Polling data is a bedrock of news reporting, from politics and health to business and sports.
Polls are also complex, which makes it easy to report on them in (unintentionally) misleading ways. This list of “best practices” is intended to help make sure numbers are used correctly and in context.
You’ll notice, we focus on how to talk about polls, not how to spot a bad one, although we recommend you spend time on that skill, too. Our emphasis is on language and communicative techniques.
Avoid reporting only results. Talk about what respondents were asked, the answers they could choose from, and who conducted and paid for the poll.
Many news organizations provide margin of error (MOE) values alongside polling. But it’s also important to define this term, to provide specific MOE data for different subgroups (not just the whole sample) when detailing their responses, and to use qualifying language that highlights imprecision. Also, when possible, visualize MOE values.
Instead of just discussing what a poll tells us, talk about what it doesn’t tell us. Talk about when polls were conducted, the response rates, and the percentage of respondents who chose “don’t know” or “not sure” answers. Also, when possible, compare results to those from other polls, and explain why these might differ.
When relaying statistical information, don’t jump from one representation (like a percentage) to another (like a fraction). Also, use absolute numbers when making comparisons. Lastly, tell audiences when differences are meaningful, and when they’re not.
Please help the project team learn how this content could be improved by answering some brief questions.
Take the survey.
- Explain the methods