Is ICE Intentionally Under-reporting Detainees? A Look at Anomalies in the Latest Detention Facility Data
ICE's detention facility data show weird math that erases previous detention stays and raises questions about the data's reliability.
ICE updated its detention spreadsheet this week on schedule, which I wrote about in today’s earlier post. In this post, I am joined by my colleague Adam Sawyer to discuss weirdness in the data that he discovered, specifically ICE’s data on individual detention facilities.
In short, ICE’s facility data includes a mathematical impossibility that has the effect of significantly under-reporting the detained population at many facilities this fiscal year. We explain the issue below in order to raise public awareness and to prompt the agency to correct its mistake (or provide a counter-explanation). These errors do not appear to affect other tabs in ICE’s data spreadsheet.
ICE’s detention data includes a tab with facility-by-facility data, including average daily population (ADP). This is the basis of the Interval ADP metric published by Adam Sawyer and I several weeks ago. Read the white paper here.
This week’s detention data release shows weird, unexplained behavior that we have not seen before, is not logically possible, and raises questions about the validity and reliability of ICE’s detention facility tab. To explain the issue, we need to first explain how ICE calculates average daily population (ADP) at each facility over a fiscal year.
Let’s look at Stewart Detention Facility in Georgia, one of the largest facilities in the country, but simplify the numbers. If Stewart holds 100 people overnight on October 1, the start of the fiscal year, ICE calculates the ADP as 100 people over 1 night (100 “stays”), which equals 100 ADP. Simple enough. But how to calculate for October 2?
If Stewart holds 100 people again overnight on October 2, ICE calculates the ADP as follows. First, ICE tallies the total number of “stays” for the fiscal year, which is 100 people on October 1 and 100 people on October 2—so 200 people. Second, ICE tallies the total days in the period—so 2 days. Third, complete the simple division: 200/2 = 100 ADP.
You already knew that answer intuitively, but now let’s mix it up. What if 50 people stay overnight on October 3? Do the math: (1) 100 + 100 + 50 = 250 “stays”; (2) 3 days in the fiscal year; (3) 205/3 = 83.3 ADP.
If you’re following along so far, you probably realize by now that every time ICE releases a new detention sheet, two things should always be true: new data should never show fewer days in a fiscal year (except in October when transitioning to a new fiscal year) and never show fewer stays in a fiscal year. That is, even if Stewart Detention closed today, the total number of stays overnight so far during this fiscal year should never decrease because they already happened.
Yet this is exactly what the most recent ICE data shows.
As Adam wrote yesterday, ICE’s latest detention spreadsheet shows a decline in total stays at 96 facilities compared to ICE’s previous detention data. This is a mathematical impossibility.
Continuing with the example of Stewart Detention Center, this facility apparently had more total stays this fiscal year as of June 23 compared with July 7.
ICE’s data as of June 23 reported an average daily population of 1,857 ADP. Since this was the 266th day in the fiscal year, we can calculate a total of 493,962 stays (1,857 * 266) reported at that time.
Yet ICE’s latest data, current as of July 7, reports an ADP of 1,748. Again, since this was the 280th day of the fiscal year, we can infer that ICE is reporting a total of 489,440 stays (1,748 * 280) as of that date.
But this is impossible. ICE can’t have fewer cumulative stays this fiscal year on July 7 than on June 23. Even if Stewart shut its doors on June 24, the cumulative stays should have remained at 493,962 until the start of the new fiscal year. In short, while it may be true that Stewart’s detained population declined between June 23 and July 7, it’s impossible for it to have declined so much as to erase the previously-completed stays by people at the facility.
It is possible, though not likely, that the June 23 data overrepresented the total number of stays—but if this is true, it only displaces, rather than resolves, the underlying question of the data’s reliability.
This aberration affects a total of 96 facilities. Among these facilities with abnormalities, Lexington County Jail shows the most significant drop. On June 23, ICE reported that this facility had an average daily population of 877 people; as of July 7, ICE reports less than 4. Similarly, New Hanover County Jail reported an ADP of 428 on June 23; now, ICE reports an ADP of 1.
ICE reports an ADP drop at Karnes County Immigration Processing Center from 881 to 787 over this time period. This might not seem as dramatic, but given Karnes’ size and high population, this actually amounts to the third largest erasure of detention stays over this fiscal year among all 96 facilities.
Adam calculated the under-reporting at each facility in order to quantify the effects of this anomaly. We prepared an interactive table in Datawrapper (which, for some reason, won’t embed into this Substack post) as well as an Excel spreadsheet below.
For further inquiries, feel free reach out to Adam Sawyer (email).
Support public scholarship.
This newsletter is only possible because of your support. If you believe in keeping this work free and open to the public, consider becoming a paid subscriber. 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. If you already support this newsletter financially, thank you.
 | A guest post by
|
Might this suggest a coordinated effort and not just sloppy reporting so as to evade scrutiny for certain detainees, as in the most egregious cases of due process denied or those who may have been harmed in detention, given this administration's sledgehammer approach to immigration?
ICE is incentivized to overcount its totals, not undercount them so It's likely this is due to mathematical incompetence. How can we find out if people are being lost (disappeared) through ICE logistical incompetence, poor record keeping, or criminal carelessness? This will make it easier to cover up fatalities due to ICE brutality and unhealthy detention conditions.