Caltech Logo
Caltech Science Exchange wordmark
How Political Polls Work cover with people taking a poll
Caltech Science Exchange  /  Topics  /  Voting and Elections  /  Polling

How Do Political Polls Work?

This article was reviewed by a member of Caltech's Faculty.

Political polls provide insights into to the preferences and beliefs of certain groups of people, often voters in a specific country, state, or district. Polling results can inform policy, track the likely outcomes of elections, and help candidates and government officials decide where to invest time, campaign funds, and other resources. Furthermore, some experts argue that the ability of independent organizations to conduct and report the results of polls, free from government interference, is emblematic of a free and fair society.

Because posing questions to every member of a community would be costly and logistically difficult, pollsters instead survey a smaller, but representative subset of a population, known as a sample. A high-quality sample includes all types of people that make up the population. While larger samples are generally more precise, a sample size of just 1,000 to 1,500 people can be enough to estimate national opinion in the United States with a high level of accuracy.

One method used to build a representative sample is random sampling, which gives everyone in the population an equal chance of being included or not. Pollsters typically draw a random sample of participants from computer-generated lists of landline and cell phone numbers (Increasingly, polls are also being conducted online). However, this alone does not ensure a random sample because some groups of people are likelier than others to have a phone or participate in polls. To avoid overrepresenting some groups and underrepresenting others, factors such as age, race, gender, educational attainment, or region are given more or less weight in the final polling results so the sample more closely matches the demographics of the full population.

How do you know if a poll is trustworthy? Credible polls report their margin of error, which shows how much the result of a poll could vary from reality. Other factors to consider are the poll's funders, the pollster's sampling methods, and the specific questions participants were asked. Also keep in mind that even if a poll relies on a good national sample, it can be difficult to generate precise results about smaller subareas or subgroups. Polling firms who participate in industry transparency initiatives have been found to be more accurate than those that do not.

Polls make it possible to gauge public opinion by surveying a relatively small but representative group of people within a given population. This smaller group is known as a sample.

Think about a bowl of soup. You don't need to eat the whole bown to know how the soup tastes, just a representative spoonful. Similarly, pollsters don't need to talk to every voter to measure national attitudes. Instead, using principles of probability and statistics, they can base their estimates on a sample, or selection, of voters.

One way of building a representative sample is to take a simple random sample in which everyone in the population has an equal chance of being included or not. While larger samples are generally more precise, a sample size of just 1,000 to 1,500 people can be enough to estimate national opinion in the U.S. with a high level of accuracy.

Although some polls are conducted online, pollsters typically draw on computer-generated lists of landline and cell phone numbers. But that isn't enough to ensure a representative sample because some groups are likelier than others to participate in polls.

To avoid overrepresenting these groups and underrepresenting others, pollsters use a statistical method called weighting: factors such as age, race, gender, educational attainment, or region are given more or less weight in the final analysis so that the sample more closely matches the demographics of the full population.

So how should you read a poll? Don't forget the fine print. Credible polls report a margin of error. This figure shows how much a polling result may vary from reality.

Case in point: A poll shows 51% of voters prefer Candidate A while 49% prefer Candidate B, with a margin of error of 4%. We can't say for certain that Candidate A is leading. That's because the margin of error indicates the actual percentage of voters who prefer Candidate A is most likely between 47% and 55%, while the percentage who prefer Candidate B is between 45% and 53%. (Nonetheless, assuming the poll was well conducted, we can say Candidate A is probably leading.)

Besides the size and quality of the sample, other factors that influence the accuracy of a poll include . . . how the questions are phrased, the order in which questions are asked, and whether people respond honestly.

Read More
Learn about other threats to election security and how they can be addressed on the
Caltech Science Exchange.

Dive Deeper

Sam Wang teaching at a Whiteboard
Why I Ate a Bug
Read Techer Article   ＞
Erin Hartman poses in front of a wall
Knowing the Vote
Read Techer Article   ＞
One small error in voter calculations
5 Key Things to Know About the Margin of Error in Election Polls
Read the Article   ＞