History
The Role of Political Polling in Elections
Quick fact
A properly conducted poll of just 1,000 randomly selected respondents can represent the views of over 300 million people with a margin of error of roughly ±3 percentage points——the same accuracy you would get from surveying tens of thousands of people if done carefully.
Why this is interesting
Every election cycle, we're bombarded with polls——but why do they sometimes get it so wrong? A poll that surveys only 1,000 people is supposed to represent millions of voters——how can that possibly work?
Read the full explanation
Understanding The Role of Political Polling in Elections
Think of a poll as a recipe tester: you don't need to eat the whole pot of soup to know if it's salty——you taste just one spoonful, assuming the soup is well stirred. Similarly, a poll takes a small sample of the population, but only if that sample is representative——mirroring the demographics and opinions of the whole. The key steps are: define the target population (e.g., likely voters), randomly select a subset, ask consistent questions, and then weight the responses to match known population characteristics like age, gender, and education. Random selection is crucial: if some groups are more or less likely to be included, the sample becomes biased, like tasting only the top layer of the soup where the salt hasn't mixed. The result is a snapshot of public opinion at that moment, complete with a margin of error that tells you how much the result might differ from the true national value.
A deeper explanation
The power of polling comes from the law of large numbers and probability theory. When you randomly sample from a population, the sample average will be close to the population average, with the spread (standard error) decreasing as sample size increases and as the population variance decreases. The margin of error is roughly 1/√n times a constant, so a sample of 1,000 gives about ±3%. But this assumes a perfect random sample——a huge challenge in practice. Many polls now use opt-in internet panels, which are not truly random, and thus rely on statistical weighting to correct for biases. Still, non-sampling errors like question wording, timing, and respondent honesty can affect accuracy. For example, the 'shy Tory factor' in UK elections or the 'Bradley effect' in US races show that some voters may not admit support for certain candidates. That's why polls are not predictions; they are snapshots that can shift. Understanding the mechanics helps you critically evaluate polls: check the sample source, sample size, margin of error, and question wording, and remember that the margin of error applies to each candidate's percentage, not just the lead. Polls are valuable for showing trends and public priorities, but they are one input among many in an election season.