The facility-by-facility data released by Immigration and Customs Enforcement (ICE) on July 17 showed an impossibility. If we are to believe their numbers, we are left with a choice of inference. Either ICE’s statistics from late June were erroneously high or the detainee population numbers published today are erroneously low.
Upon first analysis, this impossibility seems to afflict 96 detention centers. That’s a little bit more than half of all facilities ICE reported today. If you want to see which locations have the weird numbers, click here to download this xlsx file (20 KB).
The rest of this post dives into the math a bit, but the real reason I am posting here is to make that list available to the public.
Here’s the math if you are interested.
When a person is held in an authorized ICE facility overnight, that person’s detention is logged and counted toward the facility’s “Average Daily Population”.1 Statistically, you could never subtract that person’s overnight “stay” from your books.2 An average can fluctuate. A cumulative total will not.
We can calculate the total number of cumulative stays at a detention facility as the total number of people held in custody across all nights in a fiscal year. It can be figured out using some basic math.
Average Daily Population (ADP) * Fiscal Year Day = Cumulative Stays
As an example, Stewart Detention Center in Lumpkin, Georgia had apparently more people stay this year in total as of June 23 compared with July 7, at least according to ICE.
June 23: 1,857 ADP * 266th day in the fiscal year = 493,962 stays
July 7: 1,748 ADP * 280th day in the fiscal year = 489,440 stays
The number of stays within a fiscal year should never decrease.3 July 7 cumulative stays should have been greater than or equal to those reported on June 23. And yet that’s what we are seeing at Stewart Detention Center and 96 other facilities.
Here’s how the chart for Stewart would look if I processed it using ICE’s numbers as they are reported. That drop in the final column represents today’s release, using the Interval ADP method. That drop is weird and unexpected. It forces a second look at the entire facility-by-facility sheet in ICE’s June 7 data release.
On July 8, TRACReports published an article detailing the contractual capacity ceilings at numerous detention facilities across the country. Could this have affected ICE’s population numbers produced today?
Perhaps not as it’s tricky to fit on the timeline. The data ICE released today are valid as of July 7 and it’s unclear how much ICE processes their numbers after “pulling” them from their internal dashboards.
But it is conspicuous that ICE now reports facility population numbers that are much lower for several of the detention centers listed in that article.
If you have questions, you can email me at adam at relevant-research dot com.
Footnotes
Definition of ADP from the ICE detention statistics spreadsheets: “The Average Daily Population (ADP) is based on MANDAY Count. A MANDAY is based on whether a SUBJECT is in an ERO detention facility for the midnight count. For every SUBJECT in a facility for the midnight count, that corresponds to one MANDAY. The ADP is the number of MANDAYS for a given time period, divided by the number of days in that time period.”↩︎
Much like the term “detention center”, it feels like a minimization of horrible circumstances to refer to the detention of an individual as a “stay”. We would normally associate that word with a hotel reservation. I’ve thought about using the term “cumulative beds” but is everyone provided with a bed every night? I’m still struggling through different words to express the accumulated detention of human beings in these concentrated spaces. When the state commits atrocities, language fails.↩︎
Cumulative stays will drop to 0 on October 1, but only because that date marks the beginning of the new fiscal year. Otherwise, it will increase every night that ICE holds a person in custody at that location.↩︎