1. Reference Polling Details
It's important to include these basic facts:
- Who conducted the poll
- How many people were sampled
- When were they polled
- What question was asked
How many people were sampled: A statewide poll of 500 people may be statistically representative of the public at large, but 50 is probably too small to be accurate.
When were they polled: It matters if some significant event occured that could skew the results. Candidate "A" may have been in the lead until his sex scandal broke. If the poll was taken before that bombshell information, it may not represent where the race stands currently.
What question was asked: Pollsters, especially those working for groups that want to sway the results, can ask leading questions in order to get the results they want. "Do you support candidate 'A' even if he's involved in a sex scandal?" is an example of how a question can be slanted in order to manipulate the outcome.
2. Know the Margin of Error
The margin of error is also a critical component of any story on a political poll. There are many statistical-related reasons for this, but let's concentrate on the accuracy of your story.
Let's say candidate "A" is leading candidate "B" by two points, 51% to 49%. The poll has a four-point error margin.
It is not accurate to say that "A leads B" without referencing the margin of error. That's because "A" could be anywhere from 47-55% and "B" 45-53%. So it's possible that "B" actually leads "A" in the poll.
To be accurate, you could say the two candidates are statistically tied. Or you could say that "A leads B" by two points, but that the numbers are within the poll's error margin. Just as long as you reference the error margin in some way.
3. Watch Out for Illegitimate Polls
Beware of writing stories about illegitimate polls that have no statistical integrity to back them up. That would include polls on websites that allow multiple votes or would allow a user to delete their computer's cookies to be able to tilt the numbers.
Even if a respected news organization has a poll on its website, that doesn't mean the results will be useful. A website poll that asks, "Do you want to pay more taxes?" may have 100% saying no and 0% saying yes with 50,000 votes. Without careful statistical analysis, the poll is meaningless.
Writing about political polls can be complicated. But it is our responsibility as media pros to give our users the most complete, accurate information that we can without just skipping over these critical details.
