By-elections are the purest version of the problem we keep coming back to: one seat, one town, a low and unpredictable turnout, and a national story trying to write itself onto a very local mood. Clacton-on-Sea is about as sharp a version of that as exists in British politics.
So we synthesized it.
A town, not a nation
Clacton is a point on the map, not a shaded region — a coastal town of ~50k on the Essex coast. That’s deliberate here: the interesting unit is the town itself, and the globe simply flies in and pins it. We built a synthetic electorate matched to the constituency’s profile — older than the national average, more economically anxious, more detached from Westminster — and asked it a by-election’s worth of questions.
What the town said
Three patterns held across runs:
- Turnout is the whole game. Our runs disagree with each other more on who shows up than on who they’d back — a by-election tell if there ever was one.
- National anger, local reasons. Synthetic voters reach for national frustration, then justify it with intensely local specifics: the seafront, the surgery wait, the bus that stopped running.
- “Send a message” beats “pick a government.” Freed from choosing a national government, synthetic Clacton treats the vote as a signal far more than a selection.
“Nobody in Westminster has been to Clacton. That’s the whole point of my vote.” — synthetic voter, Clacton-on-Sea
Why publish it before the count
Same reason as always: a by-election is a small, fast, checkable event. We can state what the simulation believes now and grade it against the declared result. Single seats are where synthetic estimates are hardest — thin data, volatile turnout — which is exactly why they’re worth testing.
The Clacton electorate here is synthetic and the by-election framing is illustrative. Figures describe a simulated population, not a poll of real voters, and none of this is a forecast.