Posted on 25 January 2017

Writing in iPolitics, Paul Adams criticizes not polls, but reporters who don’t understand polls. Part of it is reporters who ignore the margin of error, reporting on polls showing a “clear winner” rather than a possible winner, but maybe not, because there’s a margin of error:

“If you look at the final polling forecast from Real Clear Politics, it showed Hillary Clinton with a 3.3 percentage point lead in the popular vote over Donald Trump. Guess what? She won the popular vote by almost 3 million votes; that’s just more than 2 per cent.
“An error of just over a percentage point is an error, for sure. But it’s a perfectly unsurprising error. No reasonable observer would expect the national polls to be more reliably accurate than they were.”

I ran into this during the last federal election. A group with the express purpose of getting people to vote strategically to defeat the Conservatives released a poll indicating that the NDP was the favourites to win- unless you looked at the margin of error. As I wrote:

“The more accurate way to read the results is with NDP support somewhere between 32 and 40 percent, the Conservatives between 26 and 34, and Liberals between 25 and 33 (19 times out of 20). The NDP could drop 4 points and the Liberals could go up 4, resulting in a Liberal victory, and it would still be within the margin of error. Or the Conservatives could win. Basically, it’s too close to call.”

For this I had a couple of days of (a very few) people accusing me of being a Liberal or Conservative shill.
I don’t know for sure, but I suspect I may have had less of this if there hadn’t been a bunch of headlines saying the NDP was the favourite to win.
In the end, the Conservative candidate won, followed by the Liberal and then the NDP.
If we must report on polls, let’s at least do it accurately.

Filed under: journalism