Census Bureau Proposes Excluding Immigrants Without Green Cards From Apportionment Counts and Dropping Race Questions

Two Numbers, One Rulebook, and a House Seat Nobody Can Agree How to Count

Every ten years, the Constitution asks the federal government to do one simple-sounding thing: count everyone living in the United States, then use that number to divide up the House of Representatives. On Wednesday, Sept. 9, 2026, the Commerce Department posted a new answer to a question that sounds settled but isn't — who, exactly, counts as "living" here.

The proposed rule, signed by Commerce Secretary Howard Lutnick, is set to appear in the Federal Register on Sept. 10^[2][17]. It would count U.S. citizens and green-card holders at their home address, the way the census always has. But other foreign nationals — people on work visas like H-1B, foreign students, and, in the version filed this week, approved refugees and asylees — would not be counted there at all for the purpose of divvying up House seats^[6][1][3]. A second, separate part of the same document would drop race, ethnicity and sexual-orientation questions from the census form entirely^[11].

Nothing changes yet. This is a proposed rule at the start of a 30-day public comment period, and it would first apply to the 2030 census, not any election happening in 2026^[2]. But the fight over it is already loud, because the number of seats in the House has been frozen at 435 since 1929^[16]. That means this isn't a bookkeeping tweak. It's a fight over a fixed pie, and whatever one state gains, another state loses.

A 236-Year-Old Habit, and the One Word Both Sides Agree On

Since the first census in 1790, the United States has counted every resident toward apportionment, regardless of immigration status. No one has ever been excluded for that reason^[1][10]. That's not a contested talking point — it's the undisputed starting point for everyone in this fight.

Here's the odd part: both sides also agree on the actual constitutional text. The Constitution doesn't say "citizens" get apportioned — it says "persons." The 14th Amendment repeats the same word^[10]. So the fight isn't really about what the document says. It's about a second word that isn't defined anywhere in it: "residence."

The administration's argument leans on that gap. The Constitution requires an "actual Enumeration," but it never spells out what counts as someone's usual home. Tourists and foreign diplomats have never been counted, so the government has always drawn some line^[3][13]. Lutnick's rule tries to draw that line further along the same spectrum — arguing that someone here on a temporary visa, or without legal status at all, doesn't have the kind of settled home life a citizen or green-card holder does.

Opponents say that's a rewrite dressed up as a definition. If "persons" means persons, they argue, the executive can't use a residency technicality to quietly exclude millions of people the text was written to include^[10]. The Supreme Court has actually had a version of this exact question before it, and ducked it. In 2020's Trump v. New York, the justices dismissed a similar challenge on procedural grounds — the case was filed too early — without ever ruling on the underlying constitutional question^[9][1]. So both sides are arguing from a genuinely open legal record, not settled law.

Why a Data Question Turns Into a Courtroom Question

The second part of the rule — dropping race, ethnicity and sexual-orientation questions — sounds like a separate, smaller story. Lutnick's stated reasoning is that the census's core constitutional job is counting people for apportionment, and that demographic questions add "response burden" and "risk of disclosure" without serving that core job^[3][4].

That argument has real weight on its own terms: fewer questions on a form can mean more people willing to fill it out, and less risk that someone could be identified from published data^[4]. But the change has a specific downstream effect that has nothing to do with response rates. To win a Voting Rights Act lawsuit over racial vote dilution, plaintiffs have to prove a minority group is large and geographically compact enough to form a majority in a voting district. That proof comes from census race data broken down block by block^[12].

Take away the race question, and that evidence disappears — regardless of whether the Voting Rights Act itself changes at all. It's a structural effect that runs through statistics, not through legislation. That's why voting-rights litigators see this provision as inseparable from the residency change, even though the two are, on paper, two different regulations bundled into one filing^[11].

The States That Don't Fit the Story

Coverage of this fight tends to sort into two teams, but the actual math scrambles that. The most-cited estimate comes from Pew Research Center, using 2020 data: removing unauthorized immigrants from the apportionment count would have shifted three House seats. California would have lost two seats instead of one. Texas would have gained two instead of three. Florida would have gained one instead of two. Alabama, Minnesota and Ohio would have been the winners^[7].

Notice what that means: Texas and Florida, both governed by Republicans, would have ended up with fewer new seats than they got under the current rules — not more. Their leaders tend to support the policy on principle, that representation should track people the government has actually admitted to stay. But their state governments also have a concrete interest in the biggest possible population count, since federal funding for health care, transportation and schools flows from those same numbers^[6][7]. National party math and individual state math are pulling in different directions here, which is easy to miss if you only look at who's cheering and who's suing.

The other twist: this isn't only about people without legal status. The rule's language reaches H-1B visa holders, foreign students, and now refugees and asylees who were approved to be here^[5][12]. That last detail actually shifted between drafts. Reporting on an August version of the rule said approved refugees and asylees would still be counted. The version filed this week excludes them^[11][3]. That's a real, documented change in scope, not a misreading by either side.

What Each Side Is Actually Afraid Of

Strip away the rhetoric and each side is defending something concrete. The administration's brookings-adjacent worry is political: if a state hosts a large temporary or unauthorized population, its residents get more House seats and more Electoral College votes than the same number of people in a state without that population^[19]. To supporters, that's a state getting extra political power from people it hasn't admitted to stay permanently.

Opponents' worry is representational: House members serve everyone living in their district, not just the people who can vote — kids in public schools, people who pay sales tax, anyone who calls 911^[10]. Brookings demographer William Frey put a sharper point on the stakes, saying the administration is "trying to change the idea of who is American," and warning that fear of the rule alone could scare people away from filling out the census — even if courts eventually strike it down^[6]. That's a real, separate risk: a chilling effect on response rates doesn't require the rule to survive, only for people to be afraid of it.

Then there's the Census Bureau itself, caught in the middle. Career statisticians and everyone who uses census data — school districts, public health agencies, redistricting mapmakers on both sides — care most about whether the numbers stay usable at all^[12]. Fewer questions can mean higher response rates. But a rule that makes any group more afraid of the form can also mean the Bureau has to spend more money chasing down people who didn't answer, making the count less accurate in exactly the places it matters most^[1][6].

Same Filing, Very Different Headlines

How outlets covered this split largely along the lines you'd expect, though the details of the split are worth knowing. The Washington Times led with the questions being dropped and used "illegal immigrants" in its headline — a description that's accurate for part of the rule but leaves out that it also covers legal visa holders, students, refugees and asylees^[4]. RedState called it "one more try" after the 2020 defeat, treating the legal question as though it's already settled in the administration's favor^[13].

NPR and NBC News led instead with how far the rule reaches and the break from 236 years of precedent, using words like "sweeping" and "huge swath" — accurate on scope, but placing the political consequence in the headline before the administration's stated rationale appears^[1][5]. Talking Points Memo went further, framing the story as reshaping "national political power," which states the contested outcome as if it were the rule's purpose^[12]. Al Jazeera filed the story under its "US Midterm Elections 2026" section, even though the rule can't affect any 2026 race — it targets the 2030 census^[6]. Roll Call's coverage was the most procedural of the group, hedging with "could" and focusing on the comment period and expected litigation rather than either side's moral case^[3].

None of that changes what's actually on the table: a proposed rule, a 30-day clock that starts Sept. 10, and a legal question about the word "residence" that the Supreme Court has already ducked once^[2][9]. Both sides expect this to end up back in court, and this time, a judge may not be able to duck it twice^[3].