Election analysis focuses on who people vote for — party affiliation, candidate quality, campaign spending, media coverage. The who-shows-up question is treated as a contextual factor, not a predictive variable. Turnout is the denominator, not the signal.
Pal, Kumar, and Santhanam (arXiv:2501.01896) show that voter turnout alone predicts key electoral statistics — the distribution of votes won by victors and runners-up, the margins of victory — with remarkable accuracy. Using a random voting model and decades of Indian election data across multiple geographic scales, they derive theoretical distributions that match observations when conditioned on turnout. The margin-to-turnout ratio exhibits scale-invariant behavior across constituencies of different sizes.
The structural point is that the participation rate carries more information than the individual votes do, once aggregated. A random voting model — where each person who shows up votes essentially at random — reproduces the statistical shape of real election outcomes when the turnout distribution is correct. The choices average out. The showing-up doesn't.
This doesn't mean individual votes don't matter. It means that at the population level, the statistical mechanics of elections is governed by the denominator. How many people are in the room determines the shape of the outcome distribution more reliably than what any of them think. The collective behavior emerges not from the content of the decisions but from the decision to participate at all.
The signal was in the participation, not the preference.