I Had the Wrong Denominator
My headline measured DHS's disclosure rate against DHS's claimed 3,000. The subhead had the FOIA number the whole time: 3,789.
Process note — analytical backstage for DHS Showed 9%. FOIA Showed Everyone.
The piece is a simple audit: DHS said the operation removed “the worst of the worst,” ICE’s own FOIA data shows what it actually produced. My opening draft set up that audit by measuring DHS’s published 335-name list against DHS’s own stated total: 3,000 arrests in six weeks. 335 ÷ 3,000 = 11.2%. The headline said “DHS Showed 11%.”
I didn’t register the denominator as a choice. 3,000 was the arrest figure DHS provided in its own press release, and the 335 names were the subset DHS chose to publicize; dividing one by the other felt like the obvious framing. I was measuring how much of the operation DHS told you about. What I hadn’t asked was: relative to what total?
The subhead was already carrying a different number. I’d written “335 names from 3,789 arrests” (the FOIA-documented total from ICE’s own arrest records, obtained through litigation by the Deportation Data Project). The headline was dividing by 3,000; the subhead was describing 3,789. Two denominators in the same header block, both in the first thing a reader sees. Once I saw that, the inconsistency had been sitting in the draft from version one.
The fix was a line edit: 335 ÷ 3,789 = 8.84%, rounded to 9%. The headline became “DHS Showed 9%.” But why the FOIA number is the honest baseline is the more important thing. The piece’s argument is that DHS’s self-report cannot be treated as a reliable account of what the operation did. Using DHS’s claimed 3,000 as the auditing denominator accepts that as given; it lets DHS set the measuring stick for a piece that is questioning the measuring stick. The 3,789 is what FOIA documented. That’s the auditor’s number. The 3,000 is the auditee’s.
The shift from 11% to 9% isn’t noise. It changes what the sentence means. “DHS only named 11% of the arrests it claimed to make” measures against DHS’s story. “DHS only named 9% of the arrests the evidence shows” measures against the record. Those aren’t the same sentence; they answer different questions. The first accepts DHS’s account as the baseline. The second doesn’t.
If you’re questioning someone’s numbers, you can’t measure the gap with their ruler. Use yours.
Paid-tier subscribers get this in every issue: the analytical moves behind the piece, including the ones that didn’t survive contact with the primary sources. This week is a freebie for all subscribers.