The Incarceration Trends project provides public access to current and historical data on local and state jail and prison populations in the United States, combining various sources to present comprehensive information. This page briefly describes the data sources and methods the Vera Institute of Justice (Vera) uses to process the data. See cautions on using the data below, as well as links to other resources. The Incarceration Trends website is a resource created by Vera’s In Our Backyards initiative.
The primary source for historical incarceration data presented in this website is the U.S. Department of Justice Bureau of Justice Statistics (BJS). The Incarceration Trends website also includes data collected by Vera directly from federal, state, and local agencies, as well as data collected and published by other researchers and agencies. A public version of the data backing the website and complete technical documentation will be available on Vera’s GitHub account in early 2023.
Vera researchers process each data source to check for anomalies and data entry errors, comparing them to other data sources and removing and correcting them when possible. Nonetheless, some errors may remain—if you notice anomalous data, you may report it to us using the problem data form.
Data sources and processing
Local jail populations: Vera obtained federally collected data through the BJS Annual Survey of Jails (ASJ), Census of Jails (COJ), and Mortality in Correctional Institutions (MCI) data collections.
The ASJ has been fielded 34 times between 1982 and 2021 and captures data for a sample of around 800 jail jurisdictions, which includes roughly the 250 largest jail jurisdictions and a stratified sample of the remaining jail jurisdictions. A jail jurisdiction includes city or municipal jails, county jails, county or parish prisons, and some private prisons that serve primarily as pretrial detention. In most counties, there is only one jail jurisdiction—the local county jail.
The COJ has been fielded 11 times since 1970—in 1970, 1972, 1978, 1983, 1988, 1993, 1999, 2005, 2006, 2013, and 2019—and captures data for all jail jurisdictions.
The primary purpose of the MCI data collection is to support the analysis of mortality in correctional institutions. To that end, BJS has collected annual measures of jail population and jail admissions by gender in all jail jurisdictions since 2000. BJS has published data up through 2019, and it is included in this website.
Since 2018, Vera researchers have obtained data on jail jurisdictions from other sources. Vera obtained information from the 12 states where centralized agencies collect data on jails at the local jurisdiction level: California, Colorado, Florida, Georgia, Indiana, Kentucky, Massachusetts, New York, Pennsylvania, Tennessee, Texas, and West Virginia. These state-level reports cover around 1,000 jurisdictions. Vera also collected data from a large number of jails in an automated way from jail websites and jail rosters. New York University’s Public Safety Lab’s Jail Data Initiative supplies data for around 350 jail jurisdictions from 35 states that would not otherwise have been included. Vera also collected data directly from state and local government agencies through phone calls and public records requests.
Local jail populations at the state level: For years up through 2019 (the year of the most recently published Census of Jails), Vera calculated each state’s total jail population by summing the jail populations for every county in the state. After 2019, Vera estimates the state jail population by combining the sum of jail populations reported by jail jurisdictions in a given quarter, using estimates for the jail jurisdictions that did not report. The estimates for each non-reporting jail jurisdiction are based on two equally weighted factors: the extent to which populations in jails that did report have changed in the state and national trends in the jail jurisdictions’ urban to rural classification.
Six states (Alaska, Connecticut, Delaware, Hawaii, Rhode Island, and Vermont) do not participate in the U.S. jail survey or census because they run unified state systems that combine prisons and jails. In those states, jail populations on this website comprise people who are either unsentenced or sentenced to less than a year.
State prison population: Vera obtained this data—the total number of people held under the jurisdiction of the state correctional authority (including private prisons) at year end—primarily through the BJS National Prisoner Statistics (NPS) Program. Vera researchers also collect data from state and federal authorities to supplement and provide more recent information. For discussion of those methods and sources, see the People in Jail and Prison in Spring 2021 data methodology.
State prison populations broken down to county level: Vera obtained this data through the National Corrections Reporting Program (NCRP). When NCRP data was not available or was unreliable, Vera collected data directly from state departments of correction. These data points show the number of people incarcerated under the jurisdiction of a state prison system on charges arising from a criminal case in a specific county and are not available in every state. The county of court commitment is generally where a person was convicted; it is not necessarily the person’s county of residence, and may not even be the county where the crime was committed, but nevertheless is likely to be both. In places like Georgia with multi-county court districts, smaller counties share a single court district and district attorney with their neighbors, but the county of commitment is still the specific county where the crime took place, and each county is counted individually in NCRP data.
Total incarceration population: Vera calculated this data by summing the state jail population and the total number of people held under the jurisdiction of the state correctional authority, with a correction for people who are held in local jails but serving state prison sentences—a substantial issue in states like Louisiana, Kentucky, Utah, and Tennessee.
Demographic data: Vera researchers took data for resident populations from the U.S. Census and National Center for Health Statistics’ bridged-race population estimates, which is available from 1970 to 2020. For 2021 incarceration rate calculations, Vera used the 2020 number as the denominator, as more recent data has not yet been released.
Urban-rural classification: Vera’s analysis of urbanicity collapses the six categories defined by the National Center for Health Statistics (NCHS) Urban-Rural Classification Scheme for Counties into four. Vera combined medium with small metropolitan areas and micropolitan (an urban area with a population of at least 10,000 but less than 50,000) with noncore areas (all other areas not considered metropolitan or micropolitan). Vera counts the latter pairing as “rural.” A county is labeled “urban” if it is one of the core counties of a metropolitan area with a million or more people, and a county is labeled “suburban” if it is within that surrounding metropolitan area. Rural areas are the most numerous category, with more than 1,900 counties.
For years up through 2019 (the year of the most recently published Census of Jails), Vera calculated the state total jail population by summing the jail populations for every county in each urbanicity. After 2019, Vera estimates the urbanicity totals using methods described in detail in the People in Jail report series. See People in Jail and Prison in 2021 for a description of these methods.
Regional jails: Not every county has a local jail; some local jails serve more than one county. In Virginia and West Virginia these are often formal regional jails with shared governance. In some other states, a relatively informal arrangement is often made to provide jail space as needed for a nearby county that has closed its jail. For counties participating in formal regional jails, Vera reports all jail and prison numbers for the group of counties as a whole. For informal regional jails, jail data in participating counties is parceled out to each participating county based on their share of the overall resident population. This divides jail incarceration statistics proportionally among the counties and avoids over- and under-estimating jail incarceration rates in counties that use regional jails.
Annual and quarterly data points: Many of the charts present multiple data points each year—one for each quarter, visible at the end of the quarter: March, June, September, and December. In most data sources, Vera researchers calculate a quarterly average.
Missing data approaches
> Interpolation: Vera interpolated data for years that jail jurisdictions do not supply data by assuming a constant rate of change between the years when data is provided. If there are no further data points with which to interpolate, the data series stops.
> Estimation: Due to lack of timely public data on jail populations, Vera researchers have produced estimates for state, urbanicity, and national jail populations in recent years, covering 2013 to present. For more information see the People in Jail and in Prison in Spring 2021 methodology.
> Maps: For the purpose of displaying information on maps in the most recent year, Vera carries forward the last known value if the jail jurisdiction is considered active. For state- and county-level line charts, no data is carried forward.
> Average daily population versus single day counts: The total jail population is the average daily population in years where that number is available. Jail admissions are measured annually. In the BJS data, the pretrial jail population is a snapshot, single day count from the end of June in each year of people categorized as “unconvicted.” The total jail population and pretrial population numbers are not always directly comparable because the total jail population is based on the average daily jail population for that year rather than a single day (end of June) count. Jail and prison data reported by race and gender are also single day counts from the end of June and may not reflect the ways that people self-identify. Further, race and gender data are based on limited categories used in jail and prison administrative data systems. Single day counts tend to fluctuate more than the average daily populations.
See the Incarceration Trends Codebook for further detail.
Resources
For more information, see The New Dynamics of Mass Incarceration; Out of Sight: The Growth of Jails in Rural America; and Divided Justice: Trends in Black and White Jail Incarceration 1990–2013, which provides analysis of the Incarceration Trends data. Also see Vera’s reports on the misuse of jails, the cost of jails, the statewide cost of local incarceration, and women in jail.
Notes
Incarceration and admission rates: Vera calculates incarceration rates—the number of incarcerated people per 100,000 working-age residents—using county population data from the U.S. Census Bureau. For a more accurate picture of incarceration rates, people under the age of 15 and over 64 are excluded, since these groups are at very low risk of jail incarceration. Also, because the proportion of these groups in the general population varies greatly by county—less than 25 percent in some counties to over 40 percent in others—including them could skew rates and make comparisons between counties difficult. This method differs from most other calculations of statewide and national incarceration rates, which use either the total resident population or the population aged 18 and older but do not exclude people over 64.
On making comparisons: The most common analyses of incarceration data feature historical or cross-jurisdictional comparisons. The results of these analyses, however, can be distorted by the comparability of the metric analyzed, whether the data is expressed as a rate or as a count, and the time period analyzed. Readers should keep in mind the following considerations when analyzing incarceration data:
> Rates or counts: Using incarceration rates per 100,000 residents can be useful to account for differences in jurisdiction size or for changes over time. On the other hand, one should also consider the absolute count when a population has grown (or declined) substantially. For example, the jail population in Texas increased 6 percent from 2000 to 2015, but the number of state residents increased 32 percent. Based on these numbers, the jail incarceration rate is down, but the number of people in jail is still growing at a time—and in a state—where there is an emerging consensus that too many people are already behind bars.
> Held for other jails or federal authorities: Cross-county comparisons of pretrial population data can become less reliable if one county holds a large number of people for other authorities, such as other counties, U.S. Immigrations and Customs Enforcement, or the U.S. Marshals Service. These incarcerated people are aggregated into the general pretrial population in BJS statistics, as they are awaiting the resolution of their cases in federal immigration and criminal courts, respectively.
See the codebook for more information on the data and methods.
Thank you to the following Vera staff for their contributions to this version of the Incarceration Trends project: Bea Halbach-Singh, Jacob Kang-Brown, Ari Kotler, Jason Q. Ng, Eital Schattner-Elmaleh, Joyce Tagal, James Wallace-Lee, and Jessica Zhang. The Incarceration Trends project was also supported by Jac Arnade-Colwill, Karen Ball, Léon Digard, Sara Duell, Jasmine Heiss, Chris Henrichson, Jill Hubley, Sarah Minion, Jack Norton, and Monica Smith. Additional support on data from Oliver Hinds, Rohan Krishnan, Nathan Poland, and Albert Sun.
Incarceration Trends has been made possible with the generous support of Arnold Ventures, the John D. and Catherine T. MacArthur Foundation Safety and Justice Challenge, the Robert W. Wilson Charitable Trust, and Google.org.
This project was supported by a volunteer partnership between the Vera Institute of Justice and Goldman Sachs. Vera wishes to thank the following people for their generous contribution of time and effort during the launch, design, and initial implementation phases of this project: Rashid Abdalla, Alfonso Austin-Rivera, Oluwaseun Tofunmi Ogungbaigbe, Ariel Saldana, Ines Sheppard, Luke Torjussen, Carlos Tovar, and Steve Turbek.
Web development for this data exploration tool provided by Zev Ross Spatial Analysis.
