Political Polling

Probability · Axiom Academy

How can 1,000 people represent 300 million voters? Every election season, we see polls claiming to predict voting outcomes for entire countries based on surveys of just 1,000 to 2,000 people. On the surface, this seems absurd—how can such a tiny sample represent hundreds of millions of voters? Yet professional pollsters consistently make accurate predictions. The 2020 U.S. presidential election saw national polls with an average error of just 2.3 percentage points, despite sampling less than 0.001% of all voters. The secret lies in the Central Limit Theorem and the mathematics of sample means. Let's explore how sample size affects polling accuracy. Use the slider below to adjust the number of people surveyed, and watch how the margin of error changes. Notice: Doubling the sample size doesn't halve the error—it follows the √n pattern Based on the visualization above, what happens to the margin of error when you increase the sample size from 1,000 to 4,000 people (quadrupling the sample)? The Mathematics Behind Polling The margin of error in political polling comes from the standard error of a proportion. When estimating the percentage of voters who support a candidate, the standard error is: Where p is the true proportion (often estimated as 0.5 for maximum uncertainty) and n is the sample size. For a 95% confidence interval, we multiply by approximately 1.96:

This is the written version of the interactive lesson above. See the full Probability course.