What Do We Learn From Polls?

Posted May 24, 2013

Pollsters are paid by corporations and political parties to discern customer tastes or voter preferences. It can be well-paid work.

The marvelous part is that there is rarely any way to know whether they are right or wrong. If a marketing campaign based on the poll results is unsuccessful, it must have been the campaign that was flawed.

Perhaps the client asked the wrong question. Pepsi easily beats Coke in blind taste tests. In 1985, Coke changed its formula to taste more like Pepsi and launched “New Coke” with much fanfare. Customers rebelled. Coke beat a hasty retreat, and held on to their superior market share. Apparently something other than taste drives customer preference in colas.

Likewise, political polls between elections always make for an interesting news story. Pundits weigh in on the significance of small movements, and what the parties should be doing to improve their position. These polls are usually unreliable as predictors of an actual vote, but that doesn’t seem to matter – pollsters sell their work and news agencies buy it.

The one time that pollsters go under the microscope is when their results are published shortly before an actual vote. The result in the recent British Columbia election thoroughly embarrassed them, not for the first time.

Recent elections saw them wrongly predict a Wild Rose victory in Alberta, widely underestimate Liberal strength in Quebec’s provincial election, and completely miss the federal NDP surge in Quebec.

In BC, Angus Reid told us in April that 67% of voters disapproved of Premier Christy Clark and only 25% approved. NDP leader Adrian Dix had 49% approving and 40% disapproving.

Pollsters (a total of nine) reported a narrowing of the big NDP lead about a week before the vote on May 14, but subsequently released results showing the gap widening again. In fact, the Liberals won a substantial majority.

Pollsters and pundits have offered various theories to explain the outcome. It has been asserted that this proved that negative advertising works, that the NDP’s “flip-flop” on the Kinder Morgan pipeline was the turning point, that the low voter turnout was key (though it was actually 1% higher than the 2009 vote), and that the Liberals did a better job of getting their vote out.

Most interesting of all are those who have tried to explain why there was a late Liberal surge. This of course assumes that the pre-election polls were right when they demonstrably were not.

The most honest and humble reflection came from Eric Grenier. He is an aggregator of all polling results. Because he is not affiliated with any one of them he can be more forthright about their problems.

If an aggregation of polls involving many thousands of voters can be clearly wrong, within days of an actual vote, what does that say about the polls we read about when elections are distant?

The best known political poll in Nova Scotia is that done quarterly by Corporate Research Associates. The most recent results, released on March 6th, were based on 801 phone interviews. And the winner was… undecided.

Of those that made a choice, 162 supported the Liberals, 133 supported the NDP, and 100 supported the Progressive Conservatives.

The number that deserves attention is the 384 who chose not to express a preference, more than double the number supporting any party.

It suggests that, although there are a lot of people who are not happy with the government, many have not yet concluded that they want to vote for someone else.

In fact, Premier Dexter might take small comfort from Angus Reid’s April survey showing his leadership numbers no worse than those of Premier Clark in BC.

Voters are smart. They are reluctant to commit themselves before having a clear idea of what the different options represent. Platforms have not been released and the opposition leaders have not yet had the kind of scrutiny that premiers receive.

Meanwhile, another CRA poll will be released in early June. It will make interesting reading for casual observers and will be obsessed over by political operatives.

It is unlikely to tell us much about what will happen in the next election.

Share

Related Articles

Who Represents You + Show all articles

Reference Material

Post a Comment

Most Recent Comments

  • Rob you are correct that polls are often correct. The point is that we only get to know which ones were right when an actual vote is taken. A poll months before an election where half the respondents are undecided is not a good predictor of what will happen in a vote.

    Bill B

    Bill | May 24, 2013

  • I disagree quite a bit with this post. As far as I can figure out, the main problem with ;polling in recent elections with methodology, and with rapidly changing events. The B. C. Liberal Party polling was pretty well right on. But there were some outstanding developments toward the end of the campaign that, I believe contributed to rapid change: The NDP leader declined to respond/explain his well known involvement in some fishy stuff when he was the Nigel Wright of the last NDP govt in BC. One should respond to attacks. The Conservative collapsed because, like the Wild Rose in Alberta last year, they had a couple of crazy, one of spoke kindly of Adolph Hitler. The Conservative vote then in the latter days of the campaign moved back to the Liberal Party from whence it had come. Also, likely partly because of Mr. Dix’s refusal to respond to the personal attacks, many voters, especially, but not only female female ones, took a second look at the female Premier. Also I think, because Mr. Dix’s pipeline decision, strategically taken to attract Green votes, not only did not do that, it drove some moderate NDP supports to the Liberals. Remember, Campaigns matter. Back to methodology. Some pollsters did cheap polling because their media sponsors did not wish to pay the bucks involved in proper polling. There was excessive of Internet panels, which simply involve going back to the same people time and other Internet type polls. These kinds of polls quite often miss or over rely on certain demographic groups: Regretably, the majority of young people don’t vote, but they do participate internet panels especially, because they are interested in non-political subject these panels deal with. And very older folks, the vast majority of who do vote, are underrepresented in this type of polling.
    In Alberta last year, much the same–looney Wild Rose candidates, A better PC campaign, plus the fact that it was easy for many potential rose folk to
    to go back to the Tories in the closing days, because that’s where they’d been for all of their voting lives. And also public polling companies stopped polling five or six days before the election.
    Again Campaigns matter.

    I don’t enough about the Quebec election to comment intelligently.
    In the federal election of 2011, the NDP surge was caught by the major polling companies in the second half of the campaign. I knew from following the polls that something quite dramatic was happening–didn’t know the exact, but I knew it was big.

    Regarding the local scene, I’ve been following polling for many years, and when I was provincial campaign committees, I had the privilege of working with and learning from Allan Gregg and Peter Butler, the people responsible for our polling–so I have some understanding here. Also the CRA polls have a positive record of being right in Nova Scotia, including in the 2009 election. And as Don Mills has pointed out in recent days, that record has been solid for 25 years. His polls are not Internet polls–they are telephone polls, not robotelephone polls either, but good old fashioned real person to real person telephone conversations. CRA polls have always had a very large undecided number, which seemingly over the years has tended to break the same way as the rest of the vote–I suspect a large portion of the undecided don’t vote. Again a good poll is only a snapshot of the thinking at the time of its taking. Things can change, sometimes very fast. I think the March 6 CRA poll was useful, at least in the sense that it helps explain why there has been no provincial election to date this year. I suspect the June will be even more relevant in this regard. But yes, campaigns matter–and things can change overnight. But I find polling generally a pretty good guide to what folks are thinking. I should also say, that in my experience in Nova Scotia CRA polling results have never differed substantially from the party polling that I have been privy too, admitted that has not been all that recently. I apologize for any typos in the above.

    Rob Smith | May 24, 2013

  • Dief said polls were for dogs. What about for those who pay the bill?

    gordon a.... | May 24, 2013

View all comments