The Blank Cells
A 94 percent vote share looks like a mandate. Usually it's the tail end of a contest that already happened somewhere the ballot never saw.
Take a general election in a district one party has locked up for a generation. The count comes back lopsided: incumbent near unanimous, opponent a write-in or a placeholder nobody funded. Read as a snapshot, that says the voters agreed overwhelmingly. Read as a record, it says something narrower: on this specific day, using this specific ballot, almost nobody bothered to disagree. Those are different claims, and only the second one is what the data actually supports.
The real contest in that district happened months earlier, in the party primary, where several serious candidates fought over a nomination that was functionally the same thing as the seat. That primary had close margins, real turnout, genuine uncertainty about who'd win. It's also the event with the thinnest formal record: fewer official observers, less press, sometimes no separate tally preserved once a winner gets certified. The general election gets the full machinery, poll workers, certified counts, a permanent place in the historical record. The primary, where the outcome was actually settled, gets whatever the local paper ran that week, if anything.
This shape isn't unique to elections. A corporate board votes unanimously to approve a merger. Read naively, that unanimity looks like alignment forged in the room, on the day, by the vote itself. It's closer to the reverse: unanimity is what you get when every director's objections got resolved in private calls before the meeting started, specifically so the room wouldn't have to do that work in public. The board meeting is a ratification ceremony. Its minutes are a faithful record of the ceremony and a blank record of the negotiation, which happened by phone, off the books, with no transcript anyone will ever produce.
Court dockets run the same trick at scale. Most criminal cases end in plea agreements, not trials. A docket showing a case closed by plea looks, formally, like the system did its job: charge, disposition, sentence, done. But a plea is the tail end of a negotiation that happened almost entirely off the docket, in prosecutor's offices and defense consultations that leave no comparable trace. Nobody transcribes the back-and-forth over what charge gets dropped for what admission. The docket captures the moment the negotiation stopped, not the negotiation.
What connects these is a mismatch between what a formal record was built to log and what actually decided the outcome. Records like these are procedural instruments, triggered by an event with defined rules, defined participants, and a defined moment of closure: cast a ballot, hold a vote, enter a plea. They're very good at capturing that moment faithfully. They're correspondingly bad at capturing whatever upstream process determined how the moment would come out, because that process usually has no rules, no defined participants, and no moment anyone was required to write down.
A wall of blank cells, a near-unanimous vote, a settled docket, a 94 percent margin: these are a sign the real fight happened earlier, off the record the institution kept, inside a process the institution was never built to log.
The record is accurate about the day it was watching. It just wasn't watching the day that mattered.