The Denominator Problem: How Immigration Statistics Can Mislead

September 21, 2026|QLP News Room

Every article in this series so far has run into the same underlying issue: the same real numbers can be arranged into a "70% are criminals" headline or a "70% have no conviction" headline, and both can be technically true at once, describing different populations. This piece names that problem directly. It's called the denominator problem, and once you can see it, you'll notice it in almost every immigration-statistics claim you encounter — including ones you'll find elsewhere in this series, like the "murderers and rapists" framing and the "only going after criminals" claim.

What a Denominator Is, and Why It's the Whole Argument

Interpretation, building on documented examples: A rate is a fraction: some count of things, divided by some total population. The top number (numerator) gets most of the attention in headlines. The bottom number (denominator) — what, exactly, is being divided into — usually gets left out, and it's where most of the misleading happens. "70% are criminals" and "70% have no conviction" can both come from real ICE data because they're describing two different populations at two different moments: one is a snapshot of who's in detention on a single day, the other is a cumulative count of everyone detained across many months. Same underlying dataset, different denominator, opposite-sounding headline — the exact contradiction we trace with specific numbers in the first piece in this series.

Historic postcard view of the Syracuse University campus

Syracuse University
TRAC, the research clearinghouse whose detention snapshots anchor most of this series, is based at Syracuse University — one of the few outlets that publishes ICE's own case-by-case data for anyone to check the denominator themselves. Photo: public domain postcard (pre-1931), via Wikimedia Commons.

Six Ways the Same Numbers Get Rearranged

Documented fact, using real numbers from earlier pieces in this series:

Offenses vs. people. One person can generate multiple charges. A table of "offense counts" will always be a bigger, scarier-looking number than a table of "people," because some people are counted more than once. TRAC's methodology of classifying each detained person by their single most serious offense exists specifically to correct for this — not every source does the same.
Arrests vs. convictions. As shown in the first piece in this series, ICE's own "criminal" classification includes people who were arrested and charged but never convicted. A charge is an accusation, not a legal finding of guilt.
A snapshot vs. a cumulative total. "70,766 people were in detention on January 24, 2026" (a snapshot, the single-day peak) is a very different kind of number from "393,000 people were arrested across a full year." Adding, comparing, or swapping these without saying so produces nonsense math.
An enforcement population vs. the general population. ICE arrest statistics describe people ICE specifically chose to target, using its own priorities and resources. That group is not a random sample of anyone, so its conviction rate cannot be compared directly to the entire U.S. population's crime rate — the two "denominators" aren't measuring the same universe of people, which is the exact statistical error the Light/He/Robey Texas study covered in the previous piece was specifically designed to avoid by using a real three-way population comparison.
Raw totals vs. rates. "43,000 people with an assault charge or conviction" sounds large in isolation. As a share of roughly 393,000 total arrests, it's about 11%. Both numbers are accurate. Only one of them tells you anything about proportion.
Repeat offenders vs. unique individuals. A dataset that doesn't deduplicate by person can let a small number of people with long records inflate an offense count without changing how many actual people are involved.

"I.C.E. Incarcerating Communities Everyday" Sticker

Once you've seen how many "accurate" headlines the same dataset can produce depending on the denominator, a lot of the outrage-bait framing around ICE statistics starts to look less like reporting and more like a Fuck Ice moment waiting to happen. This one's from our Anti-Ice Stickers collection.

➡ View the Sticker

➡ Browse the Anti-ICE Stickers Collection

A Rule of Thumb for Reading Any Immigration Statistic

Interpretation: Before accepting a claim built on a number like "70%" or "393,000," it's worth asking four questions: What exactly is being counted — people, or offenses? Over what time period, and is it a snapshot or a cumulative total? What population is the denominator — everyone arrested, everyone detained today, or everyone in the country? And does the source distinguish charges from convictions? If a claim can't answer those four questions, or if it's silently switching between different answers mid-argument, that's the tell that the "denominator problem" is doing the persuasive work instead of the underlying facts.

Protest against ICE family separation policy in Washington, D.C.

A Demonstration Against ICE Policy, Washington, D.C.
Whichever side of this debate a given crowd is standing on, both sides are usually citing a real number — the question worth asking is always what it was divided by. Photo: CC BY-SA 4.0, via Wikimedia Commons.

Frequently Asked Questions

Is this a partisan issue?

No — it's a data-literacy issue, and it cuts in whichever direction the person making the claim wants it to cut. We've shown examples in this series of both "ICE is arresting mostly non-criminals" and "70% are violent criminals" framings that overstated what the underlying data actually shows once the denominator is made explicit.

Which number should I trust?

All of the individual numbers in this series are sourced and, as far as we can verify, accurate. The goal isn't to pick a "correct" one among them — it's to always ask what population and time period a given percentage describes before comparing it to a different claim built on a different population or time period.

Does TRAC or DHS get this right?

Both organizations publish real numbers; both also get selectively quoted by advocates on all sides who choose whichever framing supports their argument. The responsibility for checking the denominator sits with whoever is citing the number, not just the agency publishing it.

Keeping Stolen Land Pure Since 2003 anti-ICE warning-sign sticker

"Keeping Stolen Land 'Pure' Since 2003" Sticker

For the pro immigrant reader who's made it through all six pieces in this series and wants a sticker that rewards close reading, this design's a good fit — and it's one more entry in our full Anti-Ice Stickers collection.

➡ View the Sticker

➡ Browse the Anti-ICE Stickers Collection

Every Abolish Ice sticker and magnet purchase from our Anti-ICE Stickers collection helps fund the reporting and continued development of this public resource.