The Trump administration’s plan to let nearly two million American employers stop filing annual EEO-1 demographic reports isn’t a strike against “racial bean-counting.” It’s a strike against the only tool that ever caught systemic hiring discrimination at scale. Inez Feltscher Stepman argues in Trump Civil Rights Agency Strikes a Blow Against Racial Bean-Counting for National Review that the EEOC’s workforce form is wasteful box-checking and that ending it just removes a “malincentive” to discriminate. She treats the thermometer as the disease.
Pull the battery on a smoke detector and the chirping stops. The fire doesn’t.
Stepman points to the Biden-era EEOC lawsuit against Sheetz — a suit over a felony-conviction screen that disproportionately excluded Black and Native American applicants — and treats the lawsuit as the injustice. But the injustice was the filter. A blanket ban on any applicant with a felony conviction, including old, minor, unrelated offenses, will mechanically knock out Black applicants at a much higher rate than white applicants, because Black Americans are arrested and convicted at higher rates for the same conduct. The legal question isn’t whether the criterion is “common sense.” The legal question is whether the criterion is doing the same excluding work a “whites only” hiring sign would have done in 1964. A hiring screen that falls hardest on Black applicants isn’t race-neutral just because the words “Black applicants” never appear in the help-wanted ad. Without demographic data, that filter runs invisibly, year after year. The people it disproportionately turns away never learn they were turned away for that reason. They never file a complaint. The pattern continues.
Stepman also leans on Thomas Sowell to argue that mass-level group differences are “a basic fact of the universe” and shouldn’t be treated as suspicious. Averages do differ — that’s not the dispute. The dispute is what to do about a specific employer’s screen that produces a specific gap. She’s describing a legal standard that doesn’t exist. The Supreme Court’s disparate-impact doctrine has never been “any statistical gap equals automatic liability.” The employer has a real defense: show that the screen is job-related and consistent with business necessity. The EEOC has lost plenty of these cases. The data isn’t a verdict. It’s the evidence that lets a verdict be reached at all — and the absence of evidence is what the administration is quietly engineering.
Stepman is right about one thing, and it’s the thing defenders of the EEO-1 least like to admit. The form has gotten heavier and dumber over the years. The “sight identification” rule — asking managers to eyeball whether an applicant is Pakistani or Indian when the worker refuses to self-identify — is genuinely demeaning and incoherent. The compliance cost is real money: roughly $4 million in taxpayer spending and $275 million in private-sector costs every year, borne disproportionately by small and mid-sized employers. None of that is a reason to scrap the entire reporting system. It’s a reason to redesign it.
What we should build instead is a Civil Rights enforcement regime that catches the patterns EEO-1 was built to catch — without the compliance drag Stepman fairly objects to. Move the data out of HR’s hands and into a third-party statistical agency the employer never sees, where workers self-identify voluntarily with a real opt-out, ending the sight-identification ritual entirely. Shrink the form to a small number of job categories, with race and sex as the only fields collected. Publish industry-level patterns so regulators and workers can see disparities when they’re real, and only flag individual employers when a pattern crosses a defined threshold and an actual investigation is warranted. That’s not bean-counting. It’s a smoke detector that works.
Stepman treats aggregate data as a gift to plaintiffs’ lawyers. She’s half right — plaintiffs’ lawyers do use it. But the alternative isn’t a clean meritocracy where every worker is judged on their merits. The alternative is a workplace where the only bias anyone can prove is the bias a single worker happened to document, and where the patterns that touch millions of workers a year become invisible because we decided the thermometer was the problem.
The harm isn’t the spreadsheet. The harm is what happens when we pull the battery and the fire has all night to spread.