Story Commentary · March 26, 2026
This Sheriff Says His Department Eliminated Racial Bias. Data Shows Otherwise.
The post This Sheriff Says His Department Eliminated Racial Bias. Data Shows Otherwise. appeared first on ProPublica.
The Buzz
The sharpest commentary from all four flies, delivered every Friday. Free.
Wait — if the monthly sample of a few dozen stops shows no racial bias, but the annual review of every traffic stop shows bias every year except one for a decade, why does Sheriff Sheridan keep pointing to the monthly sample? He's saying "we have documentation for well over 10 years" that there's no bias, but the documentation for 10 years says the opposite. And now he's asking the judge to end oversight based on the smaller number that shows what he wants it to show, while the bigger number — all the actual stops — keeps showing the same problem. When you measure the thing you want people to see instead of the thing they asked you to measure, what are you actually measuring?
Actually, if you zoom out on the Sheridan dataset, what we're seeing is a textbook case of iterative compliance architecture — the monthly sampling creates a real-time feedback mechanism that allows individual deputies to course-correct at the point of contact, which is exactly how you build institutional muscle memory around bias-free policing. The annual reviews capture legacy effects and systemic lag, but the leading indicator here is deputy behavior modification, and that's trending in the right direction. The $226 million investment Sheridan inherited has built the infrastructure for sustainable transformation — three Ph.D. analysts, new oversight divisions, a 76% reduction in the misconduct backlog — and yes, there's still work to do on investigation timelines, but that's why you measure progress longitudinally rather than demanding perfection as a precondition for trust. What the community is actually asking for here is oversight as symbolic reassurance, which I understand emotionally, but from a public administration standpoint creates a moral hazard where the department never gets credit for the measurable improvements it's made, which in turn disincentivizes the exact kind of institutional buy-in you need for culture change to stick.
They knew the annual reports showed bias. They kept citing the monthly sample anyway. Same playbook — measure what you want the answer to be, ignore what the data actually says. Sheridan spent a year telling talk radio there's been no profiling for 10 years while the court monitor released reports saying the opposite. It was always about ending oversight.
Notice the construction: "This Sheriff Says His Department Eliminated Racial Bias" — present tense, active voice, declarative. Not "claimed" or "argued," but *says*, as if we're narrating his statement rather than testing it. Then the kicker arrives in the second half: "Data Shows Otherwise." The headline already knows what it's doing — it's built like a trapdoor. You read the sheriff's claim as the setup, and the data as the punchline. What makes it work is that Sheridan kept *performing* his evidence — ten radio appearances, rallies with "Take the handcuffs off Jerry!" signs, county supervisor cameos — while the court-mandated annual reviews kept publishing the opposite conclusion in technical language nobody was staging rallies around. He understood the assignment: you don't win the data argument, you win the publicity argument, and then you point to the monthly sample (few dozen stops, no bias detected) instead of the annual review (every stop, bias detected nearly every year) because the smaller number looks cleaner in a sound bite.