Quinnipiac reports Donald Trump’s Total Approval is at 38%. But on the same dates Quinnipiac sampled, Rasmussen showed Trump’s Total Approval at 53%, with 37% Strongly Approving.
Why the discrepancy?
Quinnipiac skewed its data by undersampling Republicans and oversampling Democrats.
Polls are not that hard to read, but you do need to work at it.
Regardless of how you feel about the reported result, DON’T JUST TRUST A HEADLINE!
TIPS FOR READING POLLS
Check the dates the poll was actually conducted. If any major event occurred AFTER that date that would have affected how people responded, look for a more recent poll. The earlier one may be useful as a comparison, but it won’t be significant as a measure of current public opinion.
- You can track Trump’s approvals @ http://www.rasmussenreports.com/public_content/politics/trump_administration/trump_approval_index_history
Check if the respondents were randomly selected or self-selected (like an on-line poll). Ignore the latter.
Check the total sample size. Ignore any national poll whose sample is fewer than 1,000.
Check the nature of the sample. A good poll will tell you what percentage of respondents were in significant groups for that poll. E.g., a political poll will list percentages for Democrats, Republicans, and Independents. Google for the current statistics for the nation and ignore the poll if the percentages are way off.
The Quinnipiac poll has Trump’s Total Approval way too low, because its sample — 25% Republicans, 34% Democrats, and 33% Independents — was too far off from our current national breakdown of 28% Republicans, 25% Democrats, and 25% Independents.
Check if the poll reports the MOE (Margin of Error). If it doesn’t, ignore it. If it does, keep in mind that the MOE means whatever number is reported for a specific result is meaningful for the entire population within plus or minus the MOE.
E.g., A 50% result with an MOE of ±1 means that the overall truth of that result is between 49% and 51%.
If the above are good, then dive into the actual questions asked. A good poll will not ask absurd or leading questions and will rotate the order of the answers offered to avoid bias. If the poll does not offer you a look at the questions, ignore it
- TIPS via my graduate school statistics professor.