Every year, U.S. Immigration and Customs Enforcement (ICE) subjects hundreds of thousands of people to civil immigration detention—a practice that is unjust, inhumane, and completely unnecessary. It does so under a veil of secrecy, sharing only limited, often error-prone statistics with the public. Using newly obtained data, Vera’s ICE Detention Trends dashboard reveals an unprecedented level of detail about detention populations—nationally and across the 1,081 facilities in which ICE detained people—on each day of the 11 years immediately preceding the COVID-19 pandemic (October 1, 2008, through March 30, 2020).
This dashboard allows us to see, for the first time, statistics and patterns that ICE’s public reporting obscures. This is possible for two main reasons. First, the ICE detention data Vera used in this tool—obtained through Freedom of Information Act (FOIA) request—contains an unprecedented level of detail. Second, Vera developed a novel algorithm that allowed researchers to create millions of individual detention histories for the people included in the dataset.
ICE’s network of facilities is vast and ever-changing.
ICE detained people in 1,081 facilities between October 1, 2008, and March 30, 2020.
Only 106 of the 1,081 facilities (10 percent) were active for the entirety of this period.
ICE obscures the breadth of its detention network by reporting limited, incomplete statistics, and for only a fraction of facilities.
In March 2020, ICE detained people in 409 facilities but acknowledged using just 221 of them on its website. Vera’s tool features facilities ICE excludes in its public reporting: staging facilities, hold rooms, and unexpected places such as hotels, hospitals, and sections of airports.
When ICE shares its limited statistics, the agency presents one moment in time or averages across a single year, which makes it impossible to see longitudinal trends. Vera’s dashboard allows you to see detention numbers for each day of the year.
When ICE reports on detention populations, it typically provides averaged counts of people detained at midnight (“midnight population”). In contrast, 24-hour populations reflect people detained at any point on a given day, including those booked out or transferred before midnight. As a result, 24-hour and midnight populations can differ drastically. Vera’s tool provides both.
For example, during early 2012, Florence Staging Facility in Arizona detained more than twice the number of people when measured by the 24-hour population rather than the midnight population.
Comparing facility populations over time reveals trends in detention operations.
In California, as ICE gradually stopped detaining people at Mira Loma Detention Center in late 2011, it ramped up its operations at Adelanto ICE Processing Center, a facility about an hour away.
Illuminating trends in ICE detention is crucial to understanding broader incarceration trends.
In addition to using private and federal detention facilities, ICE pays local governments to hold people facing civil immigration charges in jails. This practice is so widespread that local jails comprise the most common type of detention facilities. In this way, ICE provides a perverse incentive for jurisdictions to maintain or expand their jail capacities and offset the cost of the local criminal legal system—aligning many sheriffs and local officials with punitive immigration policies.
Sometimes, efforts to reduce incarceration can be hindered when local governments refill the empty beds with people detained for ICE.
For example, Louisiana passed criminal legal system reforms in June 2017, decreasing the number of people serving state prison sentences being held in local jails by 3,900 people by fall 2019. Yet during this same period, ICE rapidly expanded its presence in the state, detaining people in many of the facilities that were previously incarcerating people for the state prison system. ICE tripled the number of people in civil immigration detention in local jails, with a population increase of 6,400 additional people.
Explore the data through this interactive dashboard or download the data through Vera’s GitHub repository.
This new tool allows the public to see the kinds of information that Vera believes ICE should share. ICE’s failure to regularly release accurate, complete, and accessible data to the public enables it to operate its multi-billion-dollar detention network with impunity and little oversight. The federal government should regularly release current, anonymized, person-level detention datasets to make its operations transparent and to enable further research on detention and its toll on the people ICE detains.
While the data in this tool cuts off on March 30, 2020, just as the COVID-19 crisis was beginning in the United States, it provides a historical view of ICE detention operations and a foreboding glimpse into the future of ICE detention without drastic intervention. After reaching an all-time high of 50,000 people detained daily in August 2019, ICE detention populations immediately began decreasing and ultimately reached a historic low with fewer than 14,000 people daily in February and March 2021. Yet, rather than seizing this opportunity to permanently scale back ICE’s detention footprint, the U.S. government allowed detention populations to increase to 30,000 people detained daily by November 2022.
Congress has continued to fund ICE at levels that permit it to operate a revolving door of apprehension, detention, and deportation—which, together, impact hundreds of thousands of people per year. ICE detains people indefinitely—often in remote locations with little to no immigration attorney capacity—despite a wealth of evidence demonstrating detention is unnecessary. ICE does all this while shielding its operations from the public. It is time to shift away from a default of detention and surveillance to one of liberty and dignity.
Through interactive maps and graphs, this tool allows users to see changes in the following metrics geographically and over time:
This dashboard primarily draws from ICE detention data obtained through a Freedom of Information Act request, shared with Vera by David Hausman, assistant professor of law, University of California, Berkeley. The data contains more than 10.5 million rows of information that form detention histories for every person in ICE custody from October 1, 2008, through March 30, 2020 (fiscal years 2009 through mid-2020). The data includes the dates and times ICE booked people into and out of detention, the corresponding facilities, and release reasons.
The structure of the original dataset poses several challenges for researchers seeking to analyze ICE operations. Most notably, the dataset does not link together separate detention stints for people whom ICE transferred to one or more facility after their initial book-in. To overcome this limitation, Vera developed a novel algorithm to construct individual detention histories from a person’s initial book-in to their final-book out, inclusive of any transfers. Doing so allowed Vera to combine records across spreadsheets, account for duplicated data, and compute detention populations. Vera drew from additional datasets to geocode facility locations, as the original dataset included facility names and codes but no information on location or facility type.
Otherwise, Vera chose to present the ICE detention data “as is,” to the greatest extent possible, including any inconsistencies or errors that may be present in the data compiled and shared by ICE. For example, to calculate midnight populations, Vera uses 11:59 p.m. as reflected in the original data, taking the timestamps ICE records at face value. Preliminary analysis did not reveal a consistent use of either local time zones or Coordinated Universal Time (UTC). Vera retained data entries that indicated lengths of stay lasting zero minutes and those that showed people as being detained in two places at once during a transfer. Other researchers have noted similar overlaps, gaps, inconsistencies, and errors in ICE detention records.1
See Vera’s technical appendix for more detailed methodology and the ICE Detention Trends GitHub repository to download accompanying data files.
1 Emily Ryo and Ian Peacock, “A National Study of Immigration Detention in the United States,” Southern California Law Review, Vol 92, No 1 (2018), https://perma.cc/RV65-XPUK; Don Kerwin, Daniela Alulema, and Siqi Tu, “Piecing Together the US Immigrant Detention Puzzle One Night at a Time: An Analysis of All Persons in DHS-ICE Custody on September 22, 2012,” Journal on Migration and Human Security, Vol 3 No 4 (2015), https://journals.sagepub.com/doi/pdf/10.1177/233150241500300402; Human Rights Watch, “A Costly Move: Far and Frequent Transfers Impede Hearings for Immigrant Detainees in the United States,” June 14, 2011, https://perma.cc/S57K-Z8CU.
Vera thanks David Hausman, assistant professor of law, University of California, Berkeley, for providing the detention-stint-level ICE detention datasets acquired through FOIA requests which are visualized in this dashboard. Vera also thanks the following organizations for making available supplemental ICE datasets Vera used to process the data and geocode facilities: Human Rights Watch, Transactional Records Access Clearinghouse, National Immigrant Justice Center, Immigrant Legal Resource Center, and the Marshall Project.
Vera would like to thank the following colleagues for their support in providing editing, design, and guidance: Nina Siulc, Chris Henrichson, Jasmine Heiss, Jacob Kang-Brown, Kica Matos, Will Snowden, and Lizzie Allen. The authors wish to acknowledge the work of Dennis Kuo, who contributed to the development of the Detention Stay ID algorithm as a former Senior Data Scientist at Vera.
For more information, contact Noelle Smart, principal research associate, at nsmart@vera.org.