrests that result in felony convictions) to estimate how many arrests led to felony convictions.3
Finally, to avoid double-counting people with multiple felony convictions for drug sale/manufacturing, we relied on the most recently reported percentage of people whose most serious arrest charge was drug trafficking who had no prior felony conviction, which was 51% (again, from 2009). We multiplied each of the annual estimates of felony convictions for drug sale/manufacturing by 51%, resulting in a conservative final estimate, since those previous felony convictions could be for any offense type, not necessarily for drug sale/manufacturing (the only felony convictions that would be disqualifying under the Thurmond Amendment). Our final estimate of first-time convictions for drug sale/manufacturing through state systems for 1986-2023 was approximately 3,023,773.
To varying degrees, these state-level estimates all rely on data from the National Judicial Reporting Program and/or State Court Processing Statistics Program, which in turn are based on data from a sample of the nation’s 75 largest counties. Crime and law enforcement patterns vary by urbanicity (that is, places that are more rural or suburban have different policing patterns than more urban jurisdictions), so our estimates are not as accurate as they would be if we had access to data from a broader range of counties. Unfortunately, that level of data collection has never existed at the national level, to our knowledge.
For convictions in federal courts:
For the years 1990-2014, the U.S. Sentencing Commission (USSC) published the number of felony convictions for drug “trafficking” (sale/manufacturing) in its Annual Reports and Sourcebooks. For the years 2015-2023, the USSC publishes the same data on its Dashboard.
For the years 2016-2023, the USSC Dashboard also reports the annual number of people sentenced for drug trafficking who had a prior conviction for drug trafficking, which we used to adjust for the double-counting of people who received more than one such conviction. We removed the reported annual number of people with prior drug trafficking convictions from the reported annual number of trafficking convictions for those years. For the previous years for which we could not find similar estimates (1990-2015), we removed 30.3% of the reported annual convictions — the average share of people with prior trafficking convictions reported for the years 2016-2023.4
Finally, we summed our annual estimates of first-time felony drug trafficking (sale/manufacturing) convictions for all years, giving us a total of approximately 483,643 people convicted through the federal system.
Our final national estimate is the sum of our state court and federal court estimates: 3,507,417 U.S. adults with felony convictions for drug distribution or manufacturing. This is likely a significant underestimate of the actual number of people denied legal protection for housing discrimination based on their conviction under the Thurmond Amendment, because at every decision point that would impact our final estimate, we erred on the conservative side.5
Understanding the Fair Housing Act and the Thurmond Amendment
The Fair Housing Act of 1968, a key piece of civil rights legislation, declared discrimination by housing providers on the basis of race, religion, sex, national origin, familial status, or disability illegal. The 1988 Fair Housing Amendments Act strengthened the Department of Housing and Urban Development’s (HUD’s) power to enforce the Act, creating a new avenue of legal recourse for tenants and homebuyers who experience discrimination when applying to rent an apartment, buy a home, or take out a mortgage.
Enacted at the height of the failed “war on drugs,” however, the 1988 law includes an amendment named for its sponsor, the segregationist Strom Thurmond, which explicitly denies fair housing protections to people who have been convicted of drug manufacturing or distribution. While this amendment may seem like an odd thing to tack onto civil rights legislation that does not explicitly protect people from discrimination based on criminal history, it creates a carveout that allows landlords to deny housing to this specific population. Because drug distribution convictions — and criminal records more broadly — disproportionately impact Black and Brown people, discrimination against all individuals with criminal records is effectively a form of racial discrimination.6 On this basis, when housing providers discriminate against people with criminal records, individuals can make “discriminatory effect” claims under the Fair Housing Act. However, when landlords discriminate against people with drug distribution convictions specifically, the Thurmond Amendment makes these claims impossible. No other category of criminal conviction is excluded in the same way.
Importantly, repealing the Thurmond Amendment would do nothing to diminish the autonomy of landlords and other housing providers. The Fair Housing Act doesn’t require anyone to provide housing to people with drug distribution convictions, or any other kind of conviction for that matter. It just encourages them to consider applications more fully instead of automatically disqualifying people on the basis of a criminal record alone.
We know better now: Lessons learned about drugs, crime, and housing
The Thurmond Amendment was an expression of “tough on crime” 1980s politics, which were rationalized by false beliefs about drugs, crime, and punishment. For instance, in a recent op-ed, Yusef Dahl, who has been leading the effort to repeal the Amendment, points to the misconception that people who sell drugs are different from people who use drugs, which was central to how Thurmond sold the amendment to the rest of Congress: “drug dealers,” he argued, don’t deserve federal protection. In reality, distributors and users are often the same people: the most recent national data show that 78% of people in state prisons whose most serious offense was drug-related 7 reported using any drug in the month before their arrest. More than half (55%) were using at the time of the offense itself. 8 Among people sentenced to local jails for drug-related offenses (i.e., those serving shorter sentences than people in prison), the most recent data show 74% met the criteria for a substance use disorder. While it may be rhetorically convenient for politicians to vilify people who sell drugs, doing so necessarily casts a wide net around drug users who are often coping with poverty and illness.
Second, the Thurmond Amendment’s complete exclusion of people convicted of drug distribution, no matter how long ago, implies a belief that people don’t change.9 Decades of criminological research on desistance (the individual process of shifting away from lawbreaking behavior) prove otherwise. In fact, among the recommendations of experts Shawn Bushway and Christopher Uggen, authors of the chapter linked above, is “eliminat[ing] most collateral consequences of criminal justice involvement,” such as barriers to housing. Putting a finer point on it, they write: “Policies that continue to center a criminal act years after that act was committed directly contradict everything we know about desistance.”
Ultimately, this is the problem with the Thurmond Amendment: while it was added ostensibly so landlords could “protect other tenants,” it actually works directly against public safety interests. Securing safe, stable housing is one of the greatest challenges for people during reentry, and it makes a difference in reentry success, as previous research has shown. Creating additional barriers to housing for people with criminal records is simply bad policy.
Drug distribution/manufacturing charges are typically felonies, but some states classify certain distribution/manufacturing charges as misdemeanors, such as when they involve marijuana. The Thurmond Amendment also excludes people with these misdemeanor convictions from fair housing protections. ↩
For the years 2007-2023, for which our state court estimates are based on arrests for drug sale/manufacturing and the proportion of those arrests that result in felony convictions, we did not attempt to disentangle arrests that might have resulted in federal court convictions. However, we believe that this likely results in an undercount of felony convictions rather than double-counting, because arrests by the Drug Enforcement Agency (DEA) — the federal agency that makes the most drug arrests .— are not reported to the FBI’s Uniform Crime Reporting program (our source for arrest data). And because most DEA arrests result in federal rather than state prosecution, the resulting convictions are captured in the federal United States Sentencing Commission conviction data. ↩
This ratio has varied over the years; in 2004, 71 per 100 arrests for drug sale/manufacturing resulted in felony convictions, and in 1986 it was 41 per 100 arrests. ↩
The percentage of people convicted of drug “trafficking” who had prior convictions for the same offense increased fairly steadily over the years it was reported, from 27.3% in 2016 to 33.3% in 2023. Because of this linear trend, we considered using estimates based on the least squares regression line for those years, but ultimately decided that those estimates were likely to underestimate the number of people with prior convictions (for example, this method would predict just 5% of people convicted in 1990 had a prior conviction for the same offense). Instead, we opted to use the larger proportion which produced a more conservative final estimate of unique convictions for drug sale/manufacturing. ↩
For example, we excluded about half of all estimated state court convictions under the assumption that they could all be subsequent convictions for the same offense, even though the data we used to rationalize their exclusion was based on how many people arrested for drug sale/manufacturing had any prior felony conviction, not specifically drug distribution or manufacturing convictions. Given the relatively small proportion of drug sale/manufacturing convictions out of all state felony convictions reported from 1986 to 2009 (an average of 19%), and the fact that more people released from prison sentences for drug offenses are rearrested for non-drug charges than for drug charges, this choice almost certainly excluded many people who did not have a prior conviction for drug sale/manufacturing. Furthermore, our estimates do not include people who have only a misdemeanor-level drug distribution/manufacturing conviction, even though these convictions also bar people from federal fair housing protections under the Thurmond Amendment. ↩
This is known as the “discriminatory effects rule,” which the Department of Housing and Urban Development (HUD) explains in this helpful Fact Sheet: “The Fair Housing Act bars more than intentionally discriminatory conduct — it also bars policies that have an unjustified discriminatory effect based on race, national origin, disability, or other protected class. As an example, a landlord’s policy of excluding people who have any criminal record… often will have a discriminatory effect based on race, national origin, and disability.” While these policies are not always unlawful, as HUD points out, “a policy that ha[s] a discriminatory effect on a protected class [is] unlawful if it [is] not necessary to achieve a substantial, legitimate, nondiscriminatory interest or if a less discriminatory alternative could also serve that interest.” ↩
Most of these people (72%) were in prison for drug trafficking (manufacturing or distribution) as opposed to possession (25%) or other drug offenses (3%). See Table 3 in the Bureau of Justice Statistics report Profile of Prison Inmates, 2016. ↩
These percentages were also high, if less dramatically so, among people in federal prisons whose most serious offense was drug-related: 61% reported using any drug in the month before the arrest that led to their incarceration, and 38% reported using at the time of the offense. See tables 4 and 5 in the Bureau of Justice Statistics report Alcohol and Drug Use and Treatment Reported by Prisoners: Survey of Prison Inmates, 2016. ↩
On this point it is worth noting that before Sen. Thurmond introduced his amendment, the bill already included language allowing discrimination against people “convicted two or more times … of illegal manufacture or distribution of a controlled substance” (emphasis added). In his rationale for that amendment, Rep. Bob Walker noted that, on the question of whether people who used drugs would be protected under the law, an earlier committee report had made clear that the intent was that “somebody who has cleaned up their act” would not be excluded from protections but that “we are not going to allow current users to have the protection.” Walker saw his amendment (excluding “dealers” with two or more convictions) as an extension of that logic, suggesting that he and other members of Congress — unlike Thurmond — recognized that people change and that they should not be treated differently under the law if they were not “currently engaged” in drug distribution (emphasis added). ↩
A recent study from a researcher at University of Pennsylvania finds that higher jail rates are associated with higher death rates, especially for Black people and women.
The health, social, and economic harms of incarceration extend far beyond the people behind bars to their children, families, and entire communities, as a large body of research has shown. New research from Anneliese Luck at the University of Pennsylvania adds to this evidence, finding strong links between higher county jail incarceration rates and higher county mortality (death) rates. While researchers have alreadyestablished the clear connections between jail incarceration, community health, and deaths, Luck’s analysis in The Distribution of Carceral Harm: County-level jail incarceration and mortality by race, sex, and age reveals how these connections vary across age, sex, and race. Generally, Luck finds that higher jail rates are associated with higher death rates for Black people and for women than for white people and men — with the notable exception of older Black men, for whom the “mortality penalty” of higher jail rates is most dramatic.
Luck’s analysis of racial differences is a particularly important contribution, given that the burden of incarceration falls disproportionately on Black communities. While Black people make up less than 14% of the U.S. population, they account for 42% of people who are incarcerated; unsurprisingly, Black people are also overrepresented among people with an incarcerated family member. Ultimately, Luck’s breakdown of the relationship between higher jail rates and elevated death rates by demographic characteristics gives us a clearer picture of the carceral system’s disproportionate impacts on the health and well-being of some of the most vulnerable populations.
Methodology
It may be useful to define some of the terms used in Luck’s study and in this briefing discussing her findings. The main variables of interest are:
County-level jail incarceration rates (more simply, “jail rates”): This is the number of people held in a local jail relative to the number of people in the jail’s jurisdiction, which is usually a county (jails are usually run by counties or municipalities). Incarceration rates are typically expressed “per 100,000 residents,” but they can also be calculated for specific groups within the larger population — for instance, Luck uses the number of Black people in a county jail relative to the number of Black people in the county to express the county-level Black jail rate. A higher jail rate indicates that the county incarcerated a larger portion of its population.
County-level mortality rates (or “death rates”): This is the number of people in the county who die within a certain time period (in this case, five years), relative to the total number of people in the county. This rate is also expressed “per 100,000 residents” or per 100,000 people who share a certain characteristic, such as race, sex, or age group. A higher mortality rate indicates that a greater portion of its residents died over a certain period of time.
Race, age group, and sex: The data used in Luck’s study include two racial categories (non-Hispanic Black and non-Hispanic white), five age groups (19 and younger, 20-34, 35-49, 50-64, and 65+), and two sexes (male and female).
As an additional note on the language used in this briefing, we (like Luck) also use the term “mortality penalty” as a kind of shorthand to describe the increase in death rates associated with increases in jail incarceration rates. While the word “penalty” often implies a causal relationship, it’s important to note that the findings we discuss describe correlations between county jail rates and county death rates, not direct causes and effects.
For data sources, Luck used the race-specific1 jail incarceration rates from the Vera Institute of Justice’s Incarceration Trends database2 for over 1,000 counties across the country.3 She calculated the county-level death rates by combining restricted vital statistics death data from the National Center for Health Statistics with publicly available population estimates from the U.S. Census Bureau by age, sex, race, ethnicity, county, and year; this gave her age-specific death rates by county and year for four groups: Black men, Black women, white men, and white women. She then pooled (merged) the county death rates for 2010-2014 and 2015-2019, which she analyzed in comparison to the 2009 and 2014 jail incarceration rates, respectively.4
In this study, the average jail incarceration rate for Black people across all counties in the sample (1,060 per 100,000 Black residents) is more than four times that of white people (260 per 100,000 white residents) in 2014.5 This is a reflection of the well-documented racial disparities in U.S. incarceration. There is also a much wider range of county jail incarceration rates for Black people than for white people: one standard deviation6 in the race-specific jail rate is 1,500 per 100,000 Black people, but just 400 per 100,000 white people.7 Because of these differences, the findings of this study are expressed differently for Black and white populations.
Luck analyzed the data using two models that each included different variables; both showed strong, statistically significant links between county-level, race-specific jail rates and death rates across both sexes and races, and across almost all age groups. The first model simply looked at the relationship between the county-level, race-specific jail rates and the county-level death rates of the various demographic groups. In the second model, Luck controlled for a number of county-level characteristics like poverty rate, violent crime rate, percent with a college education, urbanicity, and race, age, and sex distributions. This more sophisticated model accounts for factors besides jail rates that could explain the differences in death rates, to try to isolate the relationship between jail rates and death rates as much as possible. In this briefing, we focus on the results of this second model.
Limitations of the study: Some contextual notes on the findings
Like every study, this study has some limitations. Luck relies on existing data gathered for different purposes, so there were some constraints on how she could use the data to assess the relationship between jail incarceration, deaths, and demographics. She explains various data limitations in her publications, but we thought these two were worth noting here:
The sample was limited to counties with large enough Black and white populations to allow for valid comparisons between racial groups. So while the sample represents the majority of the nation’s Black and white residents, it reflects a somewhat more racially diverse and less rural subset of counties than we would see if every single county was included. In order to include as many less-populous and rural counties as possible (375 in the final sample), Luck pooled the mortality data into two five-year periods (instead of looking at just a few deaths in a county in one year, she used the larger number of deaths over a five-year period).
Previous research indicates that the association between jail incarceration rates and community death rates weakens over time: the link is strongest when you compare jail rates to mortality rates a year later, and weaker when you compare them to mortality rates five or ten years later. Because Luck uses mortality data pooled across a five-year period, her analysis likely underestimates the associations she finds in this study.
To be clear, these limitations do not diminish this study’s contribution to a robust body of evidence of the harms of incarceration on entire communities; they simply provide additional context.
Race-specific measures reveal more about the link between incarceration and death rates
By incorporating race-specific measures into the analysis, Luck’s new study is able to show how different demographic groups in the same communities fare under aggressive jailing. Importantly, it also demonstrates that using “race-neutral” measures, as mostpreviousstudiesdo, actually underestimates the scale of the “mortality penalty” — that is, the increase in death rates associated with increases in jail rates. To show this, Luck ran her analysis again using total population rates (i.e., not differentiated by race), and found increases in total jail rates were associated with death rate increases of 0.3 to 1% across the various demographic groups she studied. While this is a statistically significant difference, using race-specific measures yielded a much greater difference in death rates (1.4% to 1.9%). As a result, the new study’s use of race-specific data provides a more nuanced look at the people facing the most serious harms associated with jail incarceration.
Higher jail rates are associated with greater increases in death rates for women than men
The new study finds that both Black and white women’s death rates in almost every age group increase more dramatically with higher jail rates than men’s do. An increase of one standard deviation in the white jail rate (400 per 100,000) corresponds to an increase in the death rate by 1.4% for white men and 1.9% for white women of all ages. An increase of one standard deviation in the Black jail rate (1,500 per 100,000) corresponds to an increase in the death rate by 1.8% for Black men and 1.7% for Black women of all ages — but this is driven by a single age group of Black men (65+) for whom the increase in mortality is more than twice that of Black women (65+), as we explain below. For all other age groups, the “mortality penalty” is higher for Black women than Black men. Because men are incarcerated at rates far higher than women,9 the impact of the carceral system on women’s lives is often overlooked, but this research underscores the need to study how mass incarceration endangers women’s lives inside and outside of jails and prisons, too.
The one exception: The “mortality penalty” is greatest for Black men 65 and older
The larger increases in women’s death rates that we see with higher jail rates are consistent across all age groups and both races — except among Black people 65 or older. Across counties with higher Black jail rates, the death rates of Black women under 65 were 1% to 3.4% higher (depending on the age group) than in the counties with lower Black jail rates; these are greater differences than observed among Black men under 65 in the same counties. However, for people 65 years or older, this trend was reversed: higher Black jail rates were associated with a 2.1% increase in death rates for Black men over 65, but a smaller 1% increase in death rates for Black women over 65.10 The author explains the implications of this finding:
“The disproportionate toll absorbed by older Black populations—particularly men, and largely driven by coupling of the increasing penalties of jail incarceration with already high levels of mortality—calls attention to the ways racial inequalities in incarceration may exacerbate other forms of socioeconomic, political, social, and health disadvantage that have been historically shouldered by Black individuals in the United States.”
Because Luck’s research is among the first to examine the relationship between jail incarceration and mortality by race, sex, and age, these findings offer new and damning evidence for what we already know: the burdens associated with incarceration are disproportionately carried by those who are already most vulnerable to other socioeconomic disadvantages.
Conclusion
This study contributes to the growing body of evidence that mass incarceration has fatal consequences that extend far beyond jail cells and prison walls. Luck’s use of race-specific jail rates to assess the ways in which jail incarceration affects white men and women differently from Black men and women calls attention to the racially disparate community consequences of incarceration.
Women and older Black men are particularly vulnerable to the “mortality penalty” associated with higher jail incarceration rates, suggesting that jails contribute to the “uneven geography of health and mortality” across the United States. In addition, this research underscores the role jails and short jail stays play in public health. While much research on the association between population-level health outcomes and incarceration has been focused on policing and prisons, these new findings emphasize the need for research to incorporate local jails into our understanding of how the carceral system hurts the health of individuals, families, and communities.
Footnotes
“Race-specific” rates are rates calculated for a specific racial group. The race-specific jail incarceration rate is the number of people in jail in a specific racial group relative to the total jurisdiction-wide population of that racial group. ↩
In its Incarceration Trends dataset, Vera presents jail incarceration rates for counties per every 100,000 residents aged 15-64. Because these jail incarceration rates exclude youth under 15 years and older adults, we can assume they represent at least a slight undercount of the total number of people incarcerated in jail. In 2022, the Bureau of Justice Statistics reported that there were 1,900 people under 18 years old in local jails, but we do not know how many of those youth were under 15 years old. Vera’s Incarceration Trends dataset has been used frequently in studies of mortality and incarceration, including Kajeepeta et al. (2021), Weidner & Schultz (2019), and Nosrati et al. (2021). ↩
Luck included counties with populations of at least 5,000 people in each race-sex category (Black-male, Black-female, white-male, white-female) in her analysis. Ultimately, the study included 1,103 counties, representing more than 95% of the total U.S. Black population and approximately 76% of the total U.S. white population. Luck explains that while this sample is not entirely nationally representative — for example, the sample reflects a slightly more racially diverse and slightly less rural subset of counties — there are only “marginal” differences between this sample and all counties in the U.S. in terms of mortality, jail incarceration, and the county covariates (county-level poverty rate, violent crime rate, college education, percent of population that is Black, percent of population that is male, percent of population that is aged 20-30, and urbanicity). ↩
Luck explains that she did this to ensure the temporal ordering of the relationship between jail incarceration rates and the subsequent five-year pooled death rates. Her use of the two time periods (2010-2014 and 2015-2019) offered more stability to her mortality estimates and allowed the inclusion of more counties. For more information on this process, please see the author’s explanation in her Data and Methods section. ↩
These rates are slightly different from the race-specific jail incarceration rates calculated by the Vera Institute of Justice, which were 944 per 100,000 for Black people and 266 per 100,000 for white people in quarter two of 2014. These rates differ because Luck indirectly estimated the mean (average) incarceration rates based on a weighted average of race-specific jail rates across counties, weighted by the total population. ↩
The standard deviation is a measure of how much variation there is in a certain measure (like county-level jail rates) within a sample population (like the counties included in a study). Where there is wider variation in the individual data points, the standard deviation is greater, and where there is less variation, the standard deviation is smaller. Standard deviations can be useful to distinguish between observed measures (like county jail rates) that are more typical (closer to the average for the population) versus those that are especially low or high compared to the others in the dataset. That is how Luck uses standard deviations in this study: when she uses “an increase of one standard deviation” to describe a difference in jail rates, you might think of that as “in counties with jail rates well above average for the population” or “in counties with higher jail rates than most.” ↩
Luck expresses the standard deviations as changes in percentage points, which we reframed as change in incarceration rate: for example, an increase of 0.4 percentage points from an incarceration rate of 1,000 per 100,000 would be 1,400 per 100,000 people, and an increase of 1.5 percentage points from an incarceration rate of 1,000 per 100,000 would be 2,500 per 100,000 people. ↩
Because of the nature of the data and research included in this study, we cannot necessarily draw causal conclusions about the direct impact incarceration rates have on mortality rates. ↩
In 2022, men had a national jail incarceration rate of 345 per 100,000 male residents, while women had a jail rate of 55 per 100,000 female residents, according to Table 4 of the Bureau of Justice Statistics report Jail Inmates in 2022 — Statistical Tables. ↩
The systems meant to maintain order and safety in prisons are ripe for abuse by corrections staff, are frequently used to dole out extreme punishments, and play a key role in keeping people in prison longer, a new report shows. The Prison Policy Initiative’s report Bad Behavior: How prison disciplinary policies manufacture misconduct, released this morning, offers an overview of all 50 state prison systems’ disciplinary policies, and explains how the use of these policies as a tool for mass punishment works against prisons’ stated goal of rehabilitation.
Bad Behavior is the broadest review of disciplinary policies to date, and draws on original research as well as testimony from 47 currently incarcerated people, providing an essential look at how prisons are run in the age of mass incarceration. The report’s findings help explain why over 50% of people in state prisons in a typical year are punished at least once for misconduct:
It is nearly impossible to avoid disciplinary infractions behind bars: Prison rules cover a vast range of possible conduct, including vague conduct such as “disrespect” and redundant rules about rulebreaking, and the vast majority of disciplinary cases are for minor violations rather than interpersonal harm.
Incarcerated people have no meaningful way to defend themselves when accused of infractions: Accused people are typically not allowed to seek representation, and face significant obstacles to finding and presenting evidence (including witnesses) at disciplinary hearings, meaning that many people are severely punished based solely on the testimony of a corrections officer.
Discipline policies are exacerbating America’s mass incarceration crisis: People found guilty of disciplinary violations are frequently put in solitary confinement or have good-time credits revoked, both of which can effectively lengthen their prison sentences.
The report notes that misconduct records impact incarcerated people’s chances of earning early release in other ways as well, such as by preventing them from participating in educational or job training programs, and working against them in hearings before parole boards.
“Whether it’s being stripped of one’s ability to visit with family, getting expelled from programs, or being put in solitary, punishments for rulebreaking always come with a cost to rehabilitation and reentry,” said report author Brian Nam-Sonenstein. “State lawmakers should be paying close attention to how many minor infractions are punished in prisons, and the impact this has on people’s circumstances when they’re released.”
The report issues recommendations to departments of corrections for how to immediately make their disciplinary systems fairer and more constructive, including:
Reduce the number of misconduct rules, focusing first on rules that are redundant and that create opportunities for discrimination and abuse.
Limit the influence of misconduct records on early release decisions, recognizing that most infractions are minor and that rules are often enforced arbitrarily.
End solitary confinement and bans on family contact, and institute non-punitive responses to violations such as drug use behind bars — especially in facilities where treatment is practically nonexistent.
Improve fairness and accountability in disciplinary processes, starting with basic measures such as allowing people representation at hearings and the ability to gather and present evidence and witnesses.
“The disciplinary system behind bars is largely a mirror of the criminal legal system on the outside, with one crucial difference — departments of corrections can modify their systems at any time,” Nam-Sonenstein said. “Immediate action can — and should — be taken to stop people in prisons being thrown in solitary or having their release jeopardized because of a system that is so reactive to minor misbehavior, and often manufactures it out of whole cloth.”
This week, we released Bad Behavior: How prison disciplinary policies manufacture misconduct, the first 50-state analysis of prison disciplinary systems published in decades. Here, we discuss how journalists can dig deeper into disciplinary policies — a topic that touches many other aspects of life behind bars.
As the report explains, prison misconduct rules cover a vast range of behavior, making these systems ripe for abuse by corrections staff. The four dozen incarcerated people we corresponded with in developing our report described frequent charges for vague offenses like “not standing for count”; harsher punishments for accused people when they attempted to defend themselves against charges; and hearing officers finding people guilty on scant evidence rather than going against the word of another officer.
Our report is a high-level overview of how disciplinary systems work. But for journalists, there is a lot more room to explore how these systems are enforced in practice and how incarcerated people are impacted by the sanctions that follow a guilty plea.
A good place to start is finding out how many misconduct cases were processed in your state’s prison system in the last 1-2 years. (That will require a public records request, as we explain more below.) The following questions about disciplinary action could also jumpstart your reporting:
If your state DOC classifies different violations by severity, in how many recent cases was the most serious charge a minor or moderate one? (You may want to drill down on specific charges, and our report includes a table of common infractions.)
How many disciplinary cases processed in recent years resulted in someone being sent to solitary confinement? How many resulted in lost visitation or communication “privileges”?
How many days of good time were lost in recent years due to disciplinary action? (We suggest also asking what types of offenses are leading to people having time credits revoked.)
How do women, and racial minorities, experience disciplinary action differently? (For instance, our report shows that women are slightly more likely to be written up than men and much more likely to receive “minor” violations.)
In addition to the sanctions imposed when someone is found guilty, misconduct records also carry collateral consequences. You might want to ask these questions to get a better sense of the impact of disciplinary action:
Do misconduct records disqualify people in your state’s prisons from participating in any educational, job training, or behavioral programs?
What kinds of disciplinary offenses impact parole decisions in your state? Does the parole board ever consider minor violations enough to deny someone early release?
Does your state allow prisons to extend someone’s sentence past their original release date due to disciplinary infractions? (For example, in Wisconsin, being found guilty of infractions or being in solitary confinement can both lead to someone’s release being postponed.)
Talking to incarcerated people about disciplinary systems
Incarcerated people are essential sources for reporting on disciplinary policies, and are often eager to share their experiences and views. Be aware, though, that your sources are taking a risk in talking to you. (Ironically, unsanctioned communication with the media may be grounds for disciplinary charges.)
Anytime you’re working with incarcerated sources, setting clear expectations is key. Make sure to communicate the scope of your project and the terms of your conversations with your sources, as well as the risk of retaliation. Be clear about when you are planning to use their quotes and personal information in your story and make plans to protect their identities if necessary. For more guidance, read the Freedom of the Press Foundation’s guide to interviewing incarcerated people.
In talking to incarcerated sources about prison discipline, we suggest asking questions like:
Do you feel that the discipline system in your prison is effective at maintaining order and keeping people safe?
Are there any conduct rules (or categories of rules) that seem to be enforced more than others?
Are people accused of disciplinary infractions in your prison able to meaningfully defend themselves, such as by interviewing witnesses at hearings and presenting evidence?
How often are incarcerated people successful in defending themselves from disciplinary charges?
How would you change the system, if you could?
We urge journalists to only share biographical information about incarcerated people that pertains to the story. An incarcerated individual’s crime of conviction, for example, might be relevant insofar as it explains how long they have been behind bars. But being judicious about including biographical details will help you build trust with your sources.
Finding data on prison disciplinary systems
The bad news for journalists is that we don’t know how much data departments of corrections maintain about disciplinary action. The good news is that every disciplinary case — whether or not it goes to a hearing — begins with a write-up, and you can file public records requests for these reports and analyze them yourself if necessary.
In some states, the state government is supposed to conduct regular audits or analyses of prison disciplinary systems. The DOC’s official discipline policy should specify whether that is the case. If so, we recommend requesting any reports or audits of the disciplinary system from the last several years.
When requesting copies of actual disciplinary reports, you can ask the DOC to provide only reports from a certain time frame, where a certain charge was filed, where a certain type of penalty was imposed, and so on. If your investigation is looking at trends in prison discipline (rather than the treatment of a specific person), we suggest specifying in your records request that you are not seeking personally-identifiable information, and that the DOC can redact any details that could identify someone. This precaution can preempt records request denials based on privacy concerns.
For general advice on filing public records requests with criminal legal system agencies, check out our records request guide.
Using our report as a resource in other investigations
Disciplinary systems are relevant to many other aspects of the prison experience that journalists may find themselves reporting on. Our report may come in handy if you’re covering:
Solitary confinement. As our report notes, 35% of people who reported receiving disciplinary action in a sample year said they were put in solitary confinement as punishment. The section of our report about disciplinary processes dives into how people can end up saddled with punishments like this, with little or no opportunity to defend themselves against the charges.
Parole, “good time,” and other early release mechanisms. In nearly all states that use discretionary parole, release rates are declining. As news outlets like AL.com and Cleveland.com have reported, misconduct records play a significant role in release decisions. We also encourage reporters to dig deeper into how disciplinary action affects how much “good time” incarcerated people are able to accrue.
Protests and strikes in prison. Disciplinary codes punish several forms of dissent against prison conditions and rules, and in some states, dissenting behaviors like refusing to go to one’s prison job are treated as serious offenses. Our table of common disciplinary offenses helps explain how prison authorities can turn nonviolent protests into punishable infractions.
Prison programming. Reporters investigating the availability of educational, job training, or behavioral programs in prisons may want to find out whether misconduct records can exclude someone from participation. While we can’t say for sure, it’s possible that overactive disciplinary systems are effectively barring many people from rehabilitative programs — programs that can be a key part of securing early release.
Discipline systems are administrative processes in prisons that mimic criminal courts, and punishments for infractions are — in the words of corrections officials themselves — “the jail of the prison.” In other words, these systems are a black box within a larger, more well-known black box. Without investigative journalism, it is likely that these systems will remain largely obscure.
If you’re considering writing about prison discipline — or have any questions about our report or this guide — we encourage you to reach out to us through our contact page.
March 19, 2025: Join Prison Policy Initiative’s Advocacy Department at 1 PM EST on Wednesday, March 19, 2025, for a webinar about understanding recidivism. We will cover strategies from our recidivism toolkit, including an overview of recidivism statistics and what they really mean, how to combat the “Willie Horton” effect, and ways that advocates can transform the narrative around reentry to make it less focused on negative outcomes and more focused on the ways that returning citizens contribute to their communities.
