Robert Orben famously mused, “Do you ever get the feeling that the only reason we have elections is to find out if the polls were right?”
In this day of constant polling and the media frenzy caused by the inching of percentages, what focus should there be on the numbers?
“Any one poll probably should not be focused on too much, but the whole series of them as they go along up towards the election can give you a fairly good indication of how things are going to go,” said Nancy Reid, a professor of statistics at the University of Toronto. “Unless you get a real surprise right near the end or something changes quite suddenly[…] then there is a time gap where the polling is not done at the time of that big change.”
The use of statistical methodology maintains the appearance of accuracy. However, a poll sampling 1,000 people results in a built-in margin of error of plus or minus 3.1 percentage points. This means that if a poll places a party’s support at 40 per cent, realistically it is indicative of support that could fall anywhere between 37.9 and 43.1 per cent.
In addition, polls are only considered accurate 19 times out of 20. This means that one out of every 20 polls could be expected to provide results that lie outside the margin of error.
In polling, staticians use samples that accurately reflect the demographic of the country. Though polling companies use random digit dialing—calling landlines at random to get the right number of people—with the widespread use of cell phones, this method is proving ineffective and could skew results.
“People are starting to wonder if [random digit dialing] is going to stick because a lot of younger people just have cell phones, they don’t have a landline anymore. As far as I know they haven’t adapted for that yet,” Reid said.
Errors in polls can also be attributed to the group of voters who refuse to take part. This can result in a sampling bias if the group’s statistics are somehow different.
The National Council on Public Polls Review Board has stated, “In general, the quality of a sample improves the longer the survey is in the field. Surveys conducted on one evening, or even over two days, have more sampling biases—due to non-response and non-availability—than surveys which are in the field for three, four, or five days.”
Accuracy of polls also depends heavily on voters providing truthful answers. According to research conducted on voting behaviour, many people answer less than honestly.
Professor Reid remarked, “This could prove to be a factor in the American elections. It’s been shown people often don’t respond to the polling people honestly on certain questions about race and that can have an impact—especially when you have a black candidate.”
