Ten things journalists should find out and report about polls they are covering
1. Is this poll really a poll?
Some unscrupulous campaigns and advocacy groups conduct “push polls,” which are not polls at all. Rather than aiming to tally people’s opinions, they actively seek to change people’s opinions about issues or individuals. One clue: these efforts typically fail to ask any demographic information.
2. Who sponsored the poll and who conducted it?
Assuming you are not an expert, ask some pros for an assessment of the sponsor’s reputation. At a minimum, include the name of the sponsor in your story, to hold them responsible for the work they are backing.
3. Who was the target population?
This gives important context for interpreting results. For example: a poll of likely voters, or a survey of U.S. teens ages 13 to 17.
4. How many individuals were sampled, and where?
Location is important for context, and larger sample sizes help ensure—but by no means guarantee—more reliable results.
5. How were the interviews collected?
Methodology can point to the representativeness of the sample. For example: a poll conducted by landline and cellular telephone, or interviews were conducted online and by telephone.
6. When was the poll conducted?
The date is important for interpreting results, especially in politics or other fast-changing landscapes. For example: Interviews were conducted September 15 to November 8, 2019.
7. What was the margin of sampling error?
Poll results aren’t complete without information about the uncertainty and range of plausible results. For example: The poll had a margin of sampling error of +/- 6.0 percentage points (which means that the true results are anywhere within six percentage points on either side of the given results).
8. Was there weighting?
If so, on what? For example: Results were weighted to ensure that responses accurately reflect the population’s characteristics in factors such as age, sex, race, education, and phone use.
9. What language was used?
This hints to the effort made in collecting a diverse sample. For example: The poll was conducted in English and Spanish.
10. Consider also reminding readers of reasons why polls may not perfectly reflect reality.
For example: There are many potential sources of error in polls, including the use of charged wording and the order in which questions appear.
Excerpted from Surveys and Polling, compiled by SciLine and the American Statistical Association.