Crosstab-to-topline calculator
Poll crosstabs tell you how each group is voting. This turns those rows into one number — the turnout-weighted topline — and shows what margin implies about win probability.
| Subgroup | Margin A−B (pts) | Turnout share (%) |
|---|
Formula: topline = Σ(marginᵢ × shareᵢ) ÷ Σ(shareᵢ)
The math, written out
Shares are normalised, so they need not sum to exactly 100 — the tool divides by the total you entered. The weighted margin is:
M = Σ(mᵢ · wᵢ) / Σ(wᵢ), where mᵢ is a subgroup margin in points and wᵢ its turnout share.
Win probability assumes the final margin is normally distributed around M with standard deviation σ: P(win) = Φ(M / σ), where Φ is the standard normal CDF. A 3-point lead with σ = 3.5 is about an 80% chance; the same lead with σ = 5.25 is about 72%.
Why the range matters more than the point estimate
The single win-probability number is only as good as your σ. Because nobody knows the true polling error before the count, the honest output is a band. If the low and high ends of that band sit on opposite sides of 50%, the race is a coin flip no matter what the topline says.
Common questions
How do I turn crosstab subgroup margins into a topline number?
Multiply each subgroup's margin (candidate A minus candidate B, in points) by that subgroup's share of the expected electorate, then add the products together. The result is the turnout-weighted topline margin. It is the same arithmetic pollsters use when they reweight a sample to a likely-voter model.
Why does the topline change when I edit turnout shares without touching the margins?
The topline is a weighted average, so it moves whenever the weights move, even if every subgroup's preference is frozen. A group that favours one side by 40 points contributes twice as much to the topline at 30% of the electorate as it does at 15%. That is why turnout assumptions, not subgroup swings, drive most of the disagreement between pollsters in the same week.
How is the win probability range estimated here?
The tool treats the weighted margin as a normal distribution centred on your number with a standard deviation you set (the polling-error sigma, typically 3 to 5 points for a general election). Win probability is the chance that distribution lands above zero. The range shown uses sigma and sigma times 1.5 as low and high error assumptions, because the true error is itself unknown.