The Deportation Data Project (DDP) announced new immigration enforcement data last night as part of their ongoing FOIA lawsuit, which has served as the main source of data for reporters, analysts, and researchers over the past several months. Today I will conduct a first-pass analysis so that the data can get into circulation as quickly as possible. New data includes ICE arrests, detainers, and detentions. The DDP announcement notes that the data on removals and encounters were not published because they contained errors.
Here’s how this works. I will gradually update this post throughout the day as I prepare each section and each data visualization, and include time stamps for each update. I invite readers to post questions and comments below, and I will respond to them in as close to real time as possible. I would encourage you to wait until I mark the post as “finished” to cite it—I may tweak the text or visualizations as I go—but please do share this post throughout the day and invite others to follow along.
This is intended to be a fun and interesting way to do “open social science” by bringing people from all perspectives and knowledge levels into the conversation. I really want to build community around understanding data, rather than treating it like secret knowledge that only a few High Priests of Immigration Data have access to. Help me out by joining the discussion and sharing this post online. Let’s democratize immigration data!
My enormous thanks, as always, to the DDP team’s groundbreaking work. I would encourage you to sign up for DDP’s emails, check out their new tools that allow you to download smaller subsets of the huge datasets, and register for their upcoming webinar on December 3 at 1:00 PM eastern. The webinar description is as follows:
“We are hosting a data webinar on Wednesday, December 3, 2025, at 10am PT / 1pm ET. The webinar, which we hope will be especially useful for new users, will include information about the datasets, a quick tour of resources on our website, and ways in which you can use the data for reporting and advocacy.”
New at 10:47 AM:
I’ll start with the new arrest data file released yesterday. I’ve written about this extensively on Substack, so I won’t go into all of the nitty-gritty details—but let’s cover the basics. This data reflects administrative immigration arrests conducted by ICE Enforcement and Removal Operations (ERO) and recorded in their data systems. It does not include arrests by other components of ICE, such as Homeland Security Investigations (HSI) or other DHS components, such as Border Patrol. Nor does it include arrests made by non-federal agencies (e.g., sheriff’s offices, police departments, etc.) enrolled in immigration enforcement programs such as 287(g). This data does not capture the entire universe of immigration-related arrests, but it does represent detailed data on arrests conducted by ICE at a time when the agency is engaged in intensified operations in places like Chicago, Los Angeles, and, this week, Raleigh, North Carolina.
A quick overview of the data. The data comes from an ongoing FOIA lawsuit based on a FOIA that requested data starting at the beginning of September 2023. Each new data release brings that data more current. The previous arrest data went through the end of the day on July 28 and a partial day on July 29 for a total of 294,254 arrests. The current arrest data goes through the end of the full day of October 15 and a partial day on October 16 for a total of 377,067. The new data, therefore, adds 82,813 arrest records from the end of July to mid-October. When looking at the data on a monthly basis, I will often use projections for October, using a simple formula: (October arrests / 15 days) x 31 days. I’m going to skip a discussion of my own simple validation steps. You can find more discussion about that in previous posts.
Let’s get into the main findings.
The number of ICE arrests has increased substantially to a projected 34,000+ by the end of October based on this data. Now, because we actually have more recent national summary data through November, we know that this is likely an undercount. October arrests were over 40,000, according to my recent post based on ICE’s detention data. Nevertheless, let’s just stick with what this data shows.
Although the national data from DDP is a little behind, it’s far more valuable than ICE’s public data because it contains detailed record-by-record data. That means we can assess the criminal history breakdown of people arrested by ICE. I’m repeating myself here, but let’s say it anyway: the administration has claimed they are going after the worst of the worst, but the data repeatedly shows a more complicated picture—and that continues with this data. Extending my previous graphic of total numbers of arrests by criminal history shows a similar (actually exacerbating) story of increasing numbers of immigrants with no criminal history targeted for arrest. It’s not even that arrests are going up for all three categories—they aren’t. The total number of arrests of immigrants with criminal convictions has stayed entirely constant since May.
To view this data one more way, let’s just look at this at percentages. As I wrote for a previous analysis of DDP’s data, the percentage of ICE arrests for immigrants with no convictions or charges became the dominant group back in June during the raids in L.A.—and they have remained the largest group since, growing to 45% of all arrests for the first half of October.
New at 11:28 AM:
While ICE’s arrest data includes specific apprehension site information, I’m focusing here on state-level totals. The chart below shows the top 10 states by number of arrests in October 2025. Keep in mind that these October figures are projections based on data through mid-month, so final numbers may differ.
For this graphic and the next one, I’m using individualized y-axes. This emphasizes change for the individual state and nationality but it can be a little misleading when sitting next to other graphs. Just be sure to pay close attention to the magnitude of the y-axis. Texas (and later Mexico) are far and away the largest variables in their respective graphics groups of graphics.
What stands out clearly is that Texas continues to lead by far with the most arrests in the country. This isn’t surprising. Texas is one of the most populous states, and it has extensive enforcement infrastructure through 287(g) agreements that allow local law enforcement to function as immigration agents. The entire state apparatus has aligned itself with the Trump administration’s enforcement priorities, often pushing constitutional and legal boundaries to do so.
We’re also seeing the uptick in ICE arrests in and around Chicago reflected in Illinois’s numbers, and this is the enforcement activity that’s gotten significant press attention in recent months. Meanwhile, other high population states like California and Florida, while near the top of the list, haven’t seen nearly as much change in arrest numbers recently.
Edit at 1:18 PM. By request, I’m adding a spreadsheet with all state arrests by month below.
Looking at nationality, arrests of Mexican nationals dominate followed by the Central American countries of Guatemala, Honduras, and El Salvador. Venezuela is high on the list, too. I added South Korean in the mix just to show the data traces of the absurd mass arrests of South Koreans in Georgia several weeks ago.
New at 12:02 PM:
I decided to add one more aspect of state level analysis before moving on to detainers. I was curious to see the breakdown of criminal history by state, since that could be useful to reporters trying to localize the national data to their region. This data is cumulative from February 1 to October 15, 2025. An interesting (possible) pattern emerges: ICE arrests in more heavily Democrat areas appear focused on people with only immigration violations while more Republican areas show higher percentages of immigrants with criminal histories. This back-of-the-envelope observation really needs closer analysis, but it stuck out to me.
Of ICE arrests in Washington, D.C., 81% were for people with no criminal history; this was 61% of arrests in New York. Comparatively, in Idahoaowa, Kentucky, and Kansas (Republican states with muuuuuuch less enforcement activity) less than 10% of arrests were for people with only immigration violations. I don’t know for sure if this is something, but have a look for yourself and tell me what you think.
New at 12:33 PM:
The second dataset ICE released yesterday covers detainers. For those unfamiliar, here’s how ICE officially describes what a detainer is:
An immigration detainer is a request from ICE that asks a federal, state or local law enforcement agency, including jails, prisons or other confinement facilities, to notify ICE as early as possible before they release a removable alien and to hold the alien for up to 48 hours beyond the time they would ordinarily release them so DHS has time to assume custody.
Detainers are useful for understanding immigration enforcement activity at the local level because the data includes the specific facility where the person was originally held on local or state charges. Unlike general arrests where ICE picks people up off the street, detainers target people who have already been arrested on other charges. This means they’re more likely to have pending charges or a conviction than the general population ICE arrests.
A few important caveats here. We don’t always know what happens after a detainer is sent. ICE often complains that local agencies don’t comply with detainers, but agencies aren’t actually required to honor them. On the flip side, many law enforcement officers I’ve interviewed over the years say that even when they do comply, ICE doesn’t always show up to take custody. So a detainer doesn’t necessarily mean something happens to that person immediately. Not everyone goes directly to an immigrant detention facility, though there may be other consequences sooner or later a result of a detainer.
As a general observation, detainers have substantially increased under the Trump administration compared to the Biden administration and the first Trump administration, for sure. But unlike arrests and total detention numbers, detainers don’t stand out quite as dramatically. The number of detainers issued on a monthly basis is much higher than anytime in recent years, but it’s not as high as some months during the Obama administration. As we see in the graph below, the overall number of detainers increased between January and March, but then haven’t really budged since then, while arrest and detention numbers continue to soar. That said, we shouldn’t ignore these numbers, because detainers certainly can serve as a source of arrests, both at the time and in the future. And of course we could look at this more closely if we combined the various data sets using unique identifiers.
The historical data below which goes all the way back to 2002, puts the current detainer surge in context by combining the Deportation Data Project data with TRAC Reports’ data available online.
Since the detainer data also breaks down criminal history using the three categories I’ve discussed frequently on this Substack, let’s look at how detainers break down in this regard.
The largest number of detainers are for people with criminal charges but not convictions. This makes sense because if someone ends up in a local jail, they got there through some interaction with law enforcement, whether that’s a traffic stop, a 911 call, or a serious crime.
We should also keep in mind that there’s interplay between detainers and state policy. For instance, Georgia passed a law last year called HB 1105 that essentially forces police to book people into jails for any minor traffic offense that would normally result in a citation. The stated purpose is to check citizenship and immigration status. When states have laws like this on the books, more people get booked into jails. And that’s the entire point. Some state legislatures, like Georgia’s, want people booked into jails precisely because they know it will generate ICE detainers.
What we see in the data is that the number of detainers issued for people with nothing more than a criminal charge, but no conviction, has steadily increased. Perhaps most concerning is the growing percentage of detainers for people without any pending charges whatsoever. This means that the local criminal legal system, to put it perhaps crudely, doesn’t really have anything it wants to do with these people, but ICE is still able to pursue immigration consequences. Meanwhile, the now smallest group of detainers is for people with criminal convictions already.
New at 1:15 PM:
Let’s add the totals, not just the percentages, just for clarity and consistency.
I’m not going to reproduce the state level and nationality analysis for detainers like I did for arrests. These patterns tend to be dominated by a few major variables that show up consistently across different perspectives on the data. Instead, I want to focus on what’s new here, which is the specific local facilities receiving the most detainers during the Trump administration.
After examining facility level data for every month, I didn’t see much of a story in what’s changing over time. So I’m focusing on total detainers under Trump, which for ease I’m using data from February 1 to October 15. The table shows the total number of detainers each facility has received during this period, along with the breakdown by criminal history. That feels like one of the more important indicators of what’s happening locally.
Keep in mind that the overall percentage of detainers issued in October for people with immigration violations only was 13%. When looking at this table, some of the top facilities show very different patterns. The Georgia Department of Corrections, for example, issues just over 50% of its detainers for people with no criminal history whatsoever. Meanwhile, at Queens Central Booking in New York City, none of the detainers are for people with immigration violations only. Instead, 96% are for people who only have pending criminal charges. That pattern holds across the New York boroughs.
I’ve added the state for each facility and made the table searchable. If you’re interested in finding facilities in your state, just enter it into the search bar.
New at 2:37 PM:
The final section is based on detention data, specifically book-in data that the Deportation Data Project received from ICE. Compared to the other two datasets, this one is more difficult to work with. It’s not just that it’s much larger, though that certainly contributes to the challenges. It’s also more difficult because it contains important process level information that means you can’t just do a standard analysis of all the records or you’ll end up with duplicates. Here’s why.
When it comes to detention, ICE has what we call book-in data. This means that every time someone is booked into a detention facility, a record is created. The whole thing would be simple if each person was booked into a single facility, spent a certain period of time there, and then left that facility either to be deported or released back into the community. Reality is much more complicated. A person can be booked into a facility, transferred between facilities (typically several times during their detention), and then released for any number of reasons. And a person can be detained, released, and re-detained.
Some terminology to understand. ICE refers to an uninterrupted period in ICE custody as a “stay.” They refer to the first time a person is booked into a facility at the beginning of the stay as an “initial book-in.” They refer to the final book-out at the end of their stay as a “final book-out.” Technically, each time a person leaves a facility and enters a facility, those count as a book-out and book-in. When they’re not final, they typically fall under “transfers.”
This is all quite different from “snapshot” data like the biweekly detention spreadsheet, which just shows a single point in time, thus freezing all of the complicated moving parts of the vast detention archipelago. Snapshot data is much easier to work with but provides a less sophisticated, less dynamic view of the system.
What’s not available in the book-in dataset is data on transfers to transportation, which ICE also does keep but did not produce here. This current data only includes facility level book-in and book-out data. It’s almost as if someone is magically transported between facilities as if nothing else happens in between. This is why the transfer flight data made available by Human Rights First that I wrote about last week is so important. Be sure to check that out.
I don’t want to wade into a more complicated analysis here. Instead, I just want to focus on a few observations and explanations about the value of this data, since I haven’t written much on this data before. We already have pretty good data every two weeks on the total detained population and the breakdown by criminal history, as well as the detained population at each facility thanks to the Interval ADP method applied at DetentionReports.com.
By contrast, this dataset provides much more granular insight about where people’s initial booking took place. That often doesn’t happen at the major facilities we’re familiar with, but at smaller holding rooms or other facilities across the country where people stay for shorter periods before being transferred to the 200 or so major facilities. One technical note here is that the book-in data includes facility codes but not state information. To analyze the geographic distribution, would have to merge the detention facility codes with a reference table that includes each code and the state where the facility is located.
Just to attach a number to it, for the fiscal year 2025 data, there are a total of 763 facilities listed as book-in locations. That’s obviously far above the number of facilities we typically think about. I’ve written elsewhere about the fact that the total detained population is typically an undercount of sorts. This data potentially allows us to check the record and add more detail about all of the facilities in use on any given day, which is more than the facilities officially listed in the biweekly detention data.
Let’s illustrate how to work with the data using the example of the individual person with the unique identifier: 0013d31a6d96ba57a82a99647988124381546938. This individual has five book-ins on record. Because this is all part of one stay, every book-in has the stay book-in date listed as the same date and time. However, each subsequent book-in (each transfer) has a separate date and time and a separate facility. We can track this person through the system using this method to see that they were initially held at a holding room in Boston (perhaps the ICE field office), transferred to Chittenden (Vermont), then to Karnes (Texas), then to Laredo (Texas), then to Louisiana, which is where a lot of deportation flights leave from. All of this took place over a five month period.
As I mentioned, you’ll also notice that there is some missing time in between the book-out date and time at one facility and the book-in date and time at the following. The missing data is that transportation data I mentioned that isn’t included here. (With the transportation data, the files are humungous, by the way, even more complicated but still fun to analyze.)
Other information about this person: they are female, she is from Jamaica, she is listed as “single” (this data has marital status), she was born in 1970 (so she’s about 55 years old), she has a final removal order (issued at the end of July, while she was in detention), she falls under the category of having a criminal conviction but it’s for shoplifting (hardly a national security threat, but okay), and she does not have a book out date for the final facility which means she’s still there—or more likely has been deported in the intervening time. If we did not take into account that these five rows of data constitute a single “stay”, we would inadvertently duplicate those demographic details, i.e., we would count “5” Jamaican nationals when, in fact, this is just one person. You can see how the methodological challenges pile up the better quality administrative data you get.
I realize this isn’t even close to an analysis of the book-in data, but I have to wrap it up and come back to this question later when my colleagues or I have more time.
I think I’m going to stop there for the day. There is so much more that you can do with all of this data. If you have questions, comments, criticisms, or corrections, don’t hesitate to add them to the comments section below. I’m marking this long post FINISHED. Feel free to cite and circulate.
If you like this analysis, you’ll also find Andrew Free’s analysis of the DDP data from the perspective of deaths in detention below very insightful if also very sad. We need a lot of eyes on this data. Thanks for your work, Andrew.
This newsletter is only possible because of your support. If you believe in keeping this work free and open to the public, join the 100% of surveyed paid subscribers who said this newsletter is a good steward of their investment. You can read more about the mission and focus of this newsletter and learn why, after three years, I finally decided to offer a paid option.
